Lipid-associated PML domains regulate CCTα, Lipin1 and lipid homeostasis

Nuclear LDs (nLDs) originate at the inner nuclear membrane by a mechanism that involves the promyelocytic leukemia (PML) protein. Here we demonstrate that nLDs in oleate-treated U2OS cells are associated with Lipid-Associated PML (LAP) domains that differ from canonical PML nuclear bodies by the relative absence of SUMO1, SP100 and DAXX. nLDs were also enriched in CTP:phosphocholine cytidylyltransferase α (CCTα), the phosphatidic acid phosphatase Lipin1 and diacylglycerol (DAG). High resolution imaging revealed that LAP domains and CCTα occupy distinct polarized regions on nLDs, and that loss of LAP domains in PML knockout U2OS cells reduced the recruitment of CCTα onto nLDs by its amphipathic α-helical M-domain. The association of Lipin1 and DAG with nLDs was also LAP domain-dependent. The disruption of CCTα and Lipin1 localization on nLDs in PML knockout cells resulted in the inhibition of phosphatidylcholine and triacylglycerol synthesis indicating that LAP domains are a unique PML subdomain involved in nLD assembly and regulation of lipid metabolism.


INTRODUCTION
Lipid droplets (LDs) are cellular organelles composed of a core of triacylglyceride (TAG) and cholesterol ester (CE) surrounded by a monolayer of phospholipids and associated proteins.
In response to nutrient and hormonal signals, fatty acids and cholesterol are stored in or released from LDs to provide energy and biosynthetic precursors as well as buffer the cell from fatty acid toxicity (Sztalryd and Brasaemle, 2017). The largest neutral lipid storage depot is in adipocytes but hepatocytes, enterocytes and macrophages also have the capacity for short-term storage and release of fatty acids from LD (Walther and Farese, 2012). Accumulation of LDs in tissues, caused by an imbalance in lipid storage and hydrolysis, is linked to pathological conditions such as hepatosteatosis, obesity and lipodystrophy.
LDs are proposed to form in the endoplasmic reticulum (ER) by a process that requires the coordinated synthesis of TAG and phospholipids (Henne et al., 2018). TAG initially forms a 'lens' in the ER bilayer, expands and pinches off into the cytoplasm, and is coated with ER-derived phospholipids, which regulate surface tension and the storage capacity of LDs (Krahmer et al., 2011). Phosphatidylcholine (PC) is the most significant component of the surface monolayer of LDs (Bartz et al., 2007;Tauchi-Sato et al., 2002), and de novo PC synthesis by the CDP-choline pathway in the ER is required for LD biogenesis (Aitchison et al., 2015;Krahmer et al., 2011). In mammalian cells, the CDP-choline pathway is regulated by CTP:phosphocholine cytidylyltransferase (CCT) α and b isoforms that are activated by translocation to nuclear and cytoplasmic membranes, respectively, in response to the content of PC and specific lipid activators sch as fatty acids or type II conical-shaped lipids (e.g. diacylglycerol (DAG) or phosphatidylethanolamine (PE)) (Arnold and Cornell, 1996;Arnold et al., 1997;Xie et al., 2004).
During LD biogenesis in fatty acid-treated cells (Gehrig et al., 2008;Lagace and Ridgway, 2005) 13 and during adipocyte differentiation (Aitchison et al., 2015), PC synthesis was increased by CCTa translocation from the nucleoplasm to the inner nuclear membrane (INM). In oleate-treated insect cells, ectopically expressed CCTα and the insect homologue CCT1 translocated from the nucleus to the surface of cytoplasmic LDs (cLDs) resulting in increased PC synthesis to facilitate LD expansion and TAG storage (Guo et al., 2008;Krahmer et al., 2011;Payne et al., 2014).
Mammalian CCTα can exit the nucleus under some conditions (Gehrig et al., 2009;Northwood et al., 1999) but does not localize to the surface of cLDs in adipocytes and other cultured cells (Aitchison et al., 2015;Haider et al., 2018).
Interestingly, nuclear lipid droplets (nLDs) were recently identified as site for CCTa translocation and activation of PC synthesis in oleate-treated Huh7 hepatoma cells (Ohsaki et al., 2016). nLDs account for »10% of the total cellular LD pool in liver sections and hepatocytes, and have a unique lipid and protein composition (Lagrutta et al., 2017;Uzbekov and Roingeard, 2013).
