An RDH‐Plin2 axis modulates lipid droplet size by antagonizing Bmm lipase

Abstract

The size of lipid droplets varies greatly in vivo and is determined by both intrinsic and extrinsic factors. From an RNAi screen in Drosophila, we found that knocking down subunits of COP9 signalosome (CSN) results in enlarged lipid droplets under high‐fat, but not normal, conditions. We identified CG2064, a retinol dehydrogenase (RDH) homolog, as the proteasomal degradation target of CSN in regulating lipid droplet size. RDH/CG2064 interacts with the lipid droplet‐resident protein Plin2 and the RDH/CG2064‐Plin2 axis acts to reduce the overall level and lipid droplet localization of Bmm/ATGL lipase. This axis is important for larval survival under prolonged starvation. Thus, we discovered an RDH‐Plin2 axis modulates lipid droplet size.

An RNAi screen identifies subunits of COP9 signalosome (CSN) as negative regulators of lipid droplet formation under high fat conditions. CSN induces proteasomal degradation of RDH, which promotes Plin2 localization to the perinuclear lipid droplets.

Introduction

Cellular neutral lipids, such as triglyceride (TAG), are mainly stored in the form of lipid droplets with a neutral lipid core covered by a phospholipid monolayer coated with various proteins. The size and the number of lipid droplets determine the cellular capacity for lipid storage. In terms of lipolysis, large lipid droplets are relatively inactive compared with small lipid droplets. Under lipid‐overload conditions, the formation of large lipid droplets is beneficial to reduce lipotoxicity caused by excess lipids (Farese & Walther, 2009; Walther & Farese, 2012; Krahmer et al, 2013; Gluchowski et al, 2017; Yu & Li, 2017; Chen et al, 2019).

Lipid droplet size is determined by three factors: the neutral lipid content, the composition of the phospholipid monolayer, and the action of lipid droplet‐associated proteins. Excess cellular energy, in the form of fatty acids or of glucose, which can be converted to fatty acids through glycolysis and de novo lipogenesis, can be stored through increasing the neutral lipid content and the expansion of lipid droplets (Ruggles et al, 2013; Welte, 2015; Ding et al, 2018). Phosphatidylcholine (PC) and phosphatidic acid (PA) regulate lipid droplet size by affecting the membrane curvature and the coalescence of lipid droplets, respectively (Guo et al, 2008; Fei et al, 2011). Some lipid droplet‐associated proteins, such as DGAT2, CCT1, and GPAT4, are able to change the neutral lipid content and the composition of the phospholipid monolayer through local synthesis of phospholipids or TAG (Wilfling et al, 2013). In addition, other lipid droplet‐associated proteins, such as Fsp27, PLIN2, and PLIN5, affect lipid droplet size through regulating lipid droplet fusion, lipolysis or lipid droplet contacts with other organelles such as mitochondria or peroxisomes (Listenberger et al, 2007; Guo et al, 2008; Fei et al, 2011; Gong et al, 2011; Wang et al, 2011; Li et al, 2017; Kong et al, 2020). Despite this rapidly growing knowledge, the activities of lipid droplet‐associated proteins and the mechanisms that regulate them are far from clear.

Heterogeneity, both intercellular and intracellular, is an important but poorly studied feature of lipid droplets (Rinia et al, 2008; Herms et al, 2013). The variation in size represents just one type of lipid droplet heterogeneity. Lipid droplets may contain different neutral lipids, such as TAG, cholesterol ester (CE), or retinyl ester (RE), and these neutral lipids may also differ in the saturation of their fatty‐acyl chains (Blaner et al, 2009; Molenaar et al, 2017; Pacia et al, 2020). The composition of lipid droplet‐associated proteins also varies greatly, and this contributes significantly to lipid droplet dynamics (Rinia et al, 2008). Moreover, lipid droplets can even have heterogeneous origins. For example, lipid droplets in Drosophila fat cells can be classified into two subgroups (Diaconeasa et al, 2013; Ugrankar et al, 2019). One subgroup is called peripheral lipid droplets, which are small and near the cell surface. Peripheral lipid droplet formation depends on exogenous lipids transported from the intestine to adipose tissue. The other subgroup is medial lipid droplets, which originate from the peri‐nuclear endoplasmic reticulum (ER). Their formation and growth mainly rely on de novo fatty acid synthesis. Although the heterogeneous aspects of lipid droplets are well‐recognized, the regulation and coordination of different sub‐populations of lipid droplets have not been well studied.

Here, we report a retinol dehydrogenase (RDH)/CG2064‐Plin2‐Bmm axis which regulates lipid droplet size under high‐fat conditions. RDH/CG2064 directly interacts with Plin2. Increasing the level of RDH/CG2064 changes the localization of Plin2 from peripheral lipid droplets to small lipid droplets around the nucleus and also reduces the level and lipid droplet localization of Bmm lipase. The reduction of Bmm results in large lipid droplets under high‐fat conditions. Finally, this axis maintains larval survival under prolonged starvation.

Results

COP9 signalosome complex regulates lipid droplet size under high‐fat conditions

To study lipid storage regulation, we previously used pumpless‐GAL4 (ppl‐GAL4)‐driven gene overexpression or knockdown in Drosophila 3^rd instar salivary gland and fat body (Liu et al, 2014; Fan et al, 2017; Yao et al, 2018). Interestingly, overexpressing DGAT1 (ppl>DGAT1), which encodes a diacylglycerol O‐acyltransferase for TAG biosynthesis, did not lead to significantly enlarged lipid droplets in 3^rd instar larval fat body. The overall level of TAG is only marginally increased (Fig 1A and B). Besides genetic manipulation, we also reared larvae on a 30% coconut oil‐containing high‐fat diet. The expression of DGAT1 increases 2.8 fold under high‐fat feeding (Fig EV1A). However, similar to DGAT1 overexpression (OE), the high‐fat diet did not increase lipid droplet size in 3^rd instar larval fat body (Fig 1C). In addition, the combination of high‐fat diet and DGAT1 overexpression did not dramatically increase lipid droplet size compared with DGAT1 overexpression alone (Fig EV1B).

Figure 1. The CSN complex regulates lipid droplet size under high‐fat conditions.

3^rd instar larval fat bodies were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. Scale bars represent 25 µm.

• A
  Top: schematic diagram of the TAG biosynthetic pathway. Bottom: overexpression of DGAT1 (ppl>DGAT1) does not affect lipid droplet size.
• B
  Relative TAG levels in different genetic backgrounds were measured by glyceride assay kit. For quantification, TAG levels were normalized to protein. Error bars represent ± SEM. ***P < 0.001; NS: non‐significant.
• C
  The lipid droplet size is not increased when larvae are reared on a 30% high‐fat diet.
• D
  CSN RNAi alone does not affect lipid droplet size, but CSN RNAi dramatically increases the lipid droplet size when DGAT1 is overexpressed.
• E
  Quantification of the lipid droplet size in (A) and (D).
• F
  Lipid droplet size is increased in CSN7 RNAi larvae fed on a 30% high‐fat diet.
• G
  Quantification of the lipid droplet size in (C and F). ND: normal diet; HFD: high‐fat diet.

Data information: In (E) and (G), data were analyzed by one‐way ANOVA with a post hoc Turkey’s multiple‐comparison test. Each point represents data from one fat body, and at least 30 cells were analyzed for each fat body. Error bars represent ± SEM. ***P < 0.001; **P < 0.01; and NS: non‐significant.

Figure EV1. CSN‐mediated protein degradation regulates lipid droplet size.

• A
  Relative level of DGAT1 mRNA under normal and high‐fat diet conditions assayed by Q‐RT–PCR. Error bars represent ± SEM. ***P < 0.001. 3^rd instar larval fat bodies were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. Scale bar represents 25 µm.
• B
  DGAT1 overexpression under high‐fat feeding does not dramatically increase lipid droplet size compared with DGAT1 overexpression under normal feeding.
• C
  Knocking down Cullin1, 2, 3 or 5 under the high‐fat diet condition does not affect lipid droplet size.
• D
  Quantification of the lipid droplet size in (C).
• E
  Knocking down Cullin4 in larvae reared on a high‐fat diet increases the lipid droplet size. tub‐GAL80^ts was used to decrease the RNAi efficiency.
• F
  Quantification of the lipid droplet size in (E). HFD: high‐fat diet.
• G
  Knocking down Cand1 in DGAT1‐overexpressing larvae (ppl>DGAT1) increases the lipid droplet size.
• H
  Quantification of the lipid droplet size in (G).

