-
LETTERdoi:10.1038/nature12360
Spatiotemporal control of endocytosis
byphosphatidylinositol-3,4-bisphosphateYork Posor1, Marielle
Eichhorn-Gruenig1*, Dmytro Puchkov1*, Johannes Schoneberg2*,
Alexander Ullrich2*, Andre Lampe1,Rainer Muller3, Sirus Zarbakhsh3,
Federico Gulluni4, Emilio Hirsch4, Michael Krauss1, Carsten
Schultz3, Jan Schmoranzer1,Frank Noe2 & Volker Haucke1,5
Phosphoinositides serve crucial roles in cell physiology,
ranging fromcell signalling to membrane traffic1,2. Among the seven
eukaryoticphosphoinositides the best studied species is
phosphatidylinositol-4,5-bisphosphate (PI(4,5)P2), which is
concentrated at the plasmamembrane where, among other functions, it
is required for thenucleation of endocytic clathrin-coated pits36.
No phosphatidyl-inositol other than PI(4,5)P2 has been implicated
in clathrin-mediated endocytosis, whereas the subsequent endosomal
stagesof the endocytic pathway are dominated by
phosphatidylinositol-3-phosphates(PI(3)P)7. How
phosphatidylinositol conversion fromPI(4,5)P2-positive endocytic
intermediates to PI(3)P-containingendosomes is achieved is unclear.
Here we show that formationof phosphatidylinositol-3,4-bisphosphate
(PI(3,4)P2) by class IIphosphatidylinositol-3-kinase C2a (PI(3)K
C2a) spatiotemporallycontrols clathrin-mediated endocytosis.
Depletion of PI(3,4)P2 orPI(3)K C2a impairs the maturation of
late-stage clathrin-coatedpits before fission. Timed formation of
PI(3,4)P2 by PI(3)K C2ais required for selective enrichment of the
BAR domain proteinSNX9 at late-stage endocytic intermediates. These
findings providea mechanistic framework for the role of PI(3,4)P2
in endocytosisand unravel a novel discrete function of PI(3,4)P2 in
a central cellphysiological process.PI(4,5)P2 generation by
phosphatidylinositol phosphate-5-kinases
(phosphatidylinositol-5-kinases) is required for recruitment of
earlyPI(4,5)P2-associated coat components to mediate
clathrin-coated pit(CCP)nucleation in clathrin-mediated endocytosis
(CME)1,5. Althoughphosphatidylinositol-5-kinases canassociatewith
early coat components3,they fail to enrich at maturing CCPs8. By
contrast, CCPs contain 5-phosphatases9 that degrade PI(4,5)P2
during late stages ofCME.Giventhe identification of PI(3,4)P2 4-
and PI(3,4,5)P3 5-phosphatases aseffectors of endosomal Rab5 (ref.
10) we proposed that PI(3,4)P2 mightserve as an intermediate plasma
membrane phosphatidylinositol spe-cies en route to
PI(3)P-containing endosomes.Analysis of the cellular PI(3,4)P2
distribution using a specific anti-
PI(3,4)P2 antibody11 revealed predominant plasma membrane
label-ling that overlapped with the localization of the
PI(3,4)P2-sensingtandem PH-domain of TAPP112 (Supplementary Fig.
1a). In additionto larger PI(3,4)P2-positive structures11, akin to
circular dorsal ruffles ofmigratory cells, anti-PI(3,4)P2
antibodies decorated diffraction-limitedpuncta that partially
co-localized with plasmalemmal CCPs (Fig. 1a).To verify
specificitywe analysed cells
overexpressingPI(3,4)P2-specific4-phosphatase, type II
inositol-3,4-bisphosphate 4-phosphatase13 fusedto a
carboxy-terminal CAAX-box prenylation sequence to target it
tothemembrane (INPP4BCAAX). Overexpression of INPP4BCAAXresulted
indepletionof antibody-decoratedPI(3,4)P2,whereasPI(4,5)P2levels
remained unchanged (Fig. 1b and Supplementary Fig. 1b, c).Selective
INPP4BCAAX-mediated depletion of plasma membrane
PI(3,4)P2 but not of other phosphatidylinositols such as
PI(3)P,PI(4,5)P2, or PI(3,4,5)P3 was verified by quantitative
determinationof themembrane enrichment of specific
phosphatidylinositol-bindingdomain-based sensors using total
internal reflection (TIRF)/epifluo-rescence microscopy
(Supplementary Fig. 1c). Thus, the levels anddistribution of
PI(3,4)P2 are faithfully reported by anti-PI(3,4)P2antibodies or by
PH-TAPP1 and overexpression of INPP4BCAAXselectively depletes
plasmalemmal PI(3,4)P2.Given thepresenceofPI(3,4)P2 atCCPswe tested
its functional impor-
tance for CME. Depletion of PI(3,4)P2 by INPP4BCAAX
impairedtransferrin endocytosis and led to increased transferrin
receptor sur-face levels, similar to depletion of PI(4,5)P2 by
INPP5ECAAX, a lipidrequired for CCP nucleation (Fig. 1c).
Overexpression of membrane-targeted catalytically inactive
INPP4B(C842A13), thePI(3)P-phosphataseMTM1 (ref. 14), or the
PI(3,4,5)P3-phosphatase PTEN (see Sup-plementary Fig. 1d, e for
controls) did not affect CME of transferrin(Fig. 1c). These data
reveal a hitherto unknown regulatory role forPI(3,4)P2 in CME. To
dissect the underlying mechanism we analysedthe distribution and
dynamics of key endocytic proteins. PI(3,4)P2depletion by INPP4B
caused the accumulation of AP-2a-positiveCCPs (Fig. 1d, e) and
markedly slowed CCP dynamics (Fig. 1f,Supplementary Fig. 2 and
Supplementary Video 1), similar to dyna-min1/2-knockout (KO)15. No
such effects were observed for catalyti-cally inactive INPP4B
(C842A), MTM1 (to deplete potential plasmamembrane PI(3)P), or PTEN
(to deplete PI(3,4,5)P3) (Fig. 1e andSupplementary Fig. 2). These
data identify PI(3,4)P2 as a novel regu-lator of CME, possibly
involved in a late stage in the pathway differentfrom
PI(4,5)P2-controlled CCP initiation (Fig. 1f and SupplementaryFig.
2)4.PI(3,4)P2 canbegeneratedbywortmannin-sensitive class
IPI(3)Ksand
subsequent hydrolysis of PI(3,4,5)P3 by 5-phosphatases16
downstreamof growth factor activation. Epidermal growth factor
(EGF)-inducedincrease of PI(3,4,5)P3was abrogated bywortmannin
inhibition of classI PI(3)K (Supplementary Fig. 1f), but had only a
moderate effect onthe basal level of PI(3,4)P2 (Supplementary Fig.
1g). These data suggestthe existence of a class I
PI(3)K-independent pool of PI(3,4)P2 andare consistent with CME
being a constitutive process inmost cell types.A less well
characterized pathway for PI(3,4)P2 production is theclass II
PI(3)K-mediated phosphorylation of phosphatidylinositol-4-phosphate
(PI(4)P)16. The contribution of this pathway to cellularPI(3,4)P2
synthesis is unknown. Class II PI(3)K C2a was identified asan
interactor of clathrin17. PI(3)K C2a also binds to PI(4,5)P2 (ref.
18)and its activity is stimulated by clathrin17, but largely
refractory to inhi-bition by wortmannin19. Quantitative proteomics
showed PI(3)K C2ato be enriched in clathrin-coated vesicles (CCVs)
with about 10 copiesper vesicle20. We found endogenous PI(3)K C2a
to co-localize withclathrin in endocytic CCPs (Fig. 2a; in
agreement with ref. 17), and
*These authors contributed equally to this work.
1Leibniz Institut furMolekulare Pharmakologie (FMP)& Freie
Universitat Berlin, Robert-Roessle-Strae 10, 13125Berlin, Germany.
2Freie Universitat Berlin, DFGResearchCenterMATHEON, Arnimallee
6,14195 Berlin, Germany. 3European Molecular Biology Laboratory
(EMBL), Cell Biology and Biophysics Unit, 69117 Heidelberg,
Germany. 4Molecular Biotechnology Center, Departments of
Genetics,Biology and Biochemistry, University of Torino, Via Nizza
52, 10126 Torino, Italy. 5Charite Universitatsmedizin, NeuroCure
Cluster of Excellence, Chariteplatz 1, 10117 Berlin, Germany.