In hepatocytes, nLDs can arise from TAG-rich droplets in the ER lumen that are precursors for VLDL assembly and secretion (Soltysik et al., 2019). These luminal LDs migrate into invaginations of the INM termed the type I nucleoplasmic reticulum (NR) and are released as nascent nLDs into the nucleoplasm where they increase in size by fusion with other nLDs or by de novo TAG synthesis (Ohsaki et al., 2016;Soltysik et al., 2019). During or after release into the nucleoplasm, nLDs associate with promyelocytic leukemia (PML) protein, in what were at the time referred to as PML nuclear bodies (PML NB) (Ohsaki et al., 2016). PML NB are dynamic subnuclear domains that regulate gene expression and stress responses and are strongly associated with proteins modified by the small ubiquitin modifier 1 (SUMO1) (Ching et al., 2005;Dellaire and Bazett-Jones, 2004;Van Damme et al., 2010), the most prominent being SP100 nuclear antigen and the death associated domain protein DAXX (Ishov et al., 1999;Zuber et al., 1995) 13 However, of the seven known isoforms of PML, only the PML II isoform was associated with nLDs and required for their formation (Ohsaki et al., 2016). Moreover, the presence of CCTα and perilipin3 on nLDs suggests that these lipid-associated PML-containing domains can regulate lipid synthesis (Soltysik et al., 2019), possibly to generate a metabolic signal to accommodate the uptake and storage of excess fatty acids.
Here we sought to examine the role of PML in nLD biogenesis and function by employing CRISPR/Cas9 to knockout the PML gene in human U2OS osteosarcoma cells (PML KO). We found that nLDs in oleate-treated U2OS cells have Lipid-Associated PML (LAP) domains that are deficient in the canonical PML NB proteins SUMO1, SP100 and death-associated protein 6 (DAXX), and enriched in the lipid biosynthetic enzymes CCTa and Lipin1. Results using PML KO cells indicate that LAP domains are a novel nuclear subdomain complex required for nLD assembly and the recruitment of key enzymes for PC and TAG synthesis.

nLDs harbour non-canonical lipid-associated PML (LAP) domains
Similar to hepatoma cells, U2OS osteosarcoma cells contain abundant nLDs that are associated with mCherry-PML-II (Ohsaki et al., 2016). As additional evidence of their nucleoplasmic localization, nLDs in oleate-treated U2OS cells have endogenous PML and CCTα on their surface (Supplemental Fig. S1), and PML-positive nLDs were not encapsulated by the nuclear envelope (NE) based on immunostaining for emerin (Supplemental Fig. S2). Thus, U2OS cells offer an alternate and complementary model to investigate the structure and function of nLDs.
Initially we investigated how PML structures on nLDs differ from canonical PML NB by immunostaining control and oleate-treated U2OS cells for the PML NB-associated proteins SUMO1, SP100 and DAXX ( Fig. 1 and Supplemental Fig. S3 and S4). U2OS cells contain numerous PML NB that are positive for SUMO1, consistent with the essential role of SUMOylation in PML NB assembly and function (Zhong et al., 2000) (Fig. 1A). Oleate treatment caused the loss of SUMO1-positive PML NB puncta and the appearance of PML-positive nLDs with different levels SUMO1 expression (Fig. 1B). When quantified, a weak or non-existent SUMO1 signal was detected in 75% of PML-positive nLDs ( Figure 1C). In addition, DAXX and SP100, proteins whose interaction with canonical PML NB is SUMO-dependent (Ishov et al., 1999) or constitutive (Sternsdorf et al., 1997), respectively, were strongly localized to PML NB in untreated cells but absent or reduced in 80% of PML-positive nLDs in oleate-treated cells (Fig. 1C,Supplemental Fig. S3 and S4). Since the PML structures on nLDs are part of a large lipid complex and relatively devoid of canonical PML NB proteins, we propose that they be designated as Lipid-Associated PML (LAP) domains.
