Báo cáo khoa học: The common phospholipid-binding activity of the N-terminal domains of PEX1 and VCP/p97 doc

The ATPase activities of PEX1 and PEX6 are believed to be indispensable for normal peroxisomal biogenesis, including the critical step of Keywords AAA-ATPase; N-terminal domain; PEX1; ph

of the N-terminal domains of PEX1 and VCP/p97

Kumiko Shiozawa1, Natsuko Goda1, Toshiyuki Shimizu1, Kenji Mizuguchi2,3, Naomi Kondo4,

Nobuyuki Shimozawa4,5, Masahiro Shirakawa1,6and Hidekazu Hiroaki1

1 International Graduate School of Arts and Sciences, Yokohama City University, Tsurumi-ku, Yokohama, Kanagawa, Japan

2 Department of Biochemistry, University of Cambridge, UK

3 Department of Applied Mathematics and Theoretical Physics, Centre for Mathematical Sciences, University of Cambridge, UK

4 Department of Pediatrics, Gifu University School of Medicine, Japan

5 Division of Genomic Research, Life Science Research Center, Gifu University, Japan

6 Department of Molecular Engineering, Graduate School of Engineering, Kyoto University, Japan

The peroxisome is a single-membrane organelle

involved in various metabolic pathways [1] Biogenesis

and maintenance of the peroxisome require at least 32

proteins, known as PEX gene products or peroxins [2]

Autosomal recessive mutations in any of 12 of the

PEXgenes cause peroxisome biogenesis disorders, such

as Zellweger syndrome, neonatal

adrenoleukodystro-phy and infantile Refsum disease [3]

Peroxisomal membrane fusion and protein transloca-tion are both ATP-dependent processes, but among over 32 peroxins, only PEX1 and PEX6 are AAA-ATPases Dysfunction of the ATPase activity of either

of the two proteins results in peroxisome biogenesis disorders [4–7] The ATPase activities of PEX1 and PEX6 are believed to be indispensable for normal peroxisomal biogenesis, including the critical step of

Keywords

AAA-ATPase; N-terminal domain; PEX1;

phospholipid; valosine-containing protein

Correspondence

H Hiroaki, Division of Molecular Biophysics,

Graduate School of Integrated Sciences,

Yokohama City University, 1-7-29,

Suehirocho, Tsurumi, Yokohama, Kanagawa,

Japan 230-0045

Fax: +81 45 508 7361

Tel: +81 45 508 7214

E-mail: hiroakih@tsurumi.yokohama-cu.ac.jp

(Received 21 June 2006, revised 6 September

2006, accepted 11 september 2006)

doi:10.1111/j.1742-4658.2006.05494.x

PEX1 is a type II AAA-ATPase that is indispensable for biogenesis and maintenance of the peroxisome, an organelle responsible for the primary metabolism of lipids, such as b-oxidation and lipid biosynthesis Recently,

we demonstrated a striking structural similarity between its N-terminal domain and those of other membrane-related AAA-ATPases, such as valo-sin-containing protein (p97) The N-terminal domain of valosine-containing protein serves as an interface to its adaptor proteins p47 and Ufd1, whereas the physiologic interaction partner of the N-terminal domain of PEX1 remains unknown Here we found that N-terminal domains isolated from valosine-containing protein, as well as from PEX1, bind phosphoinos-itides The N-terminal domain of PEX1 appears to preferentially bind phosphatidylinositol 3-monophosphate and phosphatidylinositol 4-mono-phosphate, whereas the N-terminal domain of valosine-containing protein displays broad and nonspecific lipid binding Although N-ethylmaleimide-sensitive fusion protein, CDC48 and Ufd1 have structures similar to that

of valosine-containing protein, they displayed lipid specificity similar to that of the N-terminal domain of PEX1 in the assays By mutational analy-sis, we demonstrate that a conserved arginine surrounded by hydrophobic residues is essential for lipid binding, despite very low sequence similarity between PEX1 and valosine-containing protein

Abbreviations

AAA, ATPase associated with a diversity of cellular activities; ER, endoplasmic reticulum; GST, glutathione-S-transferase; IPTG, isopropyl thio-b-D-galactoside; ND, N-terminal domain; NSF, N-ethylmaleimide-sensitive fusion protein; PA, phosphatidic acid; PC, phosphatidylcholine;

PE, phosphatidylethanolamine; PhA, phytic acid; PS, phosphatidylserine; PtdIns, phosphatidylinositol; QCM, quartz crystal microbalance; SKD1, suppressor of K + transport growth defect 1; SNAP, soluble NSF attachment protein; VCP, valosine-containing protein.

peroxisomal protein import and peroxisomal

mem-brane fusion [8,9,49] The two proteins have been

shown to associate with each other both in vitro and

in vivo[5,10–12], and it has been found that ATP

bind-ing and hydrolysis are important for their interaction

in Saccharomyces cerevisiae [12] Recently, it has been

reported that the PEX1–PEX6 complex is involved in

the dissociation of ubiquitinated PEX5 from the

per-oxisomal membrane [13] PEX5 binds the perper-oxisomal

proteins (cargo proteins), which contain peroxisome

targeting signal 1 at the C-terminus in the cytosol,

tar-gets them to the protein complex machinery at the

peroxisomal membrane and releases them into the

peroxisomal lumen before returning to the cytosol

Several groups have reported that PEX5 can return to

the cytosol in a ubiquitin-dependent manner [13–17],

and it is believed that the PEX1–PEX6 complex is

indispensable to this step [15]

