Polybasic RKKR motif in the linker region of lipid droplet (LD)–associated protein CIDEC inhibits LD fusion activity by interacting with acidic phospholipids

Lipid droplets (LDs) are intracellular organelles and a central site for lipid synthesis, storage, and mobilization. The size of LDs reflects the dynamic regulation of lipid metabolism in cells. Previously, we found that cell death–inducing DFFA-like effector C (CIDEC) mediates LD fusion and growth by lipid transfer through LD–LD contact sites in adipocytes and hepatocytes. The CIDE-N domains of CIDEC molecules form homodimers, whereas the CIDE-C domain plays an important role in LD targeting and enrichment. Here, using targeted protein deletions and GFP expression coupled with fluorescence microscopy, we identified a polybasic RKKR motif in the linker region that connects the CIDE-N and CIDE-C domains of CIDEC and functions as a regulatory motif for LD fusion. We found that deletion of the linker region or mutation of the RKKR motif increases the formation of supersized LDs compared with LD formation in cells with WT CIDEC. This enhanced LD fusion activity required the interaction between CIDE-N domains. Mechanistically, we found that the RKKR motif interacts with acidic phospholipids via electrostatic attraction. Loss of this motif disrupted the protein–lipid interaction, resulting in enhanced lipid droplet fusion activity and thus formation of larger LDs. In summary, we have uncovered a CIDEC domain that regulates LD fusion activity, a finding that provides insights into the inhibitory regulation of LD fusion through CIDEC–lipid interactions.


Lipid droplets (LDs) are intracellular organelles and a central site for lipid synthesis, storage, and mobilization. The size of LDs reflects the dynamic regulation of lipid metabolism in cells. Previously, we found that cell death-inducing DFFA-like effector C (CIDEC) mediates LD fusion and growth by lipid transfer through LD-LD contact sites in adipocytes and hepatocytes. The CIDE-N domains of CIDEC molecules form homodimers, whereas the CIDE-C domain plays an important role in LD targeting and enrichment. Here, using targeted protein deletions and GFP expression coupled with fluorescence microscopy, we identified a polybasic RKKR motif in the linker region that connects the CIDE-N and CIDE-C domains of CIDEC and functions as a regulatory motif for LD fusion. We found that deletion of the linker region or mutation of the RKKR motif increases the formation of supersized LDs compared with LD formation in cells with WT CIDEC. This enhanced LD fusion activity required the interaction between CIDE-N domains. Mechanistically, we found that the RKKR motif interacts with acidic phospholipids via electrostatic attraction. Loss of this motif disrupted the protein-lipid interaction, resulting in enhanced lipid droplet fusion activity and thus formation of larger LDs. In summary, we have uncovered a CIDEC domain that regulates LD fusion activity, a finding that provides insights into the inhibitory regulation of LD fusion through CIDEC-lipid interactions.
Lipid droplets (LDs) 3 consist of the neutral lipid core surrounded by a monolayer of phospholipids and LD-associated proteins (1). The excessive storage of neutral lipids in LDs is closely associated with the development of obesity, type II diabetes, cardiovascular diseases, fatty liver diseases, and neurodegeneration (2)(3)(4)(5)(6)(7). Nascent LDs that budded from the endoplasmic reticulum (ER) (8 -11) continue to grow and mature via one of three processes, namely triglyceride (TAG) synthesis within LD (12), lipid transfer from the ER to LD through contact (13)(14)(15), or LD fusion (16 -18). The LD fusion process is mediated by the cell death-inducing DFFA-like effector (CIDE) family proteins that include CIDEA, CIDEB, and CIDEC/Fsp27 (19,20).
CIDEA is highly expressed in brown adipose tissue (BAT) and regulates LD size and lipid storage (21). Under a high-fat diet condition, CIDEA expression is greatly elevated in the liver of mice, leading to severe nonalcoholic fatty liver disease (22). CIDEB is mainly expressed in the liver and small intestine of mice (23)(24)(25). Cideb-knockout mice are resistant to high-fat diet-induced liver steatosis (23,25). CIDEC is highly expressed in white adipose tissue (WAT). Concomitantly, the adipocytes of Cidec Ϫ/Ϫ mice exhibit small multilocular LDs, reduced triglyceride storage, and increased lipolysis (26,27). A patient with a homozygous nonsense mutation in Cidec also exhibits small multilocular LDs in adipocytes and suffers from severe metabolic syndromes such as insulin resistance and partial lipodystrophy (28). In addition, the combination therapy that utilized antisense oligonucleotide to silence Cidec and fenofibrate (peroxisome proliferator-activated receptor-␣ agonist) has proven to ameliorate obesity, hypertriglyceridemia, hepatic steatosis, and inflammation induced by a high-fat diet in mouse models (29,30). Mechanistically, CIDE family proteins are enriched at the LD-LD contact sites (LDCSs) enabling lipid transfer to occur between two LDs (16, 20,31,32), thus promoting LD fusion and growth in BAT, WAT, and the liver.
Previously, we reported that CIDEC mediates LD fusion by means of directional lipid transfer from smaller (donor) to larger (acceptor) LDs (16). The enrichment of CIDEC at LDCS is a prerequisite for LD fusion prior to fusion pore formation. The latter is an essential step for lipid transfer to occur. Molec-ularly, on CIDEC, there is a conserved N-domain (CIDE-N) and a C-domain (CIDE-C) joined by a stretch of a 16-amino acid linker. The mechanistic study showed that the CIDE-C domain of CIDEC is responsible for LD targeting; in particular, aa 136 -217 are crucial for CIDEC enrichment at the LDCS (16). Other studies have also demonstrated the significance of the CIDE-C domain in LD fusion (33,34). The CIDE-N domain of CIDEC does not localize on the LD surface and has a regulatory function in promoting LD growth (16,17). The crystal structure of the CIDE-N domain, including aa 39 -119, has been resolved (17,35). The structure revealed the formation of a homodimer between two CIDEC proteins through the interaction of their CIDE-N domains. CIDEC proteins also interact with Perilipin1 (PLIN1), an activator of CIDEC-mediated LD fusion via the CIDE-N domain (17).
In the last few years, lipidomics study in mammalian cells revealed that phosphatidylcholine (PC) is the most abundant component of the phospholipids on LD. Other minor components include phosphatidylethanolamine (PE), phosphatidylinositol (PI), phosphatidylserine (PS), and ether-linked PC (36,37). PC on LD serves as the protective agent, preventing LD coalescence due to its cylindrical shape (38). However, it is phosphatidic acid (PA) that was reported to facilitate LD fusion (39,40). In fact, acidic phospholipids cluster on focal points of ϳ11 nm in diameter on the plasma membrane and associate with caveolae structures of 60 -80 nm (41,42). In addition, several groups have recently shed light on the protein structure and functional regulation through ionic interactions between proteins and lipids (43)(44)(45).
Here, we identified a unique function of CIDEC linker, which connects the CIDE-N and CIDE-C domain. This linker, in particular the polybasic RKKR motif, serves as an inhibitory domain, reducing LD fusion activity, regulating the sizes of LDs. The inhibitory function requires the interaction between CIDE-N domains in a different manner from PLIN1 activation. Mechanistically, the RKKR motif interacts with acidic phospholipids electrostatically. These data reveal a novel functional domain and provide evidence for the synergistic outcome of CIDE-N and CIDE-C domains in regulation of LD fusion and growth by interaction with phospholipids.

