Structural insights into perilipin 3 membrane association in response to diacylglycerol accumulation

Lipid droplets (LDs) are dynamic organelles that contain an oil core mainly composed of triglycerides (TAG) that is surrounded by a phospholipid monolayer and LD-associated proteins called perilipins (PLINs). During LD biogenesis, perilipin 3 (PLIN3) is recruited to nascent LDs as they emerge from the endoplasmic reticulum. Here, we analyze how lipid composition affects PLIN3 recruitment to membrane bilayers and LDs, and the structural changes that occur upon membrane binding. We find that the TAG precursors phosphatidic acid and diacylglycerol (DAG) recruit PLIN3 to membrane bilayers and define an expanded Perilipin-ADRP-Tip47 (PAT) domain that preferentially binds DAG-enriched membranes. Membrane binding induces a disorder to order transition of alpha helices within the PAT domain and 11-mer repeats, with intramolecular distance measurements consistent with the expanded PAT domain adopting a folded but dynamic structure upon membrane binding. In cells, PLIN3 is recruited to DAG-enriched ER membranes, and this requires both the PAT domain and 11-mer repeats. This provides molecular details of PLIN3 recruitment to nascent LDs and identifies a function of the PAT domain of PLIN3 in DAG binding.


50
Lipid droplets (LDs) act as energy reservoirs in cells. They contain a neutral lipid core of mainly 51 triacylglycerols (TAGs) and cholesterol esters with a phospholipid monolayer that surrounds the 52 neutral lipid core. This creates a membrane environment for LDs that is distinct from membrane 53 bilayers and recruits several LD-associated proteins [1][2][3]. In addition to a major role in energy 54 storage, LDs are also important cellular hubs that traffic proteins and lipids between organelles, 55 regulate ER stress, and contribute to viral infections [4][5][6].

57
LDs are formed in the ER where neutral lipids are synthesized [7,8]. Mechanistically, LD formation 58 involves several steps [9] including accumulation of neutral lipids in the outer ER membrane 59 leaflet, neutral lipid nucleation aided by the seipin complex and associated factors (e.g. LDAF1)

60
[10], formation of nascent LDs that bud from the ER [11,12] and LD growth and maturation [13, 61 14]. Several lines of evidence support the idea that the phospholipid composition of the ER 62 membrane is locally edited to promote LD assembly or recruit specific proteins important for LD 63 formation such as seipin and ORP proteins [7,[15][16][17]. Although there is still uncertainty regarding 64 the phospholipid composition at the initial stages of LD formation, evidence from yeast suggest

109
To systematically define how lipid composition affects PLIN3 recruitment to membrane bilayers, 110 we purified recombinant human PLIN3 from Escherichia coli and used liposome co-sedimentation 111 assays to monitor membrane binding. Liposomes were prepared using multiple freeze/thaw 112 cycles and characterized by dynamic light scattering (DLS) (Fig. S1A). Liposomes were made 113 from phospholipids with different acyl-chain combinations comprised of either palmitoyl-oleoyl 114 (PO) or di-oleoyl (DO) phospholipids (Fig. 1A). PO

122
[44] and the ratio of PC to PE has previously been shown to regulate protein distribution on LDs 123 [43,45]. Under all PC-to-PE ratios tested, PLIN3 did not bind liposomes comprised solely of PC 124 and PE (Fig. 1B).

126
We next asked whether addition of other lipids that are synthesized in the ER could recruit PLIN3 127 to membranes by generating PC/PE liposomes containing 20mol% of either phosphatidic acid 128 (PA), phosphatidylserine (PS), diacylglycerol (DAG), phosphatidylinositol (PI) or C18:1 ceramide.