Data information: In (D, F and H), data were analyzed by one‐way ANOVA with a post hoc Tukey’s multiple‐comparison test. Each point represents data from one fat body. At least 30 cells were analyzed from each fat body. Error bars represent ± SEM. ***P < 0.001; **P < 0.01; and NS: non‐significant.

To investigate why these high‐fat conditions did not result in large lipid droplets, we conducted an RNAi screen in the ppl>DGAT1 background. We found that RNAi of CSN3 led to significantly enlarged lipid droplets in 3^rd instar larval fat body (Fig 1D and E). Constitutively photomorphogenic 9 (COP9) signalosome (CSN) is an eight‐subunit protein complex that modulates the activity of an E3 ubiquitin ligase, Cullin‐Ring Ligase (CRL), by removing the ubiquitin‐like molecule Nedd8 from Cullin (Craney & Rape, 2013). We then examined whether RNAi of other CSN subunits phenocopies CSN3 RNAi. RNAi knockdown of several other CSNs, including CSN2, CSN5, or CSN7, results in enlarged lipid droplets in ppl>DGAT1 3^rd instar larval fat body (Fig 1D and E). Consistent with the increased lipid droplet size, the level of TAG is increased in CSN2 RNAi with DGAT1 OE compared with DGAT1 OE alone (Fig 1B).

Since we identified CSN3 in the DGAT1 OE background, we then tested whether CSN regulates lipid storage and lipid droplet size under normal conditions. In 3^rd instar larval fat body, CSN2 RNAi slightly, but not significantly, increased the TAG level (Fig 1B), and it had no effect on lipid droplet size (Fig 1D and E). Similarly, RNAi of other CSN subunits did not increase the lipid droplet size (Fig 1D and E), which suggests that regulation of lipid droplet size by CSN depends on high‐fat conditions. Indeed, CSN3 RNAi also increased the lipid droplet size of 3^rd instar larval fat body when the animals were fed on a 30% coconut oil‐containing high‐fat diet (Fig 1F and G). Together, these results demonstrate that CSN plays an important role in regulating lipid droplet size under high‐fat conditions.

CSN‐mediated protein degradation regulates lipid droplet size

CSN regulates CRL‐mediated protein degradation by deneddylating Cullin. Since RNAi of the deneddylase CSN5 affects lipid droplet size (Fig 1D and E), we speculated that CSN may regulate lipid droplet size through modulating CRL‐mediated protein degradation. Therefore, we examined the lipid droplet phenotype of Cullin knockdowns under high‐fat conditions. There are five Cullin genes in Drosophila. RNAi of Cullin1, 2, 3, or 5 did not affect lipid droplet size in 3^rd instar fat body of DGAT1 OE larvae (Fig EV1C and D), which suggests that these genes are unlikely to be involved in CSN‐mediated lipid droplet size regulation. ppl‐GAL4‐driven Cullin4 RNAi results in early larval lethality, which prevents us from directly observing the lipid droplet phenotype in 3^rd instar larvae. To circumvent the early lethality barrier, we used a repressible binary expression GAL4/GAL80 system to temporally control Cullin4 RNAi. We grew Tub‐GAL80^ts , ppl>Cullin4 RNAi animals at 18°C to lower the RNAi efficiency and shifted the temperature to 29°C at 96 h after egg laying to activate the RNAi effect. With that manipulation, we found large lipid droplets in 3^rd instar fat body of Cullin4 RNAi larvae fed on a high‐fat diet (Fig EV1E and F), which suggests that Cullin4 is important for lipid droplet size regulation.

Depletion of CUL4 and its “negative” regulator CSN induce the same phenotype. This apparent conflict is not surprising, because the neddylation‐de‐neddylation cycle is essential to maintain proper CRL function and disrupting the cycle often leads to a loss‐of‐function phenotype (Schwechheimer et al, 2001; Pierce et al, 2013; Bagchi et al, 2018). Along the same line, similar to the Cullin4 RNAi/high‐fat diet combination, lipid droplet size is increased in 3^rd instar fat body of DGAT1 OE larvae with Cand1 RNAi (Fig EV1G and H). Cand1 is a key player in recycling NEDD8 from CRL (Min et al, 2003). Loss of Cand1 impairs CRL‐mediated protein degradation and mimics CSN knockdown. Taken together, these results indicate that CSN‐mediated protein degradation regulates lipid droplet size under high‐fat conditions.

CSN regulates lipid droplet size through degrading CG2064, an RDH homolog

To identify downstream target(s) of CSN in this process, we performed comparative quantitative proteomic analysis on fat body samples from animals overexpressing DGAT1 alone (control), and overexpressing DGAT1 with CSN2 RNAi or CSN7 RNAi (Dataset EV1). We identified ~3,000 proteins from the proteomic analysis and we plotted the proteins according to the logarithm‐transformed ratios of different samples and P‐values. We focused on proteins with a P value below 0.05 and with a fold change of over 1.5 or below 0.67. With this standard, there are 131 up‐regulated proteins and 66 down‐regulated proteins in CSN2 RNAi compared with control, and 121 up‐regulated proteins and 66 down‐regulated proteins in CSN7 RNAi compared with control (Fig 2A). There are 90 up‐regulated proteins and 45 down‐regulated proteins in common between CSN2 RNAi and CSN7 RNAi samples (Fig 2A).

Figure 2. Elevating the level of RDH/CG2064 in CSN RNAi increases lipid droplet size.

• A
  The number of proteins changed in CSN2 or CSN7 RNAi revealed by quantitative proteomics. Red: the number of up‐regulated proteins and green: the number of down‐regulated proteins. Candidate proteins from the quantitative proteomics are listed along with protein fold changes in CSN2 or CSN7 RNAi. 3^rd instar larval fat bodies were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. Scale bars represent 25 µm.
• B
  RDH/CG2064 RNAi suppresses the enlarged lipid droplet phenotype in CSN2 RNAi under DGAT1 overexpression (ppl>DGAT1) or high‐fat diet conditions.
• C
  Quantification of the lipid droplet size in (B). HFD: high‐fat diet.
• D
  In Drosophila S2 cells expressing RDH/CG2064‐GFP, treatment with the proteasome inhibitor MG132 increases the RDH/CG2064‐GFP protein level detected by anti‐GFP antibody in a western blot.
• E
  The RDH/CG2064‐GFP protein level is increased in ppl>DGAT1 larvae with CSN2 RNAi, as detected by anti‐GFP antibody in a western blot.
• F
  RDH/CG2064 overexpression increases lipid droplet size in ppl>DGAT1 larvae.
• G
  Quantification of the lipid droplet size in (F).
• H
  Relative TAG levels in different genetic backgrounds were measured by glyceride assay kit. For quantification, TAG levels were normalized to protein.
• I
  RDH/CG2064 and CG2065 deletion does not affect lipid droplet size under normal or high‐fat diet conditions.

Data information: In (C), (G), and (H), data were analyzed by one‐way ANOVA with a post hoc Turkey’s multiple‐comparison test. Each point represents data from one fat body, and at least 30 cells were examined in each fat body. Error bars represent ± SEM. ***P < 0.001; **P < 0.01; and NS: non‐significant.

Source data are available online for this figure.

We further examined the molecular functions of proteins in the overlapping group and performed a literature search for potential lipid links. This led us to four up‐regulated proteins (CG2064, CG2065, CG9360, and Arc1) and three down‐regulated proteins (Pka‐R1, CG18135, and CG4500) (Fig 2A). CG2064, CG2065, and CG9360 belong to the NADP(H)‐dependent short‐chain dehydrogenase/reductase (SDR) family and are up‐regulated more than 3‐fold in CSN2 or CSN7 RNAi. SDR is a large protein superfamily and two SDR subfamilies, hydroxysteroid dehydrogenase (HSD) and RDH, have been reported to regulate lipid droplet dynamics (Deisenroth et al, 2011; Adams et al, 2014; Siddiqah et al, 2017; Yang et al, 2018). CG2064 and CG2065 belong to the RDH subfamily. Arc1 is a cytoskeleton protein and the cytoskeleton regulates lipid droplet morphology (Mosher et al, 2015). Pka‐R1 is a cAMP‐dependent protein kinase (PKA) regulator, and PKA is well known for its function in regulating lipase activity (Greenberg et al, 1991; Miyoshi et al, 2007). CG18135 is a glycerophosphocholine phosphodiesterase and CG4500 is a long‐chain fatty acid ligase.