Linker region of CIDEC inhibits LD fusion and growth
CIDEC comprises the CIDE-N and CIDE-C domains, connected through a linker region (Fig. 1A). The function of the linker, including aa 120 -135, remains unknown. Here, we seek to understand the role of the linker domain on CIDEC by first studying the LD phenotype upon 16-aa (linker) deletion. Fulllength CIDEC and CIDEC with linker deletion (⌬120 -135) were both tagged with GFP on the C terminus and overexpressed in 3T3-L1 pre-adipocytes. Cells expressing full-length CIDEC presented larger Bodipy C12-stained LDs compared with the GFP-expressing cells (Fig. 1B). Interestingly, in CIDEC(⌬120 -135)-GFP-expressing cells, larger and fewer LDs were found in contrast to LDs in full-length CIDEC-GFPexpressing cells (Fig. 1B). In the images acquired, the localization of CIDEC(⌬120 -135)-GFP was similar to that of full-length CIDEC with the green fluorescent signals enriched at LDCS (Fig. 1, B and C). Next, ϳ100 images of GFP-positive cells were acquired, and the diameter of the largest LD in each cell was measured (Fig. 1D). As expected, full-length CIDEC expression mediated LD fusion as shown with the larger average LD diameter compared with the control GFP-expressing cells (3.8 Ϯ 1.3 m versus 0.7 Ϯ 0.2 m, respectively). Supporting the imaging data, in the absence of the linker, the average size of the largest LD was ϳ1.5ϫ larger than that of the LDs in cells expressing full-length CIDEC (5.4 Ϯ 1.9 m, Fig. 1D). For easy characterization, LDs with a diameter larger than 4 m were defined as supersized LDs. The classification of LDs showed that 76% of CIDEC(⌬120 -135)-GFP-expressing cells exhibited supersized LDs (Fig. 1E). In contrast, only 38% of full-length CIDEC-GFP-expressing cells contained supersized LDs (Fig. 1E). Moreover, there were significantly fewer LDs in cells overexpressing CIDEC(⌬120 -135)-GFP (10 Ϯ 10 LDs) as compared with cells expressing full-length CIDEC (21 Ϯ 19 LDs) or GFP (296 Ϯ 162 LDs) (Fig. 1F). The results indicate that CIDEC(⌬120 -135) can further promote LD fusion and growth.
To understand the enlarged LD phenotype in the presence of CIDEC(⌬120 -135), we next analyzed the neutral lipid exchange between LDs using fluorescence recovery after photobleaching (FRAP) (16,17). This is a surrogate assay to demonstrate the opening status of the fusion pore during LD fusion. Here, a significantly shorter signal recovery time was observed for the LD pairs of CIDEC(⌬120 -135)-expressing cells (11.2 s for CIDEC(⌬120 -135) versus Ͼ 56 s for full-length CIDEC) (Fig. 1G). Quantitative analysis showed that the lipid-exchange rate between LD pairs for CIDEC(⌬120 -135)-GFP (2.87 Ϯ 2.03 m 3 /s) was 6.5-fold faster than that for full-length CIDEC-GFP (0.44 Ϯ 0.45 m 3 /s) (Fig. 1H). Next, we visualized and measured LD fusion process using time-lapse differential interference contrast (DIC) imaging. It is noted that without the 16-aa linker, LD fusion was completed in a much shorter time of 58 min for a donor LD of 3 m in diameter, compared with 198 min for similarly-sized LD pairs in cells expressing fulllength CIDEC (Fig. 1I). Statistically, the lipid transfer rate assay demonstrated that CIDEC(⌬120 -135)-GFP-induced neutral lipid transfer rate was 4.4-fold faster than that of full-length CIDEC in average (0.018 Ϯ 0.014 m 3 /min for full-length CIDEC versus 0.079 Ϯ 0.054 m 3 /min for CIDEC(⌬120 -135)) ( Fig. 1J). Importantly, there is no change in the total volume of LDs in CIDEC(⌬120 -135)-expressing cells with enhanced LD fusion phenotype (Fig. S1A). These data indicate that the absence of linker has no effect on total TAG content in 3T3-L1 pre-adipocytes in this short experimental period of 20 -24 h. Thus, quantitatively, we uncover that the linker region (aa 120 -135) significantly regulates LD fusion activity.

