You are on page 1of 27

ARTICLES

Syndecan–syntenin–ALIX regulates the biogenesis


of exosomes
Maria Francesca Baietti1,2,3,5 , Zhe Zhang1,3,5,6 , Eva Mortier2,5 , Aurélie Melchior1,3 , Gisèle Degeest1,3 ,
Annelies Geeraerts2,4 , Ylva Ivarsson2 , Fabienne Depoortere2 , Christien Coomans1,3 , Elke Vermeiren2 ,
Pascale Zimmermann2,7 and Guido David1,3,7

The biogenesis of exosomes, small secreted vesicles involved in signalling processes, remains incompletely understood. Here, we
report evidence that the syndecan heparan sulphate proteoglycans and their cytoplasmic adaptor syntenin control the formation of
exosomes. Syntenin interacts directly with ALIX through LYPX(n)L motifs, similarly to retroviral proteins, and supports the
intraluminal budding of endosomal membranes. Syntenin exosomes depend on the availability of heparan sulphate, syndecans,
ALIX and ESCRTs, and impact on the trafficking and confinement of FGF signals. This study identifies a key role for
syndecan–syntenin–ALIX in membrane transport and signalling processes.

Exosomes are small secreted vesicles thought to play key roles Signalling-induced ubiquitylation of membrane proteins triggers
in cell-to-cell communication1–6 . Exosome formation starts in their sorting to ILVs, thus terminating intracellular signalling by these
endosomes, by budding of the limiting endosomal membrane, away proteins. Many of the signalling events that occur at cell surfaces require
from the cytosol, generating vesicle-laden endosomes, referred to as the assistance of heparan sulphate, for instance the engagement of
multivesicular endosomes (MVEs) and multivesicular bodies (MVBs). fibroblast growth factor (FGF) with its receptor21,22 (FGFR). Most
Fusion of these late endosomes with the plasma membrane releases the of the cell-surface-bound heparan sulphate is supplied by syndecans,
intraluminal vesicles (ILVs) as exosomes1–7 . ubiquitous transmembrane proteins that are of key importance
The endosomal-sorting complex required for transport (ESCRT) ma- for development, angiogenesis, carcinogenesis, inflammation and
chinery regulates membrane budding, at cell surfaces8,9 and at the level infection23,24 . Syndecans have a short, evolutionarily conserved
of late endosomes10,11 . ESCRT-0, -I and -II contain ubiquitin-binding cytoplasmic domain, which binds cytosolic factors25–27 such as the
subunits that sort ubiquitylated membrane proteins to specific domains small PDZ scaffolding protein syntenin28,29 (also known as SDCBP).
of endosomes and into the endosomal invaginations. In vitro recon- Here, we show that the connection of the syndecans (also known as
stitution experiments suggest that ESCRT-I and ESCRT-II together SDC1-4) with syntenin, and through syntenin with ALIX supports the
support budding processes and tentatively identify ESCRT-III as the biogenesis of ILVs and exosomes, and the segregation of signalling
complex that cleaves the buds to form ILVs (refs 12,13). Yet, further cargo to these vesicles.
mechanisms must exist14–16 . ALIX (also known as PDCD6IP), a protein
that interacts with several ESCRT (TSG101 and CHMP4) proteins is RESULTS
thought to participate in the budding and abscission processes17,18 . Syntenin interacts with ALIX, mimicking retroviral late domains
Importantly, ALIX also interacts directly with Leu-Tyr-Pro-X(n)-Leu Syntenin has a simple domain structure: two PDZ domains,
(LYPX(n)L) late domains in retroviral membrane proteins, allowing flanked by a 100-amino-acid amino-terminal domain and a small
enveloped viruses to hijack cellular mechanisms of membrane-domain carboxy-terminal domain (Fig. 1a). The two PDZ domains are
biogenesis to egress from cells19,20 . Thus, both topologically and mecha- required for high-affinity interaction with the syndecans and for the
nistically, retroviral budding seems related to ILV/exosome formation. recruitment of syntenin to membranes30–33 . Screening yeast two-hybrid

