Phosphatidic Acid Is a pH Biosensor That Links Membrane Biogenesis to Metabolism

See allHide authors and affiliations

Science  27 Aug 2010:
Vol. 329, Issue 5995, pp. 1085-1088
DOI: 10.1126/science.1191026


Recognition of lipids by proteins is important for their targeting and activation in many signaling pathways, but the mechanisms that regulate such interactions are largely unknown. Here, we found that binding of proteins to the ubiquitous signaling lipid phosphatidic acid (PA) depended on intracellular pH and the protonation state of its phosphate headgroup. In yeast, a rapid decrease in intracellular pH in response to glucose starvation regulated binding of PA to a transcription factor, Opi1, that coordinately repressed phospholipid metabolic genes. This enabled coupling of membrane biogenesis to nutrient availability.

The hydrophobic portions of lipids can be sensed by hydrophobic protein domains that are often membrane inserted. Soluble protein domains recognize lipids by interacting predominately with their hydrophilic headgroups. Recruitment of proteins to membranes is dependent on the concentration of their target lipid in the bilayer. Membrane-associated transcription factors sense changes in the levels of key signaling lipids, enabling direct feedback regulation of lipid metabolism (13). In yeast, phospholipid metabolism is regulated by the transcriptional repressor Opi1, part of a lipid-sensor complex in the endoplasmic reticulum (ER) (fig. S1) (3). Opi1 is sequestered on the ER by binding both PA and the tail-anchored ER protein Scs2. Addition of inositol results in the rapid depletion of PA, release of Opi1 from the ER, and translocation of Opi1 to the nucleus (3). Nuclear Opi1 represses the Ino2/4 transcriptional activator complex, which binds a cis regulatory element, UASINO, found in many phospholipid metabolic genes (4).

Of the genes regulated by inositol and Opi1, INO1 is the most highly regulated (4). INO1 encodes the rate-limiting enzyme in inositol biosynthesis; thus, inositol auxotrophy is a sensitive measure of expression of the INO1 gene and the status of the ER lipid sensor. We screened the haploid yeast deletion collection for sensitivity to growth in the absence of inositol (5). We identified 231 mutants with notable growth defects (fig. S2 and table S1). Most of these were rescued by deletion of Opi1 (Fig. 1A and table S2). The ∆ino1, ∆ino2, and ∆ino4 mutants, which act downstream of the ER lipid-sensor, were not rescued (Fig. 1B). The ∆scs2 mutant was rescued as expected. Genes that govern intracellular pH (pHi) were enriched in our data set (fig. S2 and table S3), including all 14 subunits of the vacuolar adenosine triphosphatase (V-ATPase) complex and the four factors in the ER responsible for its assembly (6) (Fig. 1C). The V-ATPase governs pHi in part through regulation of Pma1 (7), a P-type H+ ATPase of the plasma membrane (PM) that is the master regulator of pHi (8). A hypomorphic allele of PMA1 (pma1-007) that results in a 50% reduction in expression and activity of the Pma1 protein (9) was also an auxotroph (Fig. 1D). TRK1, a K+ transporter of the PM that activates Pma1 (7), was also identified.

Fig. 1

Genome-wide screen for regulators of phospholipid metabolism. (A) Inositol auxotrophy of ~4800 deletion mutants and effect of deletion of OPI1. Plotted are log2 values of ratios of colony sizes for growth of mutants in the absence or presence of inositol (5). Single mutants are plotted on the x axis and double mutants with ∆opi1 on the y axis. (B) Inositol auxotrophy of known regulators of phospholipid metabolism and rescue by ∆opi1. Plotted are ratios of colony sizes for growth of mutants in the absence (−Ino) or presence (+Ino) of inositol. (C) Inositol auxotrophy of V-ATPase deletion mutants. Mutants are grouped by V-ATPase domain (V1, peripherally associated subunits; V0, membrane-associated subunits) or factors required for V-ATPase assembly. Genes in parentheses indicate deletion of an overlapping dubious open reading frame and may not be true nulls. (D) Inositol auxotrophy of pma1-007 and ∆trk1 mutants. Error bars indicate SD.

