A burst of auxilin recruitment determines the onset of clathrin-coated vesicle uncoating.

Abstract

Clathrin-coated pits assemble on a membrane and pinch off as coated vesicles. The released vesicles then rapidly lose their clathrin coats in a process mediated by the ATPase Hsc70, recruited by auxilin, a J-domain-containing cofactor. How is the uncoating process regulated? We find that during coat assembly small and variable amounts of auxilin are recruited transiently but that a much larger burst of association occurs after the peak of dynamin signal, during the transition between membrane constriction and vesicle budding. We show that the auxilin burst depends on domains of the protein likely to interact with lipid head groups. We conclude that the timing of auxilin recruitment determines the onset of uncoating. We propose that, when a diffusion barrier is established at the constricting neck of a fully formed coated pit and immediately after vesicle budding, accumulation of a specific lipid can recruit sufficient auxilin molecules to trigger uncoating.

Full text

Page 1

A burst of auxilin recruitment determines the onset
of clathrin-coated vesicle uncoating
Ramiro H. Massol, Werner Boll, April M. Griffin, and Tomas Kirchhausen*
Department of Cell Biology and CBR Institute for Biomedical Research, Harvard Medical School, 200 Longwood Ave, Boston, MA 02115
Edited by Pietro V. De Camilli, Yale University School of Medicine, New Haven, CT, and approved May 19, 2006 (received for review April 25, 2006)
Clathrin-coated pits assemble on a membrane and pinch off as
coated vesicles. The released vesicles then rapidly lose their clath-
rin coats in a process mediated by the ATPase Hsc70, recruited by
auxilin, a J-domain-containing cofactor. How is the uncoating
process regulated? We find that during coat assembly small and
variable amounts of auxilin are recruited transiently but that a
much larger burst of association occurs after the peak of dynamin
signal, during the transition between membrane constriction and
vesicle budding. We show that the auxilin burst depends on
domainsoftheproteinlikelytointeractwithlipidheadgroups.We
conclude that the timing of auxilin recruitment determines the
onset of uncoating. We propose that, when a diffusion barrier is
establishedattheconstrictingneckofafullyformedcoatedpitand
immediately after vesicle budding, accumulation of a specific lipid
can recruit sufficient auxilin molecules to trigger uncoating.
hsc70 ? endocytosis ? lipids
T
teins. Clathrin forms a lattice surrounding the invaginating
membrane. Various adaptor proteins provide the links required
forsortingcargoproteinsintoassemblingcoatedpits.Additional
proteins regulate coat formation and coat disassembly. Coat
assembly is relatively steady, whereas disassembly is abrupt. The
entire cycle typically lasts for 30–90 s (1).
BiochemicalandgeneticobservationsindicatethatHsc70andits
cofactor, auxilin, are directly involved in the uncoating process (2).
Hsc70, an abundant and ubiquitous cytosolic protein, has an
ATPasedomainlinkedtoaclaw-like,substrate-bindingdomain(3).
Through their J-domain, auxilins recruit Hsc70 molecules to their
substrate, clathrin-coated vesicles (4–8). Mammalian cells express
two auxilin variants, the brain-specific auxilin1 (Aux1) and the
ubiquitous cyclin G-associated kinase (GAK), also called auxilin2.
Auxilins have a region with high sequence similarity to the phos-
phatase and C2 domains of PTEN (9), a central segment with
binding sites for dynamin (10), AP-2, and clathrin (8, 11, 12) and a
C-terminal J-domain. In addition, GAK contains an N-terminal
Ser?Thr kinase domain that phosphorylates in vitro the ?-chains of
AP-1 and AP-2 clathrin adaptors (8, 13). Hsc70 promotes dissoci-
ationofclathrincoatsinvitrobyamechanismthatdependsonATP
hydrolysis and on Hsc70 recruitment by substoichiometric amounts
ofauxilin(14).TheC-terminalhalfofAux1,lackingthePTEN-like
domain,canalsosupportuncoatinginvitro(15).Thepositionofthis
fragment within the clathrin lattice, in contact with three different
clathrinlegs,hasbeendeterminedbycryoelectronmicroscopy(16).
