PA accumulates in puncta in seipin-KO cells
A previous report demonstrated a 20 % increase in phosphatidic acid (PA) in enriched ER membranes in seipin-KO (knockout) yeast cells [13]. To confirm this biochemical finding and determine whether the PA was associated with an ER subdomain, we tagged Opi1p, a phosphatidic acid sensor [19], in the genome with GFP in cells containing or lacking SEI1 (yeast seipin, formerly FLD1). Under inositol-replete nutrient conditions, Opi1p is nuclear and represses the synthesis of several genes in phospholipid biosynthesis. However, during inositol starvation, Opi1p traffics to the ER where it binds to PA, which accumulates there [19]. Wild type cells displayed weak Opi1p-GFP nuclear fluorescence when cultured in defined SCD medium (Fig. 1a). However, bright Opi1p-GFP puncta were observed in 85 % of sei1Δ cells under the same conditions, suggesting a local accumulation of PA. An identical phenotype was seen with Opi1p-mCherry, indicating the pattern was not related to a specific fluorescent tag (Fig. 1b). The bright Opi1p puncta in sei1Δ compared with the dim nuclear staining probably represents a combination of high local concentration and increase in Opi1p protein – indeed, cellular Opi1p-mCherry is increased about 80 % of the sei1Δ strain, for unknown reasons (Fig. 1c).
While membrane PA is a determinant of Opi1p binding to ER membranes, other factors come into play: Opi1p also binds to Scs2p in the ER through its FFAT sequence [20], and the Opi1p-Scs2p complex can bind to the Yet1p-Yet3p complex [21]. Furthermore, Opi1p favors PA containing esterified C16 vs. C18 fatty acids [22]. To rule out Scs2p binding, in the Opi1p puncta, we tested the effect of a Opi1p FFAT loss-of-function mutation (EEFD200-203ALLA [19]) on the punctate pattern in sei1Δ and still observed puncta in most cells (Fig. 1d). We found no difference in Scs2p-GFP localization in wild type and sei1Δ strains except a slight increase in punctate Scs2p-GFP in 10 % of stationary-phase cells cells (Additional file 1: Figure S1); this cannot account for the prominent Opi1p puncta. To confirm that the Opi1p puncta represent concentrated PA, we used another PA sensor, a fragment of Spo20p (aa 51–91) conjugated to GFP [23]. In wild type cells the Spo20p fragment bound to membranes, especially at the cell periphery. However, in the sei1Δ strain much of Spo20p(51–91)-GFP was also localized to puncta (Fig. 1e), confirming that PA is highly localized in the absence of seipin.
Sei1p localizes to ER at the junction of lipid droplets and is important for initiation of droplet formation and controlling droplet morphology, in which droplets are normally uniform in shape and size and are dispersed along the ER. We hypothesize that the concentrated PA in the absence of seipin leads to aberrant morphology observed by three groups: clustered droplets in ER tangles and a large range of sizes [11, 12, 16]. If this is the case the PA puncta may colocalize with the ER-droplet aggregates. This was clearly the case, as the puncta colocalized or were adjacent to both droplets, visualized with BODIPY 493/503, and were on the ER, detected with Kar2-CFP-HDEL (Fig. 1f) or with Sec63-CFP (not shown). These data suggested that the PA puncta and aberrant droplets in the absence of seipin were related phenomena.