To further evaluate these candidates, we manipulated their expression in different genetic backgrounds and examined the lipid droplet phenotype. Overexpression of Arc1 and RNAi of Pka‐R1, CG18135 or CG4500 did not lead to increased lipid droplet size (Fig EV2A). In addition, CG2065 RNAi or CG9360 RNAi did not rescue the enlarged fat body lipid droplet phenotype in 3^rd instar DGAT1 OE larvae with CSN2 RNAi (Fig EV2B). These results suggest that these genes do not act downstream of CSN in regulating lipid droplet size. In contrast, CG2064 RNAi suppresses the large lipid droplet phenotype caused by CSN2 RNAi in DGAT1 OE larvae (Fig 2B and C). We observed a similar rescuing effect of CG2064 RNAi on CSN2 RNAi with high‐fat feeding (Fig 2B and C). These results suggest that CG2064 acts downstream of CSN to control lipid droplet size under high‐fat conditions.

Figure EV2. Testing candidate CSN downstream targets.

3^rd instar larval fat bodies were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. Scale bar represents 25 µm.

• A
  Overexpressing Arc1 or knocking down Pka‐R1, CG18135 and CG4500 does not result in large lipid droplets in DGAT1‐overexpressing larvae (ppl>DGAT1).
• B
  CG2065 RNAi or CG9360 RNAi does not rescue the increased lipid droplet size phenotype caused by CSN2 RNAi in DGAT1‐overexpressing larvae.
• C, D
  The number and diameter of lipid droplets in different genetic backgrounds were quantified.

We therefore focused on CG2064. Since CG2064 has not been functionally studied before, we refer to it hereafter as RDH/CG2064. We expressed a C‐terminal GFP‐tagged RDH/CG2064 in Drosophila S2 cultured cells. Treatment with the proteasome inhibitor MG132 dramatically increased the level of RDH/CG2064‐GFP in S2 cells (Fig 2D), which indicates that RDH/CG2064 can be degraded through the proteasome pathway. We also generated an RDH/CG2064‐GFP knockin Drosophila reporter line and compared the level of RDH/CG2064‐GFP by western blot analysis of 3^rd instar fat body lysates of DGAT1 OE larvae with and without CSN2 RNAi. As expected, CSN2 RNAi dramatically increases the level of RDH/CG2064‐GFP (Fig 2E), which supports the hypothesis that CSN regulates RDH/CG2064 proteasomal degradation. Consistent with the post‐transcriptional regulation of RDH/CG2064 by CSN, the mRNA level of RDH/CG2064 in 3^rd instar larval fat body is not significantly changed by CSN2 RNAi (Fig EV3A). Together, these results indicate that CSN regulates lipid droplet size through proteasomal degradation of RDH/CG2064.

Figure EV3. The level of RDH/CG2064 and Plin2 in different genetic backgrounds.

1. The relative level of RDH/CG2064 mRNA detected by quantitative RT–PCR in different genotypes. NS: non‐significant.
2. Fat bodies from 3^rd instar larvae raised on a high‐fat diet were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. Scale bar represents 25 µm. In larvae consuming a high‐fat diet, RDH/CG2064 overexpression induces a large lipid droplet phenotype, which is suppressed by DGAT1 RNAi.
3. The relative level of RDH/CG2064 mRNA in larval fat body cells detected by quantitative RT–PCR in different genotypes. ***P < 0.001.
4. The RDH/CG2064‐GFP protein level in larval fat cells is decreased in RDH/CG2064 RNAi, as detected by anti‐GFP antibody in a western blot. Tubulin was used as loading control.
5. 3^rd instar larval fat bodies were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. Scale bar represents 25 µm. RDH/CG2064 RNAi does not affect lipid droplet size.
6. The level of Plin2‐GFP detected by anti‐GFP antibody in a western blot.

RDH/CG2064 overexpression increases lipid droplet size under high‐fat feeding

Since CSN RNAi increased the RDH/CG2064 level and the lipid droplet size under high‐fat conditions, we next investigated whether RDH/CG2064 overexpression is sufficient to increase lipid droplet size. In 3^rd instar larval fat body, RDH/CG2064 overexpression did not change the lipid droplet size under normal conditions, but it led to large lipid droplets with DGAT1 OE (Figs 2F and G, and EV2C and D). TAG measurement yielded a similar result (Fig 2H). In addition, in animals fed on a high‐fat diet, DGAT1 RNAi suppressed the large lipid droplet phenotype induced by RDH/CG2064 OE (Fig EV3B). This suggests that RDH/CG2064 is likely a main target of CSN in regulating lipid droplet size under high‐fat conditions.

We also examined the phenotype of RDH/CG2064 RNAi and mutants. The RNAi efficiency was verified by Q‐RT–PCR (Fig EV3C). In addition, the protein level of RDH/CG2064‐GFP was reduced upon RDH/CG2064 knockdown (Fig EV3D). We found that knockdown of RDH/CG2064 did not change the lipid droplet size in the fat body of wild‐type and DGAT1 OE 3^rd instar larvae (Fig EV3E). Considering that RDH/CG2064 and CG2065 are neighboring genes and both RDH/CG2064 and CG2065 levels are increased in CSN2/CSN7 RNAi, we generated RDH/CG2064 and CG2065 double deletion mutants to further examine the loss‐of‐function phenotype. We found that the lipid droplet size was not dramatically changed in the double mutants compared with controls under normal diet or high‐fat diet conditions (Fig 2I). This suggests that RDH/CG2064 is not required to maintain lipid droplet size under normal conditions.

RDH/CG2064 regulates lipid droplet size independent of its enzymatic activity

We next investigated the mechanism by which RDH/CG2064 regulates lipid droplet size. RDH/CG2064 has a transmembrane (TM) domain at its N‐terminus followed by a large dehydrogenase/reductase domain (Fig 3A). By BLAST, RDH/CG2064 is close to the mammalian RDH11/12 family. RDH12 is localized on the ER and its activity modulates retinol homeostasis (Keller & Adamski, 2007). ENV9, the yeast homolog of RDH12, increases lipid droplet size in a reductase activity‐dependent manner (Siddiqah et al, 2017). Therefore, we first examined whether the enzymatic activity of RDH/CG2064 is required for lipid droplet size regulation. The SDR proteins have a YXXXK motif, which is the reductase catalytic sequence (Jornvall et al, 1995). We generated two mutations in the enzymatic domain YXXXK motif, Y204F and K208A, and an upstream active site mutation, N151L (Fig 3A). Surprisingly, overexpressing these putative reductase‐dead forms of RDH/CG2064 also increased the lipid droplet size in the fat body of 3^rd instar DGAT1 OE larvae (Fig 3B and C), which suggests that RDH/CG2064 regulates lipid droplet size independent of its enzymatic activity.

Figure 3. The membrane protein RDH/CG2064 regulates lipid droplet size independent of its enzymatic activity.

1. The protein domain structure of RDH/CG2064. Grey box: predicted transmembrane region; Pink box: short‐chain dehydrogenase domain. The key enzymatic motif (YXXXK) and active site (NXXG) are shown.
2. 3^rd instar larval fat bodies were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. Overexpressing wild‐type or site‐mutated RDH/CG2064 increases the lipid droplet size in larvae with DGAT1 overexpression (ppl>DGAT1). Scale bar represents 25 µm.
3. Quantification of the lipid droplet size in (B).
4. Immunofluorescent staining of RDH/CG2064‐GFP with anti‐GFP antibody in 3^rd instar larval fat body. DAPI (blue) stains nuclei. RDH/CG2064 is enriched in the peri‐nuclear region. Scale bar represents 25 µm.
5. Immunofluorescence imaging of RDH/CG2064‐GFP, detected with anti‐GFP antibody, and the ER protein marker RFP‐KDEL in 3^rd instar larval fat body. RDH/CG2064 co‐localizes with RFP‐KDEL. Scale bar represents 10 µm.
6. Whole lysates and cytosolic and membranous fractions were prepared from fat body samples of 3^rd instar larvae expressing an RDH/CG2064‐GFP knockin reporter. RDH/CG2064‐GFP protein was detected by anti‐GFP antibody in a western blot. RDH/CG2064‐GFP is mainly present in the membranous fraction.

Source data are available online for this figure.