134
PLIN3 recruitment to DO-based liposomes depended on the surface concentration of DAG, with 135 5mol% DAG able to induce ~10% binding, and 20mol% DAG inducing 35% binding (Fig. 1E). The 136 addition of DOPA further increased PLIN3 binding in DAG-containing liposomes, which suggested 137 a synergistic effect of PA and DAG on PLIN3 membrane recruitment (Fig. 1F). Increasing the 138 total liposome concentration caused the majority of PLIN3 to bind membranes, but 5% of PLIN3 139 remained in the soluble fraction (Fig. S2A). DLS characterization indicated DAG enriched 140 liposomes had two populations based on size (Fig. S1C, S1D), which is likely due to the fusogenic 141 properties of DAG [46]. The addition of DOPA or PLIN3 protein decreased or eliminated the 142 population of the larger (>400 nm) liposomes (Fig. S1C, S1D). We concluded that PLIN3 binds 143 liposomes containing DAG and/or PA, which DO and PO liposomes,4ME phospholipids 154 significantly increased PLIN3 liposome association with ~70% binding observed with a mixture of 155 4ME-PC and 4ME-PE (Fig. 1G). Consistent with our previous observations, 4ME-PA further 156 increased PLIN3 binding, while 4ME-PS and C18:1 ceramide decreased binding (Fig. 1G)

162
As the major function of PLIN3 is to bind to emerging LDs as they bud from the ER, we next 163 assessed the ability of recombinant PLIN3 to bind artificial lipid droplets (ALDs) in vitro using a 164 flotation assay [10,28,29,49]. ALDs were generated with a triolein neutral lipid core surrounded 165 by a phospholipid monolayer of DOPC and DOPE in 1:2.5 molar ratio of phospholipids to TAG.

185
We determined the absolute exchange of PLIN3 after a 3sec pulse of deuterium incorporation at 186 0°C (equivalent to ~0.3sec at 20°C) using a fully deuterated control. After the deuterium pulse,

195
We next sought to determine any conformational changes that occur during membrane binding 196 using HDX-MS. Having established optimal conditions for PLIN3 membrane binding, we 197 measured the deuterium exchange rate over various time points (3, 30, 300, and 3000sec) in the 198 absence or presence of 4ME liposomes composed of 60mol% 4ME-PC, 20mol% 4ME-PE, and 199 20mol% 4ME-PA.

201
In the presence of membranes, striking differences in deuterium exchange were observed 202 throughout the PLIN3 sequence (Fig. 2E, Table S1, source data). The most notable differences 203 were large decreases in deuterium exchange in the PAT domain and 11-mer repeat regions. In

204
comparison to previous results that demonstrated the 11-mer repeats are sufficient for the lipid 205 droplet association [30,31], this suggests that the PAT domain and 11-mer repeats are both major 206 contributors to membrane binding. Consistently, we found that a purified PAT/11-mer repeats 207 fragment displayed similar membrane recruitment as full length PLIN3 to DO liposomes enriched 208 in DAG and PA, and 4ME liposomes (Fig. 2C, 2D).

210
The HDX-MS results suggested that membrane binding induces the formation of a secondary 211 structure in the PAT domain and 11-mer repeats. This is consistent with both AlphaFold and

212
RoseTTAFold models that predict the PAT domain and 11-mer repeats form amphipathic alpha 213 helices, with the PAT domain adopting a triangular globular structure and the 11-mer repeats 214 forming a series of extended alpha helices (Fig. 2B, 3B). Based on these results we concluded 215 that the PAT domain and 11-mer repeats of PLIN3 are intrinsically disordered in the absence of 216 membranes, with inducible amphipathic alpha helices being stabilized upon membrane binding.

218
Notably, many of the peptides with the PAT domain and 11-mer repeats displayed a bimodal 219 distribution of deuterium exchange upon membrane binding (Fig. S3A, S3B). This differs from a 220 canonical distribution where the degree of deuteriation is gaussian in individual peptides. The

224
Given our previous observation that the  repeats are intrinsically disordered in 225 the absence of membranes, the simplest interpretation is that the PAT domain and 11-mer 226 repeats cycle between two states: 1) a membrane bound alpha helical state on membranes that 227 is strongly protected from H-D exchange and 2) an intrinsically disordered solution state that 228 rapidly exchanges with deuterated solvent upon dissociation from membranes and leads to rapid

232
The protection from deuterium exchange and the non-gaussian bimodal distribution for most 233 peptides in the PAT domain persisted over a longer time course (~300sec) in comparison to 234 peptides within the 11-mer repeat regions (~30sec) (Fig. S4A). This trend of longer HDX 235 protection correlated with the degree of sequence conservation among PLINs (Fig. 3A). Taken 236 together this supports a role for the PAT domain in membrane binding and suggests a similar role 237 for the highly conserved PAT domain in other PLINs.

