As Vps20p is a coiled-coil protein and interacts with the N-terminal MIT domain of Vps4p, Vps4p ATPase activity may break coiled-coil interactions and thereby disassemble the MVB sorting
Trang 1multivesicular endosome
Parimala R Vajjhala1, Elizabeth Catchpoole1, Chau H Nguyen1, Carol Kistler1and Alan L Munn1,2
1 Institute for Molecular Bioscience and ARC Special Research Centre for Functional and Applied Genomics, University of Queensland,
St Lucia, QLD, Australia
2 School of Biomedical Sciences, University of Queensland, St Lucia, QLD, Australia
During endocytic trafficking, some integral membrane
proteins are sorted into internal vesicles which form by
invagination of the endosome limiting membrane This
process, referred to as multivesicular body (MVB)
sort-ing, is critical for a number of important biological
processes including receptor down-regulation, antigen
presentation and exosome-dependent intercellular
sig-nalling (reviewed in [1–3]) Interest in the mechanism
of MVB sorting has escalated since the discovery that
components of the MVB sorting machinery are also
utilized for virus budding, a process topologically sim-ilar to MVB sorting (reviewed in [4,5])
The process of MVB sorting was first examined in mammalian cells [6,7], but the components of the MVB sorting machinery were first characterized in Saccharomyces cerevisiae[8–10] The pathway is highly conserved from yeast to mammalian cells, although the number of components is expanded in mammalian cells because of the multiplicity of isoforms [11] Recognition of a cargo protein, usually by the presence
Keywords
endocytosis; lysosome; macromolecular
disassembly; membrane traffic; vacuole
Correspondence
A L Munn, Institute for Molecular
Bioscience, University of Queensland,
St Lucia, Brisbane, QLD 4072, Australia
Fax: +61 73346 2101
Tel: +61 73346 2017
E-mail: A.Munn@imb.uq.edu.au
(Received 17 December 2006, revised 6
February 2007, accepted 9 February 2007)
doi:10.1111/j.1742-4658.2007.05736.x
During endocytic transport, specific integral membrane proteins are sorted into intraluminal vesicles that bud from the limiting membrane of the endosome This process, known as multivesicular body (MVB) sorting, is important for several important biological processes Moreover, compo-nents of the MVB sorting machinery are implicated in virus budding Dur-ing MVB sortDur-ing, a cargo protein recruits components of the MVB sortDur-ing machinery from cytoplasmic pools and these sequentially assemble on the endosome Disassembly of these proteins and recycling into the cytoplasm
is critical for MVB sorting Vacuolar protein sorting 4 (Vps4) is an AAA (ATPase associated with a variety of cellular activities) ATPase which has been proposed to play a critical role in disassembly of the MVB sorting machinery However, the mechanism by which it disassembles the complex
is not clear Vps4 contains an N-terminal microtubule interacting and traf-ficking (MIT) domain, which has previously been shown to be required for recruitment to endosomes, and a single AAA ATPase domain, the activity
of which is required for Vps4 function In this study we have systematically characterized the interaction of Vps4 with other components of the MVB sorting machinery We demonstrate that Vps4 interacts directly with Vps2 and Bro1 We also show that a subset of Vps4 interactions is regulated by ATP hydrolysis, and one interaction is regulated by ATP binding Finally,
we show that most proteins interact with the Vps4 MIT domain Our stud-ies indicate that the MIT domain has a dual role in substrate binding and recruitment to endosomes and indicate that Vps4 disassembles the MVB sorting machinery by direct effects on multiple proteins
Abbreviations
GFP, green fluorescent protein; GST, glutathione S-transferase; HRP, horseradish peroxidase; MIT, microtubule interacting and trafficking; MVB, multivesicular body; PVDF, poly(vinylidene difluoride); Vps, vacuolar protein sorting.
