A phosphorylation-regulated amphipathic helix controls the membrane translocation and function of the yeast phosphatidate phosphatase

Results and Discussion

To investigate the regulation of lipin membrane translocation, we applied a genetic approach in yeast. Specifically, we asked whether increasing the levels of PA, the substrate of Pah1p, could enhance its recruitment onto membranes. To do this, we used a GAL-DGK1 construct to overexpress the transmembrane DAG kinase Dgk1p that localizes to the nuclear/ER membrane. Dgk1p uses CTP as a phosphate donor to phosphorylate DAG and generate PA (11, 12) (Fig. 1A), and therefore its overexpression causes phenotypes similar to those described in pah1Δ (4, 11, 12). Further genetic evidence that Dgk1p activity antagonizes that of Pah1p comes from the fact that DGK1 overexpression rescues the toxicity caused by the overexpression of Pah1p activity (11).

In media containing raffinose, where GAL-DGK1 is not induced, Pah1p-GFP expressed at endogenous levels shows a diffused soluble localization and cells display normal, round nuclei as visualized by the inner nuclear membrane reporter HEH2-CHERRY (Fig. 1B). As shown previously, in galactose-containing medium the nuclear membrane expanded in GAL-DGK1 cells (11) and similar to the nuclei of spo7Δ cells (13), expansion took place at the membrane associated with the nucleolus (Fig. 1C). Interestingly, Pah1p-GFP colocalized with the expanding nuclear membrane, although it was excluded from the rest of the nuclear envelope and ER (Fig. 1B). Time course analysis showed that in most cells recruitment was initiated and maintained onto the nucleolar-associated membrane (Fig. 1C). In contrast, Pah1p-GFP was not recruited onto membranes of cells overexpressing the catalytically dead DGK1[D177A] mutant (11, 12). Moreover, co-over-expression of GAL-PAH1 in cells expressing GAL-DGK1 resulted in complete loss of Pah1p-GFP membrane labeling and concomitantly restored normal nuclear shape (Fig. 1B). The fact that the recruitment of Pah1p-GFP depends on (i) a catalytically active Dgk1p and (ii) the levels of PAP activity suggests that PA is required for this translocation step onto membranes. To test whether structural changes to the nuclear/ER membrane, rather than the elevated PA levels, could be responsible for Pah1p-GFP recruitment, we overexpressed the nuclear envelope protein Esc1p, which causes similar structural defects to the nucleus as Dgk1p (14), or Hmg1p, which causes accumulation of stacked ER membranes to the outer nuclear envelope (15) and found that in both cases there was no Pah1p-GFP recruitment (Fig. S1). Taken together, these data indicate that Pah1p-GFP can be recruited in a PA-dependent manner onto a nuclear/ER membrane domain.

We first asked whether the membrane translocation of Pah1p-GFP could be driven only by the interaction of its catalytic site with its substrate PA. Pah1p belongs to a superfamily of Mg²⁺-dependent phosphatases that use a DXDX(T/V) catalytic motif to dephosphorylate their substrates. A Pah1p-[D398A D400A]-GFP mutant at this motif, which is catalytically dead (16), still associates with membranes, both in GAL-DGK1 or in pah1Δ cells (Fig. 1D), indicating that recruitment is independent of the catalytic activity of Pah1p and suggesting the presence of other Pah1p domains that are responsible for membrane binding.

We next examined the role of the Nem1p-Spo7p, a protein phosphatase complex that localizes to the nuclear/ER membrane and dephosphorylates Pah1p (4, 17). Nem1p is the catalytic subunit and Spo7p is the regulatory subunit that binds stably to the catalytic domain of Nem1p and is required for the activity of the holoenzyme in vivo and in vitro (4, 18). Because Nem1p and Spo7p are integral membrane proteins, we asked whether binding of Pah1p to this complex is required for its membrane recruitment. Indeed, although Spo7p-Cherry showed an ER localization in raffinose media, it colocalized with membrane-bound Pah1p-GFP in GAL-DGK1 cells (Fig. 2A). Importantly, PA-dependent translocation of Pah1p-GFP was blocked in nem1Δ spo7Δ cells (Fig. 2B), indicating that the phosphatase complex is required for this step. However, membrane translocation was restored in nem1Δ spo7Δ cells expressing PAH1-7A (Fig. 2B), a phosphorylation deficient mutant where seven phosphoresidues have been replaced by alanines (17). Therefore, once Pah1p is dephosphorylated, its membrane binding no longer requires the Nem1p-Spo7p complex. Thus, there must be other factor(s) responsible for maintaining Pah1p at its membrane-bound state.