1 1 J U LY 2 0 1 3 | V O L 4 9 9 | N A T U R E | 2 3 3
Macmillan Publishers Limited. All rights reserved2013
-
confirmed its enrichment in CCVs (Supplementary Fig. 3a).
Clathrinknockdown caused dispersal of PI(3)K C2a to the cytosol,
indicatingthat membrane targeting of PI(3)K C2a requires clathrin
(Supplemen-tary Fig. 3b). Cells depleted of PI(3)K C2a
(Supplementary Fig. 3c)showed reduced CME of transferrin and
increased transferrin receptorsurface levels (2286 23% ofmock
control; rescue, 1116 12% ofmock;s.e.m., n5 5 experiments), an
effect rescued by re-expression of shortinterfering RNA
(siRNA)-resistant PI(3)K C2a fused with enhancedgreen fluorescent
protein (EGFP; Fig. 2b).CMEofEGFwas reduced to a
lesser extent (Supplementary Fig. 3d). Defective transferrinCME
wasalso observed in mouse embryonic fibroblasts derived from
PI(3)KC2a-KOmice, an effect rescued by re-expression of wild type,
but notcatalytically inactive (Supplementary Fig. 5b) mutant PI(3)K
C2a
a
b c
d e
f
CHCPI(3,4)P2 Merge
Distance along line
Nor
mal
ized
in
tens
ity
CHC PI(3,4)P2
0.25
0.50
0.75
1.00
1.25
Lip
id s
tain
ing
inte
nsity
per
m
2
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Moc
k
INPP
4B
CAAX
PI(3,4)P2PI(4,5)P2
*
Rat
io in
tern
aliz
ed/
surf
ace-
bou
nd t
rans
ferr
in
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
mCh
erry
INPP
4BC
AAX
INPP
4B
C842
ACA
AX
MTM
1CA
AX
PTEN
CAA
X
INPP
5E
CAAX
*
**
NS
mC
herr
yIN
PP
4B
CA
AX
AP-2
Inte
nsity
of i
ndiv
idua
lA
P-2
stru
ctur
es
00.20.40.60.81.01.21.41.61.82.0
mCh
erry
INPP
4BC
AAX
INPP
4B
C842
ACA
AX
MTM
1CA
AX
PTEN
CAA
X
** **
mC
herr
yIN
PP
4BC
AA
XIN
PP
5EC
AA
X
180 s
Figure 1 | PI(3,4)P2 regulates CME. a, Partial co-localization
of PI(3,4)P2with CCPs. Confocal images of Cos7 cells stained for
PI(3,4)P2 and clathrinheavy chain (CHC). Arrowheads, structures
immunopositive for PI(3,4)P2 andclathrin. Scale bar, 10mm(inset:
2mm).b, Selective depletionof PI(3,4)P2 by thePI(3,4)P2-specific
phosphatase INPP4B (INPP4BCAAX). Levels of PI(3,4)P2or PI(4,5)P2
were quantified by immunostaining for PI(3,4)P2 or PI(4,5)P2(mean6
s.e.m.; n5 5 experiments; *P, 0.05, t-test). c, Selective depletion
ofplasma membrane PI(3,4)P2 impairs CME of transferrin. Expression
ofmCherry-tagged membrane-targeted inactive INPP4B(C842A), of the
PI(3)Pphosphatase MTM1, or of the PI(3,4,5)P3 phosphatase PTEN do
not affectCME. INPP5E-mediated depletion of PI(4,5)P2 was used as a
positive control.Bar diagrams represent ratio of internalized
(10min, 37 uC) to surfacetransferrin (45min, 4 uC) (mean6 s.e.m.;
n5 3 experiments, forINPP4B(C842A) n5 2; *P, 0.05, **P, 0.01,
t-test). d, e, Accumulation ofAP-2a-positive CCPs in
PI(3,4)P2-depleted cells. Confocal images of Cos7 cellsexpressing
mCherry or mCherryINPP4BCAAX stained for endogenousAP-2a. d, Scale
bar, 5mm. e, Mean intensity of endocytic AP-2a-containingCCPs
(mean6 s.e.m.; n5 3 independent experiments; **P, 0.01, t-test).f,
Stalled CCP dynamics in PI(3,4)P2-depleted cells analysed by TIRF
imagingof EGFPclathrin. Depletion of PI(4,5)P2 by INPP5E causing
loss of plasmamembrane CCPs was used as a control. Kymographs,
EGFPclathrinfluorescence over 180 s in cells expressing mCherry or
the indicated mCherry-tagged phosphatase. See also Supplementary
Video 1.
a b
c d
e
f
g h
PI3K C2 CHC Merge
Rat
io in
tern
aliz
ed/
surf
ace-
bou
nd t
rans
ferr
in
00.20.40.60.81.01.21.4
Moc
k
Scra
mble
d
siRNA
PI3K
C2
siRNA
PI3K
C2
siRNA
GFP GFPPI3K C2
*** ***
MockScrambled
siRNAPI3K C2siRNA 1
PI3K C2siRNA 2
GFP
cla
thrin
180 s
****
*** *******
258 60118 120176 Above 178
Lifetime (s)
Per
cent
age
of a
ll C
CP
s
0
5
10
15
20
25
3035
40
45
50
55 MockScrambled siRNAPI3K C2 siRNA 1PI3K C2 siRNA 2
1 2 3 4
PI3K C2 siRNA
Scrambled siRNA
Str
uctu
res
per
m p
erim
eter
0
0.04
0.08
0.12
0.16
0.20
1 2 3 4
MockScrambled siRNAPI3K C2 siRNA
EGFPPI3K C2mCherrySNX9
Merge60 s 48 s 36 s 24 s 12 s 0 s
Dynamin 2mCherryEGFPPI3K C2
Merge60 s 48 s 36 s 24 s 12 s 0 s
mRFPclathrinEGFPPI3K C2
Merge60 s 48 s 36 s 24 s 12 s 0 s
Nor
mal
ized
fluo
resc
ence
Lifetime (s)
PI3K C2Clathrin
SNX9
Dynamin 2
0204060801001200
0.25
0.5
0.75
1.00
1.25
Figure 2 | PI(3)K C2a controls maturation of CCPs. a, Confocal
images ofCos7 cells stained for endogenous PI(3)KC2a and clathrin
heavy chain (CHC).Scale bar, 10mm. b, PI(3)K C2a depletion impairs
CME of transferrin. Cos7cells depleted of PI(3)K C2a expressing
eGFP or siRNA-resistant EGFPPI(3)K C2a wild type were assayed for
CME of transferrin. Bar diagramsrepresent the ratio of internalized
(10min, 37 uC) to surface transferrin(45min, 4 uC) (mean6 s.e.m.;
n5 5 experiments; ***P, 0.001, t-test).c, d, PI(3)KC2a depletion
impairs CCPdynamics analysed by TIRF imaging ofEGFPclathrin
expressing Cos7 cells depleted of PI(3)K C2a. c, Kymographsshow
increased CCP-lifetimes in cells depleted of PI(3)K C2a
(seeSupplementary Videos 2 and 3). d, Lifetime distribution of CCPs
binned incategories of 60 s. Data represent mean6 s.e.m. (n5 3
experiments with.1,000 CCPs per condition; *P, 0.05, **P, 0.01,
t-test for scrambled vsPI(3)K C2a siRNA-treated cells). e, f,
Ultrastructural analysis of CCPs incontrol or PI(3)K C2a-depleted
cells. Morphological groups were shallow(stage 1), non-constricted
U-shaped (stage 2), constricted V-shaped pits(stage 3), or
structures containing complete clathrin coats (stage 4).e,
representative images from controls (top and middle) or a PI(3)K
C2a-depleted cell illustrating accumulation and clustering of
U-shaped pits(bottom). Scale bar, 100 nm. f, Bar diagram detailing
the relative abundance ofdifferent clathrin-coated structures in
control or PI(3)K C2a-depleted cells(mean6 s.e.m.; n5 10 (mock,
scrambled siRNA) or n5 11 (PI(3)K C2asiRNA) cell perimeters). g, h,
Timing of recruitment of PI(3)K C2a and SNX9to CCPs analysed by
TIRF microscopy. mRFP, monomeric red fluorescentprotein . g,
Snapshots of endocytic proteins at single CCPs (fission at t5
0).h,Mean time course of relative fluorescence intensity atCCPs
(mean6 s.e.m.; 3experiments for clathrin, dynamin 2 and PI(3)KC2a,
2 for SNX9; total numbern of CCPs: n5 58 for clathrin, n5 85 for
dynamin2, n5 248 for PI(3)K C2a,n5 100 for SNX9).