We then examined the sub‐cellular location of RDH/CG2064 with the help of the RDH/CG2064‐GFP knockin reporter. The fluorescent signal of RDH/CG2064‐GFP in 3^rd instar larval fat body is very weak, so we boosted the signal by anti‐GFP antibody staining. RDH/CG2064‐GFP is not enriched on lipid droplets. Instead, the signal is smeared and enriched around the nucleus, likely reflecting the peri‐nuclear ER region (Fig 3D). We further examined the sub‐cellular location of RDH/CG2064‐GFP by co‐labeling with the ER protein marker RFP‐KDEL. High‐resolution confocal microscopy revealed strong co‐localization between the RDH/CG2064‐GFP signal and the ER marker signal (Fig 3E). To further examine the distribution of RDH/CG2064‐GFP, we separated whole lysates of 3^rd instar larval fat body into cytosolic and membranous fractions through super‐speed centrifugation. By western blot analysis, the majority of the RDH/CG2064‐GFP was found in the membranous fraction (Fig 3F), further supporting the ER localization of RDH/CG2064‐GFP.

RDH/CG2064 interacts with the lipid droplet protein Plin2

To reveal the molecular mechanism by which RDH/CG2064 regulates lipid droplet size, we next used anti‐GFP antibody and performed immunoprecipitation‐mass spectrometry (IP‐MS) of RDH/CG2064‐GFP with GFP as control to identify the interacting protein partners of RDH/CG2064. Among the potential RDH/CG2064‐interacting proteins, the lipid droplet‐associated Perilipin protein Plin2 is the most obvious candidate (Table EV1). Drosophila has two Perilipin proteins, Plin1 and Plin2 (Blanchette‐Mackie et al, 1995; Miura et al, 2002; Welte et al, 2005). Plin2 genetically antagonizes brummer lipase (bmm), the sole Drosophila homolog of ATGL lipase (Zimmermann et al, 2004; Gronke et al, 2005), and overexpression of Plin2 increases lipid droplet size (Gronke et al, 2003; Bi et al, 2012). To validate the interaction of RDH/CG2064 and Plin2, we expressed a N‐terminal Flag‐tagged RDH/CG2064 in Plin1‐GFP, Plin2‐GFP, or Bmm‐GFP functional knockin Drosophila reporter lines. Immunoprecipitations with anti‐Flag antibody show that RDH/CG2064 pulls down Plin2, but not Plin1 or Bmm (Fig 4A). Furthermore, the physical interaction of RDH/CG2064 and Plin2 is also robust when DGAT1 is overexpressed (Fig 4B).

Figure 4. RDH/CG2064 interacts with Plin2.

• A, B
  Lysates of 3^rd instar larval fat body samples of different genetic backgrounds were immunoprecipitated with anti‐Flag antibody and then probed by anti‐GFP or anti‐Flag antibody in a western blot. Plin2, but not Plin1 or Bmm, is specifically coimmunoprecipitated with RDH/CG2064 (A). Plin2 was coimmunoprecipitated with RDH/CG2064 in samples from larvae overexpressing DGAT1 (B).
• C
  Schematic diagram of full‐length RDH/CG2064 and TM domain‐deleted RDH/CG2064 (∆TM). Lysates of Drosophila S2 cells expressing the different proteins were immunoprecipitated with anti‐GFP antibody and detected by anti‐GFP or anti‐Flag antibody in a western blot. Both full‐length and ∆TM RDH/CG2064 were coimmunoprecipitated with Plin2.
• D
  Schematic diagram of Plin1, Plin2, and the Plin2‐Plin1‐C chimeric protein. Lysates of Drosophila S2 cells expressing the different proteins were immunoprecipitated with anti‐Flag antibody and then probed by anti‐GFP or anti‐Flag antibody in a western blot. Neither Plin1 nor Plin2‐Plin1‐C chimeric protein coimmunoprecipitated with RDH/CG2064.
• E
  Lysates of Drosophila S2 cells expressing full‐length Plin1 or Plin1 with a C‐terminal deletion (Plin1ΔC) were immunoprecipitated with anti‐Flag antibody and then probed by anti‐GFP or anti‐Flag antibody in a western blot. Plin1ΔC coimmunoprecipitated with RDH/CG2064.
• F
  3^rd instar larval fat bodies were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. Under high‐fat conditions, RDH/CG2064 overexpression increases the lipid droplet size, and Plin2 RNAi suppresses the enlargement induced by RDH/CG2064 overexpression. Scale bar represents 25 µm.

Source data are available online for this figure.

We investigated the interaction of RDH/CG2064 and Plin2 further. RDH/CG2064 contains a TM domain and a dehydrogenase/reductase domain (Fig 3A). We expressed Plin2‐GFP with Flag‐tagged full‐length RDH/CG2064 or TM domain‐deleted RDH/CG2064 in Drosophila S2 cells. Both versions of RDH/CG2064 interact with Plin2 when immunoprecipitated with anti‐GFP antibody, indicating that the TM domain of RDH/CG2064 is not required for its interaction with Plin2 (Fig 4C). Plin1, the other Drosophila PAT domain protein, has an extended C‐terminal region compared with Plin2 (Fig 4D). Previous study showed that this C‐terminal region is essential for the sub‐cellular location of Plin1 and determines the functional difference between Plin1 and Plin2 (Bi et al, 2012). Considering that Plin2, but not Plin1, is able to interact with RDH/CG2064, we hypothesized that the C‐terminal region of Plin1 may prevent its interaction with RDH/CG2064. Therefore, we expressed a chimeric Plin2‐Plin1‐C‐terminal fusion protein along with Flag‐tagged RDH/CG2064 in S2 cells. This fusion protein does not bind to RDH/CG2064 when immunoprecipitated with anti‐Flag antibody (Fig 4D). Furthermore, although full‐length Plin1 does not interact with RDH/CG2064, Plin1 with a C‐terminal deletion does indeed interact with RDH/CG2064 (Fig 4E). Together, these results show that RDH/CG2064 interacts with Plin2, but not Plin1, and the C‐terminal region of Plin1 may prevent its association with RDH/CG2064. Moreover, genetic analysis results suggest that Plin2 acts downstream of RDH/CG2064. The large lipid droplet phenotype of RDH/CG2064 OE animals on a high‐fat diet is suppressed upon Plin2 knockdown (Fig 4F).

RDH/CG2064 changes the sub‐cellular localization of Plin2

Next, we investigated whether the CSN‐RDH/CG2064 axis affects Plin2. We first examined the protein level of Plin2‐GFP by western blot. In fat body samples from 3^rd instar DGAT1 OE larvae with ppl>CSN2 RNAi or ppl>RDH/CG2064 overexpression, there is no evident difference in Plin2‐GFP level compared with the DGAT1 OE control (Fig EV3F). This suggests that the CSN‐RDH/CG2064 axis does not affect the Plin2 protein level.

We then explored whether RDH/CG2064 affects the subcellular localization of Plin2. Previous protein overexpression studies show that Plin2 is usually found in small lipid droplets, in particular lipid droplets in the cortical region of larval fat cells (Bi et al, 2012), which are also called peripheral lipid droplets (Diaconeasa et al, 2013). Consistent with previous overexpression results, the Plin2‐GFP signal from the endogenous knockin reporter is also prone to localize on small cortical lipid droplets close to the plasma membrane in wild‐type 3^rd instar larval fat body. In CSN2 RNAi or RDH/CG2064 overexpression conditions, the peripheral Plin2‐GFP signal is greatly reduced and the GFP signal is increased in medial lipid droplets (Fig 5A and B). In addition, the localization of Plin2 is significantly increased on lipid droplets around the nucleus (Fig 5A and B). This finding is in line with the peri‐nuclear ER localization of RDH/CG2064. Therefore, although elevated expression of RDH/CG2064 in wild type is not sufficient to form large lipid droplets, it promotes the localization of more Plin2 to peri‐nuclear and medial lipid droplets. Such a localization change may prime the lipid droplets so that they are ready to expand.

Figure 5. The CSN‐RDH/CG2064 axis changes the sub‐cellular localization of Plin2.

Images of Plin2‐GFP in 3^rd instar larval fat body cells of different genotypes. Scale bar represents 25 µm.