AAA-ATPases are found in three domains of all

liv-ing organisms [18,19], and play an important role as

molecular chaperones, including in the dissociation of

protein complexes and protein translocation PEX1

belongs to the class of type II AAA-ATPases, which

contain two copies of the AAA cassette Extensive

structure–function studies of

N-ethylmaleimide-sensi-tive fusion protein (NSF) and valosine-containing

pro-tein (VCP) (p97), which also belong to this class, have

been reported [20–31,33–35] These enzymes form a

hexameric ring and can act as protein unfoldases Type

II AAA-ATPases are located in organelle membranes,

where they are involved in specific functions, such as

membrane fusion [20–22] and protein transport across

the membrane [23] NSF, and its yeast ortholog Sec18,

are responsible for heterotypic membrane fusion

mainly in exocytic pathways [20,24,25] a-soluble NSF

attachment protein (SNAP), b-SNAP and c-SNAP are

known to be the most important targets of NSF [21]

VCP and yeast CDC48 are involved in endoplasmic

reticulum (ER)-associated protein degradation

[23,26,27] as well as in remodeling of the Golgi and

nuclear membrane [20,28] Although the specific target

of VCP unfoldase activity remains unclear, there are

several adaptor molecules, p47 [28], Ufd1⁄ Npl4 [27],

VCIP135 [29], Derlin-1 and VIMP [30,31], which may

determine VCP functions

Previously, we have determined the crystal structure

of the N-terminal domain (ND) of mouse PEX1

(PEX1-ND) [32] It bears a striking resemblance to

those of VCP (p97) [33], NSF [34,35], the archaeal

homolog VAT [36] and Sec18 [40], despite the low

level of sequence similarity The domain architecture

of all five proteins contains an ND followed by the

tandem AAA domains D1 and D2 This architecture is

known as a ‘supradomain’, and is found in many cases where two or three domains (in this case, ND, D1 and D2) are persistently conserved in terms of their sequen-tial order and biological context [37] In the crystal structure of PEX1-ND, the characteristic crevice, sim-ilar to that of NSF, is conserved NSF-ND is assumed

to be a binding site for a-SNAP [34] In the case of VCP-ND, a flat hydrophobic surface provides an inter-face to the ubiquitin-like domain of p47 [38], and the surface might be used for binding other adaptor pro-teins, such as Npl4–Ufd1 complex and VCIP135 Inter-estingly, the ND of Ufd1, which is similar to VCP-ND

as well as PEX1-ND, also shares a common hydropho-bic interface for polyubiquitin binding [39] Ufd1 is a non-ATPase-type adaptor protein associated with VCP and Npl4 This hydrophobic surface is not conserved

in PEX1-ND, and the molecular function of this pro-tein remains to be resolved

The structural similarity between VCP, NSF and PEX1 suggested that, as VCP binds phospholipids [22,58], PEX1 could have similar properties We there-fore investigated phospholipid binding and the binding sites of PEX1 and VCP Evolutionary trace analysis revealed a conserved charged residue surrounded by hydrophobic residues in the ND of PEX1 and VCP, which may bind to the phospholipid It is shown below that both PEX1-ND and VCP-ND can bind phospho-lipids in vitro with broad specificity for phosphatidyl-inositol (PtdIns) monophosphate species By analogy

to PEX1, we examined phospholipid binding and bind-ing sites in the NDs of NSF, CDC48 and Ufd1 As their surface properties differ from those of PEX1 and VCP, no lipid-binding site was found by computa-tional analysis, although experimentally, they were all found to bind phospholipids

Results

Evolutionary trace analysis of PEX1-ND Based on the structure-restrained sequence alignment

of PEX1 orthologs with other AAA-ATPase NDs, VCP is the closest neighbor of PEX1 To identify potential protein–protein and protein–lipid interaction surfaces, we have analyzed the available structural data for PEX1 [32] and VCP [33] and searched for con-served charged residues as well as hydrophobic resi-dues exposed at the protein surface (Figs 1B and 2; supplementary Fig S1) [44,45] R135 in PEX1 and R144 in VCP are the only exposed charged residues, conserved across the two protein families, whereas K174 is relatively conserved among other PEX1 ortho-logs (Fig 1B) These positively charged residues are

located at the end of the shallow groove between the

N-lobe and C-lobe of PEX1-ND An exposed

constel-lation of hydrophobic residues is found along a

longi-tudinal line of the kidney-shaped PEX1-ND, which

surrounds the conserved basic residues In addition,

the corresponding R144 in VCP-ND, equivalent to

R135 in PEX1, was found to be surrounded by

exposed hydrophobic residues (supplementary Fig S1)

These residues are relatively well conserved among

VCP orthologs, although the shape of the hydrophobic

constellation differs (Fig 2B)

In order to unravel the characteristics of these

com-mon charged and hydrophobic interfaces, protein

interface predictions were made based on optimal

docking areas (ODAs) in PEX1-ND and VCP-ND [46]