Interaction between CIDE-N domains is required for the enhanced LD fusion phenotype mediated by CIDEC(⌬120 -135)
The striking LD phenotype caused by the deletion of the 16-aa linker is evident. Because CIDEC proteins homodimerized via the CIDE-N domain as they enriched at LDCS, we ask whether this homodimerization is required for the enhanced LD fusion mediated by CIDEC(⌬120 -135). Using

CIDEC-phospholipid interaction regulates LD fusion activity
immunoprecipitation assay, the data showed successful pulldown of CIDEC(⌬120 -135), with levels comparable with fulllength CIDEC ( Fig. 2A). In addition, in the cycloheximide (CHX) time-chase experiment, CIDEC(⌬120 -135) and fulllength CIDEC displayed a similar degradation rate (Fig. 2B). The results reveal that promotion of supersized LDs via deletion of the linker on CIDEC was not due to stronger interaction between the mutant proteins or enhanced protein stability. Taken together, the increased LD fusion rate induced by CIDEC(⌬120 -135) is neither the result of enhanced proteinprotein interaction nor increased protein level.
Next, we ask whether the linker collaborates with the N-and C-domains intramolecularly. A number of different CIDEC deletion and mutation plasmids were constructed and examined in the pre-adipocytes (Fig. 2C). These constructs include the CIDEC QQN mutant and several N-and C-terminal truncations. The former is defective in N-terminal interaction (QQN-CIDEC) and thus is impaired in LD fusion. Co-expression of PLIN1 can rescue this deficiency through its interaction with the CIDE-N domain (17). With these constructs, lipidexchange rates were measured and compared between fulllength CIDEC, CIDE-N domain (aa 120 -239) truncation, and CIDE-N-linker deletion (aa 136 -239) (Fig. 2C). Data showed that in the absence of CIDE-N, regardless of the presence of linker, lipid-exchange rates were significantly reduced (0.40 Ϯ 0.34 m 3 /s for full-length CIDEC; 0.24 Ϯ 0.18 m 3 /s for N-domain-deleted CIDEC; and 0.15 Ϯ 0.12 m 3 /s for N-domain and linker-deleted CIDEC) (Fig. 2, D and E). This is unlike the enhanced lipid-exchange rate observed when only the linker was deleted in Fig. 1H. The data indicate that the promotion of LD fusion via deletion of the linker region requires the presence of the CIDE-N domain. Next, we asked whether a functional CIDE-N is important. Using the QQN mutant that is defective in homodimerization, an additional 16-aa deletion was constructed (QQN-CIDEC(⌬120 -135)). The lipid-exchange assay performed showed that the absence of the linker region could not rescue the low lipid-exchange rate between LDs in the cells expressing CIDEC with the defective CIDEC-N domain (Fig. 2F). Our data imply that a functional N-domain is a prerequisite for the regulatory role of the linker. Interestingly, when PLIN1 was co-expressed with QQN-CIDEC(⌬120 -135), the exchange rate was increased significantly compared with the cells co-expressing QQN-CIDEC with PLIN1 (Fig. 2, F and G). This indicates that the deletion of the linker region could recover its promotion for LD fusion so long as PLIN1 can rescue the N-terminal defect of QQN-CIDEC. These data suggest that the regulatory effect of the CIDEC linker on LD fusion is a downstream step after N-domain dimerization.