• A
  CSN2 RNAi or RDH/CG2064 overexpression reduces the peripheral localization of Plin2‐GFP and increases the localization of Plin2‐GFP on small peri‐nuclear lipid droplets (boxed). DAPI (blue) stains nuclei. White arrows indicate peripheral lipid droplet. Red arrows indicate medial lipid droplets.
• B
  The peripheral localization of Plin2‐GFP is greatly reduced in fat body cells from larvae with CSN2 RNAi or RDH/CG2064 overexpression. Images were taken by focusing on the periphery of fat cells.
• C
  In fat body cells from ppl>DGAT1 larvae with CSN2 RNAi or RDH/CG2064 overexpression, the localization of Plin2‐GFP is greatly reduced at the periphery and is dramatically increased on large medial lipid droplets. White arrows indicate peripheral lipid droplet. Red arrows indicate medial lipid droplets.
• D
  Fluorescence intensity line plots of the images in (C). GFP intensity along the line across a lipid droplet was measured by ImageJ. In ppl>DGAT1 larvae with CSN2 RNAi or RDH/CG2064 overexpression, Plin2‐GFP localization on large medial lipid droplets is represented by two peaks. More than 30 lipid droplets were measured for each genotype. The plot shown is a typical image curve for the indicated genotype.
• E
  The peripheral localization of Plin2‐GFP is greatly reduced in fat body cells from ppl>DGAT1 larvae with CSN2 RNAi or RDH/CG2064 overexpression. Images were taken by focusing on the periphery of fat cells.
• F
  The localization of Plin2‐GFP is greatly reduced in peripheral lipid droplets and increased in medial lipid droplets in CSN2 RNAi larvae fed on a 30% high‐fat diet. DAPI (blue) stains nuclei. White arrows indicate peripheral lipid droplet. Red arrows indicate medial lipid droplets.
• G
  The peripheral lipid droplets in larval fat body cells from different genetic backgrounds were stained by BODIPY (green). CSN3 RNAi or RDH/CG2064 overexpression reduces the number of peripheral lipid droplets in both wild‐type and ppl>DGAT1 larvae.

We further investigated the effect of the CSN‐RDH/CG2064 axis on Plin2 localization under high‐fat conditions. The peripheral Plin2 localization is reduced in the fat body of 3^rd instar DGAT1 OE larvae with CSN2 RNAi or RDH/CG2064 overexpression (Fig 5C–E). In these animals, more Plin2‐GFP signal is present in large medial lipid droplets (Fig 5C–E). Similar changes of Plin2 localization were also observed in 3^rd instar larval fat body of CSN2 RNAi animals reared on 30% high‐fat medium (Fig 5F). Together, these results demonstrate that elevation of RDH/CG2064 level promotes the localization of Plin2 to medial lipid droplets and lipid droplets around the nucleus.

Peripheral Plin2 is important for the maintenance of small cortical lipid droplets (Ugrankar et al, 2019). In line with the reduction of peripheral Plin2, the number of small peripheral lipid droplets is greatly reduced in the fat body of 3^rd instar larvae with CSN RNAi or RDH/CG2064 overexpression in both control and DGAT1 OE conditions (Fig 5G; Movies EV1 and EV2). This phenotype is more severe in CSN3 RNAi fat body than in RDH/CG2064 fat body. It is possible that RDH/CG2064 may be a major downstream target of CSN, but not the only one. Taking the peri‐nuclear enrichment of RDH/CG2064 into account, these results suggest that the increased level of RDH/CG2064 results in the localization of more Plin2 to lipid droplets around the nucleus. These may represent newly formed lipid droplets from the peri‐nuclear ER and they may grow into large lipid droplets under high‐fat conditions.

The RDH/CG2064‐Plin2 axis reduces the level of Bmm lipase

We further explored the consequence of RDH/CG2064‐mediated changes in Plin2 localization. Genetic analysis revealed that Plin2 antagonizes the activity of Bmm (Gronke et al, 2005). Our previous study also suggested that “Plin2 may protect small lipid droplets at an early stage of lipid droplet biogenesis from Bmm‐mediated lipolysis” (Bi et al, 2012). Therefore, it is possible that the RDH/CG2064‐Plin2 axis affects Bmm.

We examined the cellular location of Bmm lipase with the help of a functional Bmm‐GFP knockin reporter. In the ppl>DGAT1 overexpression control, the fluorescent signal of Bmm‐GFP was found in the cytosol and also exhibited a punctate pattern surrounding lipid droplets in 3^rd instar larval fat body (Fig 6A and B). Interestingly, the GFP signal was greatly diminished and the peri‐lipid droplet punctate localization was absent in the fat body of 3^rd instar DGAT1 OE larvae with CSN2 RNAi or RDH/CG2064 overexpression (Fig 6A and B). The reduction of Bmm‐GFP fluorescent signal was also confirmed by western blot (Fig 6C). Furthermore, the reduction of overall Bmm‐GFP fluorescent signal and peri‐lipid droplet localization of Bmm‐GFP was similarly found in 3^rd instar fat body of CSN2 RNAi larvae fed on a 30% high‐fat diet (Fig 6D).

Figure 6. The CSN‐RDH/CG2064‐Plin2 axis affects the level and lipid droplet localization of Bmm lipase.

Confocal images of Bmm‐GFP (green) in 3^rd instar larval fat bodies of different genotypes. DAPI (blue) stains nuclei. Scale bar represents 25 µm.

• A
  Knocking down CSN2 or overexpressing RDH/CG2064 in animals with DGAT1 overexpression (ppl>DGAT1) decreases the level and lipid droplet localization of Bmm‐GFP.
• B
  The GFP intensity along the line across a lipid droplet in (A) was measured by ImageJ. The lipid droplet localization of Bmm‐GFP, represented by two peaks, is clearly visible in fat cells from ppl>DGAT1 larvae, but it is lost in fat cells from ppl>DGAT1 larvae with CSN2 RNAi or overexpression of RDH/CG2064. More than 30 lipid droplets of each genotype were measured. One typical image curve is shown for each genotype.
• C
  The level of Bmm‐GFP in 3^rd instar larval fat bodies of different genotypes was detected by anti‐GFP antibody in a western blot. In ppl>DGAT1 larvae, CSN2 RNAi or overexpression of RDH/CG2064 decreases the level of Bmm.
• D
  CSN2 RNAi decreases the level and lipid droplet localization of Bmm‐GFP in fat cells from larvae fed on a 30% high‐fat diet.
• E
  The level and lipid droplet localization of Bmm‐GFP are similar in fat cells from larvae overexpressing DGAT1 or mRFP (control).
• F
  Overexpression of Plin2 decreases the level and lipid droplet localization of Bmm‐GFP in fat cells from larvae with DGAT1 overexpression.
• G
  GFP intensity along the line across a lipid droplet from (F) was measured by ImageJ. The lipid droplet localization of Bmm‐GFP, represented by two peaks, was evident in fat cells from ppl>DGAT1 larvae, but is lost in ppl>DGAT1 larvae with overexpression of Plin2. More than 30 lipid droplets of each genotype were measured. One typical image curve is shown for each genotype.
• H
  3^rd instar larval fat bodies were stained by BODIPY (green) for lipid droplets and DAPI (blue) for nuclei. bmm RNAi increases the lipid droplet size in fat cells from larvae with DGAT1 overexpression. Scale bar represents 25 µm.
• I
  Quantification of the lipid droplet size in (H). Data were analyzed by one‐way ANOVA with a post hoc Turkey’s multiple‐comparison test. Each point represents data from one fat body. At least 30 cells were analyzed from each fat body. Error bars represent ± SEM. ***P < 0.001 and NS: non‐significant.
• J
  Relative TAG levels in different genetic backgrounds were measured by glyceride assay kit. For quantification, TAG levels were normalized to protein. Error bars represent ± SEM. **P < 0.01 and ***P < 0.001.

Source data are available online for this figure.

We also examined the localization of Bmm‐GFP in DGAT1 OE larvae with Plin2 overexpression. The localization of Bmm‐GFP is not changed in the fat body of 3^rd instar larvae overexpressing DGAT1 alone (Fig 6E). In contrast, the lipid droplet‐localized Bmm‐GFP in 3^rd instar DGAT1 OE larval fat body is dramatically reduced when Plin2 is also overexpressed (Fig 6F and G). This phenotype is very similar to that of RDH/CG2064 overexpression (Fig 6A and B). Therefore, the RDH/CG2064‐Plin2 axis likely reduces the level and the peri‐lipid droplet localization of Bmm under high‐fat conditions.