(supplementary Fig S2) ODAs are a set of continuous

surface patches with optimal protein–protein docking

desolvation energy A previous analysis correctly

located known protein–protein interfaces in 80% of

cases [46] Indeed, the known p47 interaction surface

in VCP-ND was detected with a significant low-energy

value (<) 10 kcalÆmol)1; shown in red in

supplement-ary Fig S2) In contrast, the areas near R135 in

PEX1-ND and the corresponding region around R144

in VCP-ND displayed no significant ODAs despite

their hydrophobic nature, suggesting that this surface

may be used for interactions with nonprotein

mole-cules This observation led to the hypothesis that

PEX1-ND and VCP-ND may directly associate with

phospholipids

PEX1-ND as a phosphoinositide-binding domain

One of the common features of the type II

AAA-ATP-ases is their role in organellar membrane fusion To

the best of our knowledge, there are no reports

indica-ting that PEX1-ND is responsible for recruitment of

PEX1 to the peroxisomal membrane Several lines of

evidence indicate that the N-terminal region of PEX1

may directly associate with the membrane The binding

of PEX1-ND to phospholipids as well as its specificity

have been investigated by incubating a purified

gluta-thione-S-transferase (GST)-tagged PEX1-ND on

nitro-cellulose membranes spotted with different lipid species

(PIP Strip) Binding of protein to the membrane was

quantified using anti-GST serum (Fig 3A) Mouse

PEX1-ND (residues 3–180) clearly represented the

region that binds phospholipid PEX1-ND bound most

strongly to PtdIns monophosphates (PtdIns3P,

PdsIns4P, and PtdIns5P) with approximately equal

affinity PEX1-ND weakly bound to PtdIns

bisphos-phates [PtdIns(4,5)P2, PtdIns(3,4)P2and PtdIns(3,5)P2]

with lower or negligible affinity compared to PtdIns

monophosphate Very weak binding of PEX1-ND to PtdIns, phosphatidic acid (PA) and phosphatidylserine (PS) was observed, while no binding to phosphatidyl-choline (PC) and phosphatidylethanolamine (PE) could

be detected GST alone did not bind to phospholipids under these conditions (Fig 3A, right) The same experiment was carried out in the presence of 20 mm phytic acid (inositol hexakisphosphate), which is a competitive inhibitor of various phosphoinositide-bind-ing domains with a broad specificity [47] Nonspecific electrostatic interactions between the characteristic positively charged residues of PEX1-ND and the phospholipid are likely to contribute significantly to this binding Indeed phytic acid inhibited the binding

of PEX1-ND to the phosphoinositides, thereby demon-strating that lipid binding occurs specifically to the inositol phosphate moiety (Fig 3A, middle) In addi-tion, we found that PtdIns binding to PEX1-ND is

Ca2+-independent (data not shown), in contrast to the behavior of some other phosphoinositide-binding domains, such as annexin and the C2 domain, whose lipid binding is Ca2+-dependent [69]

The lipid-binding specificity of GST–PEX1-ND was studied in detail using a liposome recruitment assay

PC + PE (1 : 1)-based liposomes containing up to 5% PtdIns3P, PtdIns4P, PtdIns5P or PtdIns were prepared (Fig 3B) The proteins that sedimented as well as those that remained in solution were analyzed

by Coomassie brilliant blue-stained SDS⁄ PAGE and quantified PEX1-ND bound most strongly to PtdIns3P and PdsIns4P, and weekly to PtdIns5P The amount of PEX1-ND recruited to the liposomes increased with the amount of PtdIns monophosphate included in the liposome preparation The apparent binding constant Kd for binding between PEX1-ND and PtdIns3P was 1 lm, as judged from its saturation curve The inhibitory effect of phytic acid is also demonstrated in Fig 3B We assumed that PEX1-ND interacts with PtdIns4P more strongly and more spe-cifically than with PtdIns3P, as PtdIns4P binding is maintained in the presence of excess phytic acids, since phytic acid is not a perfect mimic of the head-group of PtdIns4P Although PEX1-ND is recruited

to the liposomes in the presence of a large excess of PtsIns5P or PtdIns, there is no steep saturation curve, suggesting that the binding is nonspecific

The liposome interaction of PEX1-ND was further confirmed by quartz crystal microbalance (QCM) assays Figure 3C shows some typical results for GST– PEX1-ND immobilized on a sensor chip and titrated

by liposomes The frequency change was only observed when the liposome contained PtdIns4P, which is con-sistent with the results above