Coatedpitsassemblecontinuouslyuntilthecoatedvesiclespinch
off, and only then does the coat dissociate. Partially assembled
lattices should be able to recruit both auxilins and the ATP-bound
Hsc70 constitutively present in the cytosol, and therefore they
should uncoat prematurely. Premature uncoating might be pre-
vented, either by activating bound auxilin only after finishing coat
growth or by restricting auxilin recruitment to completed coated
vesicles.Toworkoutwhichofthesetwopossibilitiesdeterminesthe
onsetofuncoating,weusedlive-cellimagingtofollowthedynamics
of auxilin recruitment into assembling endocytic clathrin coats. We
find that small and variable amounts of auxilin accumulate and
he cycle of clathrin-coated vesicle formation and disassembly
requires coordinated interaction of a large number of pro-
dissociate during the growing phase, whereas much larger amounts
arrive during the rapid transition between membrane invagination
andbuddingofthecoatedvesicle.Thislateburstofauxilinrequires
its phosphatase-like domain and correlates strongly with the rup-
ture of physical continuity between the plasma membrane and the
invaginated vesicular membrane. We further demonstrate that
Aux1 binds to specific phosphoinositides in vitro and that the
PTEN-like region of auxilin is required for this binding. We
propose that the onset of uncoating is determined by a precise
timing of auxilin recruitment to the coat. This timing may be set by
a rapid change in the concentration of a specific phosphoinositide.
Results
Auxilins Are Present in All Isolated Clathrin-Coated Vesicles but only
in a Small Fraction of Clathrin-Coated Structures at the Cell Surface.
To work out what determines the onset of uncoating, we first
studied by fluorescence microscopy the association of auxilins with
clathrin-coated structures in fixed cells. Auxilins were present in
only a fraction of clathrin-coated structures at the cell surface.
Whereas all fluorescent spots containing EGFP-Aux1 and ?70%
of the EGFP-GAK spots colocalized with clathrin or AP-2, only a
small fraction (10 ? 3%; n ? 150) of the clathrin or AP-2 spots
colocalized with auxilins (Fig. 1a). We found a similar colocaliza-
tionofclathrinorAP-2withendogenousAux1inastrocytesorwith
GAK in the epithelial kidney cell line BSC1 (Fig. 1b). In the latter
case, clathrin and AP-2 were visualized by stable expression of
EGFP-LCa or ?2-EGFP. Few of the spots of endogenous auxilins
failedtocoincidewitheitherLCaor?2.Incontrasttothepresence
of auxilins in only a fraction of the structures containing clathrin or
AP-2,wedetectedtheminallisolatedcoatedvesicles(Fig.1c).This
result suggests that auxilin associates with coated pits at one
particular time of their assembly cycle, presumably at a late step of
pit formation or immediately after coated-vesicle budding.
Auxilin Is Transiently Recruited to the Plasma Membrane at Late
Stages of Coated Vesicle Formation. To characterize the temporal
behavior of auxilins as they associate with clathrin coats, we used
live-cell imaging to follow the recruitment of Aux1 and GAK to the
plasma membrane of astrocytes (see Cells) (Figs. 6–8, which are
published as supporting information on the PNAS web site). The
fluorescence signals of Aux1 (green in Fig. 2a) or GAK (green in
Fig. 2b) showed small, low-amplitude fluctuations during clathrin
coat assembly and a major burst at the end of the growth phase
(red)(Movie1,whichispublishedassupportinginformationonthe
PNAS web site). Every instance of disappearance of a clathrin-
coated pit (e.g., corresponding to the uncoating of an endocytic
vesicle) had an associated Aux1 or GAK burst; these bursts were
detected in at least three consecutive time frames, and their
Conflict of interest statement: No conflicts declared.
This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: Aux1, auxilin 1; GAK, cyclin G-associated kinase; mRFP, monomeric red
fluorescent protein; PI, phosphatidylinositol; TIRF, total internal reflection fluorescence;
Dyn2, dynamin2.
*Towhomcorrespondenceshouldbeaddressed.E-mail:kirchhausen@crystal.harvard.edu.
© 2006 by The National Academy of Sciences of the USA
www.pnas.org?cgi?doi?10.1073?pnas.0603369103PNAS ?
July 5, 2006 ?
vol. 103 ?
no. 27 ?