Trang 2of a ubiquitin molecule, is followed by sequential
recruitment of components of the MVB sorting
machinery The order of recruitment of the different
components to the endosome membrane is starting to
emerge, and structural data have recently been
obtained for several components (reviewed in [12]) A
critical process during MVB sorting is the disassembly
of the MVB sorting machinery, which allows recycling
and new rounds of vesicle budding However, the
molecular mechanisms that regulate cycling of the
MVB sorting machinery on and off endosomes is not
yet well understood
Vps4p, also known as Csc1p, End13p, Grd13p,
Vpl4p, Vpt10p, or Did6p, is the only essential
compo-nent of the MVB sorting machinery with known
enzy-matic activity It is a member of the AAA (ATPase
associated with a variety of cellular activities) family of
ATPases [13,14] Mammalian cells express two
iso-forms of VPS4, VPS4A and VPS4B, and both proteins
function in endocytic trafficking [15–17] All members
of the AAA superfamily contain one or two copies of
a conserved ATPase domain (AAA module) Although
not known for Vps4p, other AAA ATPases assemble
into oligomeric rings Distinct members of the AAA
ATPase family function in diverse cellular processes,
but a common theme is protein unfolding and
macro-molecular disassembly (reviewed in [18,19])
Loss of Vps4p function in yeast and mammalian
cells disrupts MVB sorting and results in the formation
of an aberrant multilamellar endosomal compartment
referred to as the class E compartment [8,14] As both
endocytic and biosynthetic traffic to the lysosome⁄
vacuole proceeds via the MVB, the class E
compart-ment accumulates endocytic and biosynthetic material
as well as late Golgi proteins because of defective
traf-ficking out of this compartment [9] In yeast, defective
recycling of late Golgi proteins including the receptor
that transports soluble vacuolar proteins from the
Golgi to the MVB results in missorting and secretion
of soluble vacuolar proteins to the extracellular medium
[20,21]
Loss of Vps4p function is also accompanied by the
redistribution of several components of the MVB
sort-ing machinery from the cytoplasm to endosomes [22,23]
Therefore Vps4p-dependent ATPase activity has been
proposed to be important for disassembly of the MVB
sorting machinery and release into the cytoplasm An
N-terminal microtubule interacting and trafficking
(MIT) domain is required for recruitment of Vps4p to
endosomes [22,24,25], but it is not clear precisely how
Vps4p disassembles the MVB sorting machinery
Previ-ous work from our laboratory has shown that Vps20p, a
component of the MVB sorting machinery, interacts
with Vps4p and dissociates from Vps4p upon ATP hydrolysis [26] This was the first evidence that Vps4p ATPase activity can break intermolecular interactions
As Vps20p is a coiled-coil protein and interacts with the N-terminal MIT domain of Vps4p, Vps4p ATPase activity may break coiled-coil interactions and thereby disassemble the MVB sorting machinery
Several putative interactions have been reported between Vps4p and components of the MVB sorting machinery [27–30] However, as there is a complex net-work of interactions between the components of the MVB sorting machinery, it was not clear which inter-actions with Vps4p are direct and which are indirect
In addition, it was not clear how many of these puta-tive Vps4p interactions with the MVB sorting machin-ery may be regulated by Vps4p ATPase activity That not all Vps4p interactions are regulated by ATPase activity is supported by our previous studies showing that the Vps4p interaction with Vta1p is not affected
by ATP hydrolysis [26]
Here, we demonstrate new direct interactions between Vps4p and the MVB sorting machinery in yeast We present evidence that a subset of Vps4p interactions is regulated by ATP hydrolysis, and that one interaction is regulated by ATP binding to Vps4p Finally, we also show that most Vps4p interactors interact with the MIT domain of Vps4p
Results
Vps4p binds directly to multiple components
of the MVB sorting machinery Vps4p has been shown to bind directly to a few MVB sorting machinery components including Vps20p, Vta1p and Did2p⁄ Chm1p [26,31] However, it is not known whether the function of Vps4p in disassembly