1
Laboratory for Glycobiology and Developmental Genetics, Department of Human Genetics, KULeuven, Campus Gasthuisberg, Herestraat 49, 3000 Leuven, Belgium.
2
Laboratory for Signal Integration in Cell Fate Decision, Department of Human Genetics, KULeuven, Campus Gasthuisberg, Herestraat 49, 3000 Leuven, Belgium.
3
Center for the Biology of Disease, VIB, 3000 Leuven, Belgium. 4 Laboratory for Biomolecular Dynamics, Department of Chemistry, KULeuven, 3001 Heverlee,
Belgium. 5 These authors contributed equally to this work. 6 Present address: State Key Laboratory of Molecular Biology, Institute of Biochemistry and Cell Biology,
Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200031, China.
7
Correspondence should be addressed to P.Z. or G.D. (e-mail: pascale.zimmermann@med.kuleuven.be or guido.david@cme.vib-kuleuven.be)

Received 15 June 2011; accepted 16 April 2012; published online 3 June 2012; DOI: 10.1038/ncb2502

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012 677


© 2012 Macmillan Publishers Limited. All rights reserved.
ARTICLES

a b IP c Syntenin GST–ALIX e f lys exo


5,000 5 µM

Anti-syntenin
100

Response (RU)
2.5 µM SDC1CTF
3,800 1.25 µM

Anti-ALIX

Particles (× 106 ml–1)


Syntenin Syntenin
2,600 GST 5 µM 80
N-ter PDZ-1 PDZ-2 C-ter

Input
ALIX

IgG
1 109 194 273 298
1,400 60
CD63
ALIX 0
40
--LYPSLEDL--------------LYPRLYPELSQYMGL--
Syntenin 400 800 1,200 HSP70
20
Time (s) Calnexin
d Syntenin

Normalized response
ALIX 1.0 0
GM130
CD63 50 150 250

Response (RU)
Bro1 V PRD 400
1 360 702 868 SDC2 Size (nm) LAMP-2

200 0.5
CHMP4 CD63
HIV p6 TSG101
Src SH2 EIAV p9 Src SH3 SDC2
Syntenin ALG-2 0
Endophilin 0
0 100 200 300 0.01 10 1,000
SETA/CIN85
Time (s) Syntenin (µM)
g SDC1CTF
i Percentage
Syntenin 250 j IP Control CD63 0 20 40 60 80 100 IP