Bioinformatic analysis suggests that Pah1p contains a predicted amphipathic helix in its extreme amino-terminal end (residues 1 to 18, Fig. 3A) (19). Because amphipathic helices are known to mediate reversible binding of proteins with membranes, we assayed recruitment of two Pah1p mutants, one lacking the predicted helix (Pah1p-Δ[3-16]-GFP) and a second, where three bulky residues within the hydrophobic phase of the helix (L8, V11, W15) were changed to alanines (Pah1p-Amphipathic Helix 3A, Pah1p-AH3A, Fig. 3A). Although in both cases the helix mutations did not affect the stability of Pah1p (Fig. 3C), recruitment was completely abolished in GAL-DGK1 cells (Fig. 3B). Therefore the hydrophobic phase of the helix is required for membrane recruitment. We next asked whether dephosphorylation of Pah1p would bypass the requirement of the amphipathic helix for membrane association. As seen in Fig. 3B, a Pah1p-AH3A-7A-GFP fusion was not recruited in GAL-DGK1 cells. The fact that dephosphorylated Pah1p needs the helix but not Nem1p-Spo7p for recruitment suggests that one role of the dephosphorylation is to allow the helix binding to the membrane.

Because the Pah1p-7A is more active as a PAP enzyme, its overexpression causes inositol auxotrophy due to decrease of PA levels and the Opi1p-dependent repression of the transcription of INO1 and other phospholipid biosynthetic genes (17). If the amphipathic helix is required for membrane recruitment, then mutating its hydrophobic residues should relieve the inositol auxotrophy caused by PAH1-7A overexpression. Indeed, in contrast to GAL-PAH1-7A, cells expressing GAL-PAH1-AH3A-7A grow on media lacking inositol (Fig. 3D). Consistent with this, overexpression of NEM1-SPO7 that is lethal in wild-type cells (4) is rescued in cells expressing PAH1-AH3A (Fig. 3E). Thus, in both cases, mutating the helix blocks the activity of dephosphorylated Pah1p in vivo.

We next examined the role of the helix in the membrane binding and catalytic activity of Pah1p in vitro. Because PA is not soluble, kinetic analysis was performed using Triton X-100/PA mixed micelles (20). Two assays, which monitor distinct types of interactions of Pah1p with PA, were performed: In a first set of experiments, the enzyme activities were measured as a function of the molar concentration of PA (e.g., number of micelles containing PA) by maintaining the ratio of Triton X-100 to PA in the micelle at 10∶1 (Fig. 4B). In a second set of assays, activities were measured as a function of the surface concentration of PA (mole percent, e.g., number of PA molecules on a micelle surface) while keeping the molar concentration of PA constant at 0.2 mM (Fig. 4C). K_m values for the first assay reflect the interactions of the enzyme with the micelle surface, whereas K_m values for the second assay reflect its interactions with PA during catalysis within the surface. Although in both assays the Pah1p-AH3A mutant exhibited a decrease in the V_max, its K_m value showed an increase only for the first (molar PA) but not the second (surface PA) assay. These data indicate that the affinity of Pah1p-AH3A for the initial binding step to the Triton X-100/PA surface is lower. Once Pah1p binds to the surface, however, its interaction with PA is no longer affected by the amphipathic helix.

To examine further the role of the Pah1p helix in membrane recruitment, we used a fluorescence-based assay that measures Pah1p-liposome interactions. The interaction of Pah1p with liposomes can be monitored by an increase of tryptophan fluorescence intensity with a shift from 350 to 343 nm in the wavelength of the maximum emission. This increase indicates a change from a hydrophilic to a more hydrophobic environment around the tryptophan residues (21). The native Pah1p-AH3A mutant purified from yeast showed little difference in liposome binding when compared to the wild-type Pah1p (Fig. 4D, open symbols). In vitro dephosphorylation of the wild-type Pah1p resulted in a threefold decrease in the K_d reflecting an increase in liposome binding (Fig. 4D). The dephosphorylation of the Pah1p-AH3A mutant, however, did not result in an increase in binding (Fig. 4D). Thus, as seen with the Pah1p-GFP fusions in vivo, dephosphorylation of Pah1p promotes membrane binding only in the presence of the amphipathic helix.