To investigate how LAP domains influence the lipid and protein composition and biogenesis of nLDs, we utilized U2OS cells in which the PML gene was knocked out (PML KO) by CRISPR/Cas9 gene editing ( Fig. 2A) (Attwood et al., 2019). In oleate-treated U2OS cells, CCTα was expressed in the nucleoplasm and on the surface of nLDs (Fig. 2B); whereas, in PML KO cells there were fewer BODIPY-and CCTα-positive nLDs, and CCTα was partially localize to the NE. Compared with control cells, the number of cLDs in PML KO cells was reduced slightly CCTα-positive nLDs in control and PML KO cells were similar in size suggesting the enzyme preferentially associates with larger nLDs that are formed by a PML-independent mechanism.
To identify the PML isoform involved in nLD formation in U2OS cells, PML KO cells were individually transfected with GFP-tagged version of the 7 PML isoforms. In agreement with results in Huh7 cells (Ohsaki et al., 2016), GFP-PML-II associated with nLDs and increased their abundance in PML KO cells while other PML isoforms did not associate or correct the PML KO phenotype (Supplemental Fig. S5A and B, results not shown). GFP-PML-II expression in PML KO cells significantly increased the number and size of nLDs to the level observed in U2OS cells (Supplemental Fig. S5C and E) and also restored the slight reduction in parameters for cLDs (Supplemental Fig. S5D and F). The specific requirement for PML-II in nLD formation is further evidence that LAPs are a unique nuclear PML subdomain.

CCTa and LAP domains localize to discrete regions of nLDs
CCTa associates with membranes via its amphipathic a-helical M-domain, which is antagonized by phosphorylation of 16 serine, threonine and tyrosine residues in the adjacent Pdomain ( Fig. 3A) (Cornell and Ridgway, 2015). The binding of CCTa to LD involves large hydrophobic side-chains in M-domain that insert into voids in the phospholipid monolayer (Prevost et al., 2018). To determine if the M-and P-domains regulate binding to nLDs, CCTα with truncations and point mutations in these domains was expressed in U2OS cells (Fig. 3B) and localization on nLDs was quantified by immunofluorescence microscopy ( Fig. 3C and D). CCTa localization to nLDs was completely prevented by mutation of 8 lysine residues in the M domain (CCTa-8KQ) that form electrostatic interactions with membrane lipids (Johnson et al., 2003).
Conversely, CCTa-3EQ, a M-domain mutant with enhanced membrane association (Johnson et al., 2003), was localized to nLDs as well as the NE. Deletion of the P-domain (CCTa-DP) and a dephosphorylated mimic with 16 serine residues mutated to alanine (CCTa-16SA) were strongly associated with nLDs, whereas the phosphorylated mimic with 16 serine residues mutated to glutamate (CCTa-16SE) was not detected on nLDs. Catalytic dead CCTa-K122A was localized on nLDs similarly to wild-type. These results show that interaction of CCTa with the nLD phospholipid monolayer is mediated by electrostatic interactions with the M-domain and antagonized by phosphorylation of the P-domain.
Next we used spinning disk confocal 3D and super-resolution radial fluctuation (SRRF) (Gustafsson et al., 2016) imaging to determine how endogenous CCTa and LAP domains are disposed on the surface of nLDs (Fig. 4). 3D reconstruction of a typical nLD revealed a lipid body (BODIPY) located close to the basal surface of the INM (Fig. 4A). CCTa coated most of the nLD but was absent from the top of the particle (Fig. 4A, d), which was occupied by a 'cap' of PML protein (Fig. 4A, e). SRRF imaging of this nLD in a cross-section where CCTa and PML intersect 13 (indicted by the yellow line in Fig. 4A, b-f) revealed a punctate distribution ( Fig. 4B and C) consistent with each protein occupying non-overlapping, interdigitated regions on the nLD surface (error and resolution analysis using NanoJ-SQUIRREL (Culley et al., 2018) for this image is shown in Supplemental Fig. S6). Additional 3D renderings of 8 nLD are shown in Supplemental

The DAG and Lipin1 content of nLDs is regulated by LAP domains
In addition to being a precursor for the TAG and PC that is incorporated into LD, DAG also stimulates translocation and activation of CCTα on membranes (Arnold et al., 1997). Thus, we investigated whether CCTα interaction with nLDs could be regulated by DAG as a mechanism to coordinate PC and TAG synthesis. Initially, we visualized DAG in cells cultured with or without oleate for 24 h using the DAG biosensor GFP-C1(2)d (Codazzi et al., 2001) (Fig. 5). In untreated U2OS and PML KO cells, the DAG biosensor was localized on reticular and punctate structures in the cytoplasm (Fig. 5A). Exposure of U2OS cells to oleate resulted in the appearance of the DAG biosensor on dispersed cytoplasmic structures that did not correspond to cLDs, and occasionally on LipidTox Red-positive nLDs (Fig. 5B, a-c). In contrast, the DAG biosensor was strongly associated with cLDs in PML KO cells and less frequently observed on nLDs (Fig. 5C, a-c). This is the first evidence that PML regulates the DAG content of cLDs and nLDs.