Beads

Beads
ALIX CD63

Wash

Wash
200

Control
1 2 3 4 5 6 7 8 9 10

SN

SN
SDC1CTF

h SDC1CTF Diameter (nm) 150 SDC1CTF


Syntenin
Flotillin-1
ALIX Syntenin
100 CD63

CD63
CD63 SDC1CTF
CD63 50 Syntenin

Flotillin-1 Flotillin-1
Flotillin-1 0
1 2 3 4 5 6 7 8 9 10 Beads SN Wash

Figure 1 Syntenin interacts with ALIX, adapting ALIX to syndecan, and blots, illustrating the presence of syndecan 1 CTF (SDC1CTF), syntenin,
co-fractionates with these in exosomes. (a) Schematic representation of ALIX, CD63 and HSP70 in MCF-7 cell lysate (lys) and HSC exosomes
syntenin (top) and ALIX (bottom). Numbers indicate amino-acid residues. (exo). Note that there is no detectable contamination of the exosomes by
LYPX(n)L motifs that mediate interaction with ALIX are highlighted. ALIX GM130 (Golgi), calnexin (endoplasmic reticulum) or LAMP-2 (lysosomes).
binding partners are indicated below the respective binding domains. (g,h) Exosome fractionation by density-gradient centrifugation. SDC1CTF,
(b) ALIX co-precipitates with syntenin. Detergent extracts of microsomal CD63, syntenin and ALIX occur in fractions corresponding to densities
membranes were incubated with control IgG or specific antibody. IP, of 1.094 to 1.143 g ml−1 . Note that flotillin-1 floats at higher buoyancy.
immunoprecipitate. (c) Biacore experiments showing, first, the interaction (i) Ultrastructural appearance (left) and size distribution (right) of particles
of non-tagged recombinant syntenin with immobilized syndecan 2 or co-purifying with syntenin in the density gradient. Scale bar, 50 nm. The
SDC2 (left part of the sensorgram) and, subsequently, that of GST–ALIX diameter is 80 ± 46 nm (mean, s.d.; n = 137). (j) Exosome subfractionation
with SDC2:syntenin complexes (right part of the sensorgram; perfusion by CD63 content. Adsorption to magnetic beads substituted with anti-CD63
of syntenin stops at 600 s; GST or GST–ALIX perfusion starts at 800 s). or control antibody was measured by western blotting (example on the left).
(d) Sensorgrams (left panel) comparing the binding of 2.5 µM syntenin The histogram (right) represents the mean distributions of the markers over
to SDC2 (black line) and CD63 (grey line), and normalized equilibrium the beads, supernatants (SN) and washes (the sum of the three fractions
binding isotherms (right panel) yielding apparent Kd values of 1.8 ± 0.2 µM taken as 100%), calculated from three independent immunofractionation
and 24 ± 4 µM, for SDC2 and CD63, respectively. Data were collected experiments. Note that syndecan and syntenin largely co-fractionate with
from two replicates. (e) Size distribution of the particles that compose CD63 (P < 0.001, t -test). Flotillin-1 does not. Uncropped images of
HSC exosomes, as measured by nanoparticle tracking analysis. (f) Western blots/gels are shown in Supplementary Fig. S6.

experiments, we found that the N-terminal domain of syntenin controlled by syndecan:syntenin:ALIX complexes might ‘resemble’
interacts with ALIX (Supplementary Fig. S1a). The interaction is viral budding and pertain to the formation of ILVs and exosomes.
direct, as shown by the co-immunoprecipitation of the endogenous
proteins (Fig. 1b) and by in vitro binding assays (Supplementary Fig. Syndecan, syntenin and ALIX co-accumulate in exosomes
S1b,c). Importantly, syndecans, syntenin and ALIX form tripartite Proteomic analyses have identified ALIX and syntenin in exosomes of,
complexes in vitro (Fig. 1c). for example, dendritic cells37 . Together with CD63, ALIX and syntenin
The N-terminal domain of syntenin contains three LYPX(n)L are among the 20 proteins most frequently identified in exosomes38 .
motifs (Fig. 1a), resembling ALIX-binding sites identified in HIV-1 CD63 is a tetraspanin that, similarly to syndecans, interacts with the
p6 (LYPLTSL) and in EIAV p9 (LYPDL), late domains involved PDZ domains of syntenin39 . Yet, we found the affinity of syntenin for
in viral budding19,20 . Mutating the YP residues to alanines in the CD63 (Kd = 24 µM) to be ∼10-fold lower than its affinity for syndecans
three motifs was necessary and sufficient to abolish syntenin–ALIX (Fig. 1d). We first investigated whether syndecan, syntenin and ALIX
interaction (Supplementary Fig. S1a). HIV-1 p6 and EIAV p9 interact co-accumulate in exosomes. We used MCF-7 cells, as these cells contain
with the V-domain of ALIX, with the binding centred on Phe readily detectable amounts of SDC1, SDC4 and syntenin, and release
676 (refs 34–36). Similarly, syntenin interacts with the V-domain numerous nanoparticles that are exosome-like in size (diameter =
of ALIX (Supplementary Fig. S1a), and substituting aspartate for 88 ± 13 nm; mean, s.d.; Fig. 1e). As purified by differential high-speed
Phe 676 in ALIX completely abolishes syntenin–ALIX binding centrifugation (HSC), these particles contained membrane-spanning
(Supplementary Fig. S1d). This suggests that cellular processes C-terminal syndecan fragments (CTFs), syntenin, ALIX and several