We then tested the functional significance of the Pah1p helix in lipid production. Lack of PAH1 affects the cellular levels of the major phospholipid classes (1): pah1Δ cells exhibit increased levels of phosphatidylethanolamine (PE), phosphatidylinositol (PI), and PA and a decrease in the levels of phosphatidylcholine (PC) (Fig. 5A). Expression of PAH1-AH3A in the pah1Δ mutant also led to an increase in PE, PI, and PA and decrease in PC (Fig. 5A). In addition to these changes in phospholipids, neutral lipid analysis of the PAH1-AH3A mutant showed that both DAG, generated by Pah1p, and TAG generated by acylation of DAG, decreased to levels comparable to those seen in the pah1Δ mutant (Fig. 5B). Therefore the helix is required for the lipid biosynthetic functions of Pah1p. Consistent with these effects on lipid metabolism, PAH1-AH3A cells displayed nuclear/ER membrane expansion also seen in pah1Δ cells (Fig. 5C) (4).

In summary we find that (i) the levels of PA control the localization of the lipin Pah1p, (ii) the Nem1p-Spo7p phosphatase complex and the amphipathic helix cooperate to tether Pah1p onto the nuclear/ER membrane, and (iii) the helix is essential for the biosynthetic functions of Pah1p. Although PA is important for Pah1p recruitment, additional factors (such as PA effector proteins or other phospholipids that are metabolically linked to PA) may work together to recruit Pah1p. The Nem1p-Spo7p relocalization in Dgk1p-overexpressing cells and its requirement for Pah1p recruitment suggest it may be part of a machinery that determines localized membrane translocation of Pah1p in response to PA. Because dephosphorylated Pah1p no longer depends on Nem1p-Spo7p but still requires the helix for recruitment in vivo, we support the idea that a prime function of dephosphorylation is to enable helix membrane binding. Indeed, dephosphorylation of Pah1p enhances its liposome binding in vitro, but only in the presence of a functional amphipathic helix. How dephosphorylation could control helix function is unclear, but structural changes within Pah1p or interaction with another factor could both expose the helix and stabilize recruitment. It should be noted that because PA and DAG play essential functions onto the membranes of other organelles such as mitochondria (22), the Golgi complex (23), or lipid droplets (24), it is possible that additional and organelle-specific interactions of Pah1p may operate on these membranes. Nevertheless, the effects of Pah1p-AH3A on lipid levels suggest that the helix mediates the recruitment of Pah1p to the nuclear/ER membrane where the biosynthesis of the bulk cellular pools of phospholipids and TAG takes place.

Sequence analysis suggests that the Pah1p helix is an ArfGAP1 lipid packing sensor, an amphipathic helical motif that adsorbs preferentially on curved membranes and is proposed to act as a membrane curvature sensor (19). The ER is made of flat membrane sheets that include the nuclear membrane, and tubules with high cross-sectional curvature that are stabilized by reticulons (25). This raises the intriguing possibility that Pah1p converts PA to DAG preferentially on ER tubules. However, PA-dependent recruitment of Pah1p-GFP takes place on a nuclear membrane domain (Fig. 1B) that, as expected, lacks reticulons (Fig. S2), suggesting that Pah1p is able to bind membrane sheets. An alternative potential role for curvature sensing may be related to the well-known deforming effects of PA on lipid bilayers: Changes in the curvature of PA- enriched membranes may be detected by the helix that would then mediate the binding of Pah1p.

Lipins act at a key branchpoint of lipid biosynthesis that commits the glycerol backbones to membrane phospholipids and triglyceride. Therefore, membrane recruitment of lipins could be a major regulatory step during membrane biogenesis or energy storage in lipid droplets. Mammalian lipins respond to several stimuli: Fatty acids cause translocation of a soluble PAP activity on microsomal membranes (26), whereas insulin causes phosphorylation and increase of the soluble pool of lipin 1 in 3T3-L1 adipocytes (27). Whether, as in yeast, these changes require a membrane-bound receptor, such as the mammalian Nem1p orthologue Dullard (28), is not known. Mammalian lipins 1 and 2 can functionally replace Pah1p in yeast (29) and contain predicted amphipathic helices at their amino-terminal ends although they lack the distinctive serine residues of the polar phase of the Pah1p helix. Future studies should address whether similar mechanisms operate during the membrane translocation of lipins in other systems.

Materials and Methods

Supplementary Material