Svp1p defines a family of phosphatidylinositol 3,5-bisphosphate effectors (original) (raw)

EMBO J. 2004 May 5; 23(9): 1922–1933.

Stephen K Dove,1,a Robert C Piper,2 Robert K McEwen,1 Jong W Yu,3 Megan C King,3 David C Hughes,4 Jan Thuring,5 Andrew B Holmes,5 Frank T Cooke,6 Robert H Michell,1 Peter J Parker,7 and Mark A Lemmon3

Stephen K Dove

1School of Biosciences, University of Birmingham, Birmingham, UK

Robert C Piper

2Department of Physiology and Biophysics, University of Iowa, Iowa City, IA, USA

Robert K McEwen

1School of Biosciences, University of Birmingham, Birmingham, UK

Jong W Yu

3Department of Biochemistry and Biophysics, University of Pennsylvania School of Medicine, Philadelphia, PA, USA

Megan C King

3Department of Biochemistry and Biophysics, University of Pennsylvania School of Medicine, Philadelphia, PA, USA

David C Hughes

4School of Environmental and Applied Sciences, University of Derby, Derby, UK

Jan Thuring

5Department of Chemistry, University of Cambridge, Cambridge, UK

Andrew B Holmes

5Department of Chemistry, University of Cambridge, Cambridge, UK

Frank T Cooke

6Department of Biochemistry & Molecular Biology, University College London, London, UK

Robert H Michell

1School of Biosciences, University of Birmingham, Birmingham, UK

Peter J Parker

7Protein Phosphorylation Laboratory, Cancer Research UK London Research Institute, London, UK

Mark A Lemmon

3Department of Biochemistry and Biophysics, University of Pennsylvania School of Medicine, Philadelphia, PA, USA

1School of Biosciences, University of Birmingham, Birmingham, UK

2Department of Physiology and Biophysics, University of Iowa, Iowa City, IA, USA

3Department of Biochemistry and Biophysics, University of Pennsylvania School of Medicine, Philadelphia, PA, USA

4School of Environmental and Applied Sciences, University of Derby, Derby, UK

5Department of Chemistry, University of Cambridge, Cambridge, UK

6Department of Biochemistry & Molecular Biology, University College London, London, UK

7Protein Phosphorylation Laboratory, Cancer Research UK London Research Institute, London, UK

aDepartment of Biosciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK. Tel.: +44 121 414 8513; Fax: +44 121 414 7816; E-mail: ku.ca.mahb@evod.k.s

Received 2003 Aug 26; Accepted 2004 Mar 15.

Copyright © 2004, European Molecular Biology Organization

Abstract

Phosphatidylinositol 3,5-bisphosphate (PtdIns(3,5)_P_2), made by Fab1p, is essential for vesicle recycling from vacuole/lysosomal compartments and for protein sorting into multivesicular bodies. To isolate PtdIns(3,5)_P_2 effectors, we identified Saccharomyces cerevisiae mutants that display _fab1_Δ-like vacuole enlargement, one of which lacked the _SVP1/_YFR021w/ATG18 gene. Expressed Svp1p displays PtdIns(3,5)_P_2 binding of exquisite specificity, GFP-Svp1p localises to the vacuole membrane in a Fab1p-dependent manner, and _svp1_Δ cells fail to recycle a marker protein from the vacuole to the Golgi. Cells lacking Svp1p accumulate abnormally large amounts of PtdIns(3,5)_P_2. These observations identify Svp1p as a PtdIns(3,5)_P_2 effector required for PtdIns(3,5)_P_2-dependent membrane recycling from the vacuole. Other Svp1p-related proteins, including human and Drosophila homologues, bind PtdIns(3,5)_P_2 similarly. Svp1p and related proteins almost certainly fold as β-propellers, and the PtdIns(3,5)_P_2-binding site is on the β-propeller. It is likely that many of the Svp1p-related proteins that are ubiquitous throughout the eukaryotes are PtdIns(3,5)_P_2 effectors. Svp1p is not involved in the contributions of FAB1/PtdIns(3,5)_P_2 to MVB sorting or to vacuole acidification and so additional PtdIns(3,5)_P_2 effectors must exist.

Keywords: ATG18, CVT18, AUT10, lysosome, phosphoinositide

Introduction

Phosphorylated derivatives of inositol and phosphatidylinositol fulfil a striking variety of specific functions in eukaryote cells, with their actions executed by effector proteins containing phosphoinositide-specific binding domains. Diverse protein modules can serve as phosphoinositide ‘sensors'. These include many PH, FYVE and PX (≡PhoX) domains (Stenmark and Aasland, 1999; Lemmon and Ferguson, 2000; Gillooly et al, 2001; Ellson et al, 2002).

Phosphatidylinositol 3,5-bisphosphate (PtdIns(3,5)_P_2) is the most recently identified phosphatidylinositol bisphosphate isomer (Dove et al, 1997; Whiteford et al, 1997). All eukaryotes make PtdIns(3,5)P_2 using PtdIns3_P 5-kinases related to Saccharomyces cerevisiae Fab1p (Cooke et al, 1998; Gary et al, 1998; Ikonomov et al, 2001). PtdIns(3,5)_P_2 is essential for membrane recycling from the vacuole/lysosomes (Gary et al, 1998; Ikonomov et al, 2001; Dove et al, 2002), for ubiquitin-dependent packaging of proteins into multivesicular bodies (MVBs) (Odorizzi et al, 1998; Dove et al, 2002), for growth at high temperature (Yamamoto et al, 1995; Cooke et al, 1998; Gary et al, 1998) and for vacuole acidification (Bonangelino et al, 1997). Vac14p and Vac7p are Fab1p regulators (Gary et al, 1998; Bonangelino et al, 2002; Dove et al, 2002), and PtdIns(3,5)_P_2 effector proteins that may facilitate MVB protein sorting have recently been identified (Friant et al, 2003; Whitley et al, 2003). Since deletion of these effectors does not lead to all the defects associated with loss of Fab1p, additional effectors remain to be identified. Other proteins that can bind PtdIns(3,5)_P_2 (Xu et al, 2001; Cozier et al, 2002) seem unlikely to mediate any of the known effects of this lipid.

A single unlobed vacuole that largely fills the cell is a hallmark of _fab1_Δ yeast that cannot make PtdIns(3,5)_P_2 (Yamamoto et al, 1995) so the loss of proteins that are Fab1p activators or are needed for the actions of PtdIns(3,5)_P_2 should cause a similar phenotype. Using a microscopic screen starting from this vacuole phenotype, we sought genes whose disruption phenocopies the _fab1_Δ vacuole enlargement and identified the Fab1p regulator Vac14p/Svp2p (Dove et al, 2002).