RESEARCH LETTER
2 3 4 | N A T U R E | V O L 4 9 9 | 1 1 J U LY 2 0 1 3
Macmillan Publishers Limited. All rights reserved2013
-
(Supplementary Fig. 3e). Loss of PI(3)KC 2a thus phenocopies
effectsof PI(3,4)P2 depletion on CME.Nextwe analysed the dynamics
of plasmalemmalCCPs inPI(3)KC2a
depleted cells by TIRF microscopy. Cells lacking PI(3)K C2a
showedincreased CCP lifetimes (Fig. 2c, d and Supplementary Videos
2 and 3)and this was rescued by re-expression of siRNA-resistant
EGFPPI(3)KC2a (Supplementary Fig. 4a). Although nucleation and
growth of CCPswere unaltered, they frequently failed to mature to a
fission-competentstate. Instead, many CCPs seemed to grow beyond
the size at which theywould normally undergo fission and could be
observed to split into twoor three closely neighboured CCPs
(Supplementary Fig. 4b). Attenu-ated dynamics of CCPs in PI(3)K
C2a-depleted cells were also seen influorescence recovery after
photobleaching experiments (SupplementaryFig. 4d, e).To determine
whether PI(3)K C2a regulates maturation of CCPs,
before or in conjunction with dynamin-mediated fission, we
subjectedPI(3)KC2a-depleted cells to quantitativemorphometric
analysis. Thisrevealed an increased number of U-shaped CCPs, a
stage precedingconstriction and dynamin-mediated fission, whereas
the frequenciesof early shallow CCPs, V-shaped constricted CCPs, or
of free CCVswere unaltered (Fig. 2e, f). CCPs frequently appeared
clustered (Sup-plementary Fig. 4f), as also seen by live imaging
(Supplementary Fig.4b, c). Analysis of the dynamics of endocytic
protein recruitment toCCPs showed PI(3)K C2a to follow clathrin but
to precede dynamin 2(Fig. 2g, h). We conclude that PI(3)K C2a
regulates CCP maturationby facilitating the transition from
invaginated to V-shaped CCPs.To explore whether the function of
PI(3)K C2a in CME requires its
phosphatidylinositol kinase activity we assayed catalytically
inactivemutant PI(3)K C2a (Supplementary Fig. 5b). Endocytic
proteins suchasAP-2a accumulate at CCPs following depletion of
PI(3,4)P2 (Fig. 1d,e) or PI(3)K C2a (Fig. 3a). This defect was
rescued by siRNA-resistantwild type but not catalytically inactive
mutant PI(3)K C2a, althoughboth variants localized to CCPs
(Supplementary Fig. 5a). Thus, PI(3)KC2a function in CME requires
its PI(3)K activity.Previous studies have yielded conflicting data
regarding thedominant
lipid product of PI(3)K C2a, reporting either preferential
synthesis ofPI(3,4)P2 or PI(3)P21,22. Immunopurified PI(3)KC2a
preferentially pro-duced PI(3,4)P2 as compared to either PI(3)P or
PI(3,4,5)P3 (Figs 3b,Supplementary Fig. 5b, c), in agreement with
ref. 22. If PI(3)K C2a wasto contribute to PI(3,4)P2 formation in
vivo, knockdown of PI(3)K C2ashould result in reduced PI(3,4)P2
levels. Quantitative assessment ofplasmamembrane
phosphatidylinositols by specific PI-binding domain-based sensors
revealed a selective reduction of PI(3,4)P2 in PI(3)KC2a-knockdown
cells, whereas PI(3)P, PI(4,5)P2 or PI(3,4,5)P3 remainedunchanged
(Fig. 3c). Depletion of PI(3,4)P2, but not of PI(4,5)P2,was also
detectable with PI-specific antibodies (Fig. 3d). Consist-ently,
PI(3,4)P2 largely co-localized with the plasma membrane poolof
PI(3)KC2a (Supplementary Fig. 5d). Conversely, we failed to
detectPI(3,4)P2 at CCPs in PI(3)KC2a-depleted cells (Supplementary
Fig. 5e).These results are consistent with the preferred production
of PI(3,4)P2by PI(3)K C2a in vitro and support the hypothesis that
PI(3)K C2acontributes to PI(3,4)P2 formation at CCPs in vivo.To
corroborate the preferential synthesis of PI(3,4)P2 over PI(3)P
by
PI(3)KC 2a in vivo we capitalized on the fact that the
specificity ofPI(3)Ks is encodedwithin the
phosphatidylinositol-binding activationloop23. The activation loop
of PI(3,4,5)P3-producing class I PI(3)Kscontains two basic boxes
that coordinate the phosphates of PI(4,5)P2.None of these basic
boxes is present in PI(3)P-producing class IIIPI(3)K hVps34 (Fig.
3e). PI(3)K C2a only contains basic residues thatcoordinate the
4-phosphate group, consistent with PI(3,4)P2 synthesis.To
distinguish between PI(3)K C2a-mediated formation of PI(3,4)P2or
PI(3)P at CCPs we constructed a PI(3)K C2a mutant, in which
the4-phosphate coordinating box was exchanged with the
correspondingsequence from hVps34 (Fig. 3e). This class III-like
mutant PI(3)K C2aselectively synthesized PI(3)P with wild-type
PI(3)K C2a activity, butfailed to produce PI(3,4)P2 (Supplementary
Fig. 5b, c). It was also unable
to rescue defective CME in PI(3)K C2a-depleted cells (Fig. 3f,
g). Thus,CME requires PI(3)KC2a-mediated production of PI(3,4)P2,
but not ofPI(3)P.To challenge this hypothesis by an independent
approach we made
use of cell-permeable PI-derivatives24 to exogenously supply
PI(3)Por PI(3,4)P2. Addition of cell-permeable PI(3,4)P2 partially
rescuedendocytic protein accumulation at CCPs in PI(3)K
C2a-depleted cells,whereas PI(3)P was inactive (Supplementary Fig.
5f), although it sti-mulated early endosome fusion (ref. 24 and not
shown). We concludethat PI(3)KC2a is required for local PI(3,4)P2
production at endocyticCCPs.Absence of PI(3)K C2a or depletion of
its lipid product PI(3,4)P2
causes a delay in CCP maturation, suggesting the presence of
PI(3,4)P2effectors at CCPs. To identify such effectors we monitored
CCP enrich-ment of endocytic proteins in PI(3)KC2a-depleted cells
(SupplementaryFig. 6a).Of the proteins assayed the only one that
failed to enrich at CCPs
a b c
d e
f g
Inte
nsity
of
ind
ivid
ual A
P-2
stru
ctur
es
0
1
2
3
Moc
k
Scra
mble
d
siRNA
PI3K
C2
siRNA
Moc
k
Scra
mble
d
siRNA
PI3K
C2
siRNA
Moc
k
Scra
mble
d
siRNA
PI3K
C2
siRNA
Moc
k
Scra
mble
d
siRNA
PI3K
C2
siRNA
PI3K
C2
siRNA
PI3K
C2
siRNA
GFPGFPPI3K C2 WTGFPPI3K C2 kinase inactive
***** *
NS
32P
sig
nal
0
1
2
3
4
5*** ***
TIR
F/ep
ifluo
resc
ence
rat
io
0.6
0.7
0.8
0.9
1.0
1.1
Mock ScrambledsiRNA
PI3K C2siRNA
RFP2xPHTAPP1
EGFP2xFYVE
EGFP2xPHPLC
BtkPHEGFP (basal)
*****
Fluo
ures
cenc
e p
er
m2
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6PI(3,4)P2PI(4,5)P2
*NS NS
1265
930
758
1292
957
780
PI3K C2p110Vps34
cl. III mutPI3K C2
5-phosphate 4-phosphate ProductPI(3)P,
PI(3,4)P2PI(3,4,5)P3PI(3)PPI(3)P
00.20.40.60.81.01.21.41.6
Rat
io in
tern
aliz
ed/
surf
ace-
bou
nd t
rans
ferr
in
GFP WT cl. IIImutant
*** *** Inte
nsity
of
ind
ivid
ual A
P-2
stru
ctur
es
0
1
2
3
EGFPEGFPPI3K C2 WTcl. III mutantEGFPPI3K C2
***
PI(3)
P
PI(3,
4)P2
PI(3,
4,5)P
3
Figure 3 | PI(3,4)P2 synthesis by PI(3)K C2a at CCPs. a,
Requirement forPI(3)K activity of PI(3)K C2a in CME. Mean intensity
of endocytic AP-2a-containing CCPs in PI(3)K C2a-depleted Cos7
cells expressing EGFP, siRNA-resistant wild-type (WT) or kinase
inactive EGFPPI(3)K C2a (mean6 s.e.m.;n5 3 experiments; *P, 0.05,
**P, 0.01, ***P, 0.001, t-test). b, PI(3)K C2apreferentially
synthesizes PI(3,4)P2. Enzymatic activity of
immunoprecipitated63MycPI(3)K C2a. Data, mean6 s.e.m. normalized to
level of PI(3)Psynthesis (n5 9 experiments; ***P, 0.001, t-test).