of the MVB sorting machinery is mediated solely via interactions with these proteins or whether Vps4p binds directly to and acts via other components of the MVB sorting machinery We therefore tested candidate Vps4p interactions with components of the MVB sort-ing machinery When bindsort-ing of purified Vps4p to MVB components was examined in vitro, Vps4p was found to bind directly to multiple MVB sorting machinery components, including Vps2p and Bro1p, in addition to Vps20p, Vta1p and Did2p, but not to Snf7p (Fig 1) Our data also show that, when molar equivalent amounts of Vps4p interactors are com-pared, the amount of Vps4p that binds to Did2p is greater than that bound by any of the other interac-tors These data are consistent with Vps4p having a relatively high affinity for Did2p
Trang 3ATP binding and ATP hydrolysis by Vps4p
regulates protein interactions at the MVB
Vps4p has been proposed to function as a protein
complex disaggregation machine on endosomes [22]
Consistent with this, the binding of Vps20p to Vps4p
is regulated by ATP hydrolysis [26] It was likely that
other components of the MVB sorting machinery are
also substrates for disassembly by Vps4p To test this,
we performed the in vitro binding assay in the presence
and absence of ATP In the presence of ATP, purified
wild-type Vps4 is catalytically active and will hydrolyse
added ATP Thus interactions that are regulated by
Vps4p ATPase activity are predicted to decrease under
these conditions In contrast, interactions that are not
regulated by Vps4p ATPase activity are predicted to
be unaffected The data obtained show that binding of
both Vps2p and Bro1p to Vps4p was decreased in the presence of ATP compared with binding in the absence
of ATP However, the binding of Did2p to Vps4p was not affected by the presence of ATP (Fig 2)
To determine whether the decreased binding in the presence of ATP is due to ATP hydrolysis or ATP binding, the effect of ATP on binding to a Vps4p mutant (Vps4p–E233Q) that is defective in ATP hydrolysis was also studied Binding of Vps2p and Vps20p to Vps4p–E233Q was increased in the presence
of ATP, but binding of Bro1p to Vps4p–E233Q was decreased in the presence of ATP (Fig 2)
We surmise that the decreased binding of Vps2p to Vps4p in the presence of ATP is due to Vps4p-depend-ent ATP hydrolysis In contrast, the decreased binding
of Bro1p to wild-type Vps4p and to Vps4p–E233Q in the presence of ATP may be due to competitive bind-ing or an allosteric effect
Several components of the MVB sorting machinery interact with Vps4p via the N-terminal MIT domain
To determine whether there is any correlation between the binding sites on Vps4p and the response of the interacting proteins to ATP binding and hydrolysis, we determined the region of Vps4p that mediates inter-action with Vps2p, Snf7p and Bro1p using the yeast two-hybrid technique (Fig 3) Both Vps2p and Snf7p interact with Vps4p mainly via the Vps4p N-terminal coiled-coil domain (Fig 3C) We did not detect an interaction between Bro1p and full-length Vps4p or any of the Vps4p domains using the yeast two-hybrid technique (not shown) consistent with a previous report [27]
To more precisely map the interaction sites of Vps2p, Snf7p, Vps20p, and Did2p within the Vps4p N-terminal domain, we generated two Vps4p N-ter-minal mutants (Fig 4A,B) and tested the effect of
Fig 1 All of the GST-tagged Vps4p-interacting proteins that
func-tion during MVB sorting except Snf7p can bind directly to Vps4p.
An equal amount of 6His-tagged Vps4p was incubated with
gluta-thione–agarose bearing GST–Snf7p, GST–Vps20p, GST–Bro1p,
GST–Did2p, GST–Vps2p, GST–Vta1p or GST alone Unbound
pro-tein was recovered in the supernatants Bound propro-tein was
released with Laemmli sample buffer The bound and unbound
frac-tions were subjected to SDS ⁄ PAGE and immunoblotting with a
polyclonal antibody to Vps4p The shift in the position of
Vps4p-6His bound to GST–Vps20p is due to the presence of the GST–
Vps20p, which migrates very close to Vps4p-6His The data shown
are representative of at least two independent experiments.
Fig 2 ATP binding and ATP hydrolysis by Vps4p regulates interaction with Bro1p, Vps2p and Did2p The 6His-tagged wild-type Vps4p or Vps4p–E233Q was incubated with glutathione–agarose bearing GST–Vps2p, GST–Bro1p, GST–Did2p or GST–Vps20p in the presence or absence of ATP The bound protein was released with Laemmli sample buffer and subjected to Western blotting using a polyclonal antibody
to Vps4p The data shown are representative of at least two independent experiments performed in duplicate.