We also detected puncta with Pah1p, the phosphatidic acid phosphatase that is the ortholog to mammalian lipin [24]. Most of Pah1p is cytosolic. However, more of it can be driven to ER membranes by overexpression of the DAG kinase, Dgk1p, presumably by an increase in phosphatidic acid, the product of Dgk1p [25]. Pah1p puncta also has been demonstrated very recently by new lipid droplets [26]. While membrane localization of Pah1p-GFP in living sei1Δ cells was not apparent in our hands, strong puncta occur in cells fixed with 2 % (v/v) formaldehyde (Fig. 2a); the strong puncta was not a result of higher expression of Pah1p in sei1Δ, as it was expressed (with FLAG tag) similarly in both strains (Fig. 2b). The fixative may have trapped the enzyme on the membrane, perhaps by not allowing it to metabolize the PA to which it is bound. To determine whether these spots represent PA puncta, we used a catalytically dead mutant, Pah1-D398A/D400A-GFP [25]. This protein clearly localizes to puncta in living cells, and the spots colocalize to Opi1p-mCherry (Fig. 2c). Thus, Pah1-D398A/D400A-GFP appears to be another PA sensor, albeit not as sensitive as Opi1p. A fluorescence bleaching-recovery experiment indicated that there was only slow (plateau after ~ 50 s) and incomplete replacement of bleached Pah1-D398A/D400A-GFP, suggesting that the PA pool in the puncta is relatively stable (Fig. 2d and e).
The PA puncta appear during lipid droplet formation
We then asked whether the appearance of PA puncta, which were generally adjacent to droplets on the ER in sei1Δ cells, required droplets. This was tested in strains that are deleted in all but one enzyme that synthesizes neutral lipids for droplets, and in which the expression of the remaining enzyme is under control of the galactose promoter [14, 27]. Two such strains were utilized in which TAG or SE droplets form subsequent to induction of DGA1 or ARE1, respectively, in galactose-containing medium. The endogenous copy of OPI1 was tagged in these strains with mCherry to allow detection of PA puncta. No PA puncta were observed in cells containing SEI1 when DGA1 was induced, as expected (Fig. 3a). Puncta were also absent in cells also deleted in SEI1 when growing in glucose. However, upon induction of DGA1, PA puncta were observed as droplets appeared (at 3 h). At this time point there were many droplets without clear puncta, although each punctum overlapped or was adjacent to droplets, suggesting that droplet formation in the absence of seipin may precede and even cause development of puncta.
We considered that the PA puncta in sei1Δ cells were related to TAG synthesis by Dga1p, as PA is a precursor of TAG. This was not the case. If droplet formation was driven instead by the steryl acetyltransferase Are1p, resulting in steryl ester formation and very little TAG, we saw the identical behavior: bright PA puncta upon formation of new droplets (Fig. 3b). PA puncta, therefore, form in seipin-null cells as droplets form regardless of the neutral lipid or synthetic enzyme producing them.
No single enzyme in PA metabolism is responsible for the PA puncta
We attempted to identify an enzyme related to PA that was responsible for either the synthesis of the PA in the puncta or their absence in the presence of Sei1p (Fig. 4a). PA can be generated from glycerol-6-phosphate (G-3-P) by acyltransferases in two steps: first to lysophosphatidic acid (LPA) by Gpt2p, then to PA by Loa1p, Ale1p, or Slc1p. PA can also be generated from phosphatidyl choline by the phospholipase D, Spo14p, or from diacylglycerol (DAG) by the DAG kinase Dgk1p. We deleted each of these enzymes in strains containing mCherry-tagged Opi1p with or without SEI1. As expected, there were no PA puncta in the strains containing SEI1 (Fig. 4b). Conversely, all strains in the sei1Δ contained puncta, ruling out the possibility that one PA synthetic enzyme was uniquely responsible for them.
We also attempted to eliminate the puncta by overexpressing enzymes that metabolized PA. However, high expression of Pah1p (which converts PA to DAG) or Cds1p (which converts PA to CDP-DAG) failed to attenuate the Opi1-mCherry puncta in the sei1Δ strain (Fig. 4c), suggesting that the puncta are not a result of suppression of these enzymes in the absence of seipin, and also that these enzymes do not have good access to the puncta. To attempt to target these two enzymes to the PA puncta, we generated fusion proteins between Cds1p or Pah1p and Sei1ΔNterm (Sei1p missing the first 14 amino acids, which partially targets to ER-droplet junctions [14] but does not suppress PA puncta; see below). The Cds1p-Sei1ΔNterm fusion had no effect on the PA puncta (data not shown), while the cells expressing Pah1p- Sei1ΔNterm were largely dead and were not further analyzed.