Herein we show that Svp1p, another gene identified in this screen, is a specific PtdIns(3,5)_P_2-binding protein that participates in the recycling of membrane proteins from the vacuole to the late endosome. Svp1p is the prototype member of a new family of phosphoinositide effectors.

Results

svp1_Δ_ cells have a fab1_Δ_-like vacuole defect

Each Euroscarf strain lacks one non-essential gene. We screened these for enlarged vacuoles resembling those of _fab1_Δ cells, and provisionally termed the identified genes SVP (Swollen Vacuole Phenotype) (Dove et al, 2002).

The YFR021w open reading frame (ORF) encodes SVP1, which is now termed ATG18 (Klionsky et al, 2003). Figure 1A shows DIC images of wild-type, _fab1_Δ and _svp1_Δ cells, and fluorescence images of the same cells with FM4-64-stained vacuoles. Most _fab1_Δ and _svp1_Δ cells are markedly enlarged, and both mutants usually have one large vacuole that fills much of the cell interior: normal vacuoles are smaller and multilobed. Remarkably, SVP1 is physically separated from FAB1 on chromosome VI by only one ORF (Figure 2A). FAB1 overexpression does not correct the _svp1_Δ vacuole enlargement, so suppressed Fab1p expression cannot be its cause (not shown).

An external file that holds a picture, illustration, etc. Object name is 7600203f1.jpg

The _svp1_Δ phenotype involves vacuole enlargement, and GFP-Svp1p localisation at the vacuole membrane is _FAB1_-dependent. (A) Differential interference contrast images and FM4-64 staining of wild-type, _svp1_Δ and _fab1_Δ cells, demonstrating the greatly enlarged vacuoles of _svp1_Δ cells. (B) Svp1p expression corrects the _svp1_Δ vacuolar defect (upper images, taken during methionine-repressed low-level Svp1p expression), but Svp1p overexpression induces cell vacuolation (lower images, during de-repressed expression following methionine removal). (C) When GFP-Svp1p was expressed in _svp1_Δ cells, it associated mainly with the vacuole membrane and large punctate structures. Little or none of the GFP-Svp1p was associated with the vacuole membrane in _fab1_Δ cells.

An external file that holds a picture, illustration, etc. Object name is 7600203f2.jpg

Sequence analysis and structural modelling of Svp1p and Svp1p-like proteins. (A) YFR021w/SVP1 is on the right arm of chromosome VI, adjacent to FAB1. (B) ClustalW alignment of some eukaryotic Svp1-like proteins. Light-grey bars identify sequence within the putative β-propeller. The dark-grey bar denotes the B/C insert in blade 4. The black arrow shows the deduced site of trypsin cleavage in Svp1p. The black bar indicates the C-terminal domain outside the β-propeller. (C) Threaded alignment of Svp1p, hSvp1a and Hsv2p with the β-propeller of transducin-β (Sondek et al, 1996). Assignment of blades and β-strands is based on transducin-β. Each sequence was submitted to the 3D-PSSM server (Kelley et al, 2000), and gave a significant (>80% certainty) score for alignment with transducin-β. Alignments were slightly adjusted in blades 3, 4 and 7, to maintain consistency within the Svp1p family ClustalW alignments. (D) A linear depiction of the Svp1p domain structure and a cartoon of its probable folded structure. Light ovals represent WD-40 blades, and the black arrow the point of trypsin attack.

GFP-Svp1p rescues the vacuole defects of svp1_Δ_ cells

GFP-Svp1p expression from the repressed MET25 promoter (with 5 mM methionine) corrected the _svp1_Δ vacuole enlargement. Rescue was variable and occurred at very low Svp1p expression (Figure 1B). SVP1 disruption therefore causes the vacuole defects of _svp1_Δ cells. However, overexpressed GFP-Svp1p (methionine-free medium) disrupted vacuole function (Figure 1B). Both observations indicate that Svp1p has a role in regulating vacuole morphology.

GFP-Svp1p localises to the vacuole membrane in a FAB1-dependent manner

Phosphoinositide-binding proteins frequently show inositol lipid-dependent changes in intracellular localisation. When GFP-Svp1p was expressed at low levels in wild-type cells, it localised to the vacuole membrane and to a punctate compartment (Figures 1C and 4B). In contrast, most _fab1_Δ cells lacked vacuole-associated GFP-Svp1p (Figure 1C), but the GFP-Svp1p on the punctate compartment remained (Figure 1C). Localisation of GFP-Svp1p to the vacuole membrane, but not to the punctate compartment, is therefore Fab1p-dependent, as would be expected for a PtdIns(3,5)_P_2 effector.

Previous work on the SVP1 gene

SVP1 has previously been named ATG18, AUT10 and CVT18, reflecting its involvement in AUTophagy, and in Cytoplasm-to-Vacuole protein Targeting (Barth et al, 2001; Guan et al, 2001).

None of these studies mentioned vacuole enlargement, possibly because they examined _svp1_Δ cells under starvation conditions that tend to provoke vacuole enlargement even in wild-type cells.

Svp1p-like proteins are present in all eukaryotes

The S. cerevisiae genome encodes two other _SVP1_-like proteins, YPL100w/MAI1 and YGR223c (Georgakopoulos et al, 2001; Barth et al, 2002), which we term HSV1 and HSV2 (Homologous with SVP1), respectively. Their disruption did not cause vacuole enlargement (not shown). GFP-Hsv1p/Mai1p and GFP-Hsv2p localised to a non-vacuolar punctate compartment and this localisation did not require FAB1 (not shown).

Svp1p-like proteins are widespread in all eukaryotes (Figure 2B; Barth et al, 2001). For example, the human and Arabidopsis genomes both encode at least three, and Caenorhabditis elegans and Drosophila melanogaster have two or more (Barth et al, 2001). Gene DKFZp434J154 encodes the most Svp1-like human homologue (labelled hSvp1a in Figures 2B and C).

Svp1p and its homologues are multi-WD40 proteins that fold as _β_-propellers

SVP1 encodes a serine-rich 500-amino-acid protein with two previously recognised WD-40 motifs near the centre of the sequence (residues 234–274 and 277–318, within blades 5 and 6 in Figure 2C) (Barth et al, 2001; Guan et al, 2001). The 3D-PSSM fold recognition server (Kelley et al, 2000) predicts that Svp1p, hSvp1a and Hsv2p fold as seven-bladed β-propellers. They show high similarity scores when compared with known seven-bladed β-propellers—the transducin β-subunit and the C-terminal domain of Tup1p (Figure 2C). Trypsin cleaves Svp1p at Arg377 (Figures 2B and D, black arrow), yielding Svp1p1–377 as a stable fragment (MC King and MA Lemmon, unpublished). Combining the results from the threading analysis and trypsin cleavage, we suggest that the Svp1p β-propeller ends around R377, and that the C-terminal 123 residues form an independent domain (see Figure 2D).