678 NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012


© 2012 Macmillan Publishers Limited. All rights reserved.
ARTICLES

Syntenin ΔALIX
a b c SDC1CTF Syntenin CD63
n= 3 3 3 5 7 4 22 2
– Syntenin

Syntenin
500

ALIX

ALIX

(relative to control)
RNAi

nt

nt
400

Percentage

SDC1CTF SDC1CTF 300
Syntenin
Syntenin 200

ALIX 100
ALIX

exo
CD63 CD63 Syntenin ΔALIX WT WT
exo

ALIX RNAi – – +

HSP70 HSP70 d mGFP–CD63 exosomes


TSG101 – Syntenin
TSG101
150
Flotillin-1 ALIX

Intensity (kHz)
Syntenin 100
lys

SDC1CTF
SDC1CTF
lys

Syntenin 50
β-actin

1 300 600 900


Peak rank number

Figure 2 Syntenin gain enhances exosome yields. MCF-7 cells were WT, wild type. Error bars represent s.d. (d) Exosome quantification by
sham-transfected (-), or transfected to transiently overexpress wild-type confocal fluorescence spectroscopy. MCF-7 cells were transfected with
syntenin (syntenin) or (triple LYP-LAA) mutant syntenin defective mGFP–CD63, and with or without non-tagged syntenin. HSC exosomes
for binding to ALIX (syntenin 1 ALIX), and were treated with ALIX were examined by fluorescence spectroscopy for 10 s, 50 times. The
RNAi, non-targeting RNAi (nt) or no RNAi. Total cell lysates (lys) graph shows the distributions of the fluorescent peaks (corresponding
and corresponding HSC exosomes (exo) were analysed by western to individual exosomes), ranking all collected peaks by increasing
blotting, testing for different exosomal markers, using cognate antibodies. fluorescence intensity (kHz). Note that syntenin overexpression (black
(a,b) Representative experiments, illustrating the effect of syntenin distribution) increases (more than doubles) the number of peaks, but not
overexpression (a,b) and of ALIX RNAi (b). (c) Histograms representing their intensities. Data shown are from one representative experiment;
mean signal intensities (relative to signals in corresponding controls; that similar results (fold increases) were obtained in three independent
is, cells treated with empty vector, nt RNAi or both empty vector and experiments (see Supplementary Fig. S3c). Uncropped images of blots
nt RNAi), calculated from n independent experiments (as indicated). and gels are shown in Supplementary Fig. S6.

other established exosomal markers, including CD63 (Fig. 1f). Some level of exosomal syndecan increased, along with exosomal syntenin
intact syndecan could be detected when syndecans were overexpressed, and ALIX, and other exosomal markers such as CD63 and HSP70
but CTFs predominated (Supplementary Fig. S2a). Exosome formation (Fig. 2a,c). Similar increases did not occur when syntenin harboured
depends on ceramide40 . RNA-mediated interference (RNAi) targeting point mutations in the three ALIX-binding LYPX(n)L motifs (Fig. 2a,c),
neutral sphingomyelinase-2 decreased the amounts of syndecan, CD63, or when syntenin was overexpressed in the presence of ALIX RNAi,
syntenin and ALIX sedimented by HSC, tentatively identifying the confirming a specific involvement of ALIX (Fig. 2b,c). ALIX RNAi
particles that contain these markers as exosomes (Supplementary (Figs 2b and 3d) or syntenin RNAi (Fig. 3a,d) led to a significant
Fig. S2b). During density-gradient centrifugation, exosomal syndecan decrease in exosomal syndecan, CD63 and HSP70. Conversely,
co-fractionated with syntenin, ALIX and CD63 (Fig. 1g,h), and with syndecan depletion reduced the level of exosomal accumulation
vesicles, mostly with a shape and size characteristic of exosomes of syntenin, ALIX, CD63 and HSP70 (Fig. 3b,d). Knockdown of
(Fig. 1i). Yet all were slightly offset from flotillin-1 (Fig. 1h). CD63, which depends on syndecan and syntenin for accumulating in
Immunofractionation indicated that most of the exosomal syndecan, exosomes, had little effect on exosomal syndecan, syntenin, ALIX and
syntenin and ALIX, yet very little of the exosomal flotillin-1, HSP70 (Fig. 3c,d). Of note, syndecan or syntenin RNAi had no effect
accumulates in CD63-positive vesicles (Fig. 1j; ALIX not shown). on the exosomal accumulation of flotillin-1 (Fig. 3a,b), but markedly
In turn, two-colour fluorescence cross-correlation spectroscopy reduced the recovery of total biotinylated exosomal protein from
indicated that most CD63-positive exosomes contain syntenin cell-surface-biotinylated cells (Supplementary Fig. S3a). Importantly,
(Supplementary Fig. S2c,d). Together this suggests that syndecan, the effects of syndecan RNAi were largely prevented by the expression
CD63, syntenin and ALIX largely co-segregate to a distinctive (flotillin- of exogenous syndecan (Fig. 3e,f). On the contrary, in combination
1-negative) class of vesicles. with syndecan RNAi, the overexpression of CD63 further decreased the
exosomal yields of syntenin and ALIX (Fig. 3e,f). These results reveal a
Syndecan, syntenin and ALIX support the biogenesis specific role for syndecan–syntenin–ALIX in the biogenesis of a major
of exosomes class of exosomes, the loading of exosomes with specific cargo, or both.
We then investigated the significance of syndecan, syntenin and To evaluate these possibilities, we analysed HSC exosomes by
ALIX for exosome formation. When syntenin was overexpressed, the nanoparticle tracking analysis, identifying a specific ∼2-fold increase