To better assess how PML regulated the levels of DAG on nLDs, a tandem nuclear localization signal (NLS) was inserted in GFP-C1(2)d (nGFP-DAG). When transiently expressed in oleate-treated U2OS cells, the nuclear biosensor nGFP-DAG was detected on two types of structures; small puncta that did not stain with LipidTox Red, and LAP-positive nLDs (Fig. 6A).
For the purpose of quantification, we divided LipidTox Red-positive nLDs in U2OS cells into four populations based on the presence or absence of DAG and/or PML (Fig. 6B). This revealed that 60% of total nLDs were negative for both DAG and PML (DAG/PML-). The majority of the remaining nLDs contained DAG and PML (79%), with the remainder containing only DAG (7%) or PML (14%) (Fig. 6B, insert). The cross-sectional area of DAG/PML-nLDs was significantly reduced compared to total nLDs, while the PML+ and DAG/PML+ nLDs were significantly larger ( Fig. 6C). DAG+ nLDs were virtually absent from PML KO cells compared to total and DAGnegative nLDs, both of which were also significantly reduced (Fig. 6D). The nuclear sensor also detected DAG on the abundant nLDs that form in oleate-treated Caco2 cells (Supplemental Fig.   S8A). These data indicate that DAG is primarily enriched in large LAP-positive nLDs, and that PML KO reduces DAG in nLDs but increases DAG in cLDs.
To determine whether DAG was a positive effector of CCTα association with nLDs, we determined whether CCTα was preferentially enriched in DAG-positive nLDs.
Immunofluorescence and RGB plots through nLDs in U2OS cells showed association of the nuclear DAG sensor and CCTα (Fig. 7A). In PML KO cells, there was evidence of CCTα-positive nLDs that lacked DAG suggesting poor correlation between DAG content and CCTα, a conclusion supported by quantitation of CCTa and DAG distribution on nLDs. Of the approximately 50% of nLDs in U2OS cells that contained DAG and/or CCTα, only 43% contained both DAG and CCTα suggesting that the DAG content of an nLDs is not a prerequisite for recruitment of CCTa ( Fig.  7B, insert). Quantification of DAG/CCTα distribution on nLDs in PML KO cells indicated a reduction in all four populations compared to control cells, particularly those that contained DAG and DAG/CCTa but the percent distribution of DAG and CCTα on nLDs was similar to control cells (Fig. 7B, insert). As well, there was no significant difference in the cross sectional area of nLDs containing DAG/CCTα in control versus PML KO cells (Fig. 7C). Collectively this shows that the presence of CCTa and DAG on nLDs is poorly correlated in both control and PML KO cells, a conclusion that is consistent with charged residues in the CCTa M-domain being a driving force for interaction with nLDs (Fig. 3).