RKKR motif in the linker regulates LD size
To identify possibly conserved amino acids in the linker responsible for the LD size regulation, we compared the sequences of human and murine CIDE family proteins corresponding to this domain (aa 120 -135 of murine CIDEC) (Fig. S1B). Our analysis revealed low conservation among the six proteins. To determine whether the linker of other members of the CIDE family has similar restrictions on LD fusion and growth, we compared the LD sizes in cells expressing murine full-length CIDEA-GFP, CIDEA(⌬112-120)-GFP, and CIDEA(⌬112-125)-GFP. Interestingly, instead of enhancing LD fusion and growth, the two CIDEA linker deletion mutants displayed smaller LD sizes (Fig. S1, C and D). These results indicate that CIDEC linker (aa 120 -135) deletion is unique in enlarging LD sizes.
The enhanced performance of CIDEC(⌬120 -135)-GFP in LD fusion and growth may result from one of the two possible reasons, namely close proximity between CIDE-N and CIDE-C domains and removal of inhibitory amino acids. To examine the two possibilities, two deletion CIDEC constructs, namely CIDEC(⌬120 -126)-GFP and CIDEC(⌬127-135)-GFP, were constructed. The rationale of these additional constructs is as such: if shortening the distance between CIDE-N and CIDE-C enhances the LD fusion activity, then expressing CIDEC(⌬120 -135) with a 16-aa deletion in cells will result in the strongest LD size promotion. This should be followed by CIDEC(⌬127-135) with a 9-aa deletion. The least effective construct will be CIDEC(⌬120 -126) with a 7-aa deletion. However, in the various assays, CIDEC(⌬120 -126)-GFP expression resulted in the same phenotype as the 16-aa deletion with an increase in the number of supersized LDs in the cells (Fig. S1, E and F) upon protein normalization (see Fig. S1H). CIDEC(⌬127-135)-GFP had little effect on the promotion of LD size. These data indicate that reducing the distance between the CIDE-N and CIDE-C is not the reason for the enhanced LD fusion. Next, in contrast to CIDEC(⌬127-135)-GFP, CIDEC(⌬120 -135)-GFP and CIDEC(⌬120 -126)-GFP expressions exhibited fewer LDs due to increased LD fusion (Fig. S1G). Thus, the results reveal that aa 120 -126 are likely to be the key residues in CIDEC responsible for LD size regulation.

CIDEC-phospholipid interaction regulates LD fusion activity
Next, to further dissect for the key residue, we substituted CIDEC aa 121-134 with the exception of aa 127, 129, and 133 with alanine and measured the LD sizes when overexpressed in 3T3-L1 pre-adipocytes. The result is summarized in Fig. 3A. Interestingly, we found that single mutation of any of the RKKR residues (RKKR/A) within aa 123-126 led to a comparable ability to promote LD growth and increase in the proportion of supersized LDs as CIDEC(⌬120 -135) (Fig. 3, B and C). The phenotypes caused by the point mutations of two residues, K125A and R126A, appeared to be similar to CIDEC(⌬120 -  135). Western blotting indicated the enhanced effect was not due to the differences in protein expression level (Fig. 3D). In addition, similar to the deletion of the linker (aa 120 -135), the mutation RKKR/A could individually enhance the lipid-exchange rates between the LD pairs, which were faster than that for full-length CIDEC-GFP (0.44 m 3 /s for full-length CIDEC, 1.72 m 3 /s for R123A, 1.43 m 3 /s for K124A, 1.62 m 3 /s for K125A, and 1.80 m 3 /s for R126A; Fig. 3E). Conversely, other point mutations did not induce the formation of supersized LDs (see Fig. 3, A, F, and G). The results show that the RKKR motif in the linker region of CIDEC is the key sequence that restricted LD fusion and the formation of supersized LDs.