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012 679


© 2012 Macmillan Publishers Limited. All rights reserved.
ARTICLES

a b c e RNAi nt SDC1/4

Syntenin

SDC1/4
SDC1
SDC4

SDC1

SDC1
CD63

CD63
CD63
RNAi RNAi RNAi

nt

nt

nt


SDC1CTF SDC1CTF SDC1CTF SDC1CTF
Syntenin Syntenin Syntenin Syntenin
ALIX ALIX Alix

exo
CD63

exo
CD63
exo

exo
CD63 CD63

HSP70 Alix
HSP70 HSP70
TSG101 TSG101
TSG101
Flotillin-1 SDC1
Flotillin-1 Flotillin-1
Syntenin
Syntenin
ALIX

lys
nt

SDC1CTF
SDC1CTF CD63 CD63

lys
lys

SDC4
lys

Syntenin
SDC1 ALIX
ALIX ALIX

d f n= 3 4 3 3 3 1 5 5 2 5 4 2 4 5 1
n=
Percentage (relative to control)
6 6 6 6 4 3 3 4 3 3 4 5
Percentage (relative to control)

250 SDC1CTF
100 Syntenin
SDC1CTF 200 CD63
80
Syntenin
60 150
CD63
40 100

20 50

RNAi ALIX Syntenin SDC1/4 CD63


OE SDC1 CD63 – SDC1 CD63
RNAi nt SDC1/4

Figure 3 Syndecan–syntenin–ALIX depletion impairs exosome yields. experiments (as indicated). Cell treatments are indicated at the bottom.
(a–d) Depletion experiments. MCF-7 cells were treated with various RNAi, Note that the depletion of CD63 has less effect on exosomal markers than
targeting syntenin (a,d), syndecan (b,d), CD63 (c,d) or ALIX (d, and the reductions of syndecan, syntenin or ALIX, and that extra CD63 cannot
Fig. 2b). Non-targeting RNAi (nt) was used as a control. (e,f) Rescue compensate for a loss of syndecan. Note also the significant effects of single
experiments. Cells were sham-transfected (-) or transfected to overexpress syndecan depletions, indicating that the different syndecans have either
(OE) either syndecan (mouse SDC1) or CD63 (mGFP–CD63), and were non-redundant roles in exosome formation, or are individually limiting (that
treated with nt-RNAi or RNAi targeting endogenous syndecans. Total cell is, that the remaining syndecan does not reach the required threshold).
lysates (lys) and HSC exosomes (exo) were analysed by western blotting, Note that while SDC1 levels are maintained in cells, syndecan 4 RNAi also
testing for specific proteins using cognate antibodies. Histograms (d,f) decreases the level of SDC1CTF in exosomes, indicating that the latter or a
represent mean signal intensities (and s.d.) for SDC1CTF (black), syntenin combination of both is the case. Uncropped images of blots/gels are shown
(grey) and CD63 (shaded) in exosomes, calculated from n independent in Supplementary Fig. S6.