Because the pma1-007 and ∆trk1 mutants have an impaired capacity to pump protons out of the cell, they should be sensitive to acidification of the cytosol. To test this hypothesis, we subjected the strains to acid stress by growing them on medium buffered at low pH. Unlike in wild-type cells, pHi in the pma1-007 and ∆trk1 mutants decreased with acidification of the medium (Fig. 2A). To determine whether cytosolic acidification causes derepression of Opi1, we measured transcription of Opi1-dependent genes by reporter assay (3) at pH 3, 4, and 5. We found almost complete repression in the pma1-007 mutant at lowered pH, which was alleviated by deletion of OPI1 (Fig. 2B and fig. S3). Wild-type cells showed a modest decrease in UASINO expression that was also dependent on OPI1. As expected, derepression of Opi1 resulted in inositol auxotrophy of pma1-007 and ∆trk1 cells at low pH (Fig. 2C and fig. S4). This was in contrast to the ∆scs2 mutant, which remained an inositol auxotroph at all pH values. Deletion of OPI1 in the pma1-007 mutant rescued its inositol auxotrophy at each pH. Opi labeled with green fluorescent protein (GFP-Opi1) accumulated in the nucleus of pma1-007 cells, particularly at lowered pH (Fig. 2D), consistent with the decrease in UASINO expression. The drug ebselen, found to inhibit Pma1 in vitro (10), caused an immediate drop in pHi of both wild-type and pma1-007 cells to ~6.3 (Fig. 2E). Whereas wild-type cells stabilized at pHi ~ 6.4, pHi of pma1-007 cells continued to decrease, indicating that the mutant is more sensitive, likely as a result of reduced gene dosage (10, 11). Within 5 min, ebselen caused GFP-Opi1 to translocate from the ER to the cytosol and nucleus (Fig. 2F). Thus, pHi regulates the localization and function of Opi1 and is a signal regulating transcription of phospholipid metabolic genes.

Fig. 2

pH regulates phospholipid metabolism. (A) pHi of mutants grown in medium at pH 3, 4, and 5 compared to the wild type (WT) (*, versus WT at a given pH, P < 0.001). (B) UASINO reporter expression measured in different mutants grown at pH 3, 4, and 5 (*, versus pH 5, P < 0.001). (C) Growth of mutants in the absence of inositol at varying pH at 37°C. (D) Nuclear localization of GFP-Opi1 in cells grown at pH 3 and 5 quantified by confocal microscopy (*, versus WT at a given pH, P < 0.005; **, versus pma1-007 at pH 5, P < 0.01). (E) Effect on pHi after addition of 100 μM ebselen to WT and pma1-007 cells grown in medium at pH 5 (*, versus WT at a given time point, P < 0.05). (F) Effect on the localization of GFP-Opi1 5 min after addition of 100 μM ebselen (+ebs). Arrows indicate ER localizations (straight, cortical; jagged, nuclear envelope); arrowheads indicate cytoplasmic (straight) and nuclear (jagged) localizations. Error bars indicate SEM in (A), (D), and (E) and SD (B). Scale bars, 2 μm.

We examined whether pH affected the binding of Opi1 to PA. A basic domain in Opi1, Q2, directly binds the predominant pool of yeast PA, located in the PM (3). Ebselen caused GFP-Q2 to delocalize from the PM (Fig. 3A). This was also true for Spo20 (Fig. 3B), the other verified PA-binding protein in yeast (12). PM delocalization with ebselen was not due to endocytosis (fig. S5A). To control pHi precisely, we treated yeast with the proton ionophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP) over a range of pH values (13) and quantified GFP-Q2 localization to the PM (5). GFP-Q2 delocalized as pH decreased from 7.2 to 6.4 (Fig. 3, C and D). In contrast, localization of a probe for phosphatidylserine in the PM, GFP-Lact-C2 (14), did not change (fig. S5B). Thus, the binding of proteins to PA in vivo is sensitive to pHi.