Trang 4these mutations on the different Vps4p interactions In
the Vps4p–YEE mutant, residues 26–29 in the second
helix of the MIT domain were deleted These residues
are completely conserved between the yeast and human
VPS4 isoforms In the Vps4p–IRA mutant, residues
56–71 in the third helix of the MIT domain were
dele-ted These residues are also highly conserved The
YEE and IRA mutants are named after the first three
amino acids that were deleted in each motif We also
tested the effect of a previously described
Vps4p-coiled-coil (CC) mutation [22] in which residues 50–87,
which comprises most of the second and third helices
of the MIT domain, were deleted
Yeast two-hybrid analysis (Fig 4C) revealed that
each mutation diminished but did not abolish Did2p
interaction, whereas Vps2p interaction was abolished
by the Vps4p–IRA and Vps4p-CC mutations All the
N-terminal mutations tested abolished Vps20p and Snf7p interactions As expected, the interaction of Vta1p with the Vps4p C-terminal domain was not diminished by any of the N-terminal mutants we tested
To confirm the yeast two-hybrid interactions des-cribed above and to identify the domain of Vps4p to which Bro1p binds, we performed in vitro protein-binding assays (Fig 4D) The data obtained indicate that the Vps4p–YEE and Vps4p–IRA mutations diminish direct binding of Vps2p, Vps20p and Did2p The Vps4p-CC mutation appeared to increase binding
to all interactors However, the Vps4p-CC mutant pro-tein also displayed an interaction with glutathione S-transferase (GST) alone (Fig 4D) and displayed an increased interaction with the Vps4p antibody (not shown) Thus, we cannot interpret the data obtained for this Vps4p-CC mutant protein Mutation of the
b domain abolished interaction with Vta1p as previ-ously reported [32,33] and in addition increased bind-ing to the N-terminal interactors, includbind-ing Vps2p, Vps20p and Did2p None of the N-terminal MIT domain mutations or the C-terminal b domain muta-tions diminished binding of Vps4p to Bro1p
In summary, our data indicate that most Vps4p-interacting proteins (Vps2p, Vps20p, Did2p and Snf7p) interact with the N-terminal MIT domain of Vps4p In addition, these data show that the Bro1p interaction with Vps4p is unique as it is undiminished by muta-tions in the N-terminal MIT domain or b domain
Vps4p interactions with Did2p, Vps2p and Vps20p are important for recruitment to endosomes and for MVB sorting
To determine whether the Vps4p YEE and IRA motifs play a role in Vps4p recruitment to endosomes, wild-type and mutant Vps4p tagged with green fluorescent protein (GFP) were expressed in cells in which the chromosomal VPS4 gene has been deleted (vps4D) Wild-type GFP-tagged Vps4p could be detected on punctate structures (Fig 5A) consistent with localiza-tion to endosomes, as previously reported [22] How-ever, the GFP-tagged Vps4p–YEE and Vps4p–IRA mutant proteins, like the GFP-tagged Vps4p-CC mutant protein, exhibited severely reduced punctate localization We conclude that the Vps4p YEE and IRA motifs are important for Vps4p recruitment to endosomes
To test whether the YEE and IRA motifs and the region deleted in the Vps4p-CC mutant protein are important for Vps4p function in vivo, we tested the ability of the N-terminal mutant proteins to restore
A
B
C
Fig 3 Vps2p and Snf7p interact with the N-terminal domain of
Vps4p (A) Schematic representation of Vps4p with the domain
organization inferred from structural data from mammalian VPS4A
and VPS4B (B) Constructs used for mapping the region of Vps4p
that mediates interaction (C) Yeast two-hybrid interaction analysis
of Vps2p and Snf7p with full-length wild-type Vps4p (Vps4p-full),
the N-terminal region of Vps4p (Vps4p-N), the previously predicted
AAA domain (Vps4p-AAA), and the C-terminal region (Vps4p-C).
EGY48 carrying pLexA-based bait plasmids and pB42AD-based prey
plasmids as well as the p8OpLacZ reporter plasmid were spotted
on to medium containing X-gal Plates were photographed after
overnight incubation, and two-hybrid interaction was assessed by
blue coloration Three independent transformants are shown.