B

Determination of the residues responsible for

phospholipid binding in PEX1-ND

To determine the residues responsible for

phosphoinos-itide binding, two mutants of PEX1-ND, R135A and

K174A, were generated to eliminate the conserved

pos-itive residues, which may serve as acceptors of the

phosphate moiety of phosphoinositides Both mutants

(R135A and K174A) were subjected to the PIP Strips

assay (Fig 4A) Whereas K174A was found to bind to

PtdIns monophosphates with approximately the same

affinity and specificity as the wild-type protein, R135A

had effectively lost all phosphoinositide-binding

activ-ity In order to estimate the binding ability, R135A

was subjected to the liposome recruitment assay

(Fig 4B) R135A did not bind the PC + PE-based

liposome containing PtdIns4P

Phosphoinositide-binding activity of ND

of mouse VCP

By analogy to PEX1, VCP may directly attach to the

membrane via the ND, despite the low level of

sequence similarity As mentioned above, R144 in

VCP-ND, equivalent to R135 in PEX1-ND, is

con-served among VCP orthologs and surrounded by

exposed hydrophobic residues A comparison of these

hydrophobic residues of VCP with the corresponding

residues of PEX1-ND is shown in Fig 2B In contrast,

some hydrophobic residues, colored in green, are only

conserved among VCP orthologs but not in PEX1

To test possible binding of VCP-ND to

phospholi-pids through this surface, mouse VCP-ND was purified

as a GST fusion product and subjected to a

lipid-bind-ing assay uslipid-bind-ing PIP Strips, and, as shown in Fig 4C,

significant phospholipid binding was found VCP-ND

binds PtdIns monophosphates (PtdIns3P, PtdIns4P

and PtdIns5P) as well as PtdIns bisphosphates, such as

PtdIns(3,4)P2(i.e it has a lower specificity than

PEX1-ND) By analogy to PEX1-ND, the R144A mutant of

VCP was generated to confirm the residue responsible

for phosphoinositide binding In the PIP Strips assays

(Fig 4), R144A displayed no detectable binding to

PtdIns monophosphates, indicating the importance of R144 for phospholipid binding

Other NDs also bind phosphoinositides By analogy

to PEX1, other proteins containing NDs with the same folding topology may attach to the membrane via this domain, despite the low level of sequence conservation

To test this hypothesis, we purified NDs derived from mouse NSF, yeast CDC48, mouse Ufd1 and yeast Ufd1 as GST fusion proteins, and subjected them to a lipid-binding assay using PIP Strips As shown in sup-plementary Fig S3, all GST fusion proteins of NDs bound phospholipids to some extent The specificity of binding of these NDs to phosphoinositides was again vague, as shown in supplementary Fig S3 Most of the NDs bound preferentially to PtdIns monophos-phates, whereas some bound weakly to PtdIns bis-phosphates In addition, the phospholipid-binding properties of NSF were sensitive to an excess of phytic acid, suggesting that the interaction is specific for the inositol phosphate molecules of phosphoinositides

Discussion

Phospholipid-binding interface on PEX1-ND Our results, which show that isolated PEX1-ND binds phosphoinositides with broad specificity in vitro, are consistent with the proposed function of PEX1 at the peroxisomal membrane [10,48,49,62] We have identi-fied a specific surface region of PEX1-ND implicated in phospholipid interactions Although the surface charge distribution of PEX1-ND is rather acidic, the protein does not bind phospholipids containing basic head-groups, such as PE and PC From a multiple sequence alignment analysis of PEX1 orthologs, we found a single well-conserved basic residue (R135) located at one end

of a shallow groove, which is assumed to be a substrate-binding site in NSF [34,50] Furthermore, a constella-tion of well-conserved hydrophobic residues, such as V36, V49, W60, W116, L121, L130, L131, W146, V147, L175, L176 and I177, is positioned vertically across the surface of the groove, with R135 and K174 near one end (Fig 2A) The surfaces where the hydrophobic patch

Fig 1 Sequence comparison of the N-terminal domain (ND) of PEX1 with that of other related proteins (A) Schematic representation of the domain architecture of AAA-ATPases PEX6, PEX1, N-ethylmaleimide-sensitive fusion protein (NSF) and valosine-containing protein (VCP) The ND AAA-ATPase domains (D1, D2) are shown Physical interactions among the domains are indicated by arrows (B) Multiple sequence alignment of PEX1-ND and related NDs with secondary structure elements of PEX1-ND For sequence names, see Experimental procedures The alignment was generated with CLUSTALX [60] and manually edited The secondary structure elements are shown at the top, with thick line segments for the a-helices (a1–a4) and thin line segments for b-strands (b1–b14) The conserved residues of PEX1-ND found in the putative phospholipid-binding site are shown on the second line Conserved hydrophobic residues and basic residues of PEX1-ND are indica-ted by closed circles and closed squares, respectively (see text) The conserved residues and class-specific residues of PEX1-ND and

VCP-ND are colored yellow and green, respectively Well-conserved basic residues on the surfaces of PEX1-VCP-ND and VCP-VCP-ND are colored blue.

and the conserved basic residues are found are flat The peroxisomal membrane is composed of approximately 10% PC, the head-group of which contains hydrophobic methyl groups [51] It is noteworthy that phospholipids with hydrophilic head-groups such as phosphoinositides may be sparsely distributed on the membrane, whereas

PC may form a hydrophobic cluster Thus, the juxtapo-sition of positively charged and hydrophobic residues

on a flat surface of PEX1-ND suggests an interaction with such a membrane surface The results obtained with the R135A and K174A mutants clearly demon-strate that this basic and hydrophobic surface is the phospholipid-binding interface and that arginine 135 is responsible for specific recognition of the phosphate group The importance of the conserved arginine residue

in the molecular recognition of phosphoinositides in other phospholipid-binding domains, such as FYVE [52] and PX [53], has been reported We thus conclude that PEX1-ND is one of the functional modules for membrane binding