If a reduced level of Bmm lipase lies downstream of the RDH/CG2064‐Plin2 axis, knocking down bmm should increase the lipid droplet size when DGAT1 is overexpressed. Previous studies show that bmm mutants exhibit an early onset obese phenotype in adulthood, but not at the larval stage (Gronke et al, 2005, 2007; Bi et al, 2012). Consistent with that, lipid droplets were not enlarged in the fat body of 3^rd instar larvae with bmm RNAi (Fig 6H and I). In contrast, lipid droplet size was significantly increased in the fat body of 3^rd instar DGAT1 OE larvae with bmm RNAi (Fig 6H and I). TAG measurement yielded a similar result (Fig 6J).

To follow the kinetics of CSN‐mediated RDH/CG2064 degradation and changes in the Bmm protein level, we turned to Drosophila S2 cultured cells. We treated cells expressing RDH/CG2064‐GFP or Bmm‐GFP with cycloheximide (CHX), a protein synthesis inhibitor. In control cells, the level of RDH/CG2064‐GFP protein started to decline within 1 h of CHX treatment, and no RDH/CG2064‐GFP protein was detected after 6‐h CHX treatment. In CSN5 RNAi cells, the RDH/CG2064‐GFP signal persisted longer than in control cells (Fig 7A and B). For Bmm‐GFP, the protein level started to decline within 1–2 h after CHX treatment in control cells. In CSN5 RNAi cells, the Bmm‐GFP level declined faster than in control cells (Fig 7C and D). These data are consistent with the findings in larval fat cells. Together, these results support the idea that the RDH/CG2064‐Plin2 axis reduces the level of Bmm lipase.

Figure 7. The CSN‐RDH‐Plin2 axis functions under normal conditions and is required for prolonged starvation.

1. S2 cells expressing RDH/CH2064‐GFP were treated with the protein synthesis inhibitor cycloheximide (CHX) for the indicated time (h: hours). The level of RDH/CH2064‐GFP was determined by western blotting using anti‐GFP antibody. Tubulin was used as loading control.
2. Quantification of the relative RDH/CH2064‐GFP protein levels in (A). Error bars represent ± SEM.
3. S2 cells expressing Bmm‐GFP were treated with the protein synthesis inhibitor cycloheximide (CHX) for the indicated time (h: hours). The level of Bmm‐GFP was determined by western blotting using anti‐GFP antibody. Tubulin was used as loading control.
4. Quantification of the relative Bmm‐GFP protein levels in (C). Error bars represent ± SEM.
5. The RDH/CG2064‐GFP protein level in 3^rd instar larval fat cells is increased in CSN2 or CSN7 RNAi, as detected by anti‐GFP antibody in a western blot. Tubulin was used as loading control.
6. Oil red staining of 3^rd instar larval oenocytes in different genetic backgrounds under fed or starvation conditions.
7. The survival rate of control and CSN2 RNAi larvae at various fasting time points. Error bars represent ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; and NS: non‐significant.
8. Schematic model showing how the CSN‐RDH/CG2064‐Plin2 axis inhibits Bmm and promotes the formation of large lipid droplets under high‐fat/increased TAG conditions.

Source data are available online for this figure.

The CSN‐RDH‐Plin2 axis is not affected by high‐fat feeding and is required for survival under prolonged starvation

The lipid droplet phenotype of CSN RNAi is only observed in animals fed on a high‐fat diet. This raises the possibility that CSN tunes down RDH/CG2064‐Plin2 under high‐fat conditions. Therefore, we compared the levels of RDH/CG2064, Plin2, and Bmm under normal and high‐fat conditions using GFP knockin reporters. There was no difference in the levels of RDH/CG2064‐GFP, Plin2‐GFP, and Bmm‐GFP under normal and high‐fat conditions (Fig EV4A–C), which suggests that the CSN‐RDH/CG2064‐Plin2 axis is not affected by high‐fat feeding. RDH/CG2064‐GFP remains localized in the ER under high‐fat conditions (Fig EV4D). Indeed, similar to high‐fat conditions (Fig 2E), both CSN2 RNAi and CSN7 RNAi dramatically increase the level of RDH/CG2064‐GFP under normal conditions (Fig 7E). These results, together with the regulatory effect of CSN on RDH/CG2064 and Bmm levels in S2 cells, suggest that CSN regulates the turnover of RDH/CG2064 under both normal and high‐fat conditions.

Figure EV4. High‐fat feeding does not affect the RDH‐Plin2 axis.

• A–C
  The levels of RDH/CG2064‐GFP (A), Plin2‐GFP (B), and Bmm‐GFP (C) under normal and high‐fat conditions detected by anti‐GFP antibody in western blots. Tubulin was used as loading control.
• D
  Immunofluorescent imaging of RDH/CG2064‐GFP, detected with an anti‐GFP antibody, and the ER protein marker RFP‐KDEL in the fat body of 3^rd instar larvae fed on a high‐fat diet. RDH/CG2064 is co‐localized with RFP‐KDEL. Scale bar represents 10 µm.
• E
  Cellular model showing how RDH/CH2064, located at the peri‐nuclear ER, recruits Plin2 to new‐born small lipid droplets.

Source data are available online for this figure.

The previous results raise the possibility that the CSN RNAi phenotype is subtle under normal conditions and can only be revealed under certain sensitizing conditions. Under starvation, lipolysis mobilizes lipids from fat cells to support animal survival and also promotes the formation of lipid droplets in oenocytes, hepatocyte‐like cells in Drosophila. We therefore assayed the impact of CSN RNAi on both oenocyte lipid storage and larval survival under starvation. After 6 h of starvation, there is no obvious difference in lipid storage in oenocytes of CSN2 RNAi and CSN7 RNAi compared with control under both normal and high‐fat conditions (Fig 7F). In addition, there is no significant difference in the larval survival rate in control and CSN2 RNAi animals up to 24 h of starvation (Fig 7). However, the CSN2 RNAi larvae exhibited a much lower survival rate at 36–60 h of fasting compared with controls (Fig 7G). This suggests that the CSN‐RDH‐Plin2 axis is likely required for larval survival under prolonged starvation.

Discussion

The size and the associated proteins of lipid droplets vary greatly in cells. Here, we discover that the CSN complex prevents the enlargement of lipid droplets under high‐fat conditions. CSN controls the proteasomal degradation of RDH/CG2064, which interacts with Plin2. The elevated level of RDH/CG2064 in CSN knockdown larvae changes Plin2 localization from peripheral lipid droplets to medial lipid droplets and small lipid droplets around the nucleus. Plin2 likely protects these lipid droplets from lipolysis by reducing the level of Bmm/ATGL lipase and facilitates the formation of large lipid droplets under lipid‐overload conditions (Fig 7H).

Different CSN subunits and Cullin4 have been identified in previous studies for genes that affect lipid droplet morphology and lipid storage (Guo et al, 2008; Baumbach et al, 2014; Yoshizawa et al, 2014; Sinha et al, 2020). Although the targets of CSN‐Cul4 in regulating lipid storage have not been previously reported, these studies, together with our findings here, clearly support the important role of CSN‐CUL4 in both cellular and organismal lipid storage. We further identified RDH/CG2064 as a target of CSN‐Cul4 in regulating lipid droplet size. RDH/CG2064 is a member of SDR, a large enzyme superfamily mainly defined by a similar three‐dimensional structure and catalytic amino acid tetrad. RDH and HSD, two SDR subfamilies, have been linked to lipid droplets or lipid storage‐related diseases (Sandell et al, 2007; Deisenroth et al, 2011; Jiang & Napoli, 2013; Su et al, 2014; Siddiqah et al, 2017). Some of these proteins are localized on lipid droplets and regulate lipid droplet dynamics in either an enzymatic activity‐dependent or ‐independent manner. RDH/CG2064 is closely related to the RDH11 and RDH12 subfamily. Similar to RDH12, RDH/CG2064 is likely localized in the ER. RDH12 takes part in retinoid/vitamin A production. Retinoid is reported to be negatively associated with the content of body fat in mammals (Berry & Noy, 2009). RDH11 mutation causes retinitis pigmentosa in humans (Xie et al, 2014). Besides the retina, murine RDH11 is also expressed in the liver and acts as an all‐trans‐retinaldehyde reductase to maintain the level of all‐trans‐retinol under reduced vitamin A availability (Belyaeva et al, 2018). Previous studies mainly focused on its role in retinol metabolism and it is not known whether RDH11 regulates neutral lipid storage. Our findings raise the possibility that RDH11 may have dual functions, one in lipid droplet dynamics and the other in retinoid metabolism. Further studies are required to dissect the role of RDH11 in lipid metabolism and retinitis pigmentosa.