Trang 5MVB sorting and delivery of a soluble vacuolar
pro-tein to the vacuole (vacuolar propro-tein sorting) in vps4D
yeast As a marker for MVB sorting, we used a
GFP-and ubiquitin-tagged form of the iron transporter
homologue, Fth1p (GFP-Fth1p-Ub), which is known
to undergo MVB sorting into the vacuole lumen [34]
In vps4D yeast expressing wild-type Vps4p, the
GFP-Fth1p-Ub undergoes MVB sorting and is transported
to the vacuole lumen (Fig 5B) However, in vps4D
yeast expressing the Vps4p–YEE, Vps4p–IRA or
Vps4p-CC mutant proteins, MVB sorting of GFP-Fth1p-Ub was not significantly improved compared with vps4D cells carrying empty vector To assess the ability of the Vps4p mutant proteins to restore trans-port of a soluble vacuolar protein to the vacuole in vps4D yeast, we tested their ability to correct the mis-sorting and secretion of a soluble vacuolar protein, carboxypeptidase Y Expression of wild-type Vps4p, but not Vps4p–YEE, Vps4p–IRA or Vps4p-CC mutant proteins restored vacuolar transport of
carboxy-B
C
D A
Fig 4 Interaction with Vps2p, Snf7p, Vps20p and Did2p, but not with Bro1p, is diminished by mutation of conserved residues in the Vps4p MIT domain (A) Alignment of human VPS4A, VPS4B and S cerevisiae (Sc) Vps4p sequences using CLUSTAL W [50] The conserved YEE and IRA motifs that were deleted are shown in bold, and the previously described coiled-coil (CC) mutation is shown underlined (B) Location in the VPS4A MIT domain of the conserved YEE and IRA motifs and the region deleted in the previously described Vps4p-CC mutant (C) Yeast two-hybrid interaction analysis of wild-type (WT) Vps4p and Vps4p N-terminal mutants with Vps2p, Snf7p, Vps20p, Did2p, and Vta1p Inter-action analysis was performed as described in the legend to Fig 3 Three independent transformants are shown (D) In vitro binding of 6His-tagged wild-type Vps4p and Vps4p N-terminal mutants to GST-6His-tagged Vps2p, Vps20p, Did2p, Bro1p, and Vta1p, or to GST alone Equal amounts of full-length 6His-tagged proteins were incubated with glutathione–agarose beads bearing the different GST fusion proteins or GST alone Bound protein was released with Laemmli sample buffer and subjected to Western blotting using a polyclonal antibody to Vps4p The data shown are representative of two independent experiments A Western blot of the different 6His-tagged proteins used for the in vitro binding assay (5% input) is also shown.
Trang 6peptidase Y compared with vps4D cells transformed
with empty vector alone (Fig 5C) We conclude that
the YEE and IRA motifs as well as the region deleted
in the Vps4p-CC mutant are important for Vps4p
function in MVB sorting and vacuolar protein sorting
To test whether the phenotypes of vps4D cells
expressing the Vps4p N-terminal mutants were due to
lowered expression or degradation of the mutant
pro-teins, we tested the steady-state expression of the
mutant Vps4p proteins (Fig 5D) Although both the Vps4p–IRA and Vps4p–YEE mutants are expressed, their steady-state connections are somewhat reduced compared with that of wild-type Vps4p However, such
a modest reduction in expression level is unlikely to account for the inability of these mutant proteins to restore MVB sorting and vacuolar protein sorting
in vps4D yeast Surprisingly, the expression of the Vps4p-CC mutant was significantly greater than that
C
D
Fig 5 An intact MIT domain is required for
Vps4p localization to endosomes and for
Vps4p in vivo function (A) AMY245 vps4D
yeast cells expressing GFP-tagged wild-type
(WT) Vps4p, Vps4p-CC, Vps4p–YEE, and
Vps4p–IRA or carrying empty vector
(YCplac111) were grown in SD medium and
the GFP-tagged proteins visualized by
fluores-cence microscopy The same fields of cells
are shown visualized by fluorescence (right)
and Nomarski (left) optics Scale bar, 5 lm.
(B) Ubiquitin-dependent MVB sorting of
Fth1p-GFP-Ub in AMY245 (vps4D) yeast
cells carrying plasmids expressing wild-type
Vps4p or Vps4p mutant proteins or carrying
empty vector (YCplac111) Cells were
incu-bated in SD medium containing 100 l M
bathophenanthrolinedisulfonic acid for 6 h to
chelate iron and induce Fth1p-GFP-Ub
expression Cells were then washed with
buffer containing 1% sodium azide, 1%
sodium fluoride, and 100 m M phosphate,
pH 8.0, to stop further transport The same
fields of cells are shown visualized by
fluor-escence (right) and Nomarski (left) optics.
Scale bar, 5 lm (C) Vacuolar protein sorting
in AMY245 (vps4D) yeast cells carrying
plas-mids expressing wild-type Vps4p or Vps4p
mutant proteins or carrying empty vector
(YCplac111) or no vector Cells were grown
on YPUAD solid medium for 2 days at
24 C in contact with a nitrocellulose filter.
RH1800 (wild-type) yeast cells without any
vector (boxed in both panels) was included
as a control Cells were eluted from the
filter, and carboxypeptidase Y on the filter
was detected by immunoblotting with
anti-carboxypeptidase Y serum To test for cell
lysis, the blot was stripped and re-probed
with an antibody to a cytoplasmic protein
(calmodulin) (D) Total cell lysates from
AMY245 vps4D yeast cells expressing
wild-type Vps4p, Vps4p-CC, Vps4p–YEE,
and Vps4p–IRA or carrying empty vector
(YCplac111) were subjected to Western
blotting using a polyclonal antibody to Vps4p
as well as an antibody to actin.