Sequence alignment (Figure 2C) shows that the predicted β-strands of the propeller blades are quite well conserved between Svp1p and its homologues, but the intervening loops are highly divergent. For instance, most homologues lack the large B/C loop in blade 4 of Svp1p (Figures 2B–D; Barth et al, 2001). The C-terminal domains are also highly variable.

Svp1p, Hsv1p, Hsv2p and hSvp1a bind PtdIns(3,5)P2 with high affinity and specificity

GFP-Svp1p localises to the vacuole in a _FAB1_-dependent manner, so we determined whether Svp1p might bind PtdIns(3,5)_P_2, particularly under ionic conditions like those in the cytosol (inc. millimolar Mg2+). We first used phosphoinositide ‘dot blots' (Kavran et al, 1998) and saw binding of 32P-labelled GST-Svp1p (Figure 3A) and GST-Hsv2p (not shown) only to PtdIns(3,5)P_2 and PtdIns3_P.

An external file that holds a picture, illustration, etc. Object name is 7600203f3.jpg

Svp1p and related proteins bind PtdIns(3,5)P_2 with high affinity and selectivity. (A) A dot-blot assay indicates that GST-Svp1p binds both PtdIns3_P and PtdIns(3,5)_P_2. (B) GST-Svp1p bound to PtdIns(3,5)_P_2-derivatised Affigel beads is selectively displaced by exogenous PtdIns(3,5)_P_2 and not by other Ptdins_P_2 isomers or by PtdIns_P_s. (C) Monomeric Svp1p binds to a DOPC layer ‘doped' with 3 mol% PtdIns(3,5)_P_2 but not with other phosphoinositides, as detected by Biacore analysis. (D) The affinity of monomeric Svp1p binding to PtdIns(3,5)_P_2-‘doped' lipids is similar to phosphoinositide affinities in other protein/phosphoinositide combinations. (E, F) Mg2+ at a ‘physiological' (0.5 mM) concentration is needed for Svp1p to show its full PtdIns(3,5)_P_2 selectivity. (G) Comparison of PtdIns(3,5)_P_2 binding by GST-Svp1p, GST-Hsv2p (S. cerevisiae) and GST-hSvp1a (human). All data are representative of at least three independent experiments.

Since dot-blot assays do not always report a protein's native lipid selectivity reliably, we used two more quantitative approaches to analyse phosphoinositide binding by GST-Svp1p. First, we found that GST-Svp1p bound strongly to Affigel beads bearing covalently attached PtdIns(3,5)_P_2 (Figure 3B, ‘total binding'). Washing with buffer containing PtdIns(3,5)P_2 almost completely displaced the bound GST-Svp1p, but other PtdIns_P and PtdIns_P_2 isomers had no effect (Figure 3B). Gro_P_Ins(3,5)_P_2, the hydrophilic backbone of PtdIns(3,5)_P_2, did not displace Svp1p (not shown).

Specific binding of PtdIns(3,5)_P_2 to Svp1p therefore involves at least two interactions: (a) between the anionic head group and a basic amino-acid cluster; and (b) between hydrophobic parts of the lipid and a nearby hydrophobic patch on Svp1p.

Surface plasmon resonance (SPR) analysis was undertaken to obtain a quantitative measure of Svp1p binding to mixed lipid vesicles ‘doped' with a phosphoinositide (3%, mol/mol) and immobilised on BIACore L1 chips (Yu and Lemmon, 2001). GST-Svp1p bound strongly to PtdIns(3,5)_P_2-containing membranes in the presence of 0.5–2 mM MgCl2 (apparent _K_D∼180 nM; Figure 3G). Since GST-induced dimerisation will potentiate Svp1p binding to phosphoinositides, we also examined native Svp1p that was monomeric by gel filtration. This bound PtdIns(3,5)P_2 selectively and with high affinity (half-maximal at ∼500 nM Svp1p) (Figures 3C and D). Svp1p bound weakly to PtdIns3_P, and not to PtdIns(4,5)_P_2, PtdIns(3,4)P_2 or PtdIns4_P.

We compared PtdIns(3,5)_P_2 binding by Svp1p with the interactions between other highly selective phosphoinositide sensor domains and their phosphoinositide ligands (Figure 3D). The affinity of the Svp1p/PtdIns(3,5)P_2 interaction was at least 10-fold greater than between the Hrs1 FYVE domain and PtdIns3_P (Sankaran et al, 2001), higher than for the phospholipase C-δ1 PH domain/PtdIns(4,5)_P_2 interaction (Rebecchi and Pentyala, 2000), and weaker than for PtdIns(3,4)_P_2 binding to the DAPP1 PH domain (Ferguson et al, 2000) (Figure 3D).

The PtdIns(3,5)_P_2 selectivity of Svp1p was reduced when the buffer lacked Mg2+. Although the PtdIns(3,5)_P_2 affinity was enhanced (_K_D∼18 nM), Svp1p also bound to PtdIns(3,4)_P_2 (_K_D∼190 nM) and PtdIns(4,5)_P_2 (_K_D∼200 nM) (Figures 3E and F), although less avidly and with one-third as much bound at saturation, possibly indicating an altered stoichometry.

GST-Hsv2p and GST-hSvp1a bound to PtdIns(3,5)_P_2 with high affinity (Figure 3G; 0.5 mM MgCl2 present) and selectivity (not shown): the affinity of GST-Hsv2p was similar to that of GST-Svp1p. GST-Hsv1p/Mai1p also bound specifically to PtdIns(3,5)_P_2 (_K_D∼500 nM), as did a GST-Tagg-340-residue Drosophila Svp1p homologue with a very short C-terminal tail (CG11975 in Figure 2B; _K_D 200–500 nM) (not shown).

It therefore appears that specific PtdIns(3,5)_P_2 binding is a conserved and functionally important feature of many Svp1p-like proteins.

