The middle lipin domain adopts a membrane-binding dimeric protein fold

Phospholipid synthesis and fat storage as triglycerides are regulated by lipin phosphatidic acid phosphatases (PAPs), whose enzymatic PAP function requires association with cellular membranes. Using hydrogen deuterium exchange mass spectrometry, we find mouse lipin 1 binds membranes through an N-terminal amphipathic helix, the Ig-like domain and HAD phosphatase catalytic core, and a middle lipin (M-Lip) domain that is conserved in mammalian and mammalian-like lipins. Crystal structures of the M-Lip domain reveal a previously unrecognized protein fold that dimerizes. The isolated M-Lip domain binds membranes both in vitro and in cells through conserved basic and hydrophobic residues. Deletion of the M-Lip domain in lipin 1 reduces PAP activity, membrane association, and oligomerization, alters subcellular localization, diminishes acceleration of adipocyte differentiation, but does not affect transcriptional co-activation. This establishes the M-Lip domain as a dimeric protein fold that binds membranes and is critical for full functionality of mammalian lipins. Lipins need to bind cell membranes before they can function as phosphatidic acid phosphatases. Here, the authors elucidate the structural basis of lipin membrane-association and identify a lipin domain with a novel protein fold that is critical for membrane binding and full functionality of lipins.

The architecture of mammalian lipins and Sc Pah1 differ, but all lipin/Pah homologs share two common and conserved regions called the N-Lip and C-Lip regions, which are located at the respective N-and C-termini of mammalian lipins (Fig. 1a) 6 . Recently, we determined the structure of Tetrahymena thermophila Pah2 (Tt Pah2), which revealed the N-Lip and C-Lip regions co-fold to form a catalytic unit comprised of a split immunoglobulin-like (Ig-like) domain and a haloacid dehalogenase-like (HAD-like) catalytic domain 18 . The N-Lip and C-Lip regions are separated by an extended linker that varies in length and sequence across species 17,19 . In mammalian lipins, this linker is 500 amino acids and can be hyperphosphorylated 7,20-22 , sumoylated 23 , or acetylated 24 . Phosphorylation, sumoylation, or acetylation within the linker region is reported to regulate the subcellular location and activity of mammalian lipins 7,[21][22][23][24][25] . However, it is not known if the linker region has additional roles.
As the only enzyme in the glycerol-3-phosphate pathway that is not constitutively membrane-bound, the regulation of lipin/Pah membrane association is a determinant of its enzyme activity. In vitro, purified Sc Pah1, Tt Pah2, and mammalian lipins are recruited to membranes containing their substrate, PA 18,22,26 . Lipin/Pahs lack canonical lipid binding domains (e.g., PH and PX domains) found in other membrane-binding proteins, but contain a conserved N-terminal amphipathic helix that is necessary for lipin/Pahs to bind membranes in vitro and in cells 18,26 In addition, a nuclear localization signal/polybasic region in mammalian lipins has been implicated in membrane binding 27,28 .
We sought to characterize how lipins bind membranes and in the process identified a new middle lipin (M-Lip) domain that is universally conserved in mammalian and mammalian-like lipins, but not present in Sc Pah1. Herein, we report the structural and functional characterization of the M-Lip domain. Our principal findings are that the M-Lip domain is a new protein fold that forms a dimer, binds membranes, and can affect lipin PAP activity, oligomerization, subcellular localization, and adipogenesis.

Results
Structure and dynamics of lipin 1. Full-length mouse lipin 1α (herein referred to as lipin 1) was purified from Sf9 cells and the structure and dynamics were probed using hydrogen deuterium exchange mass spectrometry (HDX-MS). HDX-MS measures the exchange rate of amide hydrogens with deuterium. The major determinant of exchange is the stability of secondary structure 29 . Thus, HDX-MS provides a readout of secondary structure dynamics. In addition, HDX experiments with extremely short exposures of D 2 O can be used to identify disordered regions within proteins when compared to a fully deuterated condition 29 .
HDX experiments were carried out with a short pulse of deuterium exposure to map regions of order/disorder in fulllength lipin 1. A total of 189 peptides spanning 83.6% of the primary sequence were identified and their deuterium content was quantified. Peptides for residues 121-256 and residue 413 were not identified by tandem MS/MS. Residues with no MS coverage were all located between the N-Lip and C-Lip regions, and near the highly basic nuclear localization signal.
The N-Lip and C-Lip regions both had low rates of deuterium exchange with a 3 s pulse of D 2 O exposure at 4°C, which indicates these regions were largely ordered into secondary structure elements (Fig. 1b). This suggests that the N-Lip and C-Lip regions co-fold to form the split Ig-like domain and HAD-like domain observed in the Tt Pah2 structure 18 , and that in vitro their interaction is most likely constitutive.