10265–10270
CELL BIOLOGY

Page 2

intensitieswerehigherthantheaveragelocalbackgroundbyoneor
more standard deviations. The probability of an incorrect assign-
ment by our tracking algorithm is ?10?4. The maximal peak signal
of the EGFP-Aux1 bursts corresponds to ?20–40 molecules,
because their intensities were four to five times lower than the
EGFP-LCasignalsfromcoatedvesiclesobservedinBSC1cellsjust
before uncoating (?100–120 molecules; ref. 1). In the majority of
cases (?90% of 148 events) a single burst started immediately
before the clathrin signal reached its maximum, hence just before
completionofcoatassembly(clathrin-coatedpitstage)(Fig.2c).In
theremainingfewcasesthereweretwoormorerecruitmentevents
just before the end of the growth phase (Fig. 2c). We observed no
significantcorrelationbetweenthesizesoftheclathrincoatsbefore
uncoating (estimated by the maximum net signal of clathrin) and
the number of auxilin molecules recruited in the final burst (Fig.
2d). We likewise found no correlation between the time required
tobuildthecoat(growthphase)andthetimeusedtouncoatit(Fig.
2e, uncoating phase).
We confirmed that the recruitment of auxilin to fully formed
clathrin-coated structures happens during the transition from a
coated structure still close to the plasma membrane (e.g., a coated
pit ready to pinch off) to a structure that moves away from it (e.g.,
a budded coated vesicle). We did so by measuring the fluorescence
intensity of EGFP-Aux1 while rapidly alternating between total
internal reflection fluorescence (TIRF) and wide-field acquisition
modes. In ?55% of the EGFP-Aux1 objects the TIRF signal
disappeared faster than the wide-field signal (Fig. 9a, which is
published as supporting information on the PNAS web site),
indicating that Aux1 was present in objects that had escaped the
evanescent field (?140 nm) by moving away from the ventral
plasma membrane toward the cell interior. In the remaining cases
the simultaneous disappearance of both fluorescence signals indi-
cated that the Aux1 association and dissociation happened in
objects that remained within the evanescent field, close to the
plasma membrane, until complete uncoating (Fig. 9b). We ob-
served a similar behavior for GAK (Fig. 9 c and d).
The PTEN Homology Region of Aux1 Is Required for the Burst of
Recruitment to a Clathrin-Coated Pit. To map the regions of Aux1
(Fig. 3a) required for coated pit?vesicle recruitment, we generated
Fig.1.
sections of selected cells (a and b) or of a clathrin-coated vesicle sample (c).
Arrows point to examples of colocalization between auxilins and the indi-
cated markers. (a) U373mg astrocytes stably expressing EGFP-Aux1 or EGFP-
GAK were stained with antibodies specific for clathrin heavy chain or ?1??2-
subunits of AP-1 and AP-2 complexes. EGFP-tagged Aux1 and GAK also gave
astrong,diffusesignal,probablyoriginatingfromexcessauxilinattheplasma
membrane and?or in the cytosol. (b) U373mg astrocytes (top two rows) and
BSC1 cells (bottom two rows) stably expressing EGFP-LCA or ?2-EGFP were
stained with antibodies specific for Aux1 or Aux1?GAK. (c) Calf brain clathrin-
coated vesicles were stained with antibodies specific for clathrin heavy chain
and Aux1?GAK. (Scale bars: 2 ?m.)
Auxilinsarepresentinclathrin-coatedstructures.Shownareconfocal
Fig. 2.
astrocytes stably expressing EGFP-Aux1 together with tomato-LCa (expressed transiently for 24 h); examples of clathrin-coated structures are shown in the
kymograph view (Left); the fluorescence intensity plot (Right) shows that Aux1 is recruited transiently to the clathrin spot in small and variable amounts during
the growth phase and in a significantly larger burst at the onset of uncoating. (b) As in a, but cells transiently coexpressing EGFP-GAK and tomato-LCa (16 h).
(c) Bar plot showing that the most prominent Aux1 burst occurs during the uncoating phase. Similar results were obtained with all other cells expressing Aux1
or GAK (data not shown). (d) Scatter plot of the maximum fluorescence intensities of Aux1 and clathrin recruited to a given spot shows no correlation (r ? 0.08)
of the number of auxilin molecules recruited with the size of the coated vesicle. (e) Scatter plot shows no correlation (r ? 0.2) between the durations of the
uncoating and growth phases of a coated vesicle (n ? 196) in cells expressing EGFP-Aux1.
Aux1 or GAK are synchronously recruited at a late stage during coated vesicle formation. (a) Confocal time series acquired every ?2 s from U373mg
10266 ?
www.pnas.org?cgi?doi?10.1073?pnas.0603369103Massol et al.

