Fig 1.
Unbiased search for lipid droplet-associated lipases involved in HCV assembly.
(a) Venn diagram of metabolic serine hydrolases [32] and the Huh-7 lipid droplet proteome [38]. (b) mRNA expression of these genes was determined in primary human hepatocytes (PHHs) from 3 donors by RNA-seq (NCBI database, GEO accession number GSE132548, (Tegtmeyer B, Vieyres G, submitted for publication)). The vertical dotted line indicates an arbitrary expression threshold of 0.5 RPKM (reads per kilobase per million mapped reads). (c) Venn diagram depicting the fraction of proteins from (a), whose lipid droplet association varies upon HCV infection according to Rösch et al. [39] (purple circle). Arrows pointing upwards, arrows pointing downwards and the equal sign indicate respectively an increased, decreased or unchanged lipid droplet association upon HCV infection [39]. (d) ATGL mRNA expression in a panel of cell lines and in PHHs from 8 individual donors, as determined by qRT-PCR. (e) ABHD5 and ATGL protein expression in the same cell lines and in PHHs from 3 random donors.
Fig 2.
ATGL and G0S2 colocalize at the lipid droplet surface.
ATGL and G0S2 were expressed by lentiviral transduction and detected by indirect immunofluorescence using antibodies against the respective epitope tag. Colocalization between the proteins or with the lipid droplet markers (BODIPY 493/503 or AUTOdot [42]) was tested in Lunet N hCD81 cells cultured with or without oleic acid (OA), as indicated. (a) Subcellular localization of HA-tagged ATGL WT or S47A relative to lipid droplets. (b) Subcellular localization of Flag-tagged G0S2 relative to lipid droplets. Note that we had to co-transduce ATGL (here HA-tagged ATGL, which was left unstained) to achieve a robust and reproducible G0S2 detection. (c) Subcellular localization of Flag-tagged G0S2 relative to HA-tagged ATGL and lipid droplets. The plots on the right side indicate the intensity profiles in the different channels along the red dotted line depicted in the merge picture.
Fig 3.
ABHD5 and ATGL colocalize at the lipid droplet surface.
ABHD5 and ATGL S47A were expressed by lentiviral transduction and detected by indirect immunofluorescence using antibodies against the respective epitope tag or by their auto-fluorescence in case of the ABHD5-mCitrine fusion protein. Colocalization between the proteins or with the lipid droplet marker AUTOdot was tested in naive (a) or HCV-infected (b) Lunet N hCD81 cells after oleic acid (OA) induction. (a) Subcellular localization of HA-tagged ATGL S47A relative to ABHD5-mCitrine (mC) and lipid droplets. (b) Subcellular localization of HA-tagged ABHD5 relative to ATGL S47A and lipid droplets, in HCV-infected cells. HCV infection was verified by immunostaining against HCV E2 glycoprotein. In (a) and (b) the plots on the right side indicate the intensity profiles in the different channels along the red dotted line depicted in the merge picture.
Fig 4.
Endogenous or ectopically expressed ATGL controls lipid droplet lipolysis and HCV assembly.
(a) Western blot verification of ABHD5, ATGL and G0S2 expression 72 hours after lentiviral transduction in Lunet N hCD81 cells. Proteins were detected with antibodies directed against ATGL, or the HA and Flag epitopes. For ATGL detection, the endogenous protein can be seen in the high contrast picture (lanes 1 and 2). (b) Effect of ABHD5, ATGL or G0S2 expression on HCV assembly and release. HCV production was determined in a whole replication cycle assay with the JcR2a virus in Lunet N hCD81 Fluc cells and normalized for replication (n = 6 for left panel, n = 7 for right panel). (c) Effect of ABHD5, ATGL or G0S2 expression on lipid droplet lipolysis. Lipid droplet content was measured by flow cytometry 72 hours post-lentiviral transduction in Lunet N hCD81 cells (n = 4 for all conditions except HA-tagged ATGL constructs and Flag-tagged G0S2 where n = 3). (d) Correlation between HCV production and lipid droplet lipolysis. This graph gathers the results plotted in panels b and c as well as the linear regression (full line), 95% confidence band (dotted lines) and r2 value. Data in panels b, c and d were normalized to empty vector- transduced cells. (e) Effect of ABHD5, ATGL or G0S2 expression on lipid droplet lipolysis in naive, bystanders and HCV-infected cells. Lunet N hCD81 cells were lentivirally transduced with the constructs indicated on the X axis and infected 48 hours later with HCV Jc1. Two days later, the cells were harvested, mixed with mRuby2-expressing reference cells and fixed, as explained in S2A Fig. In this case however, the cells were permeabilized and stained with an antibody against HCV NS5A to determine their infection status, in addition to the lipid droplet dye BODIPY. Naive, bystanders and HCV-infected cells were gated as depicted in S3B Fig. Their lipid droplet content was measured and normalized for the reference cell population, to correct for staining or measurement variations. For each condition, the results were then normalized to the respective empty vector-transduced cell population (n = 3).
