Lipid droplet and peroxisome biogenesis occur at the same ER subdomains

Nascent lipid droplet (LD) formation occurs in the ER membrane1-4. It is not known whether LD biogenesis occurs stochastically in the ER or at subdomains with unique protein and lipid composition. We previously identified ER subdomains in S. cerevisiae that contain Pex30, a reticulon-like ER-resident membrane protein5. There are ~25 Pex30-containing puncta in the ER per cell. These sites are regions where preperoxisomal vesicles (PPVs) are generated5. Here we show that Pex30 subdomains are also the location where most nascent LDs form. Mature LDs remain associated with Pex30 subdomains and the same Pex30 subdomain can simultaneously associate with a LD and a PPV. Pex30 subdomains become highly enriched in diacylglycerol (DAG) during LD biogenesis, indicating they have a unique lipid composition. We find that in higher eukaryotes multiple C2 domain containing transmembrane protein (MCTP2) is the functional homologue of Pex30; MCTP2 resides in ER subdomains where most nascent LD biogenesis occurs and that are often associated with peroxisomes. Together, these findings indicate that most LDs and PPVs form and remain associated with conserved ER subdomains and suggest a link between LD and peroxisome biogenesis.

We wondered whether LD biogenesis occurs at Pex30 domains in the ER for two reasons. First, there are about 10-fold more Pex30 domains than there are PPVs in cells 5 , suggesting these domains have other functions. Second, recent evidence suggests that some proteins play dual roles in the biogenesis of both LDs and peroxisomes; the Kapito group found that two proteins required for peroxisome biogenesis, Pex3 and Pex19, insert membrane-embedded proteins into the surface of LDs at ER subdomains 11 .
To determine whether nascent LDs mature at Pex30 subdomains, we visualized LD biogenesis in a S. cerevisiae strain in which LD formation can be controlled. Four enzymes produce neutral lipids in this yeast: Are1 and Are2, which generate steryl esters, and Lro1 and Dga1, which synthesize TAG. Cells lacking all four proteins lack neutral lipids and LDs. We used a strain in which the galactose regulatable promoter GAL1 controls expression of LRO1 and the other three neutral lipid-synthesizing enzymes are not produced (GAL1-LRO1 3D). When this strain is grown in a medium containing raffinose, it lacks LDs but begins to produce them when galactose is added 12 . The strain also expressed Pex30-2xmCherry and Erg6-GFP, a LD marker. Before LRO1 induction, Erg6-GFP is on the ER but it localizes to LDs after galactose addition (Fig. 1A). About 70% of Erg6-GFP punctae colocalize with Pex30-2xmCherry (Fig. 1B). Similar results were obtained when nascent LDs were visualized with the lipophilic dye BODIPY ( Supplementary Fig. 1A).
To verify that most nascent LDs mature at Pex30 subdomains, we induced LD production with a second method. When oleic acid is added to growing cells, they rapidly begin to produce new LDs. We added oleic acid to cells expressing Dga1-GFP from a high copy plasmid and endogenously expressed Pex30-2xmCherry. In growing cells, Dga1-GFP is in the ER but it relocates to the surface of LDs when LD production is induced. We found that Dga1-GFP puncta colocalize with Pex30-2xmCherry within 1-2 hours after oleic addition (Fig. 1C). Lro1-GFP similarly accumulates at sites containing Pex30-2xmCherry after oleic acid addition. (Supplementary Fig. 1B). Together, these findings indicate that most nascent LDs colocalize with Pex30 subdomains in the ER.
Next, we determined whether mature LDs, like nascent LDs, remain associated with Pex30 domains in the ER. Wild-type cells growing in standard growth media contain mostly mature LDs. Therefore, to determine whether Pex30 domains associate with mature LDs, we imaged cells expressing Pex30-2xmCherry and Erg6-GFP. About ~70% of Erg6-GFP puncta were closely associated with Pex30-2xmCherry ( Fig. 1D and   1G), suggesting that not only nascent LDs but also mature LDs remain associated with Pex30 subdomains. We also confirmed the presence of Pex30 at ER subdomains associated with LDs by examining the colocalization of Pex30-2xmCherry with other proteins known to be at these regions. Nem1 is part of a phosphatase complex that regulates Pah1, the yeast homologue of lipin, which plays a central role in controlling TAG levels in cells [13][14][15] . Nem1 forms puncta on the ER that are near sites of LD biogenesis 16,17 . We found that ~70% of Nem1-GFP punctae co-localized with Pex30-2xmCherry ( Fig. 1E and 1G), consistent with the idea that Pex30 subdomains remain associated with growing and mature LDs. We also determined whether Sei1, the yeast homologue of seipin, is associated with Pex30 subdomains. Seipin plays an important but still poorly understood role in LD biogenesis and it has been found that some Sei1 puncta localize at ER-LD contacts 17,18 . We found that 40% of Sei1-GFP puncta were associated with Pex30-2xmCherry domains ( Fig. 1F and 1G). Consistent with this finding, it was reported that Pex30 co-immunoprecipitates with Sei1 19 . It may be that the Sei1 puncta away from Pex30 puncta are those not associated with LDs. Taken together, these findings indicate that Pex30 subdomains frequently remain associated with mature LDs.