Fig. 3

pH governs the binding of Opi1 to PA through its protonation state. (A) Localization of GFP-Q2 after 5 min of ebselen treatment. (B) Localization of the PA-binding domain of Spo20 (GFP-Spo2061-91) after 5 min of ebselen treatment. (C) Treatment of yeast expressing GFP-Q2 with CCCP buffered at the indicated pH. (D) Quantification of PM localization of GFP-Q2 with CCCP treatment (*, versus pH 6.4; **, versus pH 6.8; ***, versus pH 7; P < 0.005). (E) GFP-Q2 localization in ∆vma2 cells. (F) pHi measured in WT and ∆vma2 cells grown in pH 5 medium (*P < 0.0001). (G) Total PA measured by mass spectrometry in WT and ∆vma2 cells grown in pH 5 medium (*P < 0.0001). (H) Binding of Q2 and Q2C3M to liposomes containing 10 mol % PA, 40 mol % phosphatidylethanolamine (PE) over a range of pH values (*, versus pH 6.4; **, versus pH 6.8; ***, versus pH 7.2; P < 0.05). (I) Binding of Q2 to liposomes (0, 100, 200 μM total lipid) containing 50 mol % PA or methyl-PA at pH 7.2. (J) Binding of Q2 and Q2C3M to liposomes containing 20 mol % methyl-PA, 40 mol % PE over a range of pH values. Error bars indicate SD except in (D) (SEM). Scale bars, 2 μm.

GFP-Q2 was delocalized in a V-ATPase mutant, ∆vma2, that had an acidified cytosol (Fig. 3, E and F). GFP-Lact-C2 localization was unaffected in this mutant (fig. S5C). Consistent with an inability of Opi1 to bind PA in ∆vma2 cells, GFP-Opi1 was translocated to the nucleus in this mutant, which also had decreased UASINO expression and was an inositol auxotroph (fig. S6). These phenotypes were not due to decreased PA levels, which were instead elevated ~70% in the mutant (Fig. 3G) (5). Thus, Opi1 failed to bind PA at lowered pHi in this mutant.

Next, we bound recombinant Q2 to liposomes containing PA at varying pH (5). We found a near-linear increase in binding between pH 6.4 and 7.6 (Fig. 3H and fig. S7D). Three basic amino acids in Q2, K136K137R138, are thought to participate in electrostatic interactions with the negatively charged headgroup of PA (3). Binding of the triple alanine substitution mutant (Q2C3M) (3) to PA between pH 6.4 and 7.6 no longer depended on pH (Fig. 3H and fig. S7D). Thus, the direct interaction between Q2 and PA is sensitive to pH.

Unlike other phospholipids, the phosphate headgroup of PA is a monoester and has a second pKa measured in model membranes to be between 6.6 and 7.9, depending on their phospholipid and protein composition (15, 16). Changes in the ionization state of PA might thus be responsible for the observed pH-dependent binding of Opi1. We tested binding of Q2 to methyl-PA, which bears a methyl-group substitution on the phosphate and lacks the second pKa (fig. S7A). Binding remained specific (fig. S7C), but was considerably weaker (Fig. 3I) and was pH independent (Fig. 3J and fig. S7D). Thus, Opi1 had higher affinity for deprotonated PA, consistent with the proposed electrostatic/hydrogen bond switch mechanism for the interaction of proteins with PA (17).