Involvement of phosphoinositides in peroxisomal biogenesis

The interaction between PEX1-ND and phospholipids could be specific to an inositol phosphate moiety, because lipid-binding activity is abolished in the pres-ence of excess phytic acid as a competitor (Fig 3A) In terms of phospholipid specificity, there is only a limited preference for PtdIns monophosphate species over other phosphoinositides (Fig 3A,B) Some of these nonspe-cific lipid-binding domains are believed to act as scaffolds for a specific membrane rather than in the PtdIns-mediated signal transduction systems There are

a few reports in the literature suggesting the involvement

of PtdIns-mediated signal transduction systems in peroxisomal biogenesis, such as those involving phos-phatidylinositol-3-kinase and phosphatidylinositol-4-phosphate-5-kinase Pexophagy, an event in the autophagic degradation of excess of peroxisomes, may

be an exceptional phenomenon [54] More recently, peroxisome fusion, the critical early step in peroxisome assembly, for which PEX1 and PEX6 are essential, has been reported to require the phosphoinositides PtdIns4P and PtdIns(4,5)P2, as well as a distinct set of

peroxisom-al membrane proteins that specificperoxisom-ally bind to these two lipids [62] This supports our finding that PEX1-ND interacts with phosphoinositides Nevertheless, phos-phoinositide-specific regulation of the peroxisome, akin

to receptor signaling systems or endocytosis, therefore appears to be unlikely, because PtdIns4P is the most abundant phosphoinositide in the cell In contrast, the number of nonspecific membrane-binding domains is

R135

V36

V49 W60

W146

L121 W116

I177

L176

L131

L175

K174

L130 L169

V147

N

C

I70 V108 Y110

Y143 A142

R144

F139 L140 V181 I182

F131

V154

F152 Y138 C

N

C

A

B

Fig 2 Constellation of hydrophobic residues with conserved basic

residues on the surface of the N-terminal domains (NDs) of PEX1

and valosine-containing protein (VCP) (A) Conserved hydrophobic

residues within the PEX1 orthologs shown in Fig 1B are colored

yellow Conserved basic residues R135 and K174 are colored blue.

(B) Hydrophobic residues conserved among PEX1 and VCP (V108,

Y138, F139, F152, V154, I182) are colored yellow Conserved

hydrophobic residues within VCP orthologs only (I70, Y110, F131,

A142, Y143, L140 and V181) are colored green The conserved

basic residue R144 is colored blue The inset is a ribbon diagram of

the molecule.

often correlated with that of other domains responsible

for specific protein–protein interactions For example,

more than 63% of pleckstrin homology-domain

con-taining proteins also contain one or more

protein-inter-acting domains such as SH3 [55] This may enhance

specificity for a particular subcellular compartment

PEX1-ND meets the criteria for such a protein, as it

forms a scaffold on the peroxisomal membrane through

specific interactions with phospholipids, whereas PEX6

interacts with either PEX26 (mammals) or PEX15

(yeast) on the membrane (Fig 5) [56,57] Some

mem-brane-binding proteins bind multiple lipids and⁄ or

pro-teins simultaneously [63] The binding of PEX1-ND to

PtdIns might also be attributed to coincident PtfIns

monophosphate binding

Common phospholipid-binding activity

of VCP-ND and PEX1-ND

We have also demonstrated that the NDs of another

type II AAA-ATPase, VCP, can directly bind

phos-pholipids in vitro Although this is the first report of direct interaction between the isolated NDs and phos-phoinositides, several lines of evidence concerning VCP and NSF support our findings Affinity purifica-tion of NSF using immobilized phospholipids, such as PtdIns(4,5)P2 and PA, has been reported [67] Purified recombinant NSF was shown to promote fusion of synthetic liposomes in the absence of a-SNAP and SNAP receptors, although the required lipid content was critical [41,42] Involvement of PtdIns4P and PtdIns(4,5)P2during Sec18-dependent homotypic vacu-ole fusion in vitro has been reported [43] All these results suggest that NSF and Sec18 possess some abil-ity to bind phospholipids VCP, which promotes homo-typic fusion of Golgi or nuclear membranes [20], in concert with its adaptor protein p47 [28] and⁄ or VCIP135 [29], can also promote the fusion of PE-based liposome vesicles in the absence of p47, with a reduced but nevertheless substantial measurable effi-ciency [58] This suggests that a specific region of VCP binds to PE-based liposomes More recently,

ATP-mPEX1 ND (3-180)

20mM PhA

B

PI5

Lipid content (%)

0 1 2 3 4 5

PI 30

25 20 15 10 5 0

C

(a)

(b)

0 -5

-10

-15

Time (s)

1 2 3 4 5 6 7 8

9 10 11 12 13 14 15 16

Fig 3 The N-terminal domain (ND) of PEX1

is a phosphoinositide-binding domain (A)

Phospholipid binding of PEX1-ND The

puri-fied glutathione-S-transferase

(GST)–PEX1-ND fusion proteins were examined by PIP

Strip assay and detected by anti-GST serum.