Fig 5.
ATGL knockdown induces lipid droplet accumulation and reduces HCV assembly.
(a, b) Effect of pools of 3 siRNAs targeting ApoE or ABHD5 or 4 siRNAs targeting ATGL (SMARTpool) on HCV replication cycle (n = 3 for 72 hours post-electroporation (h.p.e.) and n = 2 for 96 h.p.e.). Lunet N hCD81 FLuc cells were electroporated with HCV RNA and transfected 4 hours later with the siRNA pools. Panel a represents the effects on the early replication events (RLUs in the producer cells). Panel b depicts the effects on HCV production (RLUs in the target cells normalized by RLUs in the producer cells). For the statistics, we compared the results obtained with the different siRNAs to the average of the negative siRNAs 1 and 2. (c-g) Effect of single siRNAs targeting ATGL (si ATGL a, b and c) on ATGL expression, lipid droplet content and HCV replication cycle. (c) ATGL mRNA expression was assessed 96 hours (h) post-transfection by qRT-PCR and normalized to GAPDH (n = 2). (d) ATGL protein expression was verified by Western blot. The green and red arrowheads point at the ATGL and β-actin protein bands. (e) The lipid droplet content of cells transfected with ATGL-specific or control siRNAs was evaluated by flow cytometry at 24-48-72-96 hours post-siRNA transfection. (f, g) HCV JcR2a whole replication cycle was examined in the siRNA-transfected cells. (f) Entry and replication correspond to the RLuc reading in the producer cells at 48 h.p.i. normalized for the FLuc readings at the same time point, which reflect the cell viability and proliferation (n = 3, except SMARTpool where n = 2). (g) Assembly and release values are obtained by normalizing the RLuc readings in the target cells (infected with the 96 h.p.e. supernatants) by the RLuc readings in the producer cells at 48 h.p.e. (n = 4, except SMARTpool where n = 3).
Fig 6.
WT HCV also depends on ATGL-mediated lipid droplet.
Lunet N hCD81 cells were lentivirally transduced with an ABHD5-targeting or irrelevant shRNA and infected 72 hours later with HCV Jc1. We harvested 48 hours later the cell lysates and supernatants (n = 5). (a) Intracellular HCV RNA was quantified from the cell lysates by qRT-PCR and normalized to the GAPDH housekeeping gene. (b) Extracellular HCV RNA was quantified from the cell culture supernatants by qRT-PCR. (c) Released infectivity was assessed by transferring the supernatants onto Huh-7.5 cells, culturing the cells for 48 hours and immunostaining them for HCV NS5A. We quantified the infected foci automatically as described in the Methods section.
Fig 7.
Pathogenic variants of ATGL do not support HCV assembly nor lipid droplet lipolysis.
(a) Domain organization of human ATGL and description of the mutants used in this study. The patatin domain is conserved in a large family of hydrolases found in eukaryotes and bacteria [32] and is included in an α/β hydrolase domain. It comprises the two putative residues of the catalytic diad (S47 in the conserved GXSXG sequence and D166). Coding polymorphisms associated with NLSDM [35] and tested in this study are indicated in pink. Q289X corresponds to a premature stop codon. The T370D mutation is a phosphomimetics described to eliminate lipid droplet association of the mouse ATGL (T372D) without compromising the hydrolase activity [43]. (b) Western blot verification of the protein expression of the various ATGL variants in Lunet N hCD81 Fluc cells 7 days post-lentiviral transduction, at the time when HCV replication and production were measured. The full red arrowhead indicates the full-length HA-tagged ATGL whereas the empty red arrowhead points at the Q289X truncation mutant. Note that the upper band (around 56 kDa for the untagged ATGL) corresponds to the reported molecular weight of ATGL [44]. The second lower band could be a cleavage product of ATGL corresponding to the N-terminal part of the protein, since it is detected by the anti-HA antibody. We also observed this second band when staining the endogenous protein (see empty vector as well as Fig 4B, band just over the β-actin (42 kDa)). Although this band was to our knowledge not described from the literature, a fragment of around the same size could be seen in [45] (see Fig 1C for instance) as well as in the datasheets of several commercial anti-ATGL antibodies. (c) Cellular lipid droplet content in Lunet N hCD81 cells upon ectopic expression of untagged or HA-tagged ATGL variants, as measured by BODIPY 493/503 staining and flow cytometry (n = 3). (d) HCV assembly and release in Lunet N hCD81 Fluc cells upon ATGL knockdown with si ATGL a, b or c and rescue of ATGL expression with siRNA-resistant tagged or untagged ATGL variants. The horizontal dotted line corresponds to the control (neg. siRNA and empty vector). Statistics in grey highlight the knockdown effects and test differences between the different knockdowns (si ATGL a/b/c + empty vector, in grey) and the control (black bar). Statistics in black correspond to the rescue experiment and show differences to the respective grey bar (respective siRNA + empty vector) (n = 8). (e) Correlation between HCV production and lipid droplet lipolysis. This graph gathers the results plotted in panels c and d as well as the linear regression (full line), 95% confidence band (dotted lines) and r2 value. Note that the lipid droplet content is tested in an over-expression assay (X axis), whereas HCV assembly and release are assessed in the context of the rescue of an ATGL knockdown (Y axis). This is why the empty vector is set at 100% on the X axis, but at a lower value for the Y axis (presence of ATGL siRNA). (c, d, e) WT ATGL sequences (tagged or untagged) are indicated in dark red, the catalytic site mutants are in green, the clinical variants in pink and the phosphomimetics T370D mutant in light blue.