PtdIns(3,5)P2 binds to the _β_-propeller

The PtdIns(3,5)_P_2-binding site of Svp1p does not require the C-terminal sequence that lies outside the β-propeller: its removal did not change PtdIns(3,5)_P_2 binding. However, Svp1p1–170 and Svp1p171–500, each of which includes about half of the proposed β-propeller, did not bind PtdIns(3,5)_P_2. Some Svp1p homologues have quite large insertions within the β-propeller, for example, between strands 4B and 4C (Figures 2B–D). An Svp1p construct lacking this loop (Svp1pΔ4B/C) also bound PtdIns(3,5)_P_2 with wild-type specificity and affinity (Figure 4A).

An external file that holds a picture, illustration, etc. Object name is 7600203f4.jpg

The β-propeller of Svp1p binds PtdIns(3,5)_P_2. (A) PtdIns(3,5)_P_2 binding to GST-Svp1p mutants lacking the blade 4 B/C loop or with a mutated basic patch in β-sheet 2C (SPRRLR to SPSSLS) or β-sheet 5D (FRRG to FTTG). Only the conversion of Arg residues to Thr in the blade 5 basic patch substantially curtailed PtdIns(3,5)_P_2 binding. (B) Localisation of the Svp1p mutants in _svp1_Δ yeast. Wild type and Svp1pSPSSLS localised similarly, but Svp1pFTTG is no longer on vacuole membranes. The constructs were expressed as N-terminal GFP fusions from a single-copy pUG36 plasmid under control of the MET25 promoter, with 0.3 mM methionine. (C) Despite not associating with the vacuole membrane, overexpressed GFP-Svp1pFTTG causes vacuolation of 60–70% of wild-type cells (compared with 60–70% of cells when GFP-Svp1pWT is overexpressed). The constructs are N-terminal GFP fusions in pUG36, and were grown without methionine for maximal expression.

It seems that PtdIns(3,5)_P_2 binds to the β-propeller. A multiple alignment highlighted two clusters of basic residues that might be involved. There is a widely conserved basic sequence at the junction between blades 5 and 6 (QFRRG in Svp1p) that includes an invariant RRG, and the basic character of a sequence at the start of strand 2C (SPRRLR in Svp1p) is also widely conserved (Figure 2C).

Mutating 284FRRG287 to 284FTTG287 (Svp1pFTTG) decreased the PtdIns(3,5)_P_2 affinity more than 40-fold (Figure 4A). By contrast, changing 71SPRRLR76 to 71SPSSLS76 (Svp1pSPSSLS) only slightly reduced PtdIns(3,5)_P_2 binding (Figure 4A).

Svp1p constructs only correct the svp1_Δ_ vacuole enlargement if they bind PtdIns(3,5)P2

GFP-Svp1pFTTG expressed in yeast was present in the cytosol and nucleus—none was vacuole-associated (Figure 4B). In contrast, GFP-Svp1pSPSSLS localised to the vacuole normally (Figure 4B).

Expression of Svp1pFTTG under control of its own promoter did not correct the vacuole enlargement in _svp1_Δ cells (not shown), suggesting that Svp1p must bind PtdIns(3,5)_P_2 if it is to function normally in vacuole membrane trafficking. Unexpectedly, though, overexpressing GFP-Svp1pFTTG in wild-type yeast provoked a vacuole enlargement like that caused by wild-type Svp1p (Figure 4C). This suggests that GFP-Svp1pFTTG remains capable of sequestering something, probably a soluble protein, that is needed for retrograde membrane trafficking, but cannot deliver it to the PtdIns(3,5)_P_2-rich vacuole.

Membrane recycling from the vacuole fails in svp1_Δ_ cells

Vacuole enlargement in fab1_Δ, vac7_Δ and _vac14/svp2_Δ cells is caused, at least partially, by a failure of membrane recycling to the late endosome (R Piper, unpublished) (Bryant et al, 1998). As explained in Figure 5A, this process can be detected in vivo by appearance in the Golgi of proteolytically matured RS-ALP (mRS-ALP), a variant of the Pho8p alkaline phosphatase that has an FXFXD motif incorporated into its N-terminal region. In stage 1, recently synthesised pro-RS-ALP traffics from the Golgi directly to the vacuole, via an AP-3-dependent route. It is processed to mRS-ALP by vacuolar Pep4p (stage 2), and the FXFXD motif then directs mRS-ALP into the FAB1/PtdIns(3,5)_P_2-dependent retrograde pathway from the vacuole to late endosomes (stage 3), from whence it goes to the Golgi via the retromer pathway (stage 4). As a result, the wild-type Golgi contains both pro-RS-ALP and mRS-ALP, but no mRS-ALP gets to the Golgi of cells defective in step 3 (e.g. fab1, vac7, vac14 and pep12 mutants).

An external file that holds a picture, illustration, etc. Object name is 7600203f5.jpg

Svp1p is needed for the recycling of vacuole membrane proteins. (A) Scheme depicting routes of trafficking of RS-ALP trafficking (for explanation, see the text). (B) In _svp1_Δ cells, the Golgi contains only pro-RS-ALP. Vph1p serves as a marker for the vacuole membrane, and Vps10p for Golgi membrane (for details, see Materials and methods).

We detected both pro-RS-ALP and mRS-ALP in Golgi membranes from wild-type cells, but the _svp1_Δ Golgi contained no mRS-ALP (Figure 5B). This demonstration of a failure of mRS-ALP recycling, together with the vacuole enlargement in _svp1_Δ cells, suggests that the initial step that launches membrane into the vacuole-to-late-endosome retrograd trafficking pathway requires the Svp1p/PtdIns(3,5)_P_2 complex.

svp1_Δ_ cells make abnormally large amounts of PtdIns(3,5)P2

The above observations indicate that some Svp1p functions rely on Fab1p-catalysed PtdIns(3,5)_P_2 production, so we checked that _svp1_Δ cells can make PtdIns(3,5)_P_2. Lipids were extracted from unstressed 3H-inositol-labelled cells and from cells that were salt-stressed to provoke rapid PtdIns(3,5)_P_2 synthesis (Dove et al, 1997).

Remarkably, unstressed svp1_Δ cells contained 5–10 times more PtdIns(3,5)P_2 than wild type: PtdIns4_P, PtdIns3_P and PtdIns(4,5)_P_2 levels were normal (Figures 6A and B). There was four times more PtdIns(3,5)_P_2 than PtdIns(4,5)_P_2 in salt-challenged _svp1_Δ cells—making PtdIns(3,5)_P_2 comprise 6% of the phosphoinositides, more than has been seen in any other cell. The fact that vacuole defects persist in _svp1_Δ cells that contain exorbitant amounts of PtdIns(3,5)_P_2 underlines the fact that _svp1_Δ cells do not respond appropriately to PtdIns(3,5)_P_2.