The majority of the >500 residues that separate the N-Lip and C-Lip regions had high rates of deuterium exchange, indicating that they are likely disordered (Fig. 1b). One exception was a continuous stretch of~100 residues that were protected from exchange, which is indicative of secondary structure formation (Fig. 1b). This ordered region is called the middle lipin (M-Lip) domain and we discuss M-Lip in more detail below.
Lipin 1 association with membranes. In line with previous observations 22,27 , recombinant lipin 1 bound strongly to PC/PA liposomes (see below). To identify the membrane-binding regions of lipin 1 we employed HDX-MS, which has been particularly useful in examining protein-lipid interactions 30,31 . HDX-MS experiments were carried out in the presence and absence of PC/ PA liposomes and at 4 different time points of exchange (3, 30, 300, and 3000 s).
HDX-MS revealed multiple peptides that were protected from H/D exchange in the presence of liposomes, which suggests these regions associate with membranes. The regions protected by membrane were distributed throughout the primary structure and clustered into several key areas ( Fig. 1c-e, Supplementary Fig. 1a, c). This included (i) the N-terminus (residues 2-12, 13-30) that is predicted to form an amphipathic helix (Fig. 1c) as in other PAPs 18,26 , (ii) the C-terminal end of the M-Lip domain (residues 544-555) that is enriched in basic and hydrophobic residues ( Fig. 2a), (iii) peptides in the Ig-like domain (residues 651-662) that are predicted to lie at the membrane interface, (iv) the catalytic active site of the HAD-like domain (residues 660-685, 724-732) where PA hydrolysis occurs, and (v) the C-terminal end of the C-Lip (residues 840-845, 855-867) that is situated between the end of the HAD-like domain and the conserved Trp motif 32 ( Fig. 1a-e, Supplementary Fig. 1a, c). As stated above, peptides containing the NLS were not identified by tandem MS/MS. Thus, we were unable to assess the dynamics of membrane association for the NLS, which has previously been implicated in lipin 1 membrane binding 27 .
Notably, the membrane protected regions within the N-Lip and C-Lip regions of lipin 1 (Fig. 1c, Supplementary Fig. 1a, c) were nearly identical with those previously observed for Tt Pah2 18 . This suggests the catalytic core of PAP enzymes utilize a conserved mechanism for membrane binding that involves an N-terminal amphipathic helix, the HAD-like active site, and portions of the Ig-like domain.
The M-Lip domain. We next turned our attention to M-Lip, as the HDX-MS experiments suggested it may represent a third domain in lipin 1 that is involved in membrane binding. BLAST searches revealed that the M-Lip domain was selectively found in lipin homologs and was absolutely conserved in all mammalian lipins (Fig. 2a). The M-Lip domain was also detected in some plant (A. thaliana), fungal (Cryptococcus neoformans), ciliate (Tetrahymena thermophila), insect (Drosophila melanogaster), and apicomplexan (Plasmodium falciparum) lipin homologs (Supplementary Fig. 2) but was not detected in Sc Pah1. The M-Lip domain is thus one feature that distinguishes mammalian and mammalian-like lipin PAPs from Sc Pah1.
Structure of the M-Lip domain reveals a new protein fold. The M-Lip domain did not share sequence homology with any domains of known function. To determine if M-Lip was indeed a protein domain with a defined tertiary structure, we sought to determine its structure. The mouse lipin 1 M-Lip domain was purified from Escherichia coli. The detergent Triton X-100 was required to prevent the M-Lip domain from pelleting during centrifugation after cell lysis but was not required in subsequent purification steps if high salt concentrations were maintained.
Despite extensive efforts, we have yet to successfully crystallize the complete M-Lip domain.