The spots correspond to lyso-phosphatidic

acid (PA) (1), lyso-phosphatidylcholine (PC)

(2), phosphatidylinositol (PtdIns) (3),

PtdIns3P (4), PtdIns4P (5), PtdIns5P (6),

phosphatidylethanolamine (PE) (7), PC (8),

sphingosine 1-phosphate (9), PtdIns(3,4)P 2

(10), PtdIns(3,5)P2(11), PtdIns(4,5)P2(12),

PtdIns(3,4,5)P3(13), PA (14), PS (15), blank

(16) The negative control was GST only

(right panel) The middle panel corresponds

to the same experiments conducted in the

presence of 20 mM phytic acid (PhA) (B)

Liposome-binding assay of GST–mPEX1-ND.

Fraction of protein bound to PC ⁄ PE (1 :

1)-based liposomes was plotted against

increasing amounts of PtdIns3P (PI3),

PtdIns4P (PI4), PtdIns5P (PI5) and PtdIns

(PI) in the liposomes, in the absence (filled

bar) or presence (open bar) of 20 mM PhA

as a competitor Solid lines indicate standard

deviation of three experiments (C) Quartz

crystal microbalance (QCM) assay of

lipo-somes, which were added onto GST–

mPEX1-ND immobilized on the QCM

electrode A typical time-dependent drop

in frequency after injection of PC⁄ PE

(1 : 1)-based liposome is shown (a) 5%

PtdIns(4)P (b) PC⁄ PE only.

hydrolysis-deficient mutants of VCP were shown to

accumulate on reticular membranes in vivo [22] On the

other hand, colocalization of VCP and PtdIns-4-kinase

IIa on the buoyant subfraction of ER-derived

mem-brane was observed [66] The ND of VCP, rather than

the D1 or D2 domains, was assumed to act as an

interface for the phospholipid membrane The data

shown here may partly provide the molecular basis of

attachment of VCP to the ER-derived membrane It is

noteworthy that other AAA-ATPase proteins, belong-ing to the SF6 subfamily, possess transmembrane heli-ces at their N-terminus [19]

Sequence similarities in D1–D2 domains as well as the entire domain architectures of PEX1 and VCP sug-gest that PEX1 can form a stable hexameric ring struc-ture via the tandem D1–D2 AAA-ATPase domains such as VCP Although there are few data concerning direct protein–lipid interaction involving AAA-ATP-ases that would prove a common molecular function

of the ND, our results are largely consistent with pre-vious observations The apparent association of PEX1 and VCP with the subcellular membranes could, of course, be attributed not only to protein–lipid interac-tion, but also to protein–protein interactions via other membrane-associated proteins We do not rule out this possibility, as simultaneous binding of a single ND to both a lipid and a ligand may occur [63] In addition, binding of PtdIns to PEX1-ND may also coincide with binding of PtdIns phosphates Protein interface predic-tions based on ODAs [46] (supplementary Fig S2) enabled detection of the known p47 interaction site of VCP-ND, but the corresponding surface in PEX1-ND produced no significant energy values Therefore, PEX1-ND is unlikely to bind the Ubx domain or ubiquitin in a manner similar to that of the VCP-ND– Ubx (p47) complex, consistent with our previous ana-lyses [32] Interestingly, a small patch of low-energy surface was found in a-helix 3 of PEX1-ND, which is highly conserved among its orthologs This area might

be involved in protein–protein interaction As this

Fig 5 Schematic representation of the binding of the N-terminal domain (ND) of PEX1 to the peroxisomal membrane PEX1 is pre-dicted to interact with the PEX6 portion of AAA domains PEX1-ND binds to the peroxisomal membrane by recognizing specific lipids, whereas PEX6 interacts with either PEX26 or PEX15 on the mem-brane.

mPEX1 ND (3-180)

A

Wild-type

B

Lipid content (%)

PI(4)P

30

25

20

15

10

5

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

GST only

mVCP ND (1-200)

Wild-type

C

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

20m M PhA

Fig 4 Phospholipid binding of wild-type and R144A mutant of the

N-terminal domain (ND) of valosine-containing protein (VCP)-ND (A)

Two mutants of the conserved basic residues, R135A and K174A,

were analyzed by PIP Strips assay K174A retained the ability

to bind phosphatidylinositol (PtdIns) monophosphates, whereas

R135A did not (B) The R135 mutant was subjected to the

lipo-some recruitment assay as in Fig 3B The fraction of protein bound

to PC ⁄ PE (1 : 1)-based liposomes containing increasing amounts of

PtdIns4P is plotted Solid lines indicate the standard deviation of

three experiments (C) PIP Strips assay for wild-type VCP-ND and

its R144A mutant, which does not bind PtdIns monophosphates.