in the number of particles with a diameter of 30–110 nm on co-localization of mCherry–syntenin and syndecan intracellular
overexpression of syntenin, and a ∼50% reduction on knockdown domain (ICD), including their accumulation, along with ALIX, ‘inside’
of syndecan, syntenin or ALIX (Supplementary Fig. S3b). By the lumen of enlarged endosomes that were decorated with cerulean
confocal fluorescence spectroscopy, assessing the effects of syntenin (cer)-tagged RAB5Q79L (Fig. 4b and Supplementary Fig. S4). Lumina
overexpression on the detection of mGFP–CD63 in HSC exosomes, we filled with mCherry–syntenin were also filled with CD63, syndecan-
found an increase (2–3-fold) in the number, but no change in the mean ectodomain outlining only the rim of these endosomes (Fig. 4b). When
amplitude, of the fluorescent peaks (Fig. 2d and Supplementary Fig. such cells were incubated with anti-CD63, many of the enlarged
S3c). We concluded that syntenin controls the number of exosomes endosomes were filled with antibody (Fig. 4c). In contrast, without
made/released, rather than the level of (CD63) cargo the modal concomitant mCherry–syntenin expression, internalized anti-CD63
exosome is loaded with. was mostly restricted to the limiting membranes of the endosomes
Next, we examined whether syntenin–ALIX supports budding (Fig. 4c). Together, these results suggest a role for syntenin in the intra-
processes at endosomal membranes. Overexpression of RAB5Q79L , luminal budding of endosomal membrane domains that contain CD63
which impairs early intra-endosomal trafficking and causes endosomes and syndecan, the latter mostly as syndecan-CTF. The knockdown of
to enlarge41 , severely reduced the exosomal release of syndecan, CD63, RAB7, which replaces RAB5 on endosomal membranes42 , regulating
syntenin and ALIX (Fig. 4a). Confocal microscopy revealed extensive late endosomal traffic downstream of MVB (ref. 43), also severely

680 NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012


© 2012 Macmillan Publishers Limited. All rights reserved.
ARTICLES

a RAB5Q79L – + b Merge CD63 cer–RAB5Q79L mCh–syntenin e RNAi


SDC1CTF Syntenin – + + –
Syntenin RAB7 – – + +
exo

Alix Syntenin
CD63 RAB7
Flotillin-1
Merge SDC1 ICD cer–RAB5Q79L mCh–syntenin
SDC1CTF
f
RAB5Q79L
lys

Flotillin-1

n= 4 3 3
(relative to control)

100
SDC1CTF
Percentage

Syntenin Merge SDC1 ED cer–RAB5Q79L mCh–syntenin


60
CD63

20

RAB5Q79L
RAB7

Syntenin
d c Merge CD63 cer–RAB5Q79L 80 *
RNAi
nt

Percentage of
luminal CD63
SDC1CTF 70 g

Stained MVB area (× 103 nm2)


RNAi nt
Control

100
60 RNAi syntenin
Syntenin 50
40 80
CD63

60
n= 5 5 3 Merge CD63 cer–RAB5Q79L mCh–syntenin
(relative to control)