No 3-kinase activity wasdetectable in absence of induction of
PI(3)K C2a expression. c, d, Selectivereduction of PI(3,4)P2 in
PI(3)K C2a-depleted cells. c, Loss of plasmamembrane association of
the PI(3,4)P2-sensor 23TAPP1-PHbut not of probesfor other
phosphatidylinositols determined by ratiometric
TIRF/epifluorescentimaging (mean6 s.e.m.; n (experiments)5 9
(23TAPP1-PH), n5 7(23FYVE, a sensor for PI(3)P), n5 4 (PH-PLCd, a
sensor for PI(4,5)P2, andPH-Btk, a sensor for PI(3,4,5)P3); **P,
0.01, ***P, 0.001, t-test). d, Levels ofPI(3,4)P2 or PI(4,5)P2
quantified by PI(3,4)P2- or PI(4,5)P2-specific antibodies(mean6
s.e.m.; n5 6 experiments; *P, 0.05, t-test). e, Alignment
ofsubstrate-binding loop sequences of human
phosphatidylinositol-3-kinasesand a PI(3)K C2a class III-like
mutant (cl. III mut) that can only synthesizePI(3)P but not
PI(3,4)P2. f, g, Requirement for PI(3)K C2a-mediated
PI(3,4)P2synthesis in CME. f, Impaired CME in PI(3)K C2a-deficient
cells is rescued byre-expression of wild-type (WT) but not class
III-like mutant EGFPPI(3)KC2a. Bar diagrams represent the ratio of
internalized (10min, 37 uC) to surfacetransferrin (45min, 4 uC)
(mean6 s.e.m.; n5 3 experiments; ***P, 0.001,t-test compared to
scrambled siRNA). g, Mean intensity of endocytic AP-2a-containing
CCPs in PI(3)K C2a-deficient Cos7 expressingWT or class
III-likemutant PI(3)K C2a (mean6 s.e.m.; n5 3 experiments; ***P,
0.001, t-test).
LETTER RESEARCH
1 1 J U LY 2 0 1 3 | V O L 4 9 9 | N A T U R E | 2 3 5
Macmillan Publishers Limited. All rights reserved2013
-
in PI(3)K C2a-depleted cells was the PX-BAR domain protein
sortingnexin 9 (SNX9) (Supplementary Fig. 6a).We thus analysed the
ability ofSNX9 to associate with phosphatidylinositol liposomes.
EndogenousSNX9 (Fig. 4a) or its PX-BARmodule (Supplementary Fig.
6b) preferen-tially bound to phosphatidylinositol-3-phosphates
including PI(3,4)P2,PI(3)P and PI(3,4,5)P3, but also associated
with PI(4,5)P2 in vitro. Asbinding experiments with purified
proteins might poorly reflect the situ-ation in vivo we directly
compared phosphatidylinositol association ofSNX9with that of other
endocytic proteins in brain extracts. Only SNX9preferred
association with PI(3,4)P2 over PI(4,5)P2, whereas AP180,epsin 1
and AP-2a showed preferential PI(4,5)P2 binding (Fig. 4b, c).Thus,
SNX9 is a putative PI(3,4)P2 effector in CME. To test this,
weanalysed the localization of SNX9 at CCPs in cells depleted of
PI(3)KC2a or PI(3,4)P2. Loss of dynamins results in accumulation of
SNX9assemblies on elongated necks of arrested CCPs15. We confirmed
theenrichment of endogenous SNX9 at AP-2a-coated endocytic
intermedi-ates in cells depletedof dynamin2 (Fig. 4d,
e).Co-silencingofPI(3)KC2awith dynamin2 prevented SNX9 accumulation
at arrestedCCPs (Fig. 4d,e and Supplementary Fig. 7a), whereas
other endocytic proteins accumu-lated irrespective of the presence
of PI(3)KC2a (Supplementary Fig. 7b).Similar effects were caused by
INPP4BCAAX-mediated depletion ofPI(3,4)P2 (Fig. 4f, g and
Supplementary Fig. 7d). Knockdown of SNX9or PI(3)K C2a also
interfered with the formation or stability of ARP2/3-positive
tubular membrane invaginations in dynamin 2-depleted
cells(Supplementary Fig. 7c). Thus, PI(3)K C2a-mediated PI(3,4)P2
produc-tion is required for SNX9 recruitment during late stages of
CME.Previous work has shown that depletion of SNX9 interferes
with
CME in HeLa cells and we confirmed this (Supplementary Fig. 8a).
Inother cell lines (that is, Cos7) SNX9 is functionally redundant
with itsparalogue SNX1825 (ref. 25). In agreement, depletion of
SNX9 andSNX18 in Cos7 cells (Supplementary Fig. 8b) inhibited
transferrin-CME (Fig. 4h) and interfered with CCP dynamics
evidenced by AP-2a accumulation (SupplementaryFig. 8c), similar to
the effects seenupondepletion of PI(3,4)P2 or PI(3)K C2a (compare
Fig. 1e with Fig. 2c, d).Defective CME or AP-2a accumulation were
rescued by siRNA-resistantwild-type EGFPSNX9 but not mutants of
SNX9, in which key residuesrequired for binding to
phosphatidylinositol-3-phosphates (Supplemen-tary Fig. 6c, ref. 26)
had been mutated (Fig. 4h).Thus, PI(3)K C2a via its lipid product
PI(3,4)P2 facilitates enrich-
ment of PI(3,4)P2-binding effector proteins, most notably
SNX9before dynamin-mediated fission. Total internal reflection
fluorescence(TIRF) microscopy analysis indeed revealed that
accumulation ofmCherrySNX9 was delayed by about 20 s with respect
to EGFPPI(3)K C2a, but preceded arrival of dynamin 2 (Fig. 2g, h).
These dataagreewith a spatiotemporal computationalmodel that
suggests amech-anism by which PI(3,4)P2 production at CCPs triggers
selective SNX9recruitment (for details see Schoneberg et al. in
preparation, preprint
athttp://arxiv.org/find/physics/1/au:1Noe_F/0/1/0/all/0/1).The
present work identifies a novel function for PI(3,4)P2, a lipid
previously implicated in the late sustained phase of growth
factorsignalling1,2, in constitutive CME. We show PI(3)K
C2a-mediatedPI(3,4)P2 synthesis to be required for CCP maturation
and for recruit-ment of the PX-BAR domain protein SNX9 to CCPs at a
late stageprecedingdynamin-mediated fission.Our analysis of the
timing of endo-cytic protein arrival at CCPs indicates a hitherto
unknown functionalinterplay betweenPI(4,5)P2 andPI(3,4)P2 in
controllingdistinct stages ofCME in mammalian cells. We further
suggest that the combined activ-ities of PI(4,5)P2-phosphatases9
and of PI(3)K C2a catalyse phosphati-dylinositol conversion from
PI(4,5)P2 to PI(3,4)P2. Phosphatidylinositolconversion regulates
CCPmaturation and constriction andmay therebyprepare endocytic
vesicles for fusionwith PI(3)P-containing endosomes.Similar
conversion mechanisms involving Rab proteins and
phosphati-dylinositols regulate further endosomal progression27.