The source for DAG in nLDs could be Lipin1, a nuclear/cytoplasmic PA phosphatase that produces DAG for phospholipid and TAG synthesis (Csaki et al., 2013). The human and murine genes encode similar Lipin1a and b splice variants (Croce et al., 2007;Peterfy et al., 2005) that are partially localized to cLDs in COS cells and macrophages (Valdearcos et al., 2011;Wang et al., 2011), potentially recruited by seipin (Sim et al., 2012). To determine if Lipin1 is recruited to nLDs by a LAP-dependent mechanism, transiently expressed Lipin1α and b were localized in U2OS and PML KO cells by immunofluorescence microscopy. V5-tagged Lipin1α was expressed in the nucleus and cytoplasm of untreated U2OS and PML KO cells (Fig. 8A). Treatment of U2OS cells with oleate caused extensive association of Lipin1α with the surface of BODIPY-positive nLDs but not with cLDs ( Fig. 8B, a). Lipin1α was also localized to the surface of nLDs in Caco2 cells treated with oleate for 24 h (Supplemental Fig. S8B). To determine the extent of Lipin1α and DAG co-localization in U2OS cells, the distribution and cross-sectional area of Lipin1α-and DAG-containing nLDs was quantified. Of the DAG/Lipin1α-positive nLDs, 53% contained both ( Fig. 8C, insert), and Lipin1α-positive nLDs were significantly larger than other components of the population (Fig. 8D). In oleate-treated PML KO cells, Lipin1α was diffusely localized in the cytoplasm and nucleus but was virtually absent from nLDs (Fig. 8E), a conclusion that was confirmed by quantification of Lipin1α-positive nLDs in PML KO cells compared to controls (Fig.   8F). The Lipin1b isoform also associated with the surface of nLDs but not cLD in oleate-treated U2OS cells (Supplemental Fig. S9A), and was not detected on nLDs in oleate-treated PML KO cells (Supplemental Fig. S9B). The Lipin1 substrate PA was detected with the GFP-nes-2xPABP (Bohdanowicz et al., 2013) biosensor on the PM and cytoplasmic membranes of control and PML KO cells but was absent from nLDs or cLDs (Supplemental Fig. S10). These data indicate that LAP domains are required for association of Lipin1 and DAG with nLDs.

LAP domains positively regulate PC and TAG synthesis
We next tested whether ablating LAP domains in PML KO cells and reducing nLD-

DISCUSSION
The INM is contiguous with the outer nuclear membrane and cytoplasmic ER (Ungricht and Kutay, 2017) and thus could receive lipids by lateral diffusion from the ER (van Meer et al., 2008). However, the INM in yeast (Romanauska and Kohler, 2018) and mammals (Aitchison et al., 2015;Csaki et al., 2013;Haider et al., 2018;Peterson et al., 2011) harbours lipid biosynthetic enzymes and regulatory proteins suggestive of compartmentalized nuclear lipid synthesis.
Analogous to the ER, the INM is the site for de novo biogenesis of nLDs in yeast (Romanauska and Kohler, 2018) whereas in hepatoma cells, nLDs are derived from luminal ER LDs that enter the nucleus after dissolution of the INM (Soltysik et al., 2019). PML expression, in particular the PML-II isoform, aids in nLD biogenesis at the INM by disrupting lamin-free regions of the INM to allow luminal LD release (Ohsaki et al., 2016). Ohsaki and colleagues also demonstrated that more than 70% of PML-containing structures in the nuclei of hepatoma cells were associated with the surface nLDs but their function is unknown. Here we demonstrate that these structures, which we refer to as LAP domains, are enriched in DAG and lipid biosynthetic enzymes but depleted of proteins that associate with canonical PML NBs. Further, LAP domains are not essential for nLD formation in U2OS cells but are necessary for the functional maturation of the LD into structures that recruit CCTα and Lipin1 for the regulation of PC and TAG synthesis.
Coincident with the appearance of LAP domains on nLDs in oleate-treated U2OS cells was the loss of canonical PML NB containing SUMO1, DAXX and SP100. While the expression of PML protein was unaffected by oleate and it appeared to transfer quantitatively to nLDs, the resultant LAP domains were poorly SUMOylated and depleted of PML NB resident proteins.