CIDEC-phospholipid interaction regulates LD fusion activity
To examine whether the inhibitory function of murine CIDEC linker region is conserved in human CIDEC, we deleted aa 120 -134 of human CIDEC corresponding to the murine CIDEC linker region and compared the lipid-exchange rates of these three constructs when expressed in 3T3-L1 pre-adipocytes (Fig. S2, A and B). The data showed that there was no difference in activities between human and murine CIDEC, and there was no significant difference between full-length and ⌬120 -134 human CIDEC. These results indicate that the human CIDEC linker region has no inhibitory function as a murine CIDEC. Next, to confirm the difference in function between the human and murine CIDEC linker regions, Arg-125 of human CIDEC was mutated to either Ala (R125A) or Gln (R125Q) and expressed in 3T3-L1 pre-adipocytes. The lipidexchange rates were measured and compared with murine CIDEC R126A and R126Q mutants. Consistently, murine CIDEC R126A and R126Q enhanced the lipid-exchange rate significantly. This was, however, not observed when human CIDEC R125A/R125Q mutations were expressed (Fig. S2, C  and D). The data further ascertain that the human CIDEC linker region has no role in the regulation of fusion activity. Taken together, the unique inhibitory function of the RKKR motif in the regulation of LD fusion is specific to murine CIDEC protein.

Polybasic RKKR motif interacts with acidic phospholipids
The gain-of-function for LD fusion by the deletion or mutation of CIDEC is intriguing. Some researchers reported that polybasic residue clusters composed of lysine and arginine are inclined to interact with acidic phospholipids (46 -49). This kind of electrostatic attraction is important physiologically, thus enabling protein binding to membrane, protein enrichment, and protein-protein interaction (46 -49). For cellular LDs, the composition of the phospholipid monolayer depends on the cell type but mostly consists of neutral phospholipids PC, PE, and acidic phospholipids that possess negatively charged polar heads. These include PS, PA, and PI (50). Here, we hypothesized that CIDEC interacts with the phospholipids via the polybasic RKKR motif. We adopted an aromatic fluorescence emission assay to detect the interaction between the peptides and phospholipids (Fig. 4A). When peptides with aromatic amino acids were attached to a liposome composed of phospholipids, there was a decrease in their movement in the solution. This was followed by an increase in emission intensity, which can be captured in the presence of an excitation light of wavelength 280 nm. Here, WT peptide and the RKKR mutant (4A) peptide, both with a tryptophan, were synthesized and used in a tryptophan fluorescence emission assay (Fig. 4A). The calculation of their net charge shows that the WT peptide is positively charged, whereas the mutated RKKR to alanine reduces positive charges (Fig. 4A). Assay data show that the WT peptides displayed increased association with PS, PA, and PI in a dose-dependent manner (Fig. 4, B-D). Interestingly, the WT peptides had a lower saturation concentration of 20 M when incubated with PA liposomes compared with PS and PI liposomes, indicating stronger interaction with PA. It is noted here that the WT peptide was unable to interact with PC liposomes (Fig. 4E). The results show that the WT peptide can specifically bind to acidic phospholipids, especially PA. By contrast, the 4A peptides completely lost the ability to bind to PS, PA, PI, or PC liposomes (Fig. 4, F-I) with unchanged emission intensity in the presence of different concentrations of phospholipids. The data imply that the loss of the RKKR motif disrupted their electrostatic interaction with the acidic phospholipids. In addition, binding to acidic phospholipids by the control peptide with alanine replacement on Lys-134 (K134A) remained unaffected (Fig. S3, A-C). The results are summarized in table format in Fig. 4J. Next, we ask whether all four polybasic residues were required for the interaction with PA. Here, three additional mutant peptides with single, double, or triple basic amino acid replacements were constructed. They are K125A (1A), K125A/ R126A (2A), and K124A/K125A/R126A (3A). Using these peptides, tryptophan fluorescence emission experiments were performed and compared with WT and 4A peptides (Fig. 4K). The results showed that the binding activities of these mutant peptides reduced progressively with each successive point mutations to alanine. The ranking of binding activities is as follows: In other words, the K125A mutant had a weaker binding ability than the WT linker region of CIDEC. These data indicate that with more elimination of the positive charge amino acid on the polybasic motif, concomitantly, the binding activity of the peptide to PA was reduced. These data indicate that the CIDEC-lipid interaction depends on the number of polybasic residues. Every positive charge loss will gradually reduce the binding activity of the peptide to PA. Circular dichroism assay was commonly used to survey the peptide-lipid interaction. The PA molecule was reported to play an important role in CIDEA-mediated LD fusion (39). Here, we analyzed the CD spectra of both WT and 4A mutant peptides in aqueous solution or with PC or PA liposomes. The results revealed that WT or 4A mutant peptides were unordered in buffer solution and in the presence of PC liposomes. However, in the presence of PA liposomes, both peptides exhibited ␣-helix conformation (Fig. S3, D and E). Furthermore, the conformational change to ␣-helix for WT peptide was relatively higher compared with 4A mutant peptides, indicating the presence of a stronger interaction between the WT peptide and PA phospholipid. The loss of polybasic residues resulted in much weaker interaction between peptide and phospholipid. Here, it is revealed that the polybasic RKKR motif on CIDEC preferably binds to the acidic phospholipids by means of electrostatic interaction.