Trang 7of wild-type Vps4p We conclude that the
pheno-types observed in vps4D cells expressing the Vps4p
N-terminal mutants are due to loss of function of
these mutant proteins
We surmise that the interactions of the Vps4p MIT
domain with Did2p, Vps2p, Vps20p and Snf7p are
critical for Vps4p recruitment to endosomes and Vps4p
function in MVB sorting
Discussion
Here, we show that Vps4p has the ability to interact
directly with multiple components of the MVB sorting
machinery A number of these interactions are
medi-ated by the MIT domain of Vps4p, and a subset are
regulated by Vps4p-dependent ATP hydrolysis
Interes-tingly, however, two interactions had unique features
The interaction of Bro1 with Vps4p is regulated by
ATP binding rather than hydrolysis, and interaction
of Did2p with Vps4p is regulated by neither ATP
binding nor ATP hydrolysis Our data highlight the
fact that the role of Vps4p in MVB sorting is more
complex than previously assumed As there exists
mammalian orthologues of these MVB sorting
pro-teins, our findings are likely to have relevance to VPS4
function in mammalian cells
An indication that Vps4p may interact with Vps2p
and Bro1p came from previous studies [28,35]
How-ever, a network of interactions connects components
of the MVB sorting machinery, therefore it was not
clear whether these putative Vps4p interactions were
direct or indirect Ours is the first study to show that
the interactions with Vps2p and Bro1p are direct
Sev-eral lines of evidence suggest that the interactions we
have characterized are physiologically important
Firstly, the Vps4p interactions studied here are all with
proteins known to function in MVB sorting Secondly,
a subset of the interactions are regulated by ATP
bind-ing or ATP hydrolysis by Vps4p Finally, direct in vivo
evidence that these interactions are important comes
from our phenotypic analysis of vps4 mutants in which
some of these interactions are abrogated We did not
detect a direct interaction between Vps4p and Snf7p,
consistent with a previous study [36] However, while
this paper was in preparation, a report [37] was
pub-lished demonstrating a direct interaction using
differ-ently tagged constructs
Our findings in this study allow us to classify Vps4p
interactions into three types; those regulated by ATP
binding, those that are regulated by ATP hydrolysis,
and those that are not regulated by either These
dif-ferent types of interaction may contribute to Vps4p
function in different ways The interactors may be
important for recruitment of Vps4p to endosomes, they may be substrates acted upon by Vps4p during disassembly of the MVB sorting machinery, or their function at the MVB may be regulated by Vps4p For example, the high affinity of Did2p for Vps4p coupled with the fact that this interaction is not regulated by ATP binding or hydrolysis supports a role for Did2p
in recruitment of Vps4p to endosome membranes This
is consistent with previous studies showing that Vps4p
is efficiently recruited to endosomes in both its nucleo-tide-bound [22] and nucleotide-free [38] states In con-trast, Vps2p and Vps20p are likely to be substrates of Vps4p during disassembly of the MVB sorting machin-ery This process is known to require cycling of Vps4p between an ATP-bound and nucleotide-free state Finally, although Bro1p does not appear to be a sub-strate, it might be regulated by Vps4p, as it is dis-placed by ATP binding to Vps4p
The MIT domain of Vps4p appears to play a dual role in endosome localization and substrate binding Here, we have shown that motifs in the MIT domain that are highly conserved between yeast Vps4p and mammalian VPS4 isoforms are required for endosome localization These data are consistent with a previous study showing that deletion of a larger region, which includes most of the second and third helices of the MIT domain, prevents Vps4p recruitment to endo-somes Moreover, our interaction studies have shown that these conserved motifs are also important for interaction with Vps2p, Vps20p, Did2p and Snf7, indicating that some or all of these interactions may
be required for efficient targeting of Vps4p to the endosome Each of these proteins has been proposed
in various studies to be important for Vps4p targeting
to endosomes [23,26,39,40] Although the Vps4p N-terminal mutant proteins all retain interaction with Bro1p and Vta1p, these are clearly not sufficient for Vps4p recruitment to endosomes As the binding