239
In contrast to the N-terminal PAT domain and 11-mer repeats, the C-terminal 4-helix bundle 240 displayed a large increase in deuterium exchange (Fig. 2E, S4A). The most apparent increase in 241 deuterium exchange was in the middle of the 4-helix bundle. We hypothesized this was due to 242 the 4-helix bundle unfolding upon membrane binding, with this possibly contributing to membrane 243 recruitment. To test this, we examined the ability of a purified PLIN3 4-helix bundle fragment 244 (residues 197-427) to bind membranes. Consistent with our hypotheses, the isolated 4-helix 245 bundle bound to the 4ME-PC/PE/PA liposomes that were used in the HDX-MS experiment (Fig.   246 2C, 2D). However, the 4-helix bundle did not bind to 4ME-liposomes lacking PA, or to any DO-247 based liposomes even when PA and DAG were present (Fig. 2C, 2D). We concluded that the 4-

290
In general, fragments encompassing the PAT domain bound strongly to liposomes in the 291 presence of DAG, with modest liposome binding in the presence of PA (Fig. 3D). In contrast, the 292 11-mer repeats bound to liposomes containing either PA or DAG (Fig. 3D). Taken together, the 293 data suggests that the PAT region does form a domain that spans residues 22-116 in PLIN3.

294
Notably, this expanded PAT domain is capable of membrane binding and displays a preference 295 for binding DAG enriched membranes, while the 11-mer repeats of PLIN3 do not.

298
We first sought to confirm the effects of liposome binding on secondary structure of PLIN3 by 299 circular dichroism (CD) (Fig. S5). Liposomes were prepared with 4ME-PC/PE/PA, which recruited 300 all the PLIN3 constructs (full length, PAT, 11-mer repeats, PAT/11-mer repeats, 4-helix bundle) 301 in the previous liposome co-sedimentation experiments. For full length PLIN3, the presence of 302 liposomes increased overall helicity as observed by an increased negative peak around 222nm 303 in the CD spectra. In comparison, the CD spectra of the 4-helix bundle was largely unaffected by 304 the presence of liposomes and was consistent with a stable alpha helical structure. In contrast, 305 the PAT/11-mer repeats adopted a mostly random coil structure in solution with a negative peak 306 around 200nm, and a shift to alpha helices in the presence of liposomes as indicated by a large 307 negative peak at 222nm. Liposomes induced similar changes in the CD spectra for both the PAT 308 domain and 11-mer repeats alone. Taken together, this confirms that the increase in helicity 309 observed in full length PLIN3 by membranes was due to the PAT/11-mer repeats undergoing a , we attached a single spin-label to residue 37C and did not observe a 335 dipolar evolution (DEER shape) that could indicate a pair with a shorter than ~80Å separation 336 based on only slightly concave shape of the signal (Fig. 4E). The DEER data thus suggest that 337 conspicuous dipolar signals in the doubly labeled proteins result from intramolecular interactions, 338 rather than being caused by intermolecular interactions on membranes.

345
for these two sites. We also checked the spectral shape by recording field-swept echo with pulse 346 separation of 250ns and did not notice any conspicuous broadening that could indicate a shorter 347 range of distances (<15Å). We do however see from continuous wave (CW) ESR of 37C/114C 348 (Fig. S6) that there might be a sizeable fraction of spins in the 15-20Å range whose contribution 349 to the distance distribution will be significantly attenuated, since DEER has low sensitivity to 350 distances in this range. The conformations with this distance range correlate well with AlphaFold 351 predictions. The spread of the P(r) to longer distances could originate from the mobility of the C-352 terminal helix where residue 114C is located (Fig. 4C). Taken

377
5B, 5F and 5H). The PAT domain was recruited to diolein DEVs, and the largest fragment 378 (residues 1-116) bound to both the droplet and bilayer surface with similar magnitudes to full 379 length PLIN3 (Fig. 5C, 5F and 5H). The 11-mer repeats also bound to both the droplet and bilayer 380 surface of diolein DEVs, but to a lesser extent (Fig. 5C, 5G

391
[15], and cytoplasmic PLIN3 marks these sites [10,49]. We sought to test whether DAG 392 accumulation was sufficient for PLIN3 recruitment using an independent system and also assess 393 what fragments of PLIN3 were necessary and sufficient for ER recruitment.