Electron microscopy showed the surface of nLDs had the typical morphology of PML NB (Ohsaki et al., 2016), indicating that LAP domains are structurally related but have lost features that are 13 required for PML NB assembly, notably SUMO1 and SP100. Importantly, since LAP domains are deficient in these normally constitutive protein components of this nuclear subdomain, by convention and definition they cannot be called PML NBs. By making this distinction, LAP domains are place within the continuum of PML-containing cellular structures that are already described in the literature with PML NBs at one extreme, followed by LAP domains, and at the far extreme, Mitotic Accumulation of PML Protein (MAPPs) that are completely devoid of DAXX, SP100 and SUMO1 (Dellaire et al., 2006). Since PML KO prevented the increase in PC synthesis and reduced CCTα-positive nLDs, we conclude that LAP domains are essential to activate PC synthesis in response to prolonged exposure to fatty acids. Analysis of CCTα mutants showed that association with nLDs was mediated by basic and acidic residues in the M-domain and negatively regulated by phosphorylation of the P-domain. These domains also regulate CCTα association with membrane in vitro, demonstrating that interactions with nLDs are lipid-mediated. Thus LAP domains promote the formation of nLDs on to which CCTa can translocate and activate PC synthesis. To identify lipids that promote CCTα interaction with nLDs we initially focused on DAG, which induces packing defects in membranes into which the CCTα M-domain can insert (Cornell, 2016;Cornell and Ridgway, 2015). PML expression was positively correlated with large nLDs that were enriched in DAG. However, DAG and CCTα were not preferentially co-localized in nLDs nor was their relative distribution in nLDs altered by loss of PML expression. While the exclusion of CCTα and the biosensor on nLDs could be due to competition for DAG, it is more likely that factors such as surface curvature, enrichment in anionic lipids or a high PE/PC ratio promote the association of CCTα. For instance, the high PE/PC ratio in insect cLDs relative to human cLDs (Jones et al., 1992;Tauchi-Sato et al., 2002) could be responsible for nuclear export and localization of CCT1 and CCTα on cLDs during oleate loading in Drosophila S2 cells (Krahmer et al., 2011).
The striking PML-dependent redistribution of DAG between nLDs and cLDs was linked to the nuclear PA phosphatase Lipin1a, which co-localized with DAG on nLDs in oleate-treated U2OS and Caco2 cells. Since small nLDs in PML KO cells were relatively devoid of both Lipin1a and DAG, Lipin1a appears to provide DAG for nuclear TAG synthesis by diacylglycerol acyltransferase 2, which was localized to nLDs when ectopically expressed in Huh7 cells (Ohsaki et al., 2016). The factors mediating the selective association of Lipin1 with nLDs are uncertain.
PA, which promotes Lipin1 association with membranes (Ren et al., 2010), was not detected on nLDs or cLDs with a biosensor suggesting it is below the detection threshold or the biosensor is not sufficiently PA-specific, as was suggested previously (Horchani et al., 2014). Lipin1a was not detected on cLDs and associated with accumulation of DAG on cLDs in PML KO cells, which could reflect a partial block in TAG synthesis or increased TAG hydrolysis.
Metabolic labelling experiments support the concept that LAP domains and nLDs are sites for TAG synthesis and regulation. PML KO cells had significantly reduced [ 3 H]oleate incorporation into TAG, reduced nLDs, and increased levels of DAG in cLDs, indicative of a block in nuclear and cytoplasmic TAG synthesis. However, PML KO did not inhibit de novo synthesis TAG from [ 3 H]glycerol suggesting a more complex role in oleate uptake and utilization. Prior studies have indicated complex, tissue-specific roles for PML in fatty acid metabolism. PML NBdependent activation of peroxisome proliferator-activated receptor (PPAR) g co-activator-1a (PGC-1a), PPAR signaling and fatty acid β-oxidation provided a growth advantage to breast cancer cells (Carracedo et al., 2012) and controlled asymmetric division and maintenance of hematopoietic stem cells . In contrast, PML KO mice had tissue-specific enhancement of both fatty acid b-oxidation and synthesis, increased metabolic rate and resistance to diet-induced obesity (Cheng et al., 2013). Together with these findings, our results suggest that LAP domains associated with nLDs could have a multi-facetted role in lipid homeostasis involving the recruitment of CCTa and Lipin1 to promote PC and TAG synthesis and storage as well as

Analysis of PC synthesis using [ 3 H]choline incorporation
U2OS and PML KO cells were incubated in the presence or absence of 400 μM oleate/BSA for 24 h. Cells were rinsed twice with choline-free medium A and then incubated with choline-free medium A containing [ 3 H]choline (2 μCi/ml) for 2 and 4 h. After isotope labeling, cells were rinsed with cold PBS, harvested in methanol:water (5:4, v/v), [ 3 H]PC was extracted in chloroform/methanol, and the radioactivity quantified by liquid scintillation counting and normalized to total cellular protein (Storey et al., 1997).