of Vps2p and Vps20p, which both bind to the MIT domain, is regulated by ATP hydrolysis, our data suggest that the MIT domain may also function as a substrate-binding site Thus our in vivo studies with the Vps4p mutant proteins suggest that loss of Vps4p function in MVB sorting and vacuolar protein sorting may be due to both inefficient recruitment to endo-somes and interaction with substrates
Our studies together with data from a previous study [25] indicate that four MVB sorting proteins, Vps2p, Vps20p, Snf7p and Did2p, all interact with the MIT domain of Vps4p, suggesting that these proteins may have a common motif or common fold Intrigu-ingly, when we searched for a common motif in these Vps4p-interacting proteins, we identified a motif,
Trang 8VDELMD, in Vps20p that is highly conserved in
Did2p (VDELMS) and Snf7p (VDETMD) and might
be present in a degenerate form in Vps2p (ADEIVN)
Although the precise binding sites are not known,
these motifs fall within the identified regions of Did2p
and Vps20p that bind to Vps4p In contrast to these
interactors, Bro1p does not contain this motif and
does not appear to bind to the Vps4p MIT domain or
b domain, as mutations in these domains did not
diminish interaction with Bro1p Although this is
intriguing, the motif may be shared by these proteins
because it has some other function important for
MVB sorting
The N-terminal and C-terminal Vps4p domains have
been thought to function independently in Vps4p
endosome recruitment and assembly This is because
the N-terminal Vps4p-CC mutant, which is not
recrui-ted to endosomes, assembles into an oligomer with
ATPase activity [22] Also, mutation of the b domain
of Vps4p abrogates Vps4p oligomer assembly, but does
not affect Vps4p recruitment to endosomes [33]
How-ever, our data indicate that mutations in the Vps4p
MIT domain strengthen Vps4p interactions mediated
by the b domain In addition, mutation of the b
do-main strengthens interactions with the MIT dodo-main
Thus our data offer the first evidence that there may
be functional interactions between the N-terminal and
C-terminal Vps4p domains
Our finding that the interaction of Bro1p with
Vps4p is regulated by ATP binding alone is intriguing
Ubiquitinated cargo proteins are sorted and then
de-ubiquinated before their incorporation into an
intralu-minal vesicle This prevents degradation of ubiquitin
along with the cargo protein [41] Bro1p recruits the
de-ubiquitinating enzyme Doa4p to the MVB [42]
Dis-placement of Bro1p by ATP binding to Vps4p may
allow removal of the de-ubiquitinating machinery
immediately before closure of the intraluminal vesicle
and disassembly of the MVB sorting machinery This
will be an interesting topic for a future investigation
An unexpected observation made during the course
of our study indicated that Vps4p may be subjected to
ubiquitin-mediated degradation Deletion of most of
helices 2 and 3 of the MIT domain led to a
consider-able increase in expression of the mutant protein
com-pared with deletion of smaller regions within or
adjacent to this large deletion A comparison of the
sequences deleted in the various mutants suggests that
this increased expression may be correlated with loss
of the sequence SYEENAAKKS This sequence bears
some resemblance to a sequence, SINNDAKSS, which
is present in the cytoplasmic tail of the yeast mating
factor receptor, Ste2p [43] In the SINNDAKSS
sequence, the serine residues are phosphorylated by casein kinase I homologues, and this in turn is required for ubiquitination of the lysine residue by the Nedd4-like ubiquitin ligase, Rsp5p [44,45] Mono-ubiq-uitination of the SINNDAKSS sequence serves as a signal for endocytosis and subsequent MVB sorting and degradation in the vacuole [43] The similarity of this sequence and that present in the MIT domain of Vps4p suggests that Vps4p may be subject to ubiqu-itin-dependent degradation This will be interesting to investigate in the future
Our findings support a model for MVB sorting in which nucleotide-free Vps4p is recruited to endosomes via interactions that may involve Vps2p, Vps20p, Snf7p and particularly Did2p, which binds Vps4p to a greater extent than all other interactors (Fig 6A) Upon recruitment of Vps4p to endosomes, it can inter-act with Bro1p initially in the absence of bound ATP Upon ATP binding to Vps4p (Fig 6B), Bro1p is dis-placed ATP binding is predicted to contribute to Vps4p assembly into an oligomeric ring [22], which is aided by Vta1p [46] ATPase activity of Vps4p is sti-mulated by Vta1p [33,46] and even more by Vps20p [33] Our data are in agreement with the proposed role