The identificationof PI(3)K C2a as a major PI(3,4)P2-synthesizing
enzyme will pave theway for the further study of this exciting
lipid in cell physiological pro-cesses other thanCMEand indisease
including cancer13 anddiabetes21,22.
METHODS SUMMARYTotal internal reflection fluorescence (TIRF)
microscopy. TIRF imaging wasperformed using a Zeiss Axiovert200M
microscope equipped with an incubationchamber (37 uC and 5% CO2), a
3100 TIRF objective and a dual-colour TIRFsetup (Visitron Systems)
using Slidebook imaging software (3i Inc.). For analysisof CCP
dynamics, time-lapse series of 3min with a frame rate of 0.5Hz
wererecorded.
a b c
d e
f g h
Inp
ut
PI
PI(3
,4)P
2
PI(3
,5)P
2
PI(4
,5)P
2
PI(3
,4,5
)P3
Bou
ndU
nbou
nd
SNX9
SNX9
-actin
-actin
20 m
g in
put
Con
trol
PI(4
,5)P
2
PI(3
,4)P
2
SNX9
AP180
AP-2
Epsin1
CHC
Rat
ioP
I(3,4
)P2/
PI(4
,5)P
2 b
ound
00.20.40.60.81.01.21.41.61.82.0 *** ** **
Scrambled siRNA PI3K C2 siRNA Dynamin 2 siRNAPI3K C2 +
Dynamin 2 siRNA
Mer
geS
NX
9A
P-2
Tota
l int
ensi
ty o
f SN
X9
stru
ctur
es p
er
m2
05
101520253035404550
Moc
k
Scr
amb
led
PI3
K C
2
Dyn
amin
2D
ynam
in 2
+P
I3K
C2
siRNA
*
Dynamin 2 siRNA
mCherrymCherry
INPP4BCAAXS
NX
9A
P-2
mC
herr
y Tota
l int
ensi
ty o
f SN
X9
stru
ctur
es
0
5
10
15
20
25
mC
herr
y
mC
herr
yIN
PP
4B
CA
AX
mC
herr
y
mC
herr
yIN
PP
4B
CA
AX
ScrambledsiRNA
Dynamin 2siRNA
***
0
0.2
0.4
0.6
0.8
1.0
1.2
Moc
kSc
ram
bled
siR
NA SN
X9si
RN
ASN
X18
siR
NA
SNX9
+18
siR
NA
Rat
io in
tern
aliz
ed/
surf
ace-
bou
nd t
rans
ferr
in
EGFPEGFPSNX9 WTEGFPSNX9 (RYK)EGFPSNX9 (K267N, R327N)
*******
AP18
0SN
X9AP
-2
Epsin
1
Figure 4 | SNX9 is a PI(3,4)P2 effector at CCPs. a, Binding of
endogenousSNX9 from Hek293 cells to liposomes containing 5mol% of
the indicatedphosphatidylinositol in flotation assays. Input, 10mg
protein for bound (top) or30mg (bottom) for unbound fractions
(representative of 3 experiments).b, c, Association of SNX9
affinity-isolated from rat brain extracts withPI(3,4)P2- beads.
Endocytic proteins AP180, AP-2a or epsin 1 preferentiallyassociate
with PI(4,5)P2-beads. Clathrin, negative control. b,
Densitometricquantification of data in a (mean6 s.e.m.; n5 3
experiments; **P, 0.01,***P, 0.001, t-test). d, e, SNX9
accumulation at endocytic intermediatesrequires PI(3)K C2a.
Confocal images of Cos7 cells depleted of PI(3)K C2a,dynamin2, or
both, stained for AP-2a and SNX9. d, Scale bar, 10mm.e,
Quantitative analysis of SNX9 levels at endocytic intermediates as
shown ind (mean6 s.e.m.; n5 3 experiments; *P, 0.05, t-test). f, g,
PI(3,4)P2 isrequired for accumulation of SNX9 at stalled CCPs. f,
Confocal images ofendocytic protein accumulation in dynamin
2-deprived Cos7 cells depleted ofPI(3,4)P2 by mCherryINPP4BCAAX.
Depletion of PI(3,4)P2 preventsaccumulation of SNX9 but not of
AP-2a at endocytic intermediates. Scale bar,10mm. g, Quantification
of SNX9 levels at stalled CCPs as shown inf (mean6 s.e.m.; n5 3
experiments; *P, 0.05, t-test). h, Impaired CME oftransferrin in
Cos7 cells depleted of SNX9 and its close paralogue SNX18 isrescued
by re-expression of wild-type (WT) EGFPSNX9 but not of
PI-bindingdeficient PX-domain mutants RYK (SNX9(R286A, Y287A,
K288); ref. 26) orK267N, R327N (see Supplementary Fig. 6c). Bar
diagrams represent the ratio ofinternalized (10min, 37 uC) to
surface transferrin (45min, 4 uC)(mean6 s.e.m.; n5 5 experiments,
except n5 4 (EGFPSNX9(RYK) andEGFPSNX9(K267N, R327N) and n5 2
(SNX18); **P, 0.01, ***P, 0.001,t-test vs scrambled siRNA).
RESEARCH LETTER
2 3 6 | N A T U R E | V O L 4 9 9 | 1 1 J U LY 2 0 1 3
Macmillan Publishers Limited. All rights reserved2013
-
Electron microscopy. Glutaraldehyde-fixed Cos7 cells treated
with siRNAs werescraped, pelleted, and subsequently processed for
electron microscopy and mor-phometric analysis.Lipid kinase
assays.Kinase activity was assessed by a radioactivity-based assay
(inkinase buffer: 5mMHEPES/KOH pH7.2, 25mMKCl, 2.5mMMgOAc,
150mMKGlu, 10mM CaCl2, 0.2% CHAPS) using recombinant 63mycPI(3)K
C2aimmunoprecipitated from overexpressing HEK293 cells. 200mM
phosphoinosi-tides, 200mM ATP and 8mCi of [c-32P]ATP were combined
with 1 recombinant63mycPI(3)KC2a and incubated at 37 uC for 10min.
Reactionswere stopped byaddition of 500ml coldmethanol:H2O:32%HCl
(10:10:1), followed by lipid extrac-tion and
thin-layer-chromatography (TLC) analysis.
Full Methods and any associated references are available in the
online version ofthe paper.
Received 23 August 2012; accepted 6 June 2013.
Published online 3 July 2013.
1. Di Paolo, G. & De Camilli, P. Phosphoinositides in cell
regulation and membranedynamics. Nature 443, 651657 (2006).
2. Wymann,M. P. & Schneiter, R. Lipid signalling in
disease.Nature Rev. Mol. Cell Biol.9, 162176 (2008).
3. Krauss, M., Kukhtina, V., Pechstein, A. & Haucke, V.
Stimulation ofphosphatidylinositol kinase type I-mediated
phosphatidylinositol (4,5)-bisphosphate synthesis by AP-2m-cargo
complexes. Proc. Natl Acad. Sci. USA 103,1193411939 (2006).
4. Loerke, D. et al. Cargo and dynamin regulate clathrin-coated
pit maturation. PLoSBiol. 7, e57 (2009).
5. McMahon,H. T.&Boucrot,
E.Molecularmechanismandphysiological functions ofclathrin-mediated
endocytosis. Nature Rev. Mol. Cell Biol. 12, 517533 (2011).
6. Zoncu, R. et al. Loss of endocytic clathrin-coated pits upon
acute depletion ofphosphatidylinositol 4,5-bisphosphate.Proc. Natl
Acad. Sci. USA104,37933798(2007).
7. Gruenberg, J. Lipids in endocyticmembrane transport and
sorting. Curr. Opin. CellBiol. 15, 382388 (2003).
8. Antonescu, C. N., Aguet, F., o., Danuser, G. & Schmid, S.
L. Phosphatidylinositol-(4,5)-bisphosphate regulates
clathrin-coated pit initiation, stabilization, and size.Mol. Biol.
Cell 22, 25882600 (2011).
9. Chang-Ileto, B. et al. Synaptojanin 1-mediated PI(4,5)P2
hydrolysis is modulatedbymembrane curvature and facilitates
membrane fission. Dev. Cell 20, 206218(2011).