If Pex30 subdomains are indeed sites of LD biogenesis, we speculated that they might become enriched in neutral lipid precursors when nascent LD production is induced. DAG is a TAG precursor used by both TAG-synthesizing enzymes in yeast. To determine whether DAG becomes enriched at Pex30 sites, we developed a sensor to investigate the distribution of DAG in the ER. The DAG-binding tandem C1 domains of Protein Kinase D have been used to sense DAG in membranes 20 . We fused these domains to GFP and the transmembrane domain of Ubc6, a tail-anchored ER protein (ER-DAG sensor). After confirming that the sensor colocalizes with the ER marker RFP-HDEL ( Supplementary Fig. 1C), we expressed the sensor in cells expressing Pex30-2xmCherry. When the cells were grow in regular media, the sensor was all over the ER but it became highly enriched in portions of the ER within 1-2 hours after oleic acid addition, forming bright puncta. About 65% of the puncta colocalized with Pex30-2xmCherry ( Fig 1H). This finding indicates that some Pex30 subdomains become highly enriched in DAG when LD formation is induced, consistent with the idea that the subdomains are regions where LDs form. Why DAG does not become enriched at all Pex30 subdomains is unclear but suggests that these subdomains have multiple functions Interestingly, we found that the soluble C-terminal portion of the Pex30 (250-523aa) binds DAG immobilized on membranes ( Supplementary Fig. 1D), suggesting that Pex30 could be regulated by DAG accumulation at sites of LD biogenesis.
To investigate the role of Pex30p in LD biogenesis, we examined LDs in cells lacking Pex30 (pex30D). LDs in these cells were often more clustered and smaller than those in wild-type cells ( Fig. 2A-F). The decrease in LD size is probably not caused by a change in the level of neutral lipids in pex30D cells (Fig 2G and H). Although these cells had a small but significant decrease in TAG levels, this change is probably not large enough to decrease LD size. The change in LD size in pex30D cells could be because Pex30 affects membrane tension at sites of LD biogenesis. A similar role has been proposed for REEP1, a mammalian reticulon-like ER-shaping protein 21 22 .
Four proteins in yeast contain reticulon homology domains (RHDs) like that of Pex30: Pex28, Pex29, Pex31, and Pex32. We determined whether mutants lacking these proteins had changes in LD size, LD clustering, or neutral lipids levels but found they did not, though there was some LD clustering in cells lacking Pex29 ( Since LDs frequently associate with Pex30-containing portions of the ER, we wondered whether ER-LD contacts were altered in cells lacking Pex30. In yeast, LDs remain attached to the ER by a membrane neck that is thought to be formed from the cytoplasmic leaflet of the ER membrane and is continuous with the phospholipid monolayer surrounding LDs 12 . Dga1 and some other ER proteins with hairpin-like membrane domains can diffuse between the ER and LDs 12,23,24 . Thus, if ER-LD contacts are altered in pex30D cells, Dga1 movement between these organelles might be affected. We found that Dga1-GFP is primarily on the ER in late stationary growth phase but diffuses onto LDs when cells are transferred to fresh media. We determined the percent of cells that have Dga1-GFP only on LDs after cells were shifted to fresh media. Two hours after shift, there were significantly fewer pex30D than wild-type cells with Dga1-GFP on LDs (Fig. 2I). Interestingly, after 24 hours, when cells have returned to stationary growth phase, more pex30D cells than wild-type cells still have Dga1-GFP on LDs (Fig.   2I). These results suggest that Dga1-GFP movement between the ER and LDs slows in cell lacking Pex30, consistent with the idea that ER-LD contacts are altered in cells lacking Pex30.
Since seipin has been suggested to localize at sites where LDs are associated with ER 2, 25 , we wondered how LD biogenesis would be affected in cell lacking both Pex30 and seipin (Sei1). Surprisingly, these cells (sei1pex30Δ) have a substantial growth defect It is not clear why sei1pex30Δ cells have a growth defect. We found that sei1pex30Δ cells form large clusters of small and big LDs and the ER associated with the LDs is highly proliferated around the LDs ( Fig. 3C and 3D). These changes could affect ER function and cause a growth defect. Alternatively, Pex30 and seipin may share a common unknown function, perhaps related to lipid metabolism. Together, these findings provide additional evidence that Pex30 plays an important role in LD biogenesis and function and suggest that Pex30 and seipin may have partially overlapping functions.