Page 3

a series of truncation mutants of EGFP-Aux1 and studied their
cellular localization and dynamics when coexpressed with labeled
clathrin light chain (tomato-LCa) (Fig. 3b). The recruitment of
Aux1 and GAK (data not shown) lacking the J-domain showed the
same temporal pattern as full-length auxilin, indicating that this
domain is not required for association with coats in cells. Removal
of the complete PTEN homology domain strongly impaired the
burst of Aux1 recruitment (Movie 2, which is published as sup-
porting information on the PNAS web site); removal of just the
phosphatase-like domain produced a similar perturbation (Movie
3, which is published as supporting information on the PNAS web
site). It is unlikely that the targeting defects elicited by these
truncated mutants are due to a major misfolding of Aux1, because
the mutants are recruited to the clathrin microcages and other
latticesformedincellsupontransienthypertonicshockwithsucrose
(17) (Fig. 10a, which is published as supporting information on the
PNAS web site). The transient appearance of small amounts of
auxilinduringcoatassemblyisalsounimpaired.Wedidnotobserve
any detectable decrease in the endocytosis of transferrin (Fig. 7),
suggesting that these variants are not dominant, presumably be-
causetheavailableendogenousWTauxilinsaresufficienttosustain
uncoating.
How can the PTEN-like region influence recruitment? The
homologywiththephosphatasedomainofPTENandthepresence
of a C2 lipid-binding domain suggest that it might recognize lipid
head groups. In a lipid overlay assay we found that recombinant
Aux1 binds phosphatidylinositol (PI) (3)-phosphate, PI (4)phos-
phate, and PI (3, 4)-biphosphate more strongly than various other
phosphoinositidesornegativelychargedlipids(Fig.3c).Incontrast,
a truncated form of Aux1, lacking residues 1–546 (e.g., without the
PTEN-like region), did not bind any of these lipids (Fig. 11b, which
is published as supporting information on the PNAS web site).
Dynamin Function Is Required for the Major Recruitment of Auxilin.
TheabruptrecruitmentofEGFP-Aux1atthetransitionfromafully
formed coated pit to a coated vesicle suggests a relationship to
dynamin activity. We examined the correlation of dynamin and
auxilin bursts [using transiently expressed dynamin2 (Dyn2)–
monomeric red fluorescent protein (mRFP) and stably expressed
EGFP-Aux1]andfoundthatalloftheAux1correlatedwithasmall
but significant peak in the dynamin signal, generally ?1–2 s before
the auxilin maximum (Fig. 4a and Fig. 12b, which is published as
Fig. 3.
recruitmenttoclathrin-coatedstructures.(a)DomainorganizationofWTand
mutant forms of Aux1 used in our experiments. The domain organization of
Aux1 and GAK are similar, with an additional Ser?Thr kinase located at the N
terminus of GAK (data not shown). (b) Confocal time series acquired every 2 s
from U373mg astrocytes stably expressing WT EGFP-Aux1 or the correspond-
ing auxilin deletion mutants together with tomato-LCa (all expressed tran-
siently for 24 h). Selected examples are presented as equally normalized
kymographs (Left). The maximum amount of WT and truncated auxilin re-
cruited to a given clathrin coat is shown. (c) Aux1 interacts with specific
phosphoinositides in vitro. Shown is a representative chemiluminescence
imageobtainedfromthelipid–proteinoverlayassayusingAux1purifiedfrom
insect cells. The strip contained the following lipids: lysophosphatidic acid
(LPA), lysophosphocholine (LPC), PI, PI (3)-phosphate (PI3P), PI (4)-phosphate
(PI4P), PI (5)phosphate (PI5P), phosphatidylethanolamine (PE), phosphatidyl-
choline(PC),sphingosine-1-phosphate(S1P),PI(3,4)-biphosphate(PI3,4P2),PI
(3, 5)-biphosphate (PI3,5P2), PI (4, 5)-biphosphate (PI4,5P2), PI (3, 4, 5)-
triphosphate (PI3,4,5P3), phosphatidic acid (PA), and phosphatidylserine (PS).
Thepositionofthespots(redcircles)isshown.Thebarplotshowsthebinding
(average ? SE) of Aux1 determined in nine independent experiments (five
usingAux1purifiedfromfourbacterialpreparationsandfourfromtwoinsect
cell preparations).