Loss of polybasic RKKR motif promotes CIDEC-mediated LD fusion and growth
Finally, to further demonstrate the positively charged RKKR motif indeed has an effect on the LD size regulation, we replaced the RKKR with glutamine. Glutamine is similar in size to arginine and lysine but is electrically neutral (Fig. 5A). Expectantly, the RKKR/Q mutants behaved similarly to aa 120 -126 deletion, inducing the formation of supersized LDs when overexpressed (Fig. 5, B-D). These data indicate that the gain-offunction by the mutant was mainly due to the change in the charge property of the residues instead of their size. This is further corroborated; aa 123-126 (RKKR) were individually

CIDEC-phospholipid interaction regulates LD fusion activity
replaced with negatively charged glutamic acid. The results show that the RKKR/E expression displayed a similarly enhanced LD fusion and growth effect to that of CIDEC(⌬120 -126) (Fig. 5, E-G). Taken together, these data confirm that the promotion of supersized LD formation via LD fusion was primarily due to the loss of the polybasic RKKR motif.

Discussion
In this work, we first identify a unique regulation by aa 120 -135 in CIDEC-mediated LD fusion. Deletion of the linker promotes the formation of supersized LDs. This enhanced LD fusion phenotype is the result of increased lipid exchange and lipid transfer rates. The formation of supersized LDs requires the interaction between CIDE-N domains and works differently from PLIN1. Importantly, this regulation is determined by the polybasic RKKR motif in the linker. These data provide new insight into the regulation of CIDEC-mediated LD fusion through electrostatic interaction between the RKKR motif and the acidic phospholipids on the LD surface.
Our long-standing interest on the CIDE proteins mediating LD fusion drives this work to further dissect the CIDEC molecule and the function of its domains. Previously, our group reported that the CIDE-C domain possesses the ability to mediate LD fusion albeit with lower activity (16). The CIDE-N domain possesses a conserved Lys-56 acetylation site, which is important for its protein stability and fusion activity (51). Recently, two types of protein structures were found to associate with LD, amphipathic ␣-helices and hydrophobic hairpins (50). Interestingly, the CIDE-C domain of CIDEC also contains amphipathic ␣-helices and hydrophobic regions (16), in which the positively charged KKR amino acids play a critical role in CIDEC- mediated LD fusion (16, 20). Furthermore, it was reported that the interaction of the CIDE-C domain with PA might be necessary for LD fusion (39). Thus, we propose that the polybasic motif, including lysine and arginine on CIDEC, is a crucial regulator for LD fusion activity. This type of polybasic motif-phospholipid interaction is commonly found in the biological system. Well-known examples include some small GTPases (Ras, Rab, Arf, and Rho proteins) with polybasic clusters being targeted to the plasma membrane through interaction with phospholipids such as phosphatidylinositol 3,4,5-trisphosphate and phosphatidylinositol 4,5-bisphosphate (52). The polybasic domain was also reported to regulate the phosphorylation of Lipin1 (53). Separately, a polybasic KKR motif in NiV-F protein modulates cell-cell fusion (54). The polybasic region of Rac1 was proven to interact with the effector protein PRK1 (55). In our study, we confirmed that the polybasic RKKR motif in the linker region of CIDEC interacts with the acidic phospholipids on the LD monolayer. Interestingly, double mutations of the RKKR motif had a similar effect as the single mutation to promote the formation of supersized LDs (data not shown). We speculate that any change to the RKKR motif is sufficient to induce conformational change on CIDEC, making it difficult to interact with the phospholipids on the highly curved LD surface.
Based on the above observations, we propose a model to interpret the effect of the polybasic RKKR motif on CIDEC during LD fusion (Fig. 6). CIDEC comprises three domains: CIDE-N domain, CIDE-C domain, and the linker region. When two LDs are in close range, the CIDE-N domains of the CIDEC molecules homodimerize at LDCS, facilitating local enrichment of the CIDEC proteins. The RKKR motif in the linker region interacts with the acidic phospholipids of the LD mono-