10. Shin, H.-W. et al. An enzymatic cascade of Rab5 effectors
regulatesphosphoinositide turnover in the endocytic pathway. J.
Cell Biol. 170, 607618(2005).
11. Bae, Y. H. et al.Profilin1 regulates PI(3,4)P2 and
lamellipodin accumulation at theleading edge thus influencing
motility of MDA-MB-231 cells. Proc. Natl Acad. Sci.USA 107,
2154721552 (2010).
12. Dowler, S. et al. Identification of
pleckstrin-homology-domain-containing proteinswith novel
phosphoinositide-binding specificities. Biochem. J. 351, 1931
(2000).
13. Gewinner, C. et al. Evidence that inositol polyphosphate
4-phosphatase type II is atumor suppressor that inhibits PI3K
signaling. Cancer Cell 16, 115125 (2009).
14. Fili, N., Calleja, V., Woscholski, R., Parker, P. J. &
Larijani, B. Compartmental signalmodulation: endosomal
phosphatidylinositol 3-phosphate controls endosomemorphology and
selective cargo sorting. Proc. Natl Acad. Sci. USA 103,1547315478
(2006).
15. Ferguson, S. M. et al. Coordinated actions of actin and BAR
proteins upstream ofdynamin at endocytic clathrin-coated pits. Dev.
Cell 17, 811822 (2009).
16. Rameh, L. E.&Cantley, L. C. The role
ofphosphoinositide3-kinase lipidproducts incell function. J. Biol.
Chem. 274, 83478350 (1999).
17. Gaidarov, I., Smith, M. E., Domin, J. & Keen, J. H. The
class II phosphoinositide3-kinase C2a is activated by clathrin and
regulates clathrin-mediatedmembranetrafficking.Mol. Cell 7, 443449
(2001).
18. Stahelin, R. V. et al. Structural and membrane binding
analysis of the Phoxhomology domain of phosphoinositide
3-kinase-C2a. J. Biol. Chem. 281,3939639406 (2006).
19. Domin, J. et al. Cloning of a human phosphoinositide
3-kinase with a C2 domainthat displays reduced sensitivity to the
inhibitor wortmannin. Biochem. J. 326,139147 (1997).
20. Borner, G. H. H. et al.Multivariate proteomic profiling
identifies novel accessoryproteins of coated vesicles. J. Cell
Biol. 197, 141160 (2012).
21. Falasca, M. et al. The role of phosphoinositide 3-kinase C2a
in insulin signaling.J. Biol. Chem. 282, 2822628236 (2007).
22. Leibiger, B. et al. Insulin-feedback via PI3KC2a activated
PKBa/Akt1 is requiredfor glucose-stimulated insulin secretion.
FASEB J. 24, 18241837 (2010).
23. Pirola, L. et al. Activation loop sequences confer substrate
specificity tophosphoinositide 3-kinase a (PI3Ka). Functions of
lipid kinase-deficient PI3Ka insignaling. J. Biol. Chem. 276,
2154421554 (2001).
24. Subramanian, D. et al. Activation of membrane-permeant caged
PtdIns(3)Pinduces endosomal fusion in cells. Nature Chem. Biol. 6,
324326 (2010).
25. Park, J. et al. SNX18 shares a redundant role with SNX9
andmodulates endocytictrafficking at the plasma membrane. J. Cell
Sci. 123, 17421750 (2010).
26. Yarar, D., Surka, M. C., Leonard, M. C. & Schmid, S. L.
SNX9 activities are regulatedby multiple phosphoinositides through
both PX and BAR domains. Traffic 9,133146 (2008).
27. Rink, J., Ghigo, E., Kalaidzidis, Y. & Zerial, M. Rab
conversion as a mechanism ofprogression from early to late
endosomes. Cell 122, 735749 (2005).
Supplementary Information is available in the online version of
the paper.
AcknowledgementsWe thank E. Ungewickell, P. Di Fiore, P. De
Camilli, H. McMahon,E. Wancker, T. Sudhof and S. Carlsson for
antibodies, L. Cantley, T. Takenawa,M. Wymann, T. Ross, O. Daumke
and W. Yang for plasmids, and O. Daumke, B. Eickoltand F. Wieland
for critical comments. Supported by grants from the
DeutscheForschungsgemeinschaft (SFB 740/C8; SFB 740/D7; SFB
958/A04; SFB 958/A07;SFB 958/Z02).
Author Contributions Y.P., M.E.-G., D.P., M.K. performed
experiments; R.M., S.Z., C.S.provided reagents; A.L. and J.S. aided
with microscopy; Y.P., M.E.-G., J.S., F.N. and V.H.designed
research; F.G. and E.H. contributed reagents; J.S., A.U. and. F.N.
conductedsimulations. Y.P., F.N. and V.H. wrote the manuscript.
Author Information Reprints and permissions information is
available atwww.nature.com/reprints. The authors declare no
competing financial interests.Readers are welcome to comment on the
online version of the paper. Correspondenceand requests for
materials should be addressed to V.H. ([email protected]).
LETTER RESEARCH
1 1 J U LY 2 0 1 3 | V O L 4 9 9 | N A T U R E | 2 3 7
Macmillan Publishers Limited. All rights reserved2013
-
METHODSAntibodies. An overview of all antibodies used in this
study is given in Sup-plementary Table 1.siRNAs. All siRNA
sequences were used as 21-mers or 23-mers including 39-dTdT
overhangs. The sequences of the PI(3)K C2a-targeting siRNAs used in
thisstudy are as follows: siRNA 1 59-ggatctttttaaacctatt-39; siRNA
2 59-gcacaaacccaggctattt-39. The dynamin 2 siRNA sequence used is:
59-gcaactgaccaaccacatc-39.For silencing of SNX9 expression in HeLa
cells, a pool of 4 siRNAs was obtainedfrom Dharmacon (Thermo
Scientific). The SNX9 siRNA sequence used for Cos7cells lies within
the 39-UTR of the mRNA and is: 59-ggacagaacgggccttgaa-39.
Forsilencing of SNX18 expression the siRNA sequence used is:
59-caccgacgagaaagccuggaa-39. The scrambled control siRNA used in
all experiments corresponds tothe scrambled m2 adaptin sequence
59-gtaactgtcggctcgtggt-39.Lipid reagents. Phosphatidylinositols for
lipid binding assays were obtained fromAvanti Polar Lipids,
phosphatidylcholine (PC), phosphatidylserine (PS), and cho-lesterol
were from Sigma-Aldrich, L-a-phosphaditylethanolamine (PE) was
fromJena Bioscience and rhodaminePE was from Avanti Polar
Lipids.Plasmids. The sequence encoding full-length human INPP4B was
amplified fromcDNA provided by L. Cantley and inserted in frame
between the EcoRV and NotIsites of a pcDNA3.1(1)-based
HA-expression vector (sequence between NdeI andEcoRV exchangedwith
that frompcHA2)modified to encode the carboxy-terminalCAAX-box
prenylation sequence from K-ras (KSKTKCVIM-Stop28) directly
fol-lowing the Not I site. The INPP4BCAAX encoding sequence was
subcloned intoan mCherry-expression vector for live cell imaging.