The PTEN homology region of Aux1 is required for the major
Fig. 4.
namin function. (a) Selected frames from a time series acquired every 2 s of a
smallareaofU373mgastrocytesexpressingEGFP-Aux1stablyandDyn2-mRFP
transiently (24 h). Normalized fluorescence intensity as a function of time of
the selected coated pit is shown. A bar plot shows the distribution of occur-
rences of Aux1 major bursts in relation to the time of appearance of Dyn2
peaks (n ? 326). These data were obtained by first tracking the auxilin bursts,
then extending the mask back in time for an additional five time frames (?10
s). Finally, the relative timing corresponding to the appearance of the maxi-
mum fluorescence signals for Aux1 and Dyn2 was determined automatically.
Accuracyofthemethodwasverifiedmanuallyonasubsetoftheobservations
(see additional examples in Fig. 12b). (b) Dynasore, a chemical inhibitor of
dynamin, prevents the appearance of the auxilin burst. U373mg astrocytes
stablyexpressingEGFP-Aux1(green)andtransiently(24h)expressingtomato-
LCa (red) were incubated with medium containing 0.5% DMSO or 80 ?M
dynasore?0.5% DMSO at 37°C for 5 min. After this treatment, confocal time
series were recorded every 2 s (? Dynasore) or 6 s (? Dynasore), and examples
aredisplayedaskymographviews(TopandMiddle)andasplotsofnormalized
fluorescence intensities as a function of time (Bottom).
Aux1 recruitment follows dynamin peak and requires normal dy-
Massol et al. PNAS ?
July 5, 2006 ?
vol. 103 ?
no. 27 ?
10267
CELL BIOLOGY

Page 4

supporting information on the PNAS web site). We then explored
the relationship between the dynamics of these two proteins by
three complementary approaches (1). We depleted endogenous
Dyn2incellsstablyexpressingEGFP-Aux1bytreatmentwithsmall
interferingRNAoligonucleotides(18).WecouldnotdetectEGFP-
Aux1 recruitment to the plasma membrane (Movie 4, which is
published as supporting information on the PNAS web site) in any
cell that showed the expected block in transferrin uptake (Fig. 13a,
which is published as supporting information on the PNAS web
site). These latter cells still displayed punctate clathrin and AP-2
patterns (Fig. 13 b and c) (2). We transiently overexpressed the
dominant negative mutant Dyn2-K44A (19) fused to mRFP. Ex-
pression of this mutant impaired transferrin uptake and reduced
EGFP-Aux1 recruitment to spots on the plasma membrane (Fig.
10b and Movie 5, which is published as supporting information on
the PNAS web site); control cells overexpressing WT Dyn2-mRFP
showed no perturbations in clathrin-mediated endocytosis or Aux1
dynamics(Movie6,whichispublishedassupportinginformationon
the PNAS web site) (3). We treated cells with dynasore, a small
molecule that acutely, specifically, and reversibly inhibits the dy-
namin GTPase, thereby blocking transferrin uptake and locking
coated pits at stages before budding (20). Dynasore abolished the
final burst of auxilin recruitment (Fig. 4b); a substantially weaker
auxilin signal (25–50% less than normal) remained associated with
the locked pits.
Auxilin and dynamin have been reported to interact directly in
vitro through a region of auxilin that lies between the PTEN
homologydomainandtheclathrin-bindingdomain(10).Couldthis
interaction account for the correlations described in the preceding
paragraph? At least two lines of evidence suggest otherwise. First,
the actual dynamics are different. Dynamin is recruited steadily to
coated pits, with an incremental burst at the time of pinching (1),
whereas the auxilin burst begins essentially at baseline. Second,
truncated Aux1 that lacks the PTEN-like region but retains the
dynamin-binding segment exhibits no late burst in intensity,
whereas its early, low-level, transient recruitment appears to be
normal. We therefore believe that a direct interaction between the
two proteins cannot account for the major auxilin burst.
Discussion
Our principal finding is that auxilin recruitment to a coated pit
occursprimarilyinaburstjustasgrowthiscompleteandasbudding
occurs. We conclude that the timing of this recruitment is the
process that triggers rapid uncoating shortly after assembly has
finished. There is also a low level of transient auxilin incorporation
during the growth phase of a coated pit. One function for these
auxilins (and the Hsc70 that they presumably bring along) might be
toreleaseafewclathrintriskelionsfromthegrowingedgeofthepit,
perhaps to correct errors appearing during assembly. Another
might be to promote local disassembly, possibly required to accom-
modate a change in curvature of the underlying membrane in
response to incorporation of cargo of larger size. Auxilin could also
have a function distinct from its role as an Hsc70 cochaperone.