Fig 8.
Subcellular localization and lipid droplet association of ATGL variants.
(a) Immunofluorescence analysis of HA-tagged ATGL localization respectively to the lipid droplets (BODIPY 493/503) and trans-Golgi (p230 marker) in Lunet N hCD81 cells. (b) Intensity profiles of the green and red channels along the dotted line depicted in the merge pictures of panel a. (c) Quantification of ATGL accumulation in the nucleus, the cytoplasm and at the lipid droplet periphery. The heatmap shows the fold enrichment of the protein signal in each compartment as compared to the mean protein signal in the whole cell. Average of 3 experiments with at least 10 pictures per experiment.
Fig 9.
ABHD5 mutants of the predicted ATGL interface have impaired co-lipase and pro-viral functions.
(a) ABHD5 interaction to ATGL was predicted using Interactome INSIDER (http://interactomeinsider.yulab.org/, [46]) and the involved residues were depicted on the ABHD5 3D model (ModBase B2R9K0, ribbon representation with key residues as space-filling spheres). We mutated those residues that belong with a high or very high confidence to the ATGL and/or PLIN1 interface. Note that the two predicted interfaces overlap, but R114 has a higher interface potential with PLIN1 while L211 is more likely to interact with ATGL. (b) Verification by Western blot of ABHD5 protein expression upon knockdown and complementation in Lunet N hCD81 FLuc cells, 4 days post lentiviral transduction (at the time when HCV replication and production were assessed). Protein expression was quantified with the Odyssey imager and normalized to β-tubulin and to ABHD5 endogenous expression level (n = 3). Green arrowheads indicate the HA-tagged (empty arrowhead) and endogenous (full arrowhead) ABHD5 proteins. (c) Cellular lipid droplet content in Lunet N hCD81 cells upon expression of the different ABHD5 mutants. Representative flow cytometry plots are shown at the top and the average effects are plotted at the bottom (n = 5). (d) HCV assembly and release in Lunet N hCD81 FLuc cells upon ABHD5 knockdown and complementation with shRNA-resistant ABHD5 mutants of the predicted ATGL/PLIN1 interface (n = 3). (e) Correlation between HCV production and lipid droplet lipolysis. This graph gathers the results plotted in panels c and d as well as the linear regression (full line), 95% confidence band (dotted lines) and r2 value. Note that the lipid droplet content is tested in an over-expression assay (X axis), whereas HCV assembly and release are assessed in the context of the rescue of an ABHD5 knockdown (Y axis). This is why the empty vector is set to 100% on the X axis, but to a lower value for the Y axis (presence of ABHD5 shRNA).
Fig 10.
Subcellular localization of the ABHD5 mutants of the predicted ATGL interface.
(a) Immunofluorescence analysis of HA-tagged ABHD5 localization respectively to the lipid droplets (BODIPY 493/503) and trans-Golgi (p230 marker) in transduced Lunet N hCD81 cells. (b) Intensity profiles of the green and red channels along the line depicted in the merge pictures of panel a. (c) Quantification of ABHD5 accumulation in the nucleus, the cytoplasm, the Golgi apparatus and at the lipid droplet periphery. The heatmap shows the fold enrichment of the protein signal in each compartment as compared to the mean protein signal in the whole cell. The average of 3 experiments with at least 10 pictures per experiment is depicted.
Fig 11.
Transfer of the lipolytic and pro-viral functions by swapping two residues between ABHD5 and its paralog ABHD4.