of Vps4p in disassembly of the MVB sorting machin-ery, and support a role for disassembly via effects on Vps2p and Vps20p (Fig 6C), as interactions with both
of these proteins are sensitive to ATP hydrolysis (this study [26]) Current data suggest that Vps20p in partic-ular has numerous interactions with components of the MVB sorting machinery [27,30] and appears to play a key role in holding together the complex Thus unfold-ing of Vps20p is predicted to destabilize the assembled MVB sorting machinery
In summary, we have performed the first systematic study of interactions between Vps4p and the MVB sort-ing machinery Several of these interactions have dis-tinct properties Only a subset of these interactions is regulated by Vps4p ATPase activity, and one interac-tion is regulated by Vps4p ATP binding Studying the role of Vps4p in the regulation of protein–protein inter-actions at the MVB will lead to a better understanding
of the mechanism of MVB sorting and virus budding
Experimental procedures
Media, reagents, strains and plasmids YPUAD and SD minimal media were as described previ-ously [26] Lucifer Yellow (LY) carbohydrazide dilithium salt was from Fluka AG (Buchs, Switzerland) Horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG, and bathophenanthrolinedisulfonic acid were from Sigma
Trang 9(St Louis, MO, USA) HRP-conjugated anti-mouse IgG was
from Bio-Rad Laboratories (Hercules, CA, USA), and
HRP-conjugated anti-goat IgG was from Zymed (San
Fran-cisco, CA, USA) Ni⁄ nitrilotriacetate–agarose and
monoclo-nal antibody to pentaHis were from Qiagen (Hilden,
Germany) Immobilized glutathione on agarose was from
Scientifix (Melbourne, Australia) Pre-stained protein
molecular mass marker was from Fermentas (Hanover, MD,
USA) poly(vinylidene difluoride) (PVDF) membrane was
from Millipore (Bedford, MA, USA) Polyclonal
anti-(carboxypeptidase Y) and anti-calmodulin sera were gifts from
H Riezman (University of Geneva, Geneva, Switzerland),
polyclonal antibody to Vps4p was from Santa Cruz
Biotech-nology (Santa Cruz, CA, USA), and monoclonal antibody
to actin was from Chemicon (Temecula, CA, USA)
S cerevisiae strains used in this study are listed in Table 1 Transformation of yeast with plasmid DNA was performed using a modified lithium acetate protocol [47] PCR primers used for plasmid constructions were from GeneWorks (Thebarton, Australia) and are listed in Table 2 Plasmids used in this study are listed in Table 3
Fig 6 A model for Vps4p-mediated disassembly of the MVB sorting complex (A) Vps4p is recruited to the endosome membrane through interactions that may involve Did2p, Vps2p, Vps20p and Snf7p Did2p is predicted to play an important role because of its high-affinity inter-action that is insensitive to ATP binding or ATP hydrolysis The Vps4p dimers shown at the endosome membrane may form in the cyto-plasm and then be recruited to endosomes, or monomeric Vps4p may be recruited to the endosome membrane and then oligomerize on the membrane Bro1p can initially interact with Vps4p and recruits Doa4p, which de-ubiquitinates the cargo protein (B) ATP binding to Vps4p
at the endosome membrane mediates its assembly into an oligomeric ring, which is promoted by Vta1p Bro1p is displaced by ATP binding
to Vps4p (C) Upon ATP hydrolysis by Vps4p, effects on Vps2p and Vps20p are predicted to break interactions with these proteins and thus disassemble the MVB sorting complex In addition, the Vps4p high-molecular-mass oligomer disassembles (D) The soluble components of the MVB sorting machinery are ready for another round of MVB sorting The Vps4 interactors that we have studied are shown in pink, and other components are shown in grey Components of the ESCRT complexes (0–III) are circled in box A.
Table 1 Yeast strains used in this study.
EGY48 MATa his3 trp1 ura3 LexAop(· 6)-LEU2 Clontech AMY245 MATa vps4-D::KanMx leu2 ura3 his4
lys2 bar1
[33]
strain
Trang 10Table 2 Primers used for mutagenesis.
Primer
Sequence (5¢- to 3¢)
Table 3 Plasmids used in this study.
pAM 482 pET11a E coli expression vector expressing Vps4p with a C-terminal 6His tag [26]
pAM 496 Original library clone of DID2 ⁄ CHM1 in pB42AD (encoding Did2p ⁄ chmlp 41–204 ⁄ end) [33]
pAM 862 pET11d E coli expression vector expressing Vps4p-GAI with a C-terminal 6His tag [33]
pAM 990 pET11a E coli expression vector expressing Vps4p-CC with a C-terminal 6His tag This study pAM 991 pET11a E coli expression vector expressing Vps4p–YEE with a C-terminal 6His tag This study pAM 992 pET11a E coli expression vector expressing Vps4p–IRA with a C-terminal 6His tag This study