We can imagine two general ways in which the major burst of
auxilin recruitment could be timed. In one class of mechanism,
interactions with clathrin, APs, dynamin, or other proteins would
recruit auxilin, and the regulation of those interactions, e.g., by
phosphorylationorothermodification,wouldinturndeterminethe
timing of auxilin association. A model based exclusively on the
modulation of protein–protein interactions would, however, re-
quire coordinated modification of all 20–40 auxilin molecules
during the major burst or of an equivalent number of partner
proteins distributed throughout the coat. It is difficult to imagine
how local modifications can be coordinated across so large a
structure. A second kind of mechanism, based on topological
constraints,providesasimplerexplanationtoaccountforthemajor
burst of auxilin recruitment before uncoating. A specific lipid
species, generated by an enzymatic modification within the coated
membraneandperhapsrecognizedbythePTENdomain,mightact
as a recruiting signal for auxilin. For example, it is known that
synaptojanin, a phosphoinositide phosphatase that binds proteins
linked to endocytosis, is needed for the normal completion of the
coated vesicle cycle (21–28). This enzyme dephosphorylates PI (4,
5)-biphosphate phosphoinositides to create PI (4)phosphate (29),
one of the lipids shown here to bind auxilin, which could be the
recruiting signal for auxilin. During all stages of coat formation
before vesicle budding, the underlying membrane remains con-
nected to the plasma membrane, and any such lipid will rapidly
diffuse into the surrounding membrane. It will therefore fail to
reach the concentration threshold needed to capture enough auxi-
lin for productive uncoating. Only when a diffusion barrier is
established at the connecting neck or after the vesicle has been
topologicallyisolatedfromtheplasmamembrane(i.e.,afterfission)
cantheconcentrationofthepostulatedlipidincreasesufficientlyto
support the coordinated recruitment of a large number of auxilin
molecules (Fig. 5). The internal location of the Aux1 C-terminal
portion seen by cryoelectron microscopy reconstruction would
allow the PTEN region to extend inward and contact the mem-
brane (16).
Four findings support this proposal (1). Auxilin association with
the clathrin lattice is relatively weak (16), and a strong association
will probably require an additional set of contacts (2). All known
perturbationsthatblockbuddingbutstillretainclathrincoatsatthe
plasma membrane also prevent late auxilin recruitment and thus
prevent uncoating. The perturbations include removal of dynamin
by small interfering RNA treatment and inhibition of dynamin
functioneitherbyexpressionofDyn2K44Amutantorbytreatment
ofthecellswithdynasore.Cholesteroldepletion,bybriefincubation
of the cells with cyclodextrins, also retains clathrin arrays at the
plasma membrane, blocks budding, impairs auxilin recruitment
(Movie 7, which is published as supporting information on the
PNAS web site), and prevents coordinated uncoating. Thus, dis-
continuitybetweentheplasmamembraneandafullyformedvesicle
Fig. 5.
the auxilin burst and its relation to the onset of clathrin coat disassembly are
shown. The plasma membrane is represented by the orange line; the mem-
brane containing the lipid signal used to recruit auxilin is shown in gray.
During the growth phase, cargo, clathrin, adaptors, and other coat compo-
nents continuously accumulate. Continued growth results in deep membrane
invaginations, until membrane constriction and fission coordinated by a final
burst of dynamin occur. After budding, the coated vesicle moves away from
the membrane, and uncoating rapidly ensues. Small and variable amounts of
auxilin associate to and dissociate from growing pits, while a large burst of
auxilin recruitment starts during membrane constriction and ends with the
onset of uncoating. At this stage, sufficient amounts of ATP-bound Hsc70 are
captured by auxilin to drive coat disassembly. In this model, a lipid species
(indicated by the gray shadowing of the membrane), generated within the
coated membrane and perhaps recognized by the PTEN-like domain of auxi-
lin, acts as a recruiting signal for auxilin. Before vesicle budding, the under-
lying membrane remains connected to the plasma membrane, and any such
lipid will rapidly diffuse into the surrounding membrane. It will therefore fail
to reach the concentration threshold needed to capture enough auxilin to
ensue coat disassembly.
Model for the recruitment of auxilin. The principal stages leading to
10268 ?
www.pnas.org?cgi?doi?10.1073?pnas.0603369103Massol et al.

Page 5

appears essential for auxilin recruitment before uncoating (3).