RDH/CG2064 regulates lipid droplet size through Plin2. RDH/CG2064 is localized on the ER, so how does it affect the localization of the lipid droplet protein Plin2 from peripheral lipid droplets to perinuclear and medial lipid droplets? It is possible that through direct binding, RDH/CG2064 retains Plin2 in the perinuclear ER region, where the biogenesis of medial lipid droplets mainly occurs. A higher level of Plin2 facilitates the expansion of newly formed lipid droplets. The growth of medial lipid droplets is affected by both de novo lipogenesis and high‐fat feeding. In combination with high‐fat conditions, the increased level of Plin2 in medial lipid droplets and lipid droplets around the nucleus likely results in the appearance of large lipid droplets (Fig EV4E). Consistent with this, when the level of RDH/CG2064 is elevated, the size of peripheral lipid droplets is reduced, while the size of medial lipid droplets is increased.

PLINs reside in the hub for regulating lipid droplet dynamics, including lipolysis and lipid droplet fusion (Beller et al, 2010; Sun et al, 2013; Kimmel & Sztalryd, 2016). Previous studies show that different PLINs act as either activity regulators or protein localization barriers of lipolysis, in particular for the rate‐limiting lipase ATGL (Zimmermann et al, 2004; Listenberger et al, 2007). Although numerous downstream effectors of PLINs have been identified (Granneman et al, 2009; Sun et al, 2013), the upstream regulators of PLINs are relatively less well characterized. The best known one is protein kinase A (PKA), which phosphorylates PLIN1. The RDH/CG2064‐Plin2‐Bmm/ATGL axis is somewhat different from the PKA‐PLIN1‐ATGL axis. The phosphorylation of PLIN1 by PKA changes the binding partner of PLIN1 and then results in ATGL activation. The binding of RDH/CG2064 with PLIN2 changes the localization of PLIN2 and likely reduces the lipid droplet localization and the level of Bmm/ATGL. These two axes may represent two layers of ATGL activity regulation. One aims for full activation and the other limits initial basal activation. The RDH/CG2064‐Plin2‐Bmm/ATGL axis may determine the level of ATGL that is in the initial “ready‐to‐go” state. Full activation of ATGL then occurs by its interaction with the coactivator CGI‐58, which is unleashed via the PKA‐PLIN1‐ATGL axis. The coordination of these two pathways may be helpful for correct spatial or temporal ATGL activation in different types of lipid droplets.

CSN and RDH/CG2064 represent a new type of lipid droplet regulator that is usually not required under normal physiological conditions and thus cannot be easily identified. Such regulators may be poorly characterized or even missed in previous functional studies. However, with the current pandemic of obesity, they may hold great potential for understanding or treating lipid‐overload diseases. Further functional and clinical studies on this type of regulator may shed more light on systemic lipid homeostasis.

Materials and Methods

Drosophila husbandry and stocks

Unless specified, flies were raised on standard cornmeal food (8 g agar, 63.2 g glucose, 31.6 g sucrose, 30 g yeast, 77.7 g corn powder, butylparaben into 1 l water) at 25°C. The high‐fat diet contains 30% coconut oil in cornmeal food. ppl‐GAL4 was used to drive gene overexpression or knockdown in the fat body (Liu et al, 2014). For the RNAi screen, a collection of ppl‐GAL4>DGAT1 virgin flies were crossed to UAS‐RNAi males. The eggs were raised at 25°C, then transferred to 29°C at 48 h after egg laying. For RNAi, UAS‐white RNAi was chosen as the control RNAi group, while for overexpression, UAS‐mRFP (membrane myr‐RFP) was used as the UAS control. Transgenes of pUAST‐Flag‐CG2064, pUAST‐CG2064‐GFP, pUAST‐CG2064^N151L , pUAST‐CG2064^K208A , and pUAST‐CG2064^Y204F were generated. Plin1‐GFP, Plin2‐GFP, Bmm‐GFP, and CG2064‐GFP knockin flies were generated by the CRISPR/Cas9 technique. Plin1/CG10374 has four transcripts, which share the same C‐terminal region. We designed guide RNAs which target the shared C‐terminal region. Using CRISPR/Cas9, the endogenous Plin1 C‐terminal was replaced by a donor sequence which was fused in frame with GFP. Plin2/CG9057 has five transcripts, which share the same C‐terminal region. Bmm/CG5295 has two transcripts, which share the same C‐terminal region. CG2064‐GFP has only one transcript. Based on a similar approach to Plin1‐GFP, we generated Plin2 C‐terminal GFP, Bmm C‐terminal GFP, and CG2064 C‐terminal GFP knock‐in flies. Other Drosophila stocks were obtained from Bloomington Drosophila Stock Center and Tsinghua Fly Center. Fly lines used in this study are listed (Table EV2).

Tissue staining, microscopy, and image analysis

Fat bodies of wandering 3^rd instar larvae were dissected in 1x PBS buffer. The dissected samples were fixed with 4% paraformaldehyde for 30 min at room temperature. To stain lipid droplets, fixed fat bodies were incubated for 1 h at room temperature in the neutral lipid dye BODIPY 493/503 (Invitrogen, 1 mg/ml diluted 1:500 with PBS). The washed samples were incubated with 2 ng/µl DAPI to label the nuclei. The images were taken in the middle plane of the fat body cells, where the nucleus is largest, to estimate lipid droplet size according to the same criterion. To quantify the lipid droplet size, the diameters of the top three largest lipid droplets per cell from more than 50 cells were measured by ImageJ. A line was drawn along the maximum diameter in each individual lipid droplet and the diameters were ranked for size distribution analysis. The top three measurements were selected as the largest three lipid droplets in one cell. The images of peripheral lipid droplets were taken by focusing on the surface of the fat body cell. For antibody staining, fixed samples were washed in 0.1% PBST (Tween‐20 diluted with PBS). Then, samples were blocked in 5% BSA (Sigma, diluted with 0.1% PBST) for 30 min at room temperature. After another wash in 0.1% PBST, samples were incubated with anti‐GFP antibody (Abcam, diluted 1:10,000 with 0.1% PBST) at 4°C overnight. The fluorescent secondary antibody Rb488 was also diluted with 0.1% PBST. GFP knockin flies were used to examine the localization of fluorescent proteins on lipid droplets. The fluorescence intensity on the lipid droplet surface of 30 fat body cells was measured by ImageJ. For fluorescence intensity analysis, a line was plotted across a single lipid droplet. The fluorescence intensity value along the line was recorded. The peak in the curve represents the fluorescence intensity value on the lipid droplet surface. The fluorescence intensity results were normalized to the cytosolic intensity. The images were obtained by a Leica TCS SP8 and a LSM980 high‐resolution confocal microscope.

Molecular biology and quantitative RT–PCR

For pUAST‐Flag‐CG2064 and pUAST‐Flag‐CG2064ΔTM, the CG2064 coding region was amplified from Drosophila cDNA and inserted in frame into the pUAST‐attB‐Flag vector through the NotI and XbaI sites. For pUAST‐CG2064‐GFP, the CG2064 coding region (without the stop codon) was amplified and inserted in frame into the pUAST‐attB‐GFP vector through the NotI and KpnI sites. For pUAST‐CG2064^N151L , pUAST‐CG2064^Y204F , and pUAST‐CG2064^K208A , site‐direct mutagenesis was conducted by fusion PCR. For pUAST‐Plin1‐GFP and pUAST‐Plin2‐GFP, the Plin1 or Plin2 coding region (without the stop codon) was amplified and inserted in frame into the pUAST‐attB‐GFP vector through the NotI and KpnI sites. For pUAST‐Plin2‐Plin1‐C‐GFP and pUAST‐Plin1‐C deletion, the sequences of Plin2‐Plin1‐C and Plin1‐C deletion (without the stop codon) were amplified and inserted in‐frame into the pUAST‐attB‐GFP vector through the NotI and KpnI sites. Plin2‐Plin1‐C and Plin1‐C deletion were first generated by PCR. All constructs were confirmed by sequencing. The efficiency of gene overexpression or knockdown was verified by quantitative RT–PCR. All quantitative RT–PCRs were performed on a Stratagene Mx3000P QPCR cycler (Applied Biosystems) using Transstart Green qPCR super Mix kit (Transgen). The quantitative RT–PCR results were normalized with rp49 gene.