We, therefore, truncated the M-Lip domain of mouse lipin 1 to remove the C-terminal cluster of hydrophobic and basic residues (Fig. 2a) implicated in membrane binding by HDX-MS (Fig. 1d). We refer to this construct as the M-Lip xtal domain. The M-Lip xtal domain could be purified without detergent and was extremely stable with a melting temperature of 65°C ( Supplementary Fig. 3). The structure of the M-Lip xtal domain of mouse lipin 1 was determined to resolutions of 1.5 Å and 1.9 Å in two unique space groups (Table 1, Supplementary Fig. 4a, b). Phases were obtained using single-wavelength anomalous diffraction from selenomethionine (SeMet)-derivatized protein (Table 1). We also determined the structure of the mouse lipin 2 M-Lip xtal domain to 2.5 Å resolution (Table 1, Supplementary Fig. 4c). The M-Lip xtal domain from mouse lipin 3 was easily purified but did not crystallize.   d   100  50  150  300  250  350  500  450  550  700  650  750  900  850   100  300  500  700   The structure of the M-Lip xtal domain revealed a protein fold with a three-stranded anti-parallel β-sheet at the core (Fig. 2b). Flanking the β-sheet were a set of two alpha helices (α1 and α2) and a short 3-10 helix (η1) that were oriented perpendicular to the β-sheet and two C-terminal α-helices (α3 and α4) that were oriented parallel to the β-sheet. A Dali search 33 did not identify any similar existing protein structures. Thus, the M-Lip domain represents a previously unrecognized and novel protein fold.  Mammalian lipins are known to form both homo and heterooligomers 34 . We hypothesized that the M-Lip domain may be involved in lipin oligomerization. We, therefore, deleted the M-Lip domain in the context of full-length mouse lipin 1 (ΔM-Lip lipin 1) and purified ΔM-Lip lipin 1 from Sf9 cells. In comparison to wild-type lipin 1, the size exclusion profile of ΔM-Lip lipin 1 was shifted towards a lower molecular weight (Fig. 2f, Supplementary Fig. 5). This is consistent with a role for the M-Lip domain in lipin 1 dimerization.
Analysis of the dimer interface of the M-Lip revealed an extensive network of interactions between the two subunits with residues within the α3 helix, which was located in the center of the dimer, mediating the majority of these interactions (Fig. 2c). Notably, the residues involved in dimerization were highly conserved among human and mouse lipin 1, lipin 2, and lipin 3 paralogs (Fig. 2a, g). Thus, we suspected that the M-Lip domain might also be involved in lipin hetero-oligomerization. Consistent with this hypothesis, deletion of the complete M-Lip domain or the M-Lip xtal domain modestly reduced the ability of lipin 1 to co-immunoprecipitate with lipin 1, lipin 2, and lipin 3 (Fig. 2h, Supplementary Fig. 6).
The M-Lip is not necessary for lipin transcriptional coactivator activity. The high conservation of the M-Lip region among mammalian lipins suggested that it plays a role(s) in lipin function. Mammalian lipins function both as PAP enzymes and as transcriptional co-activators for PPARα and PGC1α with an LxxLL motif in the C-Lip critical for this latter function 15 . Since Sc Pah1 has neither an M-Lip domain nor transcription coactivator activity, we hypothesized M-Lip may be necessary for lipin co-activation function.
Using HEK293 cells and the established luciferase-based assay 8,15 , lipin 1 increased the transcription of luciferase under the control of three peroxisome proliferator response elements (PPREs) in the presence of peroxisome-proliferator-activated receptor alpha (PPARα) and retinoic X receptor alpha (RXRα) (Fig. 3a). Transcription was further increased in the presence of peroxisome proliferator-activated receptor gamma co-activator 1alpha (PGC1α) (Fig. 3a). Deletion of either the M-Lip or M-Lip xtal sequences had no significant effect on PGC1α coactivation (Fig. 3a). We concluded that the M-Lip domain is not necessary for lipin 1 to function as a transcriptional coactivator. Deletion of the M-Lip domain reduces lipin PAP activity. We next tested whether deletion of the M-Lip domain affected lipin 1 PAP activity in vitro using purified protein (Fig. 3b) and the fluorescent substrate nitrobenzoxadiazole-phosphatidic acid (NBD-PA). Wild-type and ΔM-Lip lipin 1 had near identical PAP activities when the substrate NBD-PA was incorporated into Triton X-100 mixed micelles (Fig. 3c, Supplementary Fig. 7). We next assessed PAP activity with NBD-PA incorporated into liposomes composed of palmitoyl-oleoyl-phospholipids, which represent a more physiologically relevant in vitro system. Liposomes composed solely of phosphatidylcholine (PC) or a mixture of PC and phosphatidylethanolamine (PE) were used, as the membrane lipid PE has previously been shown to increase PAP activity of lipin 1 22,28 . The PAP activity of wild-type and ΔM-Lip lipin 1 were similar in PC liposomes (Fig. 3d). However, unlike wild-type lipin 1, the activity of ΔM-Lip lipin 1 did not increase with the addition of PE (Fig. 3d). Thus, PAP activity in PC/PE liposomes was~50% lower for ΔM-Lip lipin 1 compared to wild-type, which suggests that the M-Lip domain is involved in the PE-mediated effects on lipin 1 PAP activity.