Pex30 does not have a mammalian homologue but we wondered whether there is an RHD-containing protein in higher eukaryotes that plays a similar role. Using the structural homology prediction program HHpred 26 , we found that the human protein MCTP2 has a RHD similar to that of Pex30 (Fig. 4A). MCTP2 is an 878 amino acid ER resident protein with a membrane embedded regions containing the RHD near the Cterminus and three C2 domains 27,28 . MCTP proteins are conserved in higher eukaryotes.
Drosophila and C. elegans have one MCTP whereas humans contain MCTP1 and MCTP2 27 . Here we show that human MCTP2 is a functional homologue of Pex30. We expressed the C-terminal 237 amino acids of MCTP2, which contain the RHD, fused to complements the growth defect of sei1pex30D cells (Fig. 4B). Cells lacking the reticulons, Rtn1 and Rtn2, and the reticulon-like protein Yop1 (rtn1rtn2yop1D), have a defect in ER structure 29 that is corrected by overexpression of Pex30 5 . We found that YFP-MCTP2 similarly restores ER structure. The cortical ER forms large sheet-like structures in rtn1rtn2yop1D cells that are not present in wild-type cells, which contains largely tubular ER in the cortex. We found that cortical ER structure in rtn1rtn2yop1D cells becomes tubular when YFP-MCTP2 is expressed in these cells (Fig. 4C). We previously found that rtn1rtn2yop1D cells that also lack the lipid regulator Spo7 are not viable but grow when Pex30 is overexpressed 5 . Similarly, overexpression of YFP-MCTP2 also rescued the rtn1rtn2yop1spo7D mutant ( Supplementary Fig. 3A). Together, these findings indicate that YFP-MCTP2 is an ER-shaping protein that can functionally replace Pex30 in yeast.
We found that YFP-MCTP2 localizes to ER subdomains in mammalian cells.
YFP-MCTP2 was transiently expressed in COS7 cells together with the ER marker Sec61-mCherry. When YFP-MCTP2 was expressed at low levels, it was found in puncta in the ER (Fig. 4D). These puncta are stable and remain associated with the same region of the ER over time (Supplementary video 1). When expressed at high levels, YFP-MCTP2 localized to ER tubules and the edges of ER sheets ( Supplementary Fig. 3B), a localization shared with the reticulons 30 and consistent with the idea that the C-terminal region of MCTP2 contains a RHD.
We next determined whether MCTP2-subdomains are sites of LD biogenesis and associate with LDs. COS7 cells were transiently transfected with YFP-MCTP2 and LiveDrop-mCherry, a fusion protein demonstrated to target nascent LDs forming in the ER and mature LDs 3 . We found that most MCTP2 and LiveDrop-mCherry punctae colocalized ( Fig. 4E and 4F). LiveDrop-mCherry punctae that did not colocalize with YFP-MCTP2 were largely not associated with the ER (Fig. 4E). These findings indicate that YFP-MCTP2 localizes to ER sites where new LDs form and suggest that ER domains containing YFP-MCTP2 are often associated with mature LDs. Thus, MCTP2 may play a role in LD biogenesis in mammalian cells like that of Pex30 in yeast.
We wondered whether Pex30/MCTP2 sites in the ER not only associate with LDs but also with PPVs and peroxisomes, since we previously found that PPVs, are generated at the Pex30 subdomain 5 . To visualize PPVs in S. cerevisiae, we used strains that lack the proteins Pex3 and Atg1 (pex3atg1D). Pex3 is required for peroxisome biogenesis and it had been thought that cells lacking this protein do not contain PPVs 31 . However, it was subsequently discovered that PPVs are present in pex3D cells when they also lack Atg1, which is required for autophagy 32 . These PPVs contain Pex14-GFP and there are typically 1 or 2 PPVs per cell. We found that some Pex14-GFP puncta are on vesicles while others are on the ER, at Pex30 subdomains 5 . To colocalize PPVs, LDs and Pex30 subdomains, we expressed Pex14-GFP, Pex30-2xmCherry, and the LD marker Erg6-BFP in pex3atg1D cells. Remarkably, most PPVs are closely associated or colocalized with Pex30 subdomains and LDs (Fig 5A and Supplementary Fig. 4). The association between PPVs and LDs is even more pronounced in pex3atg1D cells that also lack seipin ( Supplementary Fig. 4). These findings reveal that PPVs and LDs often remain associated with the same Pex30 subdomain and suggest that a single subdomain may generate both PPVs and LDs. It is also possible that both PPVs and LDs form at different regions in the ER but subsequently associate with the same Pex30 subdomain.
Pex30 subdomains are also associated with mature peroxisomes in yeast 33 . We found that MCTP2 subdomains similarly associate with both LDs and mature peroxisomes in mammalian cells. YFP-MCTP2, LiveDrop-mCherry, and the peroxisome marker CFP-SKL were co-expressed in COS7 cells. LDs and peroxisomes often associate with the same MCTP2 subdomain (Fig. 5B). About 30 percent of the MCTP2 and LiveDrop punctae that colocalize are also associated with peroxisomes (Fig. 5C). These