An external file that holds a picture, illustration, etc. Object name is 7600203f6.jpg

svp1_Δ cells accumulate abnormally large amounts of PtdIns(3,5)P_2. Anion-exchange HPLC chromatograms of the PtdIns_P and PtdIns_P_2 complements of wild-type and svp1_Δ cells, and their responses to hyperosmotic stress. The deacylated phosphoinositides eluted in the order: PtdIns3_P, PtdIns4_P, PtdIns(3,5)_P_2 (filled peak) and PtdIns(4,5)_P_2 (see lower left panel). The relative amounts of phosphoinositides in _svp1_Δ cells are also shown in the table. The total phosphoinositide complement is unchanged, but in _svp1_Δ cells an abnormally large proportion of this is PtdIns(3,5)_P_2. Cells were labelled to isotopic equilibrium, so relative phosphoinositide concentrations match the relative levels of labelling. The data are representative of at least four experiments.

Deletion of HSV1 or HSV2, or both, did not change the cellular PtdIns(3,5)_P_2 complement (not shown).

Svp1p is not a PtdIns(3,5)P2 phosphatase

Might PtdIns(3,5)_P_2 accumulate in _svp1_Δ cells because the missing Svp1p is a PtdIns(3,5)_P_2 phosphatase? Although the Svp1p sequence includes no recognisable phosphatase-like motifs, we compared the ability of biologically active GST-Svp1p to hydrolyse PtdIns(3,5)_P_2 with that of the PtdIns(3,5)_P_2 3-phosphatase MTMR3 (Walker et al, 2001). Under conditions in which GST-MTMR3 rapidly dephosphorylated PtdIns(3,5)_P_2, GST-Svp1p showed no activity (not shown). Moreover, wild-type cells that overexpressed GFP-Svp1p retained a normal PtdIns(3,5)_P_2 complement and showed a normal increase in PtdIns(3,5)_P_2 content following hyperosmotic stress (not shown).

The autophagic role of Svp1p does not need FAB1 or PtdIns(3,5)P2

The observation that _svp1_Δ cells fail to initiate autophagy correctly (Barth et al, 2001) led to its previous designation as AUT10. To determine whether Svp1p must bind PtdIns(3,5)_P_2 to play its part in autophagy, we investigated this process in _fab1_Δ cells, which express Svp1p but contain no PtdIns(3,5)_P_2.

We assessed autophagy by following the maturation of a truncated Pho8p pro-enzyme (Pho8Δ60p) to active alkaline phosphatase. Pho8Δ60p lacks a transmembrane domain needed for trafficking to the vacuole and is made as an inactive cytosolic protein. Its processing needs intravacuolar Pep4p, so it only becomes activated when bulk-sequestered cytosol is transferred into the vacuole during starvation-induced autophagy (Noda et al, 1995; Huang and Klionsky, 2002; Noda et al, 2002).

Pho8Δ60p trafficking was suppressed in _svp1_Δ cells (Figure 7A) to the same degree as in cells lacking Apg1p, another essential protein (Harding et al, 1996; Matsuura et al, 1997; Abeliovich et al, 2003). However, Pho8Δ60p matured normally in _fab1_Δcells, so Svp1p does not need to interact with Fab1p and/or PtdIns(3,5)_P_2 to fulfil its role in autophagy.

An external file that holds a picture, illustration, etc. Object name is 7600203f7.jpg

The role of Svp1p in autophagy is PtdIns(3,5)_P_2-independent, and S. cerevisiae Svp1p-like proteins are not needed for several FAB1/PtdIns(3,5)_P_2-dependent processes. (A) Autophagy processed Pho8Δ60p normally in _fab1_Δ cells, but not in _svp1_Δ cells or in the autophagy mutant _apg1_Δ (for details, see Materials and methods). (B) Svp1p-related proteins are not needed to maintain growth at elevated temperatures (for details, see Materials and methods). All strains grew at 23°C, so only the 42°C plate is shown. Data are representative of those from three or four experiments that gave similar results. (C) Svp1p-related proteins are not needed for vacuole acidification, as assessed by accumulation of the fluorescent weak base quinacrine. (D) Svp1p-related proteins are not needed for the sorting of proteins into MVB, as assessed by the trafficking of GFP-Phm5p.

Some FAB1/PtdIns(3,5)P2-dependent functions do not require Svp1p, Hsv1p or Hsv2p

Heat tolerance

_fab1_Δ yeast grow poorly and lyse at elevated temperatures (Yamamoto et al, 1995) (Figure 7B). However, _svp1_Δ cells were almost as heat-tolerant as wild-type cells, and their slight sensitivity at 44°C was ameliorated by simultaneous deletion of HSV1 and HSV2 (data not shown). None of the Svp1p-like proteins is therefore needed for PtdIns(3,5)_P_2 to confer heat tolerance to cells.

Vacuole acidification

Inactivation of VAC14/SVP2, VAC7 or FAB1 causes a failure of vacuole acidification, even though the vacuolar H+-ATPase (V-ATPase) localises correctly. It has been suggested that PtdIns(3,5)_P_2 might positively regulate V-ATPase or somehow stabilise the proton gradient (Bonangelino et al, 1997; Gary et al, 2002).

Quinacrine, which is fluorescent and accumulates in acidified intracellular compartments, was used to show the severe acidification defect of _fab1_Δ cells (Gary et al, 1998). When we stained wild type, _fab1_Δ, _svp1_Δ and _svp1_Δ/_hsv1_Δ/_hsv2_Δ triple deletion mutants similarly, vacuole acidification was only compromised in the _fab1_Δ cells (Figure 7C). Control of acidification must therefore employ a PtdIns(3,5)_P_2-dependent pathway that needs none of the known Svp1p-related proteins.

Protein sorting to the MVB

Cells that lack Fab1p or its activator Vac14p do not correctly sort certain proteins into MVBs, and irreversible ubiquitination of cargo proteins corrects this defect (Odorizzi et al, 1998; Dove et al, 2002).

This process can be monitored by assessing traffic to the vacuole of GFP-Phm5 (Dove et al, 2002). This normally goes into the lumen, but defective sorting mislocates it at the membrane. GFP-Phm5p sorting was normal in svp1_Δ and triple-deletion svp1_Δ//hsv1_Δ/hsv2_Δ cells, but defective in _fab1_Δcells (Figure 7D)—so none of the Svp1p-like proteins contribute to this PtdIns(3,5)_P_2-dependent process.