Expression plasmids encodingINPP4BCAAX(C842A), full-length human
MTM1CAAX14, full-length humanPTENCAAX and residues 214614 of human
INPP5ECAAX29 were designedidentically.TheRFP23PHTAPP1 constructwas
a gift fromT.Takenawa and theGFPPH domain of Brutons tyrosine
kinase was provided by M. Wymann. Foranalysis of recruitment of
proteins to CCPs in live cells, a fusion of mRFP to ratclathrin
light chain inserted between KpnI and ApaI of pcDNA5/FRT/TO and
amouse dynamin 2mCherry construct provided by O. Daumke were used
in con-junction with a pEGFPC3PI(3)K C2a construct encoding human
full-lengthPI(3)K C2a assembled from HeLa cDNA (verified by
sequencing). A kinase-inactive mutant of PI(3)K C2a was obtained by
mutating the ATP-binding site(K1138A,D1157A) and the catalytic loop
(D1250A)30. Constructs ofwild-type andkinase-inactive PI(3)KC2a
resistant to siRNA1were generated by creating 4 silentmutations:
59-agatctattcaaaccgatt-39. The PI(3)P-restricted class III mutant
ofPI(3)K C2a resulted from the mutation of 1283KRDR1286 to
1283KPLP1286.For visualization of proteins at CCPs, a 33HAHip1R
construct was provided byT. Ross and a clone encoding epsin 1GFP
was purchased from OriGene Tech-nologies. All constructs encoding
full-length SNX9 or domains thereof werederived from human SNX9
cDNA provided by W. Yang. For GSTPXBAR,cDNA encoding amino acids
204 to 595 of human SNX9 (ref. 31) was cloned inbetween the EcoRI
and NotI sites of pGex-4T-1.Cell lines. All cell lines used (Cos7,
HEK293, HeLa) were obtained from ATCCandnot used beyond passage 30
from original derivation byATCC. Cell lines wereroutinely tested
for mycoplasma contamination.siRNA and plasmid transfections. HeLa
or Cos7 cells seeded on day 0 weretransfectedwith siRNAs
usingOligofectamine (Invitrogen) according to themanu-facturers
instructions on day 1, expanded on day 2, transfected a second time
onday 3, seeded for the experiment onday 4, andused for the
experiment on day 5. Forexpression of recombinant proteins in
knockdown cells, plasmids were transfectedusing lipofectamine 2000
(Invitrogen) according to themanufacturers instructionson day 4. In
the case of INPP4BCAAX constructs, cells were sequentially
trans-fected first with siRNAs and then with plasmid both on day 3
to allow for a totalexpression time of 40h.Upon plasmid
transfection of untreated cells, cells were generally allowed
to
express protein overnight and analysed the next day except for
INPP4BCAAXconstructs where expression for two days was found to
give better results.Transferrin uptake and surface labelling. HeLa
cells treated with siRNAs ortransfected with mCherryINPP4BCAAX were
seeded on Matrigel (BD bios-ciences)-coated coverslips. On the day
of the experiment, cells were serum-starvedfor 1.5 h and used for
either transferrin uptake or transferrin receptor surfacelabelling.
For transferrin uptake, cells were incubated with 25mgml21
transferrinAlexa568 or transferrinAlexa647 (Molecular Probes,
Invitrogen) for 10min at37 uC. After two washes with ice-cold PBS
cells were acid washed at pH5.3 (0.2Msodiumacetate,
200mMsodiumchloride) on ice for 2min to remove
surface-boundtransferrin, washed 2 times with ice-cold PBS and
fixed with 4% paraformaldehyde(PFA) for 45min at room temperature.
For surface labelling, cells were incubatedwith 25mgml21
transferrinAlexa568 at 4 uC for 45min to block endocytosis andlabel
transferrin receptors on the cell surface. Cellswerewashed 3
timeswith ice-coldPBS on ice for onemin and fixed with 4% PFA for
45min at room temperature.
Transferrin labelling was analysed using a Zeiss
Axiovert200Mmicroscope andSlidebook imaging software (3i Inc.).
Internalized transferrin per cell was quan-tified and normalized to
the amount of surface-bound transferrin determined inthe same
experiment as a measure for the efficiency of
internalization.Immunocytochemistry. Staining of proteins in
cultured cells seeded on glasscoverslips was performed as
described32. For lipid antibody stainings, Cos7 cellswere fixed in
2% PFA at room temperature for 20min and permeabilized withsaponin
(30min at room temperature in 0.5% saponin, 1% bovine serum
albumin(BSA) in PBS). Cells were labelled with lipid-specific
antibodies (see Supplemen-tary Table 1) diluted in 1%BSA in PBS for
2 h at room temperature. After washingthree times for 5min with
PBS, cells were incubated with appropriate fluorescentsecondary
antibodies for 1 h andwashed three times 10minwith PBS. Protein
andlipid immunocytochemistry stainings were routinely analysed and
quantifiedusing a spinning disk confocal microscope (Ultraview ERS,
Perkin Elmer) andVolocity imaging software (Improvision, Perkin
Elmer).Total internal reflection fluorescence (TIRF) microscopy.
TIRF imaging wasperformed using a Zeiss Axiovert200M microscope
equipped with an incubationchamber (37 uC and 5% CO2), a 3100 TIRF
objective and a dual-colour TIRFsetup from Visitron Systems using
Slidebook imaging software (3i Inc.). For ana-lysis of CCP
dynamics, time-lapse series of 3min with a frame rate of 0.5Hz
wererecorded. CCP lifetimes were assessed by arbitrarily selecting
50 or 25 CCPs percell in the centre frame of the time-lapse series
and determining the frame ofappearance and disappearance. In case
CCPs already existed in the first frameor persisted until the last
frame, these frames were counted. For the analysis ofrecruitment
time courses of proteins to CCPs, only CCPs were used that
bothappeared and disappeared within the time lapse series. From
these, fluorescenceintensities over time were quantified and
aligned on the time axis by the last frameof GFPPI(3)K C2a presence
(t5 0, fission). Fluorescence intensities for all timepoints in
relation to t5 0 were averaged over all CCPs in the analysis and
renor-malized to the resulting peak value. For analysis of GFPPHBtk
membrane asso-ciation, TIRF and epifluorescence images of the same
cell were acquired and theTIRF fluorescence intensity was
normalized to the epifluorescence signal in orderto achieve
intrinsic correction for expression level variations between
cells.Fluorescence recovery after photobleaching (FRAP). FRAP
experiments wereperformed using a spinning disk confocal microscope
equipped with an incuba-tion chamber (37 uC) and a photokinesis
unit (Ultraview ERS, Perkin Elmer). Oneto three regions of interest
in the peripheral, flat part of an EGFPCLC expressingcell were
selected. A time-lapse series at 0.5Hzwas recorded with 10 frames
beforeand 60 frames after bleaching. For quantification, the sum
EGFPCLC fluor-escence intensity at CCPs was quantified over time
and normalized to the meansum intensity during the pre-bleaching
period.PIP/AM experiments. Cell-permeable acetoxy methylester
(AM)-protectedphosphatidylinositol derivatives were synthesized as
described33. For treatmentof cells, PI(3)P/AM or PI(3,4)P2/AM
dissolved in dry DMSO were mixed with anequal volume of 10%
pluronic F127 in DMSO (Sigma-Aldrich) to enhance solu-bility in
aqueous buffers and diluted inDMEM to a final concentration of
200mM.Cells on coverslips were treated with DMSO1 pluronic
(control) or PIP /AMs for10min at 37 uC and then processed for
immunocytochemistry as described above.Electron microscopy.
Ultrastructural analysis was performed as
described34.Glutaraldehyde-fixed Cos7 cells treated with siRNAs
were scraped, pelleted, andsubsequently processed for electron
microscopy and morphometric analysis aspreviously described15.