Deletion of M-Lip reduces membrane association and alters subcellular localization. Given our findings that M-Lip reduces PAP activity on PC/PE liposomes, we hypothesized that the M-Lip domain affects lipin 1 binding to membranes. Using a liposome sedimentation assay, we found that lipin 1 bound to liposomes containing the neutral lipids PC or PC/PE, and liposome association was further enhanced by the presence of 20 mol% PA (Fig. 3e, f) composition (Fig. 3e, f). As a negative control, lipin 1 sedimentation was not affected by lipid-induced aggregation (Supplementary Fig. 8), as previously observed for PAP from rat liver 35,36 .
To assess if the M-Lip domain affects the subcellular distribution of lipin 1, we transiently transfected Cos-7 cells with wild-type and ΔM-Lip lipin 1 fused at their C-terminus with monomeric enhanced GFP (mEGFP) and assessed their subcellular localization using confocal microscopy. Lipin 1 co-localized with the ER marker mApple-Sec61b (Fig. 3g, h, Supplementary Fig. 9a). In contrast, ΔM-Lip lipin 1 also co-localized with the ER but accumulated in the nucleus (Fig. 3g, h, Supplementary Fig. 9b). We concluded that the M-Lip domain is necessary for proper subcellular localization and full-membrane binding in vitro.
The isolated M-Lip domain binds membranes. To verify that the M-Lip domain directly binds membranes, we conducted liposome sedimentation assays using the isolated M-Lip and M-Lip xtal domains purified from Escherichia coli. The M-Lip domain bound to PC liposomes, and membrane association was enhanced by the presence of the anionic lipids PA, phosphatidylserine, and phosphatidylinositol ( Fig. 4a, b). In contrast, the M-Lip xtal domain, which lacks the conserved C-terminal hydrophobic and basic residues, exhibited weak association with liposomes that was not significantly affected by the presence of anionic lipids (Fig. 4a, b).
Next, we employed HDX-MS to identify the regions of the M-Lip domain that interact with membranes. HDX-MS experiments were conducted in the presence and absence of PC/PA/PE liposomes. PE was included as we and others have observed differences in lipin PAP activity in the presence of PE 22,37 . Several peptides were protected from deuterium exchange in the presence of liposomes. The strongest protection was observed for peptides containing a WWF motif (Fig. 4c, Supplementary Fig. 1), which are part of the conserved hydrophobic and basic residues at the C-terminal end of the M-Lip (Fig. 2a). Overall, HDX changes observed in the M-lip domain using the M-lip domain alone versus full-length lipin 1 were very similar with slight differences potentially due to variation in membrane binding parameters. Two additional peptides present in the crystallized M-Lip xtal domain were also protected (Fig. 4c). These peptides mapped to the same surface, which suggests that the core of the M-Lip xtal domain can also interact with membranes but is not the major determinant of membrane binding (Fig. 4c).
To determine if the isolated M-Lip domain localized to membranes within cells, Cos-7 cells were transfected with M-Lip fusions with mEGFP. A mEGFP M-Lip fusion strongly colocalized with the ER marker mApple-Sec61b (Fig. 4d, e,  Supplementary Fig. 10a). In contrast, the mEGFP M-Lip xtal fusion accumulated in the nucleus with a minor co-localization with the ER, suggesting a role for the hydrophobic and basic residues at the C-terminus of M-Lip in membrane targeting (Fig. 4d, e, Supplementary Fig. 9b). Taken together, these results identify the M-Lip domain as a new type of membrane-binding domain.
M-Lip role in adipogenesis. Lastly, we sought to characterize the role of the M-Lip domain in lipin 1 function in adipocytes. Lpin1 is expressed in a bi-phasic manner during adipocyte differentiation, and has specific roles in both the early stages of adipogenesis and in the formation of mature, lipid-laden adipocytes 5 . In preadipocytes, lipin 1 PAP activity regulates PA signaling to promote expression of a key adipogenic transcription factor, peroxisome proliferator-activated receptor γ (PPARγ) 5,38 . Lipin 1 PAP activity is also required in mature adipocytes for triglyceride synthesis and lipid hydrolysis 38,39 . We investigated whether the M-Lip domain is required for the optimal activity of lipin 1 in the regulation of gene expression and lipid accumulation during adipocyte differentiation.