ODA area and the putative phospholipid-binding

sur-face do not overlap, simultaneous binding of a single

ND to both a membrane and a protein substrate might

occur We are currently testing this hypothesis

In the present study, we verified that the NDs of

PEX1, VCP, NSF and Ufd1 bind PtdIns Ufd1-ND,

which has structural similarity with PEX1-ND, is the

only known example of a protein with an ND and no

AAA-ATPase domain Furthermore, there are several

reports indicating that other AAA-ATPases bind

phos-phoinositides (e.g in the ER and Golgi) [43,64,65,68]

Examples include phosphoinositide binding of SKD1,

which is a type I AAA-ATPase, whose ND (MIT

domain) is structurally different from that of PEX1

[68,70] The binding mode and binding, however,

remain unknown and may differ from those of PEX1

Experimental procedures

Expression and mutational study of PEX1-ND

The construction of a GST fusion protein containing the

mouse PEX1 gene (encoding residues 3–180) in

pGEX-4T3-PRESAT, according to the PRESAT vector methodology

[59], has been previously described [32] Ala-substituted

mutants R135A and K174A were prepared with a

Gene-Editor site-directed mutagenesis system (Promega, Madison,

WI, USA), according to the manufacturer’s instructions

The coding regions were sequenced after introduction of

the mutations Mouse PEX1-ND and their mutants were

expressed as GST fusion proteins in LB medium containing

1% glucose (LBG) containing ampicillin (50 lgÆmL)1) The

cells were grown to a D600 of 0.3, and heterologous gene

expression was induced by addition of 1 mm isopropyl

thio-b-d-galactoside (IPTG) The cells were collected

12–16 h after IPTG induction, washed, and disrupted by

sonication Recombinant proteins were purified by a

single chromatography step using glutathione Sepharose

(Amersham Bioscience, Uppsala, Sweden) The GST fusion

proteins were dialyzed against buffer containing 150 mm

KCl and 50 mm Hepes (pH 7.5) prior to the

phospholipid-binding assays

Expression and mutational study of VCP-ND

The plasmids for GST fusion proteins containing the mouse

VCP (residues 1–200) were constructed by a standard

pro-tocol using PCR, and subcloned into pGEX-4T3

(Amer-sham Bioscience) Ala-substituted mutants R144A were

prepared with a Gene-Editor site-directed mutagenesis

sys-tem (Promega), according to the manufacturer’s

instruc-tions The coding regions were sequenced after introduction

of the mutations Mouse VCP and mutant were expressed

as GST fusion proteins in LBG containing ampicillin

(50 lgÆmL)1) The cells were grown to a D600 of 0.5, and heterologous gene expression was induced by addition of

1 mm IPTG The cells were collected after 3–5 h of IPTG induction, pelleted, washed, and disrupted by sonication Recombinant proteins were purified by a single chromato-graphy step using glutathione Sepharose The GST fusion proteins of the NDs were dialyzed against buffer containing

150 mm KCl and 50 mm Hepes (pH 7.5) prior to the phospholipid-binding assays

Expression of other NDs of NSF, CDC48 and Ufd1

A GST fusion expression construct containing the mouse NSF gene (encoding residues 5–200) was constructed in a manner similar to that for PEX1-ND The plasmids for GST fusion proteins containing yeast CDC48 (residues 1–210), mouse Ufd1 (residues 14–194) and yeast Ufd1 (resi-dues 1–210) were constructed by a standard PCR protocol, and subcloned into pGEX-4T3 (Amersham Bioscience) The cells were grown to a D600 of 0.5, and heterologous gene expression was induced by addition of 1 mm IPTG The cells were collected after 3–5 h of IPTG induction, pel-leted, washed, and disrupted by sonication Recombinant proteins were purified by a single chromatography step using glutathione Sepharose The GST fusion proteins of the NDs, except mouse⁄ yeast Ufd1, were dialyzed against buffer containing 150 mm KCl and 50 mm Hepes (pH 7.5) prior to the phospholipid-binding assays Mouse and yeast Ufd1 were dialyzed against buffer containing 85 mm KCl,

50 mm Hepes (pH 7.5) and 10% glycerol

Phospholipid-binding assays For assessment of phospholipid-binding properties, PIP Strips (Echelon Bioscience Inc., Salt Lake City, UT, USA) were blocked with binding buffer containing 150 mm NaCl,

10 mm Hepes (pH 7.4), supplemented with 3% fatty acid-free BSA for 1 h at room temperature The strips were then incubated with purified GST fusion proteins at a concentra-tion of 300 lgÆmL)1in blocking buffer at room temperature for 3 h After three washes in the binding buffer, PIP Strips were incubated for 3 h at room temperature with anti-GST (Nacalai Tesque, Kyoto, Japan) serum in the same buffer Secondary antibody incubation and 3,3¢,5,5¢-tetra-methylbenzidine staining were performed to detect GST-tagged proteins bound to the phospholipid spots on the membrane

Liposome recruitment assay Liposome recruitment reactions were performed in 150 mm KCl and 50 mm Hepes (pH 7.5) Liposomes were made from a 1 : 1 mixture of PC and PE (Sigma Aldrich, Tokyo, Japan) in the presence or absence of PtdIns3P, PtdIns4P,

PtdIns5P, or PtdIns (5% weight ratio to the base liposome;

Sigma Aldrich) Purified GST fusion proteins (50 lgÆmL)1)

were incubated with the liposomes (1 mg lipidÆmL)1) After

5 min of incubation at room temperature, membranes were

recovered by centrifugation at 35 000 g at 4C with a

Beckman TL-100 ultracentrifuge, rotor type TLA-100

(Beckman Coulter, Fullerton, CA, USA) The reaction was

carried out on a 50 lL scale Supernatants and pellets were

resuspended in SDS sample buffer and analyzed by

SDS⁄ PAGE, followed by Coomassie brilliant blue staining

The gel images were analyzed using the imagej 1.31v

soft-ware (http://rsb.info.nih.gov/ij/)