CIDEC-phospholipid interaction regulates LD fusion activity
layer through electrostatic attraction. With these, the conformational freedom of the CIDE-N domain is further restricted by the CIDEC-phospholipid interaction and enables the assembly of protein complexes at the LDCS for the formation of fusion pore, allowing neutral lipid exchange. The internal pressure difference derived from LD surface tension leads to lipid transfer from the donor to the acceptor LD. When the polybasic RKKR motif is mutated or deleted, the interaction of the linker region with acidic phospholipids is disrupted, and CIDEC proteins are unrestrained, achieving conformational freedom between the CIDE-N and CIDE-C domains. With the CIDE-N dimerization holding the CIDEC proteins in place, the newly formed fusion pore at the LDCS is enlarged and leads to enhanced LD fusion and supersized LD.
This investigation provides insight into CIDEC's role in LD fusion and growth. However, the regulatory role of the RKKR motif in vivo remains to be determined. In addition, future works are required to determine whether the motif directly influences the fusion pore size and whether the motif works collaboratively with the two CIDE-domains. In summary, more detailed studies are needed to uncover the relationship of the CIDEC-lipid interaction with the enhanced LD fusion ability.

Reagents and antibodies
M2 beads and sodium oleate (OA) were purchased from Sigma. WT and mutant CIDEC peptides were synthesized at GenScript. Phospholipids were purchased from Avanti Polar Lipid. Bodipy C12 was purchased from ThermoFisher Scientific. The antibody against CIDEC was used as described previously (16, 17). Antibodies against ␤-actin (mouse, A5441) and FLAG (mouse, F1804) were purchased from Sigma. Antibodies against GFP (rabbit, sc-8334) and HA (mouse, sc-7392) were purchased from Santa Cruz Biotechnology.

Plasmid DNA construction
Full-length cDNAs encoding CIDEC, CIDEA, and PLIN1 were amplified from cDNA of mouse adipose tissue and cloned into pEGFPN1, pCMV5-HA, or pCMV5-FLAG vectors. Deletion fragments of CIDEC and CIDEA were amplified from fulllength CIDEC or CIDEA by appropriate primers. Full-length cDNA encoding human CIDEC was amplified from cDNA of human adipose tissue and cloned into the pEGFPN1 vector. The deletions were subcloned into XhoI-EcoRI sites of pEGFPN1 or NdeI-BamHI sites of pCMV5-HA and pCMV5-FLAG vectors. Point mutations of CIDEC were generated using PCR site-directed mutagenesis from full-length CIDEC. All plasmid DNA constructs were sequenced to guarantee the fidelity.

Fluorescence imaging
Cells were transfected with the indicated GFP vectors and grown on a coverslip. Subsequently, the coverslips of cells were immersed in medium supplemented with 1 g/ml Bodipy 558/ 568 C12 fatty acid (Molecular Probes) and 200 M OA for 20 -24 h. Cells were washed twice with PBS, fixed with 4% paraformaldehyde, followed by incubation with Hoechst in PBS. For fixed cells, images were collected using a Nikon A1Rϩ confocal microscope with a ϫ100 oil-immersion objective and three lasers (405/488/561 nm in wavelength). For z-stack images, images with a step interval of 0.2 m were acquired. Live cell imaging was performed as described previously (17).

Size measurement of lipid droplets
Cells with GFP fluorescent signals were randomly selected from transfected cells. Images were taken using a Carl Zeiss inverted microscope (Axiovert 200M) with a ϫ63 oil-immersion objective. The diameter of a largest LD in each cell was measured by using ImageJ software. The sizes of largest LDs were analyzed and plotted using GraphPad Prism 5.