Discussion

The first insight into the biology of PtdIns(3,5)P_2 came with the discovery that Fab1p is the PtdIns3_P 5-kinase that makes this lipid (Cooke et al, 1998; Gary et al, 1998). Cells lacking FAB1, or expressing a kinase-inactive version of Fab1p in a _fab1_Δ background (F Cooke, unpublished), display a complex spectrum of dysfunctions, including vacuole enlargement, mis-sorting of proteins into MVBs, heat-sensitive cell lysis, faulty vacuole acidification and problems with vacuole inheritance (Yamamoto et al, 1995; Odorizzi et al, 1998; Dove et al, 2002). Why do _fab1_Δ cells display such a disparate set of dysfunctions, most or all of which are caused by a lack of PtdIns(3,5)_P_2? Do cells have one PtdIns(3,5)_P_2 effector protein that contributes to multiple cell functions, or does PtdIns(3,5)_P_2 interact with several effectors, with each fulfilling a different function(s)?

With the aim of identifying PtdIns(3,5)_P_2 targets, we screened the EUROFAN deletion mutant collection for gene deletions that cause _fab1_Δ-like vacuole enlargement (Dove et al, 2002). From this screen, YFR021w/SVP1 was particularly intriguing—it is very near YFR019w/FAB1 on chromosome VI and all eukaryotes have homologues.

That GFP-Svp1p localises to the vacuole in a _FAB1_-dependent manner suggested that Svp1p might be a PtdIns(3,5)_P_2 effector. However, Svp1p and Fab1p also seem to interact directly (Georgakopoulos et al, 2001), so that might also contribute to the vacuole localisation. Others have also concluded that some Svp1p is localised to a punctate compartment near the vacuole (Barth et al, 2001). Very gentle lysis disrupted the vacuole localisation of GFP-Svp1p but not its punctuate localisation (Guan et al, 2001), and our evidence also indicates that different mechanisms regulate Svp1p targeting to each compartment.

Under approximately ‘physiological' conditions of ionic strength and Mg2+ concentration, monomeric Svp1p is a very selective and high-affinity PtdIns(3,5)_P_2-binder, as are two related yeast proteins (Hsv1p and Hsv2p), the Drosophila homologue CG11975 and the human homologue hSvp1a. This suggests that Svp1p and its homologues constitute a substantial family of PtdIns(3,5)_P_2 specific effector proteins. These are the first proteins known to bind phosphoinositides through small basic amino-acid patches on a β-propeller structure, and so it is possible that a subset of the many proteins encoding β-propeller/WD-40 motifs may also bind to phosphoinositides. This localisation of the Svp1p PtdIns(3,5)_P_2 binding on its β-propeller also suggests a possible structural analogy with the recently discovered binding of a phosphothreonine-containing peptide to the eight-bladed β-propeller of Cdc4p (Orlicky et al, 2003).

Cellular roles of Svp1p

The normal yeast vacuole comprises an array of large organelles that undergo homotypic fusion with other vacuole elements (Wickner, 2002) and heterotypic fusion with smaller organelles, including MVBs (Odorizzi et al, 1998). Following the recruitment of membrane by fusion, the vacuole and MVB elements may normally re-segregate in a manner similar to that seen in cell-free studies with mammalian lysosome/late endosome hybrid organelles (Luzio et al, 2000) or a vesicular intermediate may be involved in vacuole to late endosome trafficking. It is likely that Svp1p is involved in a PtdIns(3,5)_P_2-dependent prebudding step in this vacuole membrane segregation process, as suggested in the model in Figure 8A. When this retrograde trafficking step is blocked, as in _fab1_Δ and _svp1_Δ cells, far more vacuole membrane accumulates than in normal cells and the multiple vacuole elements fuse (Yamamoto et al, 1995).

An external file that holds a picture, illustration, etc. Object name is 7600203f8.jpg

The involvement of Fab1p, PtdIns(3,5)_P_2 and downstream effector proteins in yeast cell functions. (A) Outline of the cycle of vacuole membrane addition and retrieval for which PtdIns(3,5)_P_2 appears to be essential. It is not clear whether retrograde vacuole-to-late-endosome trafficking occurs by re-segregation of the vacuole and late endosome or by the traffic of a vesicular intermediate between these structures. (B) A tentative synthesis of how the actions of multiple PtdIns(3,5)_P_2 effector proteins may contribute to various cell functions (see Discussion). All of these, except the involvement of Svp1p in autophagy, require the presence in cells of FAB1 and/or PtdIns(3,5)_P_2.

Studies of trafficking of the v-SNARE, Vti1p, have confirmed a defect in this segregation/reformation process when PtdIns(3,5)_P_2 production is impaired. Vti1p is involved in the fusion of late endosomes/MVBs and the vacuole, and a retrograde pathway normally recycles Vti1p to late endosomes. This is blocked in _vac7_Δ cells (Bryant et al, 1998), which lack the Fab1p activator Vac7p and make little PtdIns(3,5)_P_2 (Gary et al, 2002). Second mutations that restore PtdIns(3,5)_P_2 to normal levels rescue the _vac7_Δ defects: for example, a suppressive mutation of FIG4, which encodes a phosphoinositide phosphatase (Gary et al, 2002). Our RS-ALP trafficking results show that _svp1_Δ cells phenocopy this _vac7_Δ defect, almost certainly because Svp1p is the PtdIns(3,5)_P_2 effector required for this membrane recycling.

Our working hypothesis is that the Svp1p/PtdIns(3,5)_P_2 complex participates in a specific interaction with some other protein (or complex), and so directs budding from the vacuole surface. Folded as a β-propeller, Svp1p is well suited to form a platform for protein–protein interactions, probably of the peptide-in-groove type (ter Haar et al, 2000). Our autophagy studies make it clear that Svp1p also performs PtdIns(3,5)_P_2-independent vacuole-related functions, and involvement of the same binding partners in these functions might constitute a mechanistic link between the PtdIns(3,5)_P_2-dependent and -independent functions of Svp1p (Barth et al, 2001; Guan et al, 2001).

Since _svp1_Δ cells accumulate abnormally large amounts of PtdIns(3,5)_P_2, it also seems likely that the Svp1p/PtdIns(3,5)_P_2 complex either restrains PtdIns(3,5)_P_2 synthesis or activates PtdIns(3,5)_P_2 degradation. The known interaction between Fab1p and Svp1p (Georgakopoulos et al, 2001) suggests a possible mechanism for feedback regulation of PtdIns(3,5)_P_2 synthesis.

Svp1p as one of several PtdIns(3,5)P2 effectors?