Briefly, after epoxy resin embedding and sectioning,micro-graphs
were taken along the cell perimeter at320,000magnification. Images
werecombined to reconstruct the cell perimeter and numbers of
clathrin-coated inter-mediates were determined.Purification of
clathrin-coated vesicles. CCVs were purified essentially
asdescribed35. Briefly, calf brain was homogenized and the
cytosolic andmicrosomalfraction was obtained by sequential
centrifugation at 17,000g and 30,000g. Lightmembranes were pelleted
at 150,000g, resuspended and mixed with an equalvolume of 12.5%
Ficoll, 12.5% sucrose solution to adjust the density of the
solutionto that of CCVs. Contaminating, heavier membranes were
removed by centrifu-gation at 90,000g and the CCV-containing
supernatant was diluted in order toallow sedimentation of CCVs at
150,000g. For stripping of coat proteins, purifiedCCVswere
incubated over night at room temperaturewith 0.8MTris-HCl pH7.4to
disrupt proteinmembrane interactions. Vesicles including integral
membraneproteins were sedimented at 250,000g.In vitro kinase
activity assays. Kinase activity of recombinant PI(3)K C2a
wasassessed using a radioactivity-based assay essentially as
described36. In brief, one10-cm dish of HEK293 cells transiently
overexpressing 63myc-PI(3)K C2a waslysed in immunoprecipitation
(IP) buffer (20mM HEPES, 100mM KCl, 2mMMgCl2, 1% CHAPS, 1mM PMSF,
0.3% protease inhibitor cocktail (Sigma)), andcentrifuged for 10min
at 20,500g at 4 uC. The resulting supernatant was centri-fuged at
65,000 r.p.m. in a TLA-110 rotor (Beckman Coulter). PI(3)K C2a
was
RESEARCH LETTER
Macmillan Publishers Limited. All rights reserved2013
-
immunoprecipitated from the extract using,15mg c-myc antibody
and 1.5mg ofprotein as IP input. The IP was washed twice with IP
buffer and once in kinasebuffer (5mM HEPES/KOH pH7.2, 25mM KCl,
2.5mM magnesium acetate(Mg(CH3COO)2), 150mM KGlu, 10mM CaCl2, 0.2%
CHAPS). Phosphati-dylinositols were dissolved in kinase buffer
(note that presence of 0.2% CHAPSwas required for full solubility
of PI(4)P), incubated on ice for 30min, sonicatedfor 1min using a
small tip sonicator (Bandelin Sonoplus) 1 s on 1 s off at
70%intensity. 200mMphosphoinositides, 200mMATP and 8mCi of
[c-32P]ATP werecombined with 1/8 of one IP sample and incubated at
37 uC for 10min. Thereactions were stopped by the addition of 500ml
cold methanol:H2O:32%HCl(10:10:1) and lipid extraction and
thin-layer-chromatography (TLC) analysis wereperformed as
described36.Liposome flotation assay. Liposome preparation. A total
of 600800mg of lipidswere dissolved in a mixture of
CHCl3:methanol:1N HCl (2:1:0.01) to the desiredconcentration,
combined in a glass vial and dried under pressurized N2 followedby
vacuum for 30min. Liposomes were rehydrated in 300ml HEPES
buffer(50mM HEPES pH7.4, 100mM KCl (140mM KCl for experiments with
GFPSNX9WT or K267N, R327N; 200mMKCl for experiments with GSTSNX9
PXBAR)) for 1 h at room temperature under frequent vortexing. After
the addition of1.7ml H2O the liposomes were centrifuged in a TLA110
rotor at 20,000 r.p.m. at4 uC for one hour. The resulting pellet
was resuspended in HEPES buffer to a finallipid concentration of
3mgml21. Liposome mixtures were extruded 14 timesthrough an
800-nmpolycarbonatemembrane (Whatman) using amanually oper-ated
extruder (LiposoFast, Avestin, Inc.). The final concentration of
lipid species inmol% were: 50%PC, 20% cholesterol, 19%PE, 1%
rhodaminePE, 5% PS, and 5%phosphatidylinositols.Flotation assay.
450mg liposomes in HEPES buffer were combined with either
2mg of purified GSTSNX9 PX-BAR protein or with 30mg of HEK293
cell extractcontaining overexpressed HASNX9 PX-BAR or GFPSNX9
full-length proteinand incubated for 15min at 4 uC on a rotating
wheel. The mixture was thenadjusted to 30% sucrose by adding 75%
sucrose in HEPES buffer and transferredto a TLS-55 centrifuge tube.
This was overlaid with 200ml of 25% sucrose inHEPES buffer followed
by 50ml of HEPES buffer. Liposomes were floated bycentrifuging one
hour at 55,000 r.p.m. (,240,000g) in a TLS 55 swing out
rotor(Beckman Coulter). The fractions were collected using a
blunt-ended needleattached to a calibrated syringe by removing the
bottom layer first (,250ml totalvolume), followedby themiddle layer
(200ml) and in the end the top layer contain-ing the liposomes and
any bound protein. Top and bottom fractions were sepa-rated on 8%
acrylamide gels and stained with Coomassie for GSTSNX9 PX-BARand
immunoblotted for HASNX9 PX-BAR or GFPSNX9 full length.PIP
bead-based affinity purification. Agarose PIP Beads (Echelon
Biosciences)containing 10 nanomoles of bound PI(4,5)P2 or PI(3,4)P2
were used to pull down
proteins fromrat brain extract. Rat brain extractwas prepared
from2.5 g frozen ratbrain homogenized in homogenization buffer (4mM
HEPES pH7.4, 320mMsucrose, 1mM PMSF, 0.3% protease inhibitor
cocktail) using 13 strokes of aglassTeflon-homogenizer at 900
r.p.m. The homogenate was centrifuged at900g for 10min at 4 uC. To
the supernatant PIP pull-down buffer (20mMHEPES pH7.4, 50mM NaCl,
0.25% NP-40) was added to 13 concentrationand incubated on ice for
30min followed by centrifugation at 43,500g at 4 uC.The supernatant
was centrifuged again at 265,000g for 15min at 4 uC to
removeaggregated proteins. 68mgof proteinwere added to 100ml of 1:1
washed agarose-bead slurry and incubated at 4 uC for 1.5 h on a
rotatingwheel. Beadswere pelleted,washed 3 times with PIP pull-down
buffer and bound proteins were eluted twotimes in 30ml of 13Laemmli
sample buffer. 30ml of the pooled eluate were thenloaded onto an 8%
acrylamide gel for SDSPAGE followed by immunoblotting.Statistical
methods. For analyses comprising multiple independent
experiments(n), sample size within each experiment was chosen to
provide statistically signifi-cant estimates for each sample,
corresponding to 20 to 40 images per sample formicroscopy-based
quantifications. In all experiments, cells were arbitrarily
chosenbased on the signal in a separate channel independent from
the signal to bequantified. All statistical tests performed were
two-tailed, unpaired t-tests asjudged appropriate for the
respective experiments.
28. Malecz, N. et al. Synaptojanin 2, a novel Rac1 effector that
regulates clathrin-mediated endocytosis. Curr. Biol. 10, 13831386
(2000).
29. Varnai, P., Thyagarajan, B., Rohacs, T. & Balla, T.
Rapidly inducible changes inphosphatidylinositol 4,5-bisphosphate
levels influence multiple regulatoryfunctions of the lipid in
intact living cells. J. Cell Biol. 175, 377382 (2006).
30. Gaidarov, I., Zhao, Y. & Keen, J. H. Individual
phosphoinositide 3-kinase C2adomain activities independently
regulate clathrin function. J. Biol. Chem. 280,4076640772
(2005).
31. Pylypenko, O., Lundmark, R., Rasmuson, E., Carlsson, S. R.
& Rak, A. ThePX-BARmembrane-remodeling unit of sorting nexin 9.
EMBO J. 26, 47884800(2007).
32. Maritzen, T. et al. Gadkin negatively regulates cell
spreading and motility viasequestration of the actin-nucleating
ARP2/3 complex. Proc. Natl Acad. Sci. USA109, 1038210387
(2012).
33. Laketa, V. et al.Membrane-permeant phosphoinositide
derivatives asmodulators of growth factor signaling and neurite
outgrowth. Chem. Biol. 16,11901196 (2009).
34. von Kleist, L. et al. Role of the clathrin terminal domain
in regulating coated pitdynamics revealed by small molecule
inhibition. Cell 146, 471484 (2011).
35. Campbell, C., Squicciarini, J., Shia, M., Pilch, P. F. &
Fine, R. E. Identification of aprotein kinase as an intrinsic
component of rat liver coated vesicles.Biochemistry23, 44204426
(1984).
36. Wieffer,M.,Haucke,
V.&Krauss,M.Regulationofphosphoinositide-metabolizingenzymes by
clathrin coat proteins.Methods Cell Biol. 108, 209225 (2012).
LETTER RESEARCH
Macmillan Publishers Limited. All rights reserved2013
TitleAuthorsAbstractMethods SummaryTotal internal reflection
fluorescence (TIRF) microscopyElectron microscopyLipid kinase
assays
ReferencesMethodsAntibodiessiRNAsLipid reagentsPlasmidsCell
linessiRNA and plasmid transfectionsTransferrin uptake and surface
labellingImmunocytochemistryTotal internal reflection fluorescence
(TIRF) microscopyFluorescence recovery after photobleaching
(FRAP)PIP/AM experimentsElectron microscopyPurification of
clathrin-coated vesiclesIn vitro kinase activity assaysLiposome
flotation assayPIP bead-based affinity purificationStatistical
methods
Methods ReferencesFigure 1 PI(3,4)P2 regulates CME.Figure 2
PI(3)K C2a controls maturation of CCPs.Figure 3 PI(3,4)P2 synthesis
by PI(3)K C2a at CCPs.Figure 4 SNX9 is a PI(3,4)P2 effector at
CCPs.