QCM assay of in vitro protein–lipid interaction

The in vitro liposome binding of a mutant of PEX1-ND

was determined by frequency change in a 27-MHz QCM

using an AFFINIX-Q (QCM2000) instrument (Initiam Co.,

Tokyo, Japan) A GST fusion protein of wild-type

PEX1-ND was immobilized on a QCM gold electrode (diameter

4.5 mm, QCMST27) according to the manufacturer’s

instructions A 2 lL aliquot of a 50 lgÆmL)1 solution of

GST–PEX1-ND was used The electrode was washed

sev-eral times in buffer containing 150 mm KCl and 50 mm

Hepes (pH 7.45), soaked in the same buffer (2 mL cuvette),

and then monitored continuously for QCM frequency

chan-ges at 25C Once the frequency had stabilized, 8 lL of

the PC + PE (1 : 1) liposomes (1 mgÆmL)1) with or

with-out PtdIns4P (5% weight ratio to the base liposome) were

injected

Evolutionary trace analysis

The four sequences of PEX1-ND orthologs and six

sequences of VCP-ND orthologs were obtained from NCBI;

they are PEX1_HUMAN (human PEX1; AAB99758),

PEX1_MOUSE (Mus musculus PEX1; XML131895),

PEX1_ARATH (Arabidopsis thaliana PEX1; AAG44817),

PEX1_YEAST (yeast PEX1; CAA82041), VCP_MOUSE

(Mus musculus VCP; NP_033529), VCP_HUMAN (human

CAH70993), VCP_XENOPUS (Xenopus VCP; AAH74716),

VCP_ZEBRAFISH (zebrafish VCP; NP_958889), CDC48_

YEAST (yeast cdc48; CAA98694) and VAT_THEAC

(Thermoplasma acidophilum VAT; AAC45089) The

multiple sequence alignment was produced using the

pro-gram clustalx [60] and then refined manually The

evolu-tionary trace analysis [45] was then carried out on the

Evolutionary Trace Server (tracesuite II) [46] Here, the

evolutionary trace method defined groups of proteins using

an evolutionary time cut-off in a phylogenetic tree, and

divided all residues of aligned sequences into three classes:

conserved (residues invariant throughout the groups),

class-specific (residues invariant within each group but that

change between groups), and neutral (all others)

(supple-mentary Fig S1) Gaps were counted as an extra residue

type The class-specific residues are the most interesting in terms of the development of functional innovation Trace residues were mapped onto the structure by pymol [61] The surfaces of PEX1-ND and VCP-ND were further subjected

to ODA analysis by using program pydock (J Fernandez-Recio, unpublished results)

Acknowledgements This work was supported by grants to MS and HH from the Japanese Ministry of Education, Science, Sports and Culture (Protein3000) We thank Dr Y Fujiki for a vector containing PEX1 cDNA We thank

Dr M H J Koch and Dr W Schliebs for many valu-able discussions and for critical reading of the manu-script

References

1 van den Bosch H, Schutgens RB, Wanders RJ & Tager

JM (1992) Biochemistry of peroxisomes Annu Rev Biochem 61, 157–197

2 Eckert JH & Erdmann R (2003) Peroxisome biogenesis Rev Physiol Biochem Pharmacol 147, 75–121

3 Wanders RJ & Waterham HR (2005) Peroxisomal dis-orders I: biochemistry and genetics of peroxisome bio-genesis disorders Clin Genet 67, 107–133

4 Imamura A, Tamura S, Shimozawa N, Suzuki Y, Zhang Z, Tsukamoto T, Orii T, Kondo N, Osumi T & Fujiki Y (1998) Temperature-sensitive mutation in PEX1 moderates the phenotypes of peroxisome defi-ciency disorders Hum Mol Genet 7, 2089–2094

5 Geisbrecht BV, Collins CS, Reuber BE & Gould SJ (1998) Disruption of a PEX1–PEX6 interaction is the most common cause of the neurologic disorders Zellwe-ger syndrome, neonatal adrenoleukodystrophy, and infantile Refsum disease Proc Natl Acad Sci USA 95, 8630–8635

6 Tamura S, Okumoto K, Toyama R, Shimozawa N, Tsukamoto T, Suzuki Y, Osumi T, Kondo N & Fujiki Y (1998) Human PEX1 cloned by functional

com-plementation on a CHO cell mutant is responsible for peroxisome-deficient Zellweger syndrome of

complementation group I Proc Natl Acad Sci USA 95, 4350–4355

7 Reuber BE, Germain-Lee E, Collins CS, Morrell JC, Ameritunga R, Moser HW, Valle D & Gould SJ (1997) Mutations in PEX1 are the most common cause of per-oxisome biogenesis disorders Nat Genet 17, 445–448

8 Dodt G & Gould SJ (1996) Multiple PEX genes are required for proper subcellular distribution and stability

of Pex5p, the PTS1 receptor: evidence that PTS1 protein import is mediated by a cycling receptor J Cell Biol

135, 1763–1774