Lipid-exchange rate assay
Lipid-exchange rate assay was performed as described previously (17) with some modifications. 3T3-L1 pre-adipocytes were transfected with the indicated plasmids (CIDEC and PLIN1) and were incubated with 200 M OA and 1 g/ml Bodipy 558/568 C12 fatty acid (Molecular Probes) for 18 h. Cells were transferred to fresh medium 1 h before FRAP experiments. Cells were cultured in a living cell station (Oko laboratory) and viewed under a confocal microscope (A1Rsi, Nikon) using a ϫ100 oil-immersion objective. Digital detector gain and laser power were set to avoid overexposure and to ensure accurate quantification of fluorescence. The bleaching area of a LD was set to ensure that at least 70% of its total area was covered and bleached by 120 ms at 100% laser power of a 561-nm solidstate laser. This was followed by time-lapse scanning of 50 frames with an interval of 1.12 s. The bleaching process was performed on the smaller LD three times for statistical analyses. As there are more fluorescent molecules in the larger LD of the LD pair, bleaching these large LDs more than twice will result in enormous reduction in fluorescent intensity. This in turn may cause the fluorescent intensities within the cores of the LD pair to decline rapidly and pose difficulty in the detection of the fluorescence recovery. Mean optical intensities within the LD core regions were measured simultaneously. A set of LD pairs was analyzed in each experiment, and unbefitting data were filtered out based on the criteria in the exchange-rate assay.

Lipid transfer rate assay
Live cells overexpressed with the indicated CIDEC-GFP plasmid were incubated on a living cell station (Oko laboratory). Differential interference contrast (DIC) images were acquired using a Nikon A1Rϩ confocal microscope with a CIDEC-phospholipid interaction regulates LD fusion activity ϫ100 oil-immersion objective. Time-lapse images were taken with an interval of 2 min for at least 4 h.

Image processing and analysis
Images were processed using NIS-element analysis (Nikon) or ImageJ. Cropping of images was done using Photoshop 5 (Adobe). The LD number per cell was determined using Imaris 9 (Bitplane) with a spot mode. The LD volume per cell was measured using NIS-element software (Nikon).

Protein stability assay
The stability of CIDEC proteins was analyzed based on CHX protein-chase experiment as described previously (56). 293T cells were transfected with FLAG-tagged full-length CIDEC or its deletion. Twenty hours after transfection, the medium was replaced with fresh medium supplemented with CHX (100 g/ml). Cells were harvested at the indicated time points (0, 15, 30, and 60 min).

Immunoprecipitation
293T cells were transfected with HA-and FLAG-tagged CIDEC plasmids and cultured in an incubator for 20 -24 h. Cells were lysed with IP buffer (20 mM HEPES, 150 mM NaCl, 1% EDTA, 1% EGTA, 1% Trion X-100, 1% phenylmethylsulfonyl fluoride) and centrifuged to remove nuclei. The supernatant was incubated with M2 beads (Sigma) for 2-3 h followed by three washes. The M2 beads were added with loading buffer and separated on SDS-PAGE.

Liposome preparation
1 ml of liposome stock consisting of 5 mM of the indicated phospholipid was prepared by dissolving appropriate weight of lipids according to their molecular weight in 1 ml of chloroform. The lipid/chloroform solution was dried under vacuum by rotary evaporation to form a homogeneous lipid layer. The lipid thin film was dissolved in 1 ml of distilled water, frozen, and thawed for five times with liquid nitrogen to form large unilamellar vesicles (LUVs). The LUVs were homogenized with an extruder through a 0.1-m filter (Whatman) 15 times to form liposomes.

Tryptophan fluorescence emission
WT and mutation peptides composed of CIDEC aa 115-135 at a final concentration of 2 M were incubated with liposomes, which comprise electroneutral phospholipids or acidic phospholipids, in Tris-HCl buffer (100 mM, pH 7.4) at room temperature for 5 min and then detected on a Varian Cary Eclipse machine. The concentrations of phospholipids in the form of liposomes were 10, 20, 40, 50, and 100 M, respectively. The excitation wavelength was 280 nm, and the scanning wavelength of emission light was set at 300 -450 nm to cover the emission peak. The spectra were acquired with a scan rate of 120 nm/min, a data interval of 1 nm, and a photomultiplier tube voltage of medium. The results were analyzed by subtraction of the appropriate buffer or liposome fluorescence spectrum.

CD
Circular dichroism (CD) experiments were performed using a Jasco J-715 spectrometer (Jasco, UK) with indicated peptides (GenScript) at a concentration of 20 M in 2 mM Tris-HCl, pH 7.4. A 2-mm path-length quartz cuvette (Hellma) was adopted in the experiments. The spectra were recorded between 190 and 250 nm with a data pitch of 0.2 nm, a bandwidth of 2 nm, a scan rate of 200 nm/min, and a response time of 1 s. CD spectra were recorded with peptide samples incubated with or without PC or PA liposomes of 200 M at 22°C. Data shown were averaged from three independent spectra after subtraction of the appropriate buffer or liposomes CD spectrum.

Statistics
Statistical analyses of all data were done using GraphPad Prism 5. The two-tailed Student's t test was performed for a single comparison, and the one-way ANOVA with Tukey post hoc test was performed for multiple comparisons. Differences were considered to be significant at p Ͻ 0.05. p values were indicated in each figure as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.