(a) Swapping two residues between ABHD4 and ABHD5 confers ATGL activation properties to ABHD4 and results in a loss-of-function for ABHD5. ABHD4/5 chimeras were designed based on the results obtained with the murine homologs by Sanders and colleagues [47]. The human ABHD4 D290R-S319G mutant was also shown not to activate ATGL [47]. Here we further engineered a loss-of-function human ABHD5 mutant encoding the mouse ABHD4 residues instead of the R297 and G326, shown in the ABHD5 mouse homolog to be important for ATGL activation (R299 and G328) [47]. These residues are depicted in blue on ABHD5 predicted 3D model ABHD5 (ModBase B2R9K0) with the Interactome INSIDER predicted ATGL interface in yellow to red (see Fig 9A). (b) Verification by Western blot of ABHD4/5 protein expression in Lunet N hCD81 Fluc cells upon knockdown and complementation, 4 days post lentiviral transduction (at the time when HCV replication and production were assessed). Protein expression was quantified with the Odyssey imager and normalized to β-tubulin and to ABHD5 endogenous expression level, directly by comparing the ABHD5 antibody signal intensities or indirectly with the HA antibody signals for those constructs that are not detected by the ABHD5 antibody (n = 3). Green arrowheads indicate the HA-tagged (empty arrowhead) and endogenous (full arrowhead) ABHD5 proteins. Red arrowheads indicate the HA-tagged ABHD5 (empty arrowhead) and ABHD4 (full arrowhead) proteins. Note that ABHD5 has 349 residues whereas ABHD4 is 7 amino acids shorter, hence the size difference. (c) Cellular lipid droplet content in Lunet N hCD81 cells upon expression of the different ABHD4/5 mutants. Representative flow cytometry plots are shown at the top and the average effects are plotted at the bottom (n = 5). (d) HCV assembly and release in Lunet N hCD81 Fluc cells upon ABHD5 knockdown and complementation with shRNA-resistant ABHD4/5 mutants (n = 4). (e) Correlation between HCV production and lipid droplet lipolysis. This graph gathers the results plotted in panels c and d as well as the linear regression (full line), 95% confidence band (dotted lines) and r2 value. Note that the lipid droplet content is tested in an over-expression assay (X axis), whereas HCV assembly and release are assessed in the context of the rescue of an ABHD5 knockdown (Y axis). This is why the empty vector is set to 100% on the X axis, but to a lower value for the Y axis (presence of ABHD5 shRNA).
Fig 12.
Subcellular localization of the ABHD4/5 mutants.
(a) Immunofluorescence analysis of HA-tagged ABHD4/5 localization respectively to the lipid droplets (BODIPY 493/503) and trans-Golgi (p230 marker) in transduced Lunet N hCD81 cells. (b) Intensity profiles of the green and red channels along the line depicted in the merge pictures of panel a. (c) Quantification of ABHD4/5 accumulation in the nucleus, the cytoplasm, the Golgi apparatus and at the lipid droplet periphery. The heatmap shows the fold enrichment of the protein signal in each compartment as compared to the mean protein signal in the whole cell. The average of 3 experiments with at least 10 pictures per experiment is depicted.
Fig 13.
Interaction between ABHD5 and ATGL depends on ABHD5 residues R297 and G326.
We transduced Lunet N hCD81 cells to co-express HA-tagged ABHD5 constructs and ATGL (WT or S47A) and harvested the cell lysates 48 hours post-transduction. We incubated the lysates on an anti-HA resin and detected ABHD5 and ATGL with specific antibodies by Western blot (mouse anti-ABHD5 and rabbit anti-ABHD5) in the inputs and the eluates. Note that roughly 5% of the inputs were loaded on the gels. (a) Immunoprecipitation of ATGL and ATGL S47A with HA-tagged ABHD5, with or without oleic acid induction of the cells. When indicated, we treated the cells overnight with 100 μM oleic acid before harvesting the lysates. (b) Three more examples depicting the co-immunoprecipitation of ATGL S47A together with HA-tagged ABHD5, in oleic acid-treated cells, in different experiments. Note that in two additional experiments, the interaction was under the detection limit. (c) Co-immunoprecipitation assay with the Chanarin-Dorfman syndrome mutant Q130P and the ABHD5 TBLC mutant. (d) Co-immunoprecipitation assay with the ABHD5 mutants of the predicted ATGL interface or with the ABHD4/5 chimeras. (c, d) Since the TBLC mutant and the ABHD4 constructs are not recognized by the anti-ABHD5 antibodies ([30] and Fig 11B), we also stained the membranes with an anti-HA antibody (rabbit) after the initial detections with the anti-ABHD5 and anti-ATGL antibodies.