Analysis of fatty acid and TAG synthesis
U2OS and PML KO cells were incubated with 2.5 μCi/ml [ 3 H]acetate for 4 h to measure fatty acid synthesis (Brown et al., 1978). Briefly, lipid extracts from cells were saponified in ethanol and 50% potassium hydroxide (w/v) for 1 h at 60°C and extracted with hexane. The hydrolysate was then acidified with HCl (pH <3) and radiolabeled fatty acids were extracted with hexane and quantified by liquid scintillation counting. Incorporation of [ 3 H]acetate into fatty acids was normalized to total cellular protein.

13
TAG and cholesterol ester (CE) synthesis from exogenous fatty acid was determined by incubating cells with 100 μM [ 3 H]oleate/BSA. After 4h, cells were rinsed twice with cold 150 mM NaCl, 50 mM Tris-HCl (pH 7.4) with 2 mg/ml BSA and once with the same buffer without BSA. Cells were rinsed with cold PBS and radioactive lipids were extracted and quantified as described above.
For super resolution radial fluctuation (SRRF) imaging (Gustafsson et al., 2016), immunostained cells (described above) were observed using a 100 X Plan-Apochromat (1.46 NA) oil immersion objective lens (Zeiss) by wide-field imaging on a Marianis microscope (Intelligent Imaging Innovations, 3i) based on a Zeiss Axio Cell Observer equipped with LED-based illumination via a SPECTRA III Light Engine (Lumencor), and a Prime BSI back-illuminated scientific complementary metal-oxide-semiconductor (sCMOS) camera (Teledyne Photometrics).
This imaging configuration resulted in an effective pixel size of 65 x 65 nm in the captured images.
To generate super resolution images, 100 wide-field images were captured at 10 ms/frame using Slidebook 6 software and then exported to ImageJ (version 1.52, NIH) in a 16-bit Open Microscopy Environment-Tagged Image File Format (OME-TIFF). Image processing used a custom SRRF algorithm (NanoJ-LiveSRRF, available on request from Ricardo Henriques, University College London/Francis Crick Institute, UK) with the following settings: radius 3; sensitivity 2; magnification 4; average temporal analysis; and with intensity weighting and both vibration and macro-pixel correction turned on. NanoJ-LiveSRRF is the newest implementation of NanoJ-SRRF within the ImageJ software, available upon request as above. However, NanoJ-SRRF is already freely available (Gustafsson et al., 2016).
For confocal imaging and 3D volume rendering, cells were immunostained as above and imaged using a 100 X Plan-Apochromat (1.46 NA) oil immersion objective lens (Zeiss) by    A large nLD was identified in association with CCTa and PML (panel a, red frame). The image was rotated and zoomed to produce a side view of the nLD (panel b, red frame). The framed structure in b was assessed by removing the DAPI channel (panels c-f) to reveal how the nLD/BODIPY (blue), CCTa (green) and PML (red) are associated. B, Widefield images of the same nLD in A imaged at the level of the yellow line indicated in panels b-f (image rotated ~90° clockwise). The nLD is highlighted and magnified in the inset (bar, 5µm; Inset scale bar, 1µm). C,

13
Images of the same field of view in B were acquired by NanoJ SRRF. The nLD is highlighted and magnified in the inset (bar, 5µm; inset bar, 1µm).  Results in panels B, C and D are presented as box and whisker plots showing the mean and 5 th -to-95 th percentile from analysis of 50-100 cells in 3 separate experiments. Significance was 13 determined by one-way ANOVA and Tukey's multiple comparison to total nLD area (C) or twotailed t-test compared to U2OS controls (D) (*p<0.05; **p<0.01; ***p<0.001). Figure 7. CCTa association with nLDs is DAG independent. A, U2OS and PML KO cells transiently expressing a nuclear DAG-GFP sensor were exposed to oleate (400 µM) for 24h, fixed and immunostained with a CCTa antibody, and incubated with LipidTox Red to visualize LDs.