Proteomic analysis

To identify the proteins downstream of CSN, 30 pairs of 3^rd instar larval fat bodies were dissected in PBS buffer. Three biological repeats were performed for each genotype. Protein digestion with sequencing grade trypsin (Promega) was performed using the filter‐aided proteome preparation (FASP) method. Resultant tryptic peptides were labeled with acetonitrile‐dissolved TMT reagents (Thermo Fisher Scientific). Equal amounts of labeled samples were mixed together before prefractionation with reversed phase (RP)‐high performance liquid chromatography (HPLC). The peptides were fractionated on a Phenomenex Gemini‐NX 5 u C18 column (250 × 3.0 mm, 110 Å, Torrance) with a Waters e2695 separations HPLC system. All samples were dried with a Speed‐Vac concentrator and desalted by StageTip before use. The LC‐MS/MS analysis was performed with an LTQ Orbitrap Elite mass spectrometer (Thermo Fisher Scientific) coupled online to an Easy‐nLC 1,000 in the data‐dependent mode. The peptides were injected into a capillary analytic column (length: 25 cm and inner diameter: 75 μm) packed with C18 particles (diameter: 5 μm). The positive ion mode was used for MS measurements. The spectra were acquired across the mass range of 300–1,800 m/z. The database search was performed for all raw MS files with MaxQuant (version 1.5).

To identify RDH/CG2064 interacting partners, 100 pairs of 3^rd instar larval fat bodies of RDH/CG2064‐GFP‐overexpressing line were dissected in cold PBS buffer and homogenized on ice in RIPA buffer. The cell lysate was spun at 16,060 g for 15 min at 4°C to remove the cellular debris. The supernatant was incubated with anti‐GFP affinity beads (Sigma‐Aldrich) at 4°C overnight. Subsequently, the beads were washed three times with RIPA buffer. The eluted proteins were then digested with trypsin then for MS analyses with an LTQ Orbitrap Elite mass spectrometer (Thermo Fisher Scientific) coupled online to an Easy‐nLC 1000 (Thermo Fisher Scientific) in the data‐dependent mode.

Western blot and quantification

Drosophila S2 cells were cultured with Sf‐900 II serum‐free medium (GIBCO) at 26°C. The overexpression plasmid and the actin‐GAL4 plasmid were transfected into S2 cells with Cellfectin II (Invitrogen) or PLUS reagent (Invitrogen). Forty‐eight hours after transfection, the cells were treated with DMSO or 40 µM proteasome inhibitor MG132 (Sigma‐Aldrich) for 7 h. After removing the medium, cells were lysed with radioimmunoprecipitation assay (RIPA) buffer for 20 min on ice. The lysates were centrifuged at 13,680 g for 10 min at 4°C. The supernatants were retained for SDS–PAGE and western blots.

For larval samples, 3^rd instar larval fat bodies were dissected in PBS buffer, transferred to 200 µl cold 1% SDS lysis buffer, and homogenized with an electronic pestle. Samples were then centrifuged at 13,680 g for 15 min at 4°C. The supernatants were retained for SDS–PAGE and western blots. Ten 3^rd instar larvae per group and three groups for each genotype were used.

For immunoprecipitation, 20 µl IP beads (Sigma‐Aldrich) were pre‐cleaned with lysis buffer by centrifuging at 860 g for 1 min at 4°C. This step was repeated three times. Lysates of cultured cells or larval samples were incubated with IP beads at 4°C overnight. Subsequently, the beads were washed three times with RIPA buffer. 60 µl 1x SDS‐loading buffer (125 mM Tris–HCl, pH 6.8, 20% glycerol, 2% SDS, 0.05% bromophenol blue, 10% 2‐mercaptoethanol) was added to the sample. The samples were boiled at 100°C for 10 min and centrifuged at 13,680 g for 2 min. The supernatant was kept for western blots. The antibodies for western blotting were anti‐GFP (Abcam), anti‐Flag (Sigma‐Aldrich), and anti‐α‐tubulin (Sigma‐Aldrich).

For cell fraction, 3^rd instar larval fat bodies were dissected in PBS buffer. Larval fat body lysates were diluted with subcellular fraction buffer (250 mM sucrose, 20 mM HEPES, 10 mM KCl, 1.5 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, and 1 mM DTT) to 5 mg/ml final protein concentration. The cytosolic fraction and the membrane fraction were separated by super‐speed centrifugation. The soluble (liquid) fraction contained the cytosolic protein and the precipitate contained the membranous fragment. The soluble fraction was concentrated from 1 ml to 40 µl. The precipitate was dissolved in 40 µl buffer.

Glyceride quantification

To measure glyceride content, fat bodies of wandering 3^rd instar larvae were dissected in 200 µl PBST buffer and then transferred into liquid nitrogen. Glyceride quantification was performed based on previous study (Palanker et al, 2009). For total glyceride, ten 3^rd instar larvae (three groups of ten for each genotype) were used. Glyceride level was measured using TAG determination kits (Zhongshengbeikong). Glyceride measurement was normalized by total protein level.

Starvation test

For oenocyte staining, fed and 6‐h starved larvae were dissected and stained with Oil Red. To test the larval survival rate after starvation, embryos for testing were collected within 4 h and raised at low density on standard fly food at 25°C. Seventy‐two hours after hatching, larvae were either fed with standard food or starved on PBS‐soaked paper for 12, 24, 36, 48, and 60 h. For the survival assay, three groups of 60 larvae for each genotype were starved. Then larvae were transferred back to standard food. Survival rate is based on counting the number of surviving adult flies.

RNAi in S2 cell

Based on RNAi stock information from flybase (CSN2:BL28290, VDRCGD4682; CSN3:DRSC32997, CSN5:DRSC16592; and CSN7:DRSC06808), 300–500 bp dsRNA fragments of CSN genes were selected. Target dsDNA fragments were amplified by PCR using the T7 promotor. The PCR fragments were purified and extracted. dsDNA was transcribed to dsRNA using the T7 Ultra Kit (Life Technologies, AM1345). S2 cells were resuspended at 1 × 10⁶ cells/ml in serum‐free medium and 1 ml cells was plated into each well of a 6‐well tissue culture plate. 30 µg dsRNA were added to each well. The plates were incubated at room temperature for 30 min, before adding 3 ml complete media with 10% FBS to each well. The cells were incubated for 3 days for further analysis.

Pulse‐chase experiments in S2 cells

To follow protein kinetics in S2 cells, RDH/CG2064‐GFP or Bmm‐GFP overexpression was driven by actin‐Gal4 by transfection at 1 × 10⁶ cells/ml in serum‐free medium. To inhibit protein synthesis, cells were treated with 50 µM cycloheximide (Cayman) for different times (0.5–6 h). After 48 h overexpression, protein was extracted to measure the levels of RDH/CG2064 or BMM protein with or without CSN knockdown by western blot.

RDH/CG2064 and CG2065 deletion mutants

The RDH/CG2064‐CG2065 double knockout (DKO) mutant was generated by Cas9 precise excision (Qidong Fungene Biotechnology Co., Ltd.). Cas9 excision leads to a 3‐kb deletion, spanning the coding regions of the RDH/CG2064 and CG2065 genes.

Statistical analysis

Unless specified, all data are shown as mean ± SEM. All the experiments were performed three times. Based on different types of data, statistical analyses were performed with Student’s t test (two group comparison) or one‐way ANOVA with a post hoc Tukey’s multiple comparison test (> 2 group comparison). **P < 0.001, **P < 0.01, and *P < 0.05, ns: not significantly different.

Author contributions

Xuefan Zhao: Conceptualization; Investigation; Writing – original draft. Wei Wang: Methodology. Yan Yao: Resources; Methodology. Xia Li: Formal analysis. Xiahe Huang: Methodology. Yingchun Wang: Supervision. Mei Ding: Supervision. Xun Huang: Conceptualization; Supervision; Writing – original draft; Project administration; Writing – review & editing.

In addition to the CRediT author contributions listed above, the contributions in detail are:

Experiments were designed by XZ and XH. Samples preparation and most experiments were done by XZ. IP of RDH/CG2064 interaction proteins was performed by WW. Plin1, Plin2, and Bmm Knock‐in Drosophila was first analyzed by YY. Drosophila screen was performed by XL. Proteomic analysis was performed by XHH and YW. XZ, MD, and XH analyzed the results. XZ and XH wrote the manuscript.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Dataset EV1

Movie EV1

Movie EV2

Source Data for Expanded View

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 6

Source Data for Figure 7

EMBO reports (2022) 23: e52669.