395
To best visualize PLIN3 recruitment to ER membranes, we generated intracellular giant ER 396 vesicles (GERVs) by submitting cells to hypotonic medium [72,76,77] (Fig. 6A). After exchange 397 to hypotonic medium, cells were pretreated with DMSO or the DGAT inhibitors (DGATi), followed 398 by oleic acid to induce TAG or DAG synthesis [77] (Fig. 6A). Confocal microscopy was used to visualize cells prior and after exchange to hypotonic medium, and after oleic acid treatment.

400
Imaging was done in the following minutes after the treatments.

402
As expected, the majority of subcellular localization of GFP-tagged PLIN3 was cytoplasmic in 403 both normal and hypotonic media without addition of oleic acid (Fig. 6B, upper panels), and co-404 localized with LDs (Fig. 6D). Under conditions of DGAT inhibition, PLIN3 co-localized to the outer 405 periphery of GERVs (Fig. 6B, yellow arrow, 6E and 6F), which suggests DO-DAG accumulation 406 is sufficient to recruit PLIN3 to ER membrane bilayers.

408
In line with our previous results, a construct containing both the PAT domain and 11-mer repeats 409 showed a similar subcellular localization as full length PLIN3 under all conditions (Fig. 6C-6F).

410
This suggests the PAT domain and 11-mer repeats are sufficient for PLIN3 recruitment to both 411 DAG enriched membrane bilayers and TAG containing LDs, which might also contain DAG whose 412 concentration could increase the protein binding level. Interestingly, constructs containing only 413 the PAT domain (residues 1-116) or only the 11-mer repeats (residues 114-204), which bound to 414 DEVs in a DAG-dependent manner, remained cytoplasmic under all conditions (Fig. 6D, 6E, S7   bundle. Therefore, PLIN3 membrane association may not only be determined by membrane 451 packing defects, but could also involve selective physical interactions between PLIN3 and DAG 452 or PA. This idea is also supported by a previous study that found that the LD binding properties 453 of PLINs are sensitive to the polar residue composition of their amphipathic helices [83].

455
In this study, we attempted to clarify the function and boundaries of the PAT domain. Our HDX-

456
MS results using full length PLIN3 clearly implicate both the PAT domain and 11-mer repeats in 457 membrane binding. In addition, we were able to define a functional PAT domain that 458 encompasses all of the conserved residues within PLINs and is longer than previously suggested.

459
This expanded PLIN3 PAT domain is sufficient to bind DAG enriched membranes, but not LDs.

460
In contrast, the 11-mer repeats display some affinity for DAG enriched membranes and are 461 necessary to bind LD monolayers. Our overall conclusion is that the PAT domain and 11-mer 462 repeats serve synergistic functions, as the individual regions are necessary for both LD and DAG 463 recruitment in cells.

465
The PAT domain is predicted to adopt a triangular tertiary structure by both AlphaFold and

497
The genes encoding PLIN3 constructs were codon-optimized for expression in Escherichia coli 498 and cloned into pTHT, which is a derivative of pET28 that contains a TEV cleavable N-terminal 499 6x His tag. For GFP tagged PLIN3 constructs, monomeric superfolder GFP (msfGFP) was 500 inserted between the 6x His tag and the N-terminus of PLIN3. PLIN3 plasmids were transformed 501 into BL21(DE3) RIPL cells and protein expression was induced with 1mM isopropyl β-D-1-502 thiogalactopyranoside (IPTG) at 37°C for 3h. Cells was harvested by centrifugation at 3,320 x g 503 for 20mins and stored at -80°C until use. Cell pellets were resuspended with buffer A containing 504 500mM NaCl, 20 mM Tris-HCl pH 7.5, 5% glycerol, 5mM 2-mercaptoethanol (BME) and lysed by 505 sonication. Cell lysates were centrifuged at 81,770 x g for 1h at 4°C and the resulting supernatant 506 was incubated with 5mL of Ni-NTA resin for 2h at 4°C prior to loading onto a gravity column. The

642
The raw %D incorporation for the fully deuterated sample is included in the source data, with the 643 average back exchange being 33%, and ranging from 10-60%.

1232
to induce DAG accumulation on the ER in Cos7 cells.