3T3-L1 preadipocytes were transfected with expression vectors for wild-type or ΔM-Lip lipin 1. We titrated expression levels of wild-type and ΔM-Lip lipin 1 to ensure that they were expressed at comparable levels ( Supplementary Fig. 11a). Expression of adipocyte genes was monitored at intervals during differentiation, and neutral lipid accumulation was examined after 5 days (Fig. 5a). Wild-type lipin 1 expression, but not ΔM-Lip lipin 1 expression, increased the levels of oil red O-stained lipids beyond those observed in cells transfected with a vector control when assessed at day 5 ( Fig. 5b and Supplementary Fig. 11b). Both wildtype and ΔM-Lip lipin 1 expression led to enhanced Pparg expression compared to vector controls, but ΔM-Lip was less effective than wild-type lipin 1 in promoting expression of the transcription factor CCAAT/enhancer binding protein α (Cebpa), or mature adipocyte genes such as fatty acid binding protein 4 (Fabp4) and adiponectin (Adipoq) (Fig. 5c). An independent replicate experiment also showed that ΔM-Lip lipin 1 was less effective at inducing lipogenic genes encoding acetyl-CoA carboxylase (Acaca) and diacylglycerol acyltransferase 1 (Dgat1) (Supplementary Fig. 11c).

Discussion
This study identifies the M-Lip domain as a new protein fold that dimerizes and binds membranes. The M-Lip domain is functionally important for lipin subcellular localization, and enhancing adipogenesis, and can directly affect PAP activity in vitro in a manner that is dependent on membrane lipid content. With the exception of lipin 1 transcriptional co-activator function, which was unaffected, deletion of the M-Lip had a negative effect but did not completely abrogate any lipin function. This is consistent with the sufficiency of the N-Lip and C-Lip regions for lipin PAP activity 18 and all known disease mutations residing within the N-Lip or C-Lip regions.
We could identify the M-Lip domain in lipin homologs from several evolutionarily distant organisms, but not in Sc Pah1, nor in any non-lipin proteins from any species. Thus, the M-Lip domain appears to represent a unique feature of lipins and an evolutionary branch point that differentiates lipin PAP enzymes from mammals, invertebrates, ciliates, and plants from Sc Pah1.
Membrane association is the main regulatory mechanism that controls lipin PAP activity. Our HDX-MS studies reveal that membrane binding involves regions distributed throughout lipin 1. We propose mammalian lipins associate with membranes through a series of multi-valent interactions involving the Nterminal amphipathic helix, nuclear localization signal, Ig-like domain, HAD-like phosphatase domain, and the M-Lip domain. In this model, the M-Lip domain contributes one site for membrane binding and simultaneously doubles the number of membrane binding interactions through dimerization (Fig. 6).
All of the individual membrane-binding regions in lipins that have been characterized to date are responsive to the presence of anionic lipids, in particular PA 18,22,26,27,37 . This suggests that lipins may be recruited to cellular membranes in response to elevated PA levels. Given that PA is also the lipin enzyme substrate, this may reflect a mechanism to maintain low levels of PA. A role for PE may also exist, as PE has been hypothesized to impact the electrostatic charge of PA 22,37 . While we observed increased PAP activity in the presence of PE, intriguingly PE modestly decreased lipin 1 membrane association, which has also been observed for Sc Pah1 40 . In addition, the effects of acyl-chain composition and membrane fluidity on lipin activity warrants future study. This hypothesis derives from the similar specific activities we observed in comparison to other groups using Triton X-100 mixed micelles 22,41 (Fig. 3c), which was diminished in liposomes containing palmitoyl-oleoyl phospholipids (Fig. 3d). In contrast, a previous report found no major differences in specific activity when comparing mixed micelles and liposomes composed of di-oleoyl phospholipids 22 .
As revealed by HDX-MS, the N-Lip and C-Lip regions are predominantly ordered in solution with substantial secondary structure. This finding suggests the N-Lip and C-Lip co-fold to form a split Ig-like domain and the HAD-like catalytic domain observed in Tt Pah2 18 . This is consistent with previous findings that a fusion of the N-Lip and C-Lip regions in both mouse lipin 2 18 and Sc Pah1 32 is sufficient for PAP activity in vitro. Notably, while the association of the N-Lip and C-Lip regions with one another appears to be constitutive in vitro, we cannot rule out that their association is transient or regulated in cells.