Because glucose-starved yeast exhibit a rapid drop in pHi to ~ 6 (7) and our screen identified several major regulators of glucose signaling (fig. S8) (5), we hypothesized that glucose might be a physiological pH signal. Glucose starvation resulted in translocation of GFP-Opi1 from the ER to the nucleus (Fig. 4, A and B), which correlated with the drop in pHi (Fig. 4C). INO1 transcription was also repressed on a similar time scale (Fig. 4D). Translocation of GFP-Opi1 was not dependent on known modulators of PA levels (fig. S9) (5), suggesting that glucose acted independently of changes in the concentration of PA. Consistently, PA levels measured in ER microsomes isolated after 20 min of glucose starvation did not change significantly (fig. S10).

Fig. 4

Nutrient sensing and phospholipid metabolism are co-regulated by pH. (A) Localization of GFP-Opi1 in WT cells during glucose starvation. Time after removal from glucose (−Dex) is shown. Arrows and arrowheads as in Fig. 2. (B) Quantification of nuclear GFP-Opi1 after glucose starvation in WT and ∆reg1 cells (*, versus WT at a given time point, P < 0.0001). (C) Change in pHi measured in WT and ∆reg1 cells during glucose starvation (*, versus t = 0 for ∆reg1, P < 0.001). (D) INO1 mRNA levels during glucose starvation measured by Northern blot (+ Ino, cells grown in medium with inositol). (E) Growth of mutants at pH 4 in the presence or absence of inositol at 37°C. (F) Pma1 specific activity measured in WT and ∆reg1 cells before and 20 min after glucose starvation (*, versus WT +Dex, P < 0.05; **, versus WT +Dex, P < 0.005). Scale bar, 2 μm.

Reg1 is the glucose-signaling–specific regulatory subunit for Glc7, yeast’s protein phosphatase type 1. Glc7 is implicated in repression of Pma1 (18), and ∆reg1 yeast fail to repress INO1 and overproduce inositol (19), suggesting that Reg1/Glc7 regulates pHi through Pma1. Deletion of REG1 rescued the inositol auxotrophy of the pma1-007 mutant (Fig. 4E). Pma1 activity in the ∆reg1 mutant was higher and failed to repress upon glucose starvation (Fig. 4F). Deletion of REG1 attenuated the rapid drop in pHi in glucose-starved cells (Fig. 4C). Thus, Reg1 repressed Pma1 in response to glucose starvation. GFP-Opi1 translocation was delayed in ∆reg1 cells (Fig. 4B and fig. S9). In both wild-type and ∆reg1 cells, translocation coincided with a drop in pHi below 6.9, consistent with the reduced affinity of Opi1 for protonated PA.

Phosphatidic acid signaling can be dynamically regulated by changes in pH (fig. S11). This involves a change in the protonation state of the phosphate headgroup, making the lipid a pH biosensor. A pH-sensing role for lipids may not be limited to PA because phosphatidylinositol phosphates and ceramide-1-phosphate have pKa values within the physiological range (20, 21). Given the established roles for these lipids in signaling and the universality of pH regulation in biology, pH-dependent protein-lipid interactions may be important in a wide variety of systems.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S11

Tables S1 to S3


References and Notes

  1. See supporting material on Science Online.
  2. We are grateful to T. Levine for insight and critical reading of the manuscript; C. Boone for access to SGA technologies; M. Davies and L. Conibear for a preliminary inositol screen; and J. Church for discussions. This research was supported by grants from the National Sciences and Engineering Council of Canada (NSERC), the Canadian Institute of Health Research (CIHR), the Michael Smith Foundation for Health Research (MSFHR), the Canada Foundation for Innovation, and the British Columbia Knowledge Development Fund (C.J.R.L. and E.J.); the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases (W.A.P.); and the Singapore National Research Foundation under CRP Award 2007-04, the Biomedical Research Council of Singapore (R-183-000-211-305), the National Medical Research Council (R-183-000-224-213), and the RTD project LipidX (M.R.W.). J.J.H.S. is the recipient of a NSERC Alexander Graham Bell Canada Graduate Scholarship. C.J.R.L. is a CIHR New Investigator, a MSFHR Scholar, and a Tula Foundation Investigator.
View Abstract

Navigate This Article