Formation of supersized droplets that accumulate in sei1Δ cells can be suppressed by addition of 75 μM inositol to the medium, which results in an increase in phospholipid synthesis [13]. We found that suppression of supersized droplets by inositol addition did not inhibit the accumulation of PA puncta in sei1Δ cells (Additional file 1: Figure S2).
In summary, none of our genetic manipulations of single genes to lower PA levels eliminated the puncta, suggesting that PA from multiple sources was trapped there, and that it was not easily accessible as substrate to enzymes for further metabolism.
PA puncta disappear in the presence of human seipin or a combination of Sei1pΔNterm and Ldb16p
Recently Ldb16p was identified as a binding partner for Sei1p [15]. Knocking out LDB16 produced a droplet phenotype similar to sei1Δ. The double knockout could be rescued by human seipin, suggesting that Sei1p-Ldb16 formed a functional seipin complex [15]. Consistent with this report we also observed similar droplet phenotypes of sei1Δ and ldb16Δ, namely a combination of droplet clusters (diffuse BODIPY pattern) and supersized droplets (Fig. 5a); we detected no droplet morphological difference between these two strains. Interestingly, in a strain in which both genes were deleted (“ΔΔ”), we observed significantly more cells with supersized droplets (Fig. 5b), suggesting that the frequent droplet clusters observed in the single knockouts may be a gain-of-function phenotype of expression of one of the partners alone. Expression of both proteins in the strain under the strong PGK1 promoter resulted in a different phenotype: a proliferation of very small droplets (too many to count) dispersed throughout the cytoplasm (Fig. 5a). Yeast Sei1p has been shown to be important for initiation of droplet assembly [14]. This experiment suggests there are no other limiting factors for droplet production other than Sei1p and Ldb16p.
We recently showed that removal of the amino terminal 14 amino acids from seipin (generating Sei1ΔNterm) resulted in a partial phenotype compared to the null strain: droplet initiation was slower, resulting in larger droplets, but the droplets appeared otherwise normal [14]. Similarly, overexpression of Sei1ΔNterm with Ldb16p in the double deletion strain resulted in a homogeneous population of droplets larger than found in the full-length Sei1p control (Fig. 5a).
PA puncta were observed in the ΔΔ strain (Fig. 5c). Replacement of either protein alone did not result in their disappearance. However, no puncta were observed upon expression of human seipin (BSCL2). As human seipin can complement yeast seipin null strains (sei1Δ and sei1Δ ldb16Δ) in promoting droplet formation [15], our data suggest that preventing PA puncta is a conserved role of seipin.
Sei1ΔNterm, even when overexpressed, was unable to cause the disappearance of PA puncta in the sei1 null strain (Fig. 5d), and caused the appearance of PA puncta even when expressed in the wild type strain (not shown); this dominant negative effect was seen before in regards to droplet morphology [14]. Surprisingly, PA puncta were suppressed when Sei1ΔNterm was co-overexpressed with Ldb16, either as independent proteins or as a tethered unit, Ldb16- Sei1ΔNterm (Fig. 5e). Therefore, while the two proteins together produce larger droplets (consistent with the role of the Sei1p amino terminal domain in droplet initiation), this phenotype can be dissected away from the appearance or PA puncta.
In our system, deleting SEI1, LDB16, or both resulted in a strong decrease in cellular TAG with no effect of SE levels (Fig. 5f and data not shown). Replacement with overexpressed Sei1p or Sei1pΔNerm in the double deletion mutant did not increase TAG. However, overexpression of Ldb16p alone in the double deletion strain restored TAG to normal levels. This result indicates that Ldb16 can function independently of Sei1p to affect droplet size and TAG accumulation.