Here we expand the domain architecture of mammalian lipins to include M-Lip as a third protein domain that forms a stable dimer. Therefore, mammalian lipins must, at a minimum, be dimeric with the respective N-Lip and C-Lip regions co-folding either in cis (from the same subunit, Fig. 6) or in trans (from different subunits). We note that if the N-Lip and C-Lip interaction occurs in trans, this creates the potential to form larger oligomers when 4 or more lipin subunits combine. Lipin homo-and hetero-oligomerization has been demonstrated previously 34,42 , and the size exclusion profile of recombinant lipin 1 yielded two peaks consistent with oligomerization beyond a dimer. Lastly, our data suggest that the conserved dimer interface of the M-Lip may contribute to both homo and hetero-dimerization of mammalian lipins. This provides a foundation for more detailed experiments to unravel the functional consequences of lipin oligomerization in the regulation of phospholipid and triglyceride synthesis.    Supplementary Fig. 11b. c Expression of genes encoding adipogenic transcription factors PPARγ (Pparg) and C/EBPα (Cebpa), adipocyte fatty acid binding protein FABP4 (Fabp4), and the adipocyte hormone, adiponectin (Adipoq), at days 0, 2, and 4 of differentiation. Gene expression was analyzed by 2-way ANOVA, followed by paired t-tests when ANOVA results were significant. *p < 0.05; **p < 0.01; ***p < 0.001. Data are presented as mean values ± SD. n = 4 biologically independent experiments. Source data are provided as a Source data file. in Ultra-High Yield Flasks (Thompson Instrument Company) to an OD 600nm of 1.5 and then cooled at 10°C for 2 h. Protein expression was induced with isopropyl β-D-1-thiogalactopyranoside (IPTG) at 15°C for 12 h, cells were harvested by centrifugation and stored at −80°C. Frozen cells were resuspended in buffer A (50 mM Tris-HCl, 60 mM imidazole, 500 mM NaCl, 5%(v/v) glycerol, 1% v/v Triton X-100, pH 7.4), lysed by sonication, centrifuged at 58,540 × g at 4°C for 1 h, and applied to a gravity column with pre-equilibrated Ni-NTA resin (GoldBio). The column was washed with buffer A without Triton X-100, and protein was eluted with buffer B (50 mM Tris-HCl, 300 mM imidazole, 500 mM NaCl, 5% (v/v) glycerol, pH 7.4). The protein was diluted 4 fold in buffer C (50 mM HEPES, 50 mM NaCl, pH 7.35), applied to a HiTrap SP HP cation exchange column, washed with 5 column volumes of buffer C, and eluted with a linear gradient with buffer C supplemented with 1 M NaCl. Protein was aliquoted, flash frozen, and stored at −80°C.
Mouse lipin 1 M-Lip xtal domain in ppSUMO. The mouse lipin 1 M-Lip xtal domain that was crystallized was expressed and purified from the construct cloned into the ppSUMO plasmid. Expression conditions and the Ni-NTA purification protocol were identical as that described here for the M-Lip xtal domain in pET28a. However, after elution from the Ni-NTA column, the protein was digested with purified ULP-1 overnight at 4°C. The digestion mixture was diluted 10 fold in buffer A, and re-applied to a Ni-column to remove the His-SUMO fusion. After, the M-Lip xtal protein was further purified by SEC using a HiLoad Superdex 75 26/600 column (GE Healthcare) equilibrated with SEC Buffer. Pooled fractions from SEC were concentrated to 10-20 mg/ml, aliquoted, flash frozen, and stored at −80°C.  cycles, then centrifuged at 9168 × g for 30 min. The supernatant was collected as lipid-BSA complexes. 20 μL lipid-BSA complexes were incubated with 20 μL lipin 1 at 4°C for 30 min, then centrifuged at 100,000 × g for 1 h. The supernatant was collected carefully from the tube. The pellet (P) and the supernatant (S) fractions were analyzed by SDS-PAGE.
HDX-MS to determine ordered and disordered regions of full length lipin. HDX reactions were conducted in 20 µL reaction volumes with a final concentration of 0.63 µM full length lipin 1 per sample. Exchange was carried out in triplicate for a single time point (3 s on ice). All tips and tubes were pre-chilled at 4°C. Hydrogen deuterium exchange was initiated by the addition of 18 µL of icecold D 2 O buffer solution (20 mM HEPES pH 7, 100 mM NaCl) to the protein solution, to give a final concentration of 84.8% D 2 O. Exchange was terminated by the addition of 50 µL acidic quench buffer, giving a final concentration 0.6 M guanidine-HCl and 0.9% formic acid (pH~2.5). Samples were immediately frozen in liquid nitrogen at −80°C. A denatured protein sample for back exchange correction was generated by first denaturing the protein in 6 M guanidine for 1 h at 18°C. Following denaturation, 18 μL of D 2 O buffer was added to the denatured protein and allowed to incubate for 15 min at 18°C before quenching as described above. Samples were flash frozen in liquid nitrogen and stored at −80°C until injection onto an ultra-performance liquid chromatography system for proteolytic cleavage, peptide separation, and injection onto a QTOF for mass analysis. The reaction allowed to proceed for 3 s, 30 s, 300 s, or 3000 s at 18°C before being quenched with 50 µL diluted ice-cold acidic quench buffer resulting in a final concentration of 0.6 M guanidine-HCl and 0.9% FA post quench to a final pH of~2.5 (as described above). All conditions and timepoints were created and run in triplicate. Samples were flash frozen in liquid nitrogen and stored at −80°C until injection onto an ultra-performance liquid chromatography system for proteolytic cleavage, peptide separation, and injection onto a QTOF for mass analysis.