Yeast strains and plasmids
The strains and plasmids used in this study are listed in Tables S1 and S2.

Media and growth conditions
Yeast cells were grown at 30 o C, unless otherwise indicated, in YPD medium (1% Bacto yeast extract, 2% Bacto Peptone, and 2% glucose) or in synthetic complete (SC) media containing 2% glucose, 0.67% yeast nitrogen base without amino acids (United States Biological) and an amino acid mix (United States Biological). In some cases the glucose in SC was replaced with 2% raffinose or 2% galactose. When inducing LD production, cells were washed with sterile water twice and transferred to SC containing 1mM oleic acid and 1% Brij58. Images were taken within 1-2 hours. When staining LDs with BODIPY, cells in early stationary growth phase were washed with phosphate buffered saline and incubated with 0.5µg/ml BODIPY 493/503 (Invitrogen) for 10 minutes.
Prior to live cell imaging, the medium was changes to CO 2 -independent medium (Gibco) containing 10% FBS and 2 mM L-glutamine.

Fluorescence microscopy
For Fig

Purification of MBP-Pex30 (250-523) protein.
The portion of PEX30 encoding amino acids 250-523 was cloned into the plasmid pMAL-c2x. The plasmid was expressed E.coli grow to mid-logarithmic growth phase at 37 o C, 0.2mM IPTG was added to the medium, and the cell were grown at 30 o C for 3 hours. The cells were harvested by centrifugation, resuspended in lysis buffer (200mM NaCl, 20mM TRIS pH 8.0, 1mM EDTA, 1mM DTT and protease inhibitor), and lysed using french press. The soluble protein was purified from the lysate using amylose resin (New England Biolabs) on an FPLC column (GE healthcare). Protein was eluted using 10mM maltose. The purified protein was separated using HiLoad 16/60 Superdex 200 prep grade (GE healthcare). Peak fractions with purified protein were collected, pooled and analyzed using SDS-PAGE stained with Coomassie. Purified protein (0.5µg/ml) was used for in vitro lipid binding assay.

Electron microscopy (EM)
Sample preparation and visualization was performed as described previously 6 . In brief, yeast cells were grown to mid-logarithmic growth phase, and 10 OD 600 units of cells were pelleted and fixed in 1 ml of fixative media (2.5% glutaraldehyde, 1.25% PFA, and 40 mM potassium phosphate, pH 7.0) for 20 min at room temperature. Cells were pelleted, resuspended in 1 ml fresh fixative media, and incubated on ice for 1 h. The cells were pelleted, washed twice with 0.9% NaCl, once with water, incubated with 2%

Neutral Lipids detection
Cells were grown to early stationary growth phase in SC glucose containing 10 µCi/ml [ 3 H] acetate (American Radiolabeled Chemicals). 10 OD 600 units of cells were harvested, lysed using a Precellys24 homogenizer, and lipids were extracted as described previously 36 . To quantitate TAG and SE, the lipids were spotted onto silica gel 60 TLC plates (EMD Millipore) and developed with hexane-diethylether-acetic acid (80:20:1).
Lipids on TLC plates were quantified with a RITA Star Thin Layer Analyzer (Raytest).

DGA1-GFP localization assay
Cells expressing Dga1-GFP from the DGA1 promoter in the centromeric plasmid YCplac111 were grown in SC. These cells were diluted in fresh medium and incubated for 24 hours followed by dilution to 0.3 OD 600 units/ml. These cells were then imaged at the indicated times to determine the localization of Dga1-GFP.