Removal either of the entire PTEN homology domain of auxilin or
of just the phosphatase-like domain completely prevents the final
burst of auxilin, even though binding sites for dynamin, AP-2,
clathrin, and Hsc70 are still present on the truncated auxilins. The
C2 domain of PTEN has been reported to confer nonspecific
recognition of lipid head groups (30), and it seems likely that the
auxilin C2 domain functions similarly. The phosphatase-like do-
main of auxilin lacks the residues in PTEN essential for catalytic
activity toward phosphoinositides, but the extent of sequence
identity is sufficiently high that recognition of phosphoinositides is
possible (4). Purified full-length Aux1 (but not a truncated form
lacking its PTEN-like region) displays preferential binding to
phosphoinositides in a lipid overlay assay. Treatment with wort-
mannin or LY-294002 (Fig. 12a) did not alter coat dynamics in a
noticeable way; this lack of an effect could be due to incomplete PI
3-kinase inhibition (31) or to the presence of a pool of PI (3)-
phosphate lipids with a relatively slow turnover. Future studies will
be needed to identify the lipid species recognized by auxilin and to
test its role in triggering uncoating.
Materials and Methods
Plasmids, Oligos, Transfections, and Protein Purification. Full-length
bovine Aux1 (910 residues) was amplified by PCR from a
full-length cDNA clone and inserted into pEGFP-C1 (CLON-
TECH), resulting in pEGFP-Aux1. The C-terminal J-domain of
Aux1 was removed by using an internal BamHI site to create a
truncation after M812 (pEGFP-Aux1?J). The N-terminal
PTEN homology region or the phosphatase-like domain was
removed by QuikChange mutagenesis (Stratagene) by incorpo-
rating a HindIII site after Q419 or a BglII site after L216. The
resulting HindIII?SacII or BglII?SacII fragment was transferred
into pEGFP-C1, resulting in pEGFP-?PTEN Aux1 or pEGFP-
?PDAux1. Full-length rat Dyn2 WT or K44A (874 residues)
EGFP fusions were a generous gift of Sandy Schmidt (The
Scripps Research Institute, La Jolla, CA). mRFP was amplified
from pRSETb and substituted for EGFP in WT and K44A
versions of Dyn2-EGFP. Full-length LCa and ?2 fused to EGFP
(pEGFP-LCa and p?2-EGFP, respectively) were previously
described (1). The sequence corresponding to the tandem repeat
tomato ORF (476 residues) was amplified by PCR and substi-
tuted for EGFP in pEGFP-LCa using Age1?BsrGI sites (pTom-
LCa). All constructs were verified by restriction digest and
sequence analysis. All transfections were carried out by using
FuGENE 6 (Roche Diagnostics, Indianapolis), and cells were
evaluated 16–48 h later or selected with geneticin to obtain
stable cell lines. The sequence of the double-stranded small
interfering RNA oligo (Dharmacon, Lafayette, CO) targeting
human Dyn2 was the following: sense strand, 5?-GGAGAUU-
GAAGCAGAGACCTT-3?; antisense strand, 5?-GGUCU-
CUGCUUCAAUCUCCTG-3?. Small interfering RNA oligo
targeting CD4 was used as a negative silencer control. Trans-
fections of oligos were performed by using Oligofectamine
(Invitrogen) following the manufacturer’s recommendations.
Cells were tested 96 h after transfection.
For bacterial expression of auxilin we transferred Aux1 from
pEGFP-Aux1 into pGEX-4T-1 using BglII and SalI restriction
enzymes, resulting in full-length Aux1 fused to GST (pGEX-4T-1-
Aux1), which was then used to transform BL21-pLys cells. Expres-
sion was achieved after overnight growth at 37°C in LB medium
supplemented with chloramphenicol?kanamycin, followed by in-
oculation of 10 ml into 1 liter of fresh LB medium and further
allowed to grow for 2–3 h at room temperature (OD560? 0.5–0.7)
before adding isopropyl-13-D-thiogalactoside (0.1 mM final). Cells
were harvest by centrifugation after a further incubation for 16 h,
washed with cold PBS, resuspended, and lysed by sonication at 4°C
(four pulses of 1 min, 30 s of cooling) in the presence of complete
protease inhibitors (Roche Molecular Biochemicals) and 1 mM
PMSF. The lysate was clarified by centrifugation (34,000 ? g, 20
min, 4°C), and the GST-Aux1 protein was isolated by using gluta-
thione affinity columns. The yield is consistently very low (0.07–
0.15 mg?liter of culture), and the preparation consists of full-length
Aux1 and several fragments.