Taken with recent reports that Ent3p, Ent5p and Vps24p are effectors that contribute to the MVB trafficking functions of PtdIns(3,5)_P_2 (Friant et al, 2003; Whitley et al, 2003), our results support the idea that the complex _fab1_Δ phenotype is caused by malfunctions in several, probably independent, PtdIns(3,5)_P_2 effector pathways. It is clear that neither Ent3p, Ent5p nor Vps24p (Friant et al, 2003; Whitley et al, 2003), nor any of the Svp1p-related proteins, participates in the effects of PtdIns(3,5)_P_2 on vacuole acidification or heat tolerance. Elsewhere, we shall describe Svp3p, an unrelated putative PtdIns(3,5)_P_2 effector that influences PtdIns(3,5)_P_2 metabolism and appears to contribute to MVB sorting and to heat tolerance (SK Dove, unpublished; see Figures 7B and D).

Other observations support the notion that PtdIns(3,5)_P_2 exerts its effects through multiple effector pathways. For example, low overexpression of the FAB1 gene in the vac14-1 background restores vacuole acidification, but the vacuole remains enlarged. However, vacuole enlargement is only corrected if FAB1 is greatly overexpressed (Bonangelino et al, 2002; Dove et al, 2002). Similarly, the catalytic site mutant _FAB1_G2042V/G2045V normalises MVB sorting in _fab1_Δ cells but does not correct the vacuole enlargement (Odorizzi et al, 1998).

Figure 8B summarises our current understanding of PtdIns(3,5)_P_2 effector pathways. It highlights the fact that none of the currently known PtdIns(3,5)_P_2 effectors mediates PtdIns(3,5)_P_2-dependent vacuole acidification, and therefore suggests that at least one S. cerevisiae PtdIns(3,5)_P_2 effector still awaits discovery. Furthermore, Hsv1p and Hsv2p are likely to be PtdIns(3,5)_P_2 effectors, but their loss causes none of the known FAB1/PtdIns(3,5)_P_2-dependent phenotypes—it is therefore probable that undiscovered PtdIns(3,5)_P_2-dependent cellular processes remain to be found.

Materials and methods

Most materials were from sources defined previously, and methods for the growth of yeast, for FM4-64 vacuolar staining, for the GFP-Phm5p MVB sorting assay, for quinacrine staining and for lipid have been described (Dove et al, 1997, 2002; Dove and Michell, 1999; McEwen et al, 1999). The source of all yeast strains used in this study is indicated in Table I. PtdIns4_P_ and PtdIns5_P_ were from Echelon Inc. (Salt Lake City, UT), and PtdIns(3,4)_P_2, PtdIns(4,5)_P_2, PtdIns(3,4,5)P_3, and PtdIns3_P were from Matreya Inc. (Pleasant Gap, PA): all were dipalmitoyl. Dioleoyl-PtdCho (DOPC), dipalmitoyl-PtdSer and PtdIns were from Sigma-Aldrich. Phosphoinositides and Affigel-linked PtdIns(3,5)_P_2 were synthesised as described (Krugmann et al, 2002). For Biacore analysis, PtdIns(3,5)_P_2 was from CellSignals Inc. (Lexington, KY).

Table 1

Yeast strains used in this study

Yeast strain Source Reference
BY4742 mat α ura3_Δ_0 his3_Δ_1 leu2_Δ_0 lys2_Δ_0 EUROSCARF N/A
BY4742 svp1_∷_KANMX4 EUROSCARF N/A
BY4742 svp1_∷_KANMX4 hsv1_∷_LEU2 hsv2_∷_HIS3 This study N/A
BY4742 hsv1_∷_LEU2 hsv2_∷_HIS3 D Alexandraki Georgakopoulos et al (2001)
BY4742 hsv1_∷_KANMX4 EUROSCARF N/A
BY4742 hsv2_∷_KANMX4 EUROSCARF N/A
BY4742 fab1_∷_KANMX4 EUROSCARF N/A
BY4742 fab1_∷_KANMX4 pho8_∷_LEU2 This study N/A
BY4742 hsv1_∷_KANMX4 pho8_∷_LEU2 This study N/A
BY4742 hsv2_∷_KANMX4 pho8_∷_LEU2 This study N/A
BY4742 svp1_∷_KANMX4 pho8_∷_LEU2 This study N/A
BY4742 svp2_∷_KANMX4 EUROSCARF N/A
BY4742 svp3_∷_KANMX4 EUROSCARF N/A
NA: not applicable.

Vacuole morphology screen

This screen was as described (Dove et al, 2002). Subsequent analysis of the identified genes included bioinformatic sifting to eliminate genes unrelated to vacuole function and to prioritise genes with homologues in many organisms.

Plasmid construction

GFP-Svp1p and GFP-Hsv2p were constructed by PCR amplification of the respective ORFs from yeast genomic DNA using the expand proof-reading DNA polymerase (Roche). The Met-regulated construct pUG36-SVP1 was created by ligating the SVP1 ORF cut with _Eco_RI and _Hin_dIII into pUG36. pUG36-HSV2 was created by ligating the complete HSV2 ORF, excised with _Bam_HI and _Sal_I, into pUG36. Transformed yeast were grown to 1 × 107 cells/ml in SC-Ura-Met and visualised as described (Dove et al, 2002). Mutations in the SVP1 gene were carried out using serial overlap extension PCR (Warrens et al, 1997). Mutants were sequenced and then moved into GST expression vectors.

Expression of GST-Svp1p and GST-Hsv2p

The SVP1 and HSV2 ORFs were cloned into pGEX-6P-1 (Amersham-Pharmacia) (with a rhinovirus-2C protease cleavage site between GST and the target protein) and pGSTag (Ron and Dressler, 1992). Both plasmids were transformed into Escherichia coli BL21 and expressed as follows: cells were grown in LB+ampicillin (100 μg/ml) to an OD600 of 0.5–0.8, induced with IPTG (200–500 μM, 2–4 h, 30°C), harvested, washed in ice-cold PBS containing protease inhibitors (pepstatin A, 5 μg/ml; leupeptin 5 μg/ml; E-64, 3 μg/ml; aprotinin, 5 μg/ml; PMSF, 100 μM), lysed (100 mg/ml lysozyme, 4°C, 30 min) and disrupted with glass beads in a BeadBeater (6 × 30 s, 4°C). Triton X-100 (1%) supernatants from the disrupted cells (8000 g, 15 min; then 100 000 g, 1 h) were filtered (0.2 μm) and applied to glutathione–sepharose 4B beads (30 min, 4°C), which were eluted following the manufacturer's recommendations. Proteins were made 50% in ice-cold glycerol, snap-frozen and stored at −80°C.