Protein digestion and MS/MS data collection. Protein samples were rapidly thawed and injected onto an integrated fluidics system containing a HDx-3 PAL liquid handling robot and climate-controlled chromatography system (LEAP Technologies), a Dionex Ultimate 3000 UHPLC system, as well as an Impact HD QTOF Mass spectrometer (Bruker). The protein was run over either one immobilized pepsin column at 10°C (order-disorder and full-length lipin) or two (M-Lip experiment) immobilized pepsin columns at 2°C and 10°C (Applied Biosystems; Poroszyme Immobilized Pepsin Cartridge, 2.1 mm × 30 mm; Thermo-Fisher 2-3131-00; Trajan; ProDx protease column, 2.1 mm × 30 mm PDX.PP01-F32) at 200 µL/min for 3 min. When using two pepsin columns, the protein is first run over the column in the 10°C box, followed by running over the 2°C column. The resulting peptides were collected and desalted on a C18 trap column at 2°C (Acquity UPLC BEH C18 1.7 µm column (2.1 mm × 5 mm); Waters 186003975).
The trap was subsequently eluted in line with a C18 reverse-phase separation column (2˚C) (Acquity 1.7 µm particle, 100 mm × 1 mm 2 C18 UPLC column, Waters 186002352), using a gradient of 3-35% B (Buffer A 0.1% formic acid; Buffer B 100% acetonitrile) over 11 min immediately followed by a gradient of 35-80% over 5 min. A blank method is run after every MS method to limit carryover, using a 3-35% B gradient over 4 min followed by a gradient of 35-80% over 2 min. Mass spectrometry experiments acquired over a mass range from 150 to 2200 m/z using an electrospray ionization source operated at a temperature of 200°C and a spray voltage of 4.5 kV.
Peptide identification. Peptides were identified using data-dependent acquisition following tandem MS/MS experiments (0.5 s precursor scan from 150 to 2000 m/z; twelve 0.25 s fragment scans from 150 to 2000 m/z). MS/MS datasets were analyzed using PEAKS7 (PEAKS), and peptide identification was carried out by using a false discovery-based approach, with a threshold set to 1% using a database of known contaminants found in Sf9 and Escherichia coli cells 44 . The search parameters were set with a precursor tolerance of 20 ppm, fragment mass error 0.02 Da, charge states from 1 to 8, and allowing for phosphorylation at Serine, Threonine, and Tyrosine.
Mass analysis of peptide centroids and measurement of deuterium incorporation. HDExaminer Software (Sierra Analytics) was used to automatically calculate the level of deuterium incorporation into each peptide. All peptides were manually inspected for correct charge state and presence of overlapping peptides. Deuteration levels were calculated using the centroid of the experimental isotope clusters. The results for the experiment comparing lipin with and without liposomes are presented as relative levels of deuterium incorporation, with no control for back exchange, and the only correction was for the level of deuterium present in the buffer (either 84.9% or 80.1%). The fully deuterated sample in the orderdisorder experiment allowed for a back-exchange correction in this specific experiment during digestion and separation. Differences in exchange in a peptide were considered significant if they met all three of the following criteria: >4% change in exchange, >0.4 Da difference in exchange, and a p-value <0.01 using a two-tailed Student's t-test. We arrived at the %D and #D criteria partially based on the average error of the HDX experiments, with the %D value being at least 5-fold greater than the average error. All compared samples were set within the same HDX experiment. The raw HDX data are shown in four different formats. The raw peptide deuterium incorporation graphs for a selection of peptides with significant differences are shown, with the raw data for all analyzed peptides in the source data. To allow for visualization of differences across all peptides, we utilized number of deuteron difference (#D) plots. These plots show the total difference in deuterium incorporation over the entire H/D exchange time course, with each point indicating a single peptide. We also generated #D plots showing the difference in deuterium incorporation for each individual timepoint. Butterfly plots were used to visualize the % deuterium incorporation of each individual peptide at each timepoint. The data analysis statistics for all HDX-MS experiments are in Supplementary Table 2 according to the guidelines of Masson et al. 29 . The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository 45 with the dataset identifier PXD022172.