For insect cell expression of Aux1 we transferred Aux1 from
pEGFPAux1 into the pFastBac HT vector (Invitrogen) using BglII
and PstI. This plasmid was used to transform DH10Bac cells to
produce, upon recombination, the bacmids coding for full-length
N-terminalHis6-taggedAux1.Sf9insectcellsweretransfectedwith
theresultingbacmids,andtheamplifiedviruswasusedtotransduce
High Five insect cells (1–2 ? 106cells per ml; volume, 500 ml
infected with a 1:600 dilution of viral stock). Cells were harvested
56–72 h after infection, and Aux1 was purified by using TALON
metal affinity resin (CLONTECH) as described (32). This proce-
dure provides a moderate yield of mostly full-length His6-tagged
Aux1 (0.5–0.7 mg?liter of culture) and with no detectable contam-
inants observed by Coomassie blue staining.
Cells. Sf9 and High Five insect cells were grown in ExCell 420
medium at 28°C in spinner flasks. BSC1, HeLa, HEK-293, and
U373mgastrocytesweregrowninDMEMsupplementedwith10%
FBS, 2 mM L-glutamine, penicillin (50 units?ml), streptomycin (50
mg?ml), and nonessential amino acids (0.1 mM). The human
U373mg astrocytes offer the proper cellular context to study the
brain-specific endogenous Aux1. U373mg cell lines, stably express-
ing EGFP fusion chimeras of Aux1, EGFP-clathrin light chain
(LCa), or ?2-EGFP were generated by selection with 0.5–0.7
mg?ml geneticin. The rates of the growth and uncoating phases of
clathrin-coated structures observed by using LCa and AP-2 fused
to fluorescent proteins is very similar (Fig. 8 a–c). The relative
expression level of EGFP-Aux1 in the crude microsomal fraction
containing clathrin-coated vesicles was 4–5 times higher than the
expression level of endogenous Aux1; there was no significant
replacement of endogenous Aux1 associated with this fraction.
Simultaneous expression of EGFP-Aux1 with tomato-LCa or with
mRFP-dynamin was performed by transient expression of the red
fluorescent proteins in cells stably expressing EGFP-Aux1 or by
transient coexpression of the green and red fluorescent proteins.
Cells were seeded on glass coverslips and imaged 16–72 h later at
?50–70% confluency. EGFP-Aux1 was recruited to diffraction-
limited spots at the ventral (Fig. 6 a and b and Movie 8a, which is
published as supporting information on the PNAS web site) and
dorsal (Movie 8b) surfaces of the cell with a residence time of 8 ?
1.5s(n?150–270particlespercellanalyzedfromsixcells;Fig.6c).
We obtained comparable results upon transient expression of
EGFP-Aux1 in HeLa, HEK-293, and BSC1 cells (Movie 8c) and of
EGFP-GAK in astrocytes (Fig. 6 d–f and Movie 9, which is
published as supporting information on the PNAS web site).
Expression of EGFP-tagged Aux1 or GAK did not affect the
uptake of transferrin (Fig. 7) or the duration of coat formation and
uncoating(Fig.8dande).Theredfluorescentproteintomatofused
toclathrinLCadoesnotdisturbthedynamicbehaviorofendocytic-
coated structures or the endocytosis of transferrin (Fig. 8f).
Immunofluorescence. Cells were fixed in 3% paraformaldehyde in
PBS for 10 min at room temperature, washed with PBS, blocked
with 50 mM NH4Cl in PBS for 15 min at room temperature, and
finally incubated with 2% BSA in PBS for 30 min at room
temperature.AllantibodiesweredilutedinPBSsupplementedwith
2% BSA and 0.1% saponin. Aux1 was labeled with the mouse
monoclonal antibody 100?4 that recognizes only Aux1 (33). Aux1
and GAK were simultaneously labeled with a rabbit polyclonal
antibody. Clathrin heavy chain, ?-adaptins, and dynamin were
labeled with the mouse monoclonal antibodies X22 (34), 10A (35),
and Hudy-1 (Upstate Biotechnology), respectively. Alexa Fluor
488-, Alexa Fluor 594-, or Alexa Fluor 647-labeled goat anti-mouse
Massol et al.PNAS ?
July 5, 2006 ?
vol. 103 ?
no. 27 ?
10269
CELL BIOLOGY