Phosphoinositide dot blots

SVP1 and HSV2 cloned into the pGSTag vector (with a PKA phosphorylation site in the GST linker) were expressed in E. coli BL21 (see above) and 32P-labelled (Kavran et al, 1998). [32P]proteins were filtered (0.2 μm) and used to probe lipid ‘dot blots' of serial two-fold dilutions of phosphoinositides (from 1000 to 15.6 pmol, in CHCl3:CH3OH:H20, 400:56:4, v/v/v) on Hybond C membrane (BioRad). Blots were blocked (3 h, room temperature, 10 ml PBS, 0.5 mM MgCl2, 5% ECL blocking reagent) and probed with 1–2 × 106 dpm of [32P]protein (30 min, room temperature). Blots were extensively washed (PBS/0.5 mM MgCl2) and radioactivity detected by autoradiography. Concentrations of the lipid stocks were quantified immediately before spotting by wet ashing and phosphate determination (Baginski et al, 1967).

Protein binding to Affigel-linked PtdIns(3,5)P2

GST-Svp1p (5 μM) was incubated (1 ml PBS containing 0.1% Triton X-100, 10 mM DTT and 50 mM PMSF) with 20 μl of Affigel-20 beads covalently linked to PtdIns(3,5)_P_2, in the presence of micellar dispersions of various phosphoinositides (10 μM) for 30–60 min at 4°C. In 10 min or less, the beads were sedimented (12 000 g, 1 min) and washed in the above buffer (twice) and in 5 mM Hepes/KOH, pH 7.5. The remaining bead-bound protein was solubilised in 40 μl SDS–PAGE sample buffer and resolved on 10% SDS–PAGE minigels (made with piperazine diacrylamide (BioRad) rather than _N_′_N_′-methylene bis-acrylamide, to reduce the silver-staining background). Gels were fixed and stained (BioRad Silver Stain Plus).

Biacore analysis of phosphoinositide binding

This was carried out exactly as described (Yu and Lemmon, 2001). Positive controls were performed alongside Svp1p experiments, and included the PLCδ1 PH domain (binds PtdIns(4,5)P_2), the DAPP1 PH domain (binds PtdIns(3,4)P_2), the FAPP1 PH domain (binds PtdIns4_P) and the Hrs1 FYVE domain (binds PtdIns3_P). Lipid surfaces were used within 8 h of generation, since signal strength began to decrease within ∼12 h.

GST was quantitatively removed from GST-Svp1p with PreScission protease. When gel-filtered on Superose-6 (Amersham-Pharmacia), the liberated Svp1p eluted as a 60–70 kDa monomer. Svp1p binding was measured by simultaneously passing it over a phosphoinositide-containing sensor surface and a control DOPC sensor surface, with the control signal subtracted from that from the phosphoinositide surface. Binding data are plotted as per cent of maximum binding against protein concentration injected. _K_D values were calculated as described (Yu and Lemmon, 2001).

Estimating KD value

For Svp1p binding to PtdIns(3,5)_P_2, PtdIns(3,4)_P_2 and PtdIns(4,5)_P_2, data were fitted to the following equation:

equation image

where [Prot] is the flowing protein concentration (assumed unaffected by binding to the surface), _K_D is the dissociation constant and Y corresponds to a residual or background signal. Fitting was performed using ORIGIN (MicroCal), with floating _K_D and Y.

[3H]inositol labelling and phosphoinositide analysis

Yeast were labelled in inositol-free media for 5–6 cell divisions, so that changes in [3H] would parallel changes in lipid concentration, and phosphoinositides were extracted and analysed as described (Cooke et al, 1998).

Assay of Pho8_Δ_60p maturation

This used a method described by Noda et al (1995). PHO8 was disrupted in appropriate strains using a pho8_∷_LEU2 cassette, and the disruptants were transformed with a plasmid that overexpresses Pho8Δ60p (pTN3, a gift from Dr T Noda, National Institute for Basic Biology, Okazaki, Japan). These strains were grown to ∼1 × 107 cells/ml, washed and either processed immediately or incubated in a nitrogen-free medium for 8 h at 30°C, to induce autophagy and then processed. They were suspended in 1 ml of 50 mM Tris–HCl, 5 mM MgCl2, 1 mM PMSF and 1 μg/ml pepstatin A, pH 9, cooled (15 min) and vortexed with glass beads (8 × 30 s, with 30 s on ice between cycles). Lysates were centrifuged (10 000 g, 15 min, 4°C). The supernatant protein (5 μg) was incubated at 25°C for 3 min in 1 ml of 10 mM _p_-nitrophenolphosphate, 0.5 M Tris and 2.5 mM MgCl2, pH 8.8, with continuous monitoring of OD420. Relative rates of dephosphorylation rates were calculated for the induced and uninduced cells.

Dilution assays for temperature sensitivity of growth

Yeast were diluted to 1 × 107 cells/ml, serially diluted four-fold, and 5 μl samples spotted on replicate plates. One was incubated at 23°C and the other at 42°C, until wild-type cells formed distinct colonies at all dilutions.

Assay for retrograde vacuole to late endosome trafficking

Cells expressing RS-ALP were spheroplasted as previously described (Urbanowski and Piper, 2001) and lysed in 300 mM sorbitol, 20 mM HEPES, pH 7.2, and 1 mM EDTA (HES buffer) containing a protease inhibitor cocktail (Complete tm, Boehringer Mannheim). Postnuclear supernatants were adjusted to 30% Optiprep and layered beneath 5 ml of HES buffer in an SW41 Beckman ultracentrifuge tube. Linear Optiprep gradients were overlaid (BioComp gradient mixer) and the samples were centrifuged (40 000 rpm, 18 h). Fractions were collected downwards from the top. Proteins were separated by SDS–PAGE and immunoblotted using monoclonal anti-ALP and anti-Vph1p antibodies (Molecular Probes) and a polyclonal anti-Vps10p antiserum (Piper et al, 1995).

Acknowledgments

We thank Drs Paul Whitley (University of Bath, UK), Gerald Hammond (UCL, London), Despina Alexandraki (University of Crete), Lois Weisman (Iowa) and Geraint Thomas (UCL, London) for valuable discussions and reagents. This work was funded by the Royal Society (SKD and RHM), the Wellcome Trust (SKD and RHM), the BBSRC (to ABT and JT) and grant 5-ROI-GM56846 from the National Institutes of Health (to MAL). SKD is a Royal Society University Research Fellow. RHM is a Royal Society Research Professor.

References

Articles from The EMBO Journal are provided here courtesy of Nature Publishing Group