Crystallization and data collection. All crystals were grown using the hangingdrop method by mixing 1.5 μL of reservoir solution mixed with 1. Co-Immunoprecipitation. Co-immunoprecipitation experiments were performed to detect lipin 1 dimer formation. To unambiguously identify the interaction between two lipin monomers, equal numbers of Hepa1-6 cells were transfected with independent lipin 1 constructs that were fused to different epitopes (V5 or BirA*-HA). To detect lipin heterodimer formation, lipin 1 constructs with V5 epitope (wild-type lipin 1-V5, ΔMLip-V5, or ΔM-Lip xtal -V5) were co-transfected with lipin 1-BirA*-HA, lipin 2-BirA*-HA or lipin 3-BirA*-HA. Two days after transfection, cells were harvested and lysed in 0.1% NP-40 with phosphatase inhibitor cocktails (Sigma-Aldrich, St. Louis, MO). After brief sonication, the supernatants were collected by centrifugation at 13,400 × g for 10 min at 4°C. 10% of the resulting protein was reserved as "input" and the remainder was immunoprecipitated. For immunoprecipitation of V5-tagged proteins, cell lysates were incubated with a 1:500 dilution of anti-V5 antibody (ThermoFisher Scientific, R960-25, Waltham, MA) at 4°C overnight. Cell lysate/antibody mixture were then incubated with protein A/G-agarose beads (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) for 2 hr at 4°C. The precipitates were washed three times with lysis buffer, boiled in sample buffer for 5 min and subjected to immunoblot assay with a 1:1000 dilution of anti-HA antibody (Cell Signaling Technology, Inc., #3724, Danvers, MA) or 1:5000 dilution of anti-V5 antibody. Densitometric quantitation was performed via a ChemiDox XRS+ using the manufacturer's software (Bio-Rad, Hercules, CA). Independent co-immunoprecipitation experiments were performed with lipin proteins having different epitope tags. For these experiments, full-length or ΔM-Lip lipin 1 tagged with V5 were immunoprecipitated with lipin 1-FLAG or lipin 3-Myc using a 1:500 dilution of anti-V5 antibody. When cells were harvested, 10% of protein was reserved as "input" and the rest was subjected to immunoprecipitation as described above. Immunoblots were detected with a 1:500 dilution of anti-FLAG antibody (#PA1-984B, Invitrogen, Carlsbad, CA) or a 1:1000 dilution of anti-Myc antibody (#2278, Cell Signaling, Danvers, MA). Gels were designed to allow direct comparison of the levels of protein input, flow-through, and immunoprecipitation of full-length and ΔM-Lip lipin 1. Densitometric quantitation was performed via a ChemiDox XRS+ using the manufacturer's software (Bio-Rad, Hercules, CA).
Oil red O stain. At day 5 of adipocyte differentiation, cells were fixed in 10% formalin for 20 min, washed with 60% isopropanol, and stained with oil red O solution (0.2% w/v in isopropanol). Cell images were captured at 100x magnification. Pictures from 4 independent wells per treatment group were used. PAP activity assays. Micelles containing 5 mol% nitrobenzoxadiazolephosphatidic acid (NBD-PA) (Avanti Polar Lipids, #9000341) and Triton X-100 (Research Products International Corp, #400001) were generated in a buffer containing 50 mM Tris pH 7.5, 100 mM NaCl, 10 mM 2-mercaptoethanol, and 4 mM MgCl 2 , to give a final bulk concentration of 80 µM NBD-PA. Liposomes used for PAP assays were composed of 10 mol% NBD-PA with 0 mol% or 40 mol% POPE with the remaining as POPC (90 mol% or 50 mol%) and prepared in buffer containing 50 mM Tris, pH 7.5, 100 mM NaCl, 10 mM 2-mercaptoethanol, and 4 mM MgCl 2 to give a final bulk concentration of 150 µM NBD-PA.
Both the mixed-micelle and liposome samples were analyzed by HPLC using a Spectra 3 µm C8SR column (3 µm particle, 150 × 3.0 mm, Peeke Scientific, #S-3C8SR-FJ) under the following conditions: solvent A: water containing 0.2% formic acid and 1 mM ammonium formate; solvent B: methanol containing 0.2% formic acid and 1 mM ammonium formate; flow rate: 0.5 ml/min. The gradient profile started at 80% for solvent B and was increased to 98% B after 7 min, then kept at 98% for 3 min. 10 µL samples were injected onto the column, which was kept at 35°C at all runs. The fluorescent signal was detected at excitation and emission wavelengths of 470 and 530 nm, respectively. The detector signal was recorded and integrated by using Agilent technology OpenLAB CDS ChemStation edition software.