Syntenin OE

100
SDC1CTF
Percentage

40
Syntenin
60 CD63
20
20 20 40 60 80 100
Total MVB area (× 103 nm2)
RNAi RAB7

Figure 4 Syntenin exosomes are of endosomal origin. (a) Western blots are shown at larger magnifications. The histogram represents the filling of
(top) illustrating the effect of RAB5Q79L on the accumulation of the cer–RAB5Q79L -decorated vacuoles with antibody. For each endosome, the
indicated proteins in HSC exosomes, relative to levels in control cells luminal fluorescence is expressed as percentage of the total fluorescence
(-). The histogram (below) shows means (and s.d.) from n independent present in the endosome. In the presence of mCherry–syntenin (open
experiments (as indicated). (b) Confocal micrographs of cells expressing triangles), endosomes contain more luminal fluorescence/anti-CD63 than
cer–RAB5Q79L and mCherry–syntenin. Cells are stained with anti-CD63 in the control (filled triangles). Scale bars, 10 µm. ∗ P < 0.05. (d,e) Western
(upper row), with monoclonal antibody 2E9, detecting the syndecan-1 blots (d, top and e) of HSC exosomes (d) and total cell lysates (e) derived
ICD (middle row), or polyclonal antibodies detecting the syndecan-1 from cells treated with control (nt), syntenin and/or RAB7 RNAi, and
ectodomain (ED; lower row). Molecules as indicated top right were histogram (d, bottom), illustrating the effectiveness of the RNAi and
identified based on their different emission wavelength. In the merge the effect of RAB7 RNAi on the accumulation of syntenin, syndecan-1
image, the first molecule is in green, the second in blue and the third CTF and CD63 in HSC exosomes (means and s.d.). (f) Representative
in red. Outlined areas are shown at larger magnifications. Note the electron micrographs, illustrating the morphology of the late endosomal
restriction of syndecan-ED to the limiting membranes, and the extensive compartment in MCF-7 cells treated with control, RAB7 and/or syntenin
co-localization of CD63 and syndecan-ICD with mCherry–syntenin inside RNAi, as indicated. Structures were revealed by peroxidase-conjugated
the lumina of the enlarged endosomes outlined by cer–RAB5Q79L . Scale anti-CD63 (internalized for 30 min), and staining with DAB. Scale bar,
bars, 10 µm. These lumina are also filled with ALIX (see Supplementary 200 nm. Note the more peripheral localization of the stain, decrease
Fig. S4). (c) Confocal micrographs of cells expressing cer–RAB5Q79L , in stained ILVs and smaller size of the endosomes in cells treated with
alone (upper row) or with mCherry–syntenin (lower row), and that have syntenin RNAi. (g) Endosome filling. For each peroxidase/DAB-marked
been incubated with anti-CD63 antibody. Cells are stained with anti-IgG. endosome, the total sectional area occupied by the endosome is plotted
OE, overexpression. Molecules as indicated in the top right were identified versus the area occupied by the stain in that endosome. Further data
based on their different emission wavelength. In the merge image, the first are shown in Supplementary Fig. S5. Uncropped images of blots/gels are
molecule is in green, the second in blue and the third in red. Outlined areas shown in Supplementary Fig. S6.

reduced the level of exosomal release (Fig. 4d). Importantly, RAB7 association with ILVs, filling up the lumen of endosomes. In contrast,
RNAi also prevented the enhanced exosomal recoveries induced by in cells treated with syntenin RNAi the staining was associated with
syntenin overexpression (Supplementary Fig. S5a). These data indicate limiting membranes and on ILVs that were peripherally located, the
that exosomes supported by syndecan–syntenin–ALIX originate from lumen of the endosomes remaining largely empty. Similar results were
endosomal membranes and from the fusion of MVBs with the cell obtained in cells treated with both RAB7 RNAi and syntenin RNAi.
surface, bypassing lysosomes. Whereas RAB7 RNAi resulted in large endosomes, filled with ILVs,
This was further explored by electron microscopy, monitoring the in cells with a double knockdown of RAB7 and syntenin, staining
localization of internalized anti-CD63:HRP conjugates (Fig. 4e–g and remained largely peripheral. Loss of syntenin also significantly reduced
Supplementary Fig. S5b,c,d). In control cells, diaminobenzidine (DAB) the average sizes of the stained endosomes (Supplementary Fig. S5c).
stain was found both on limiting endosomal membranes and in close Empty, yet smaller late endosomes imply that the knockdown of

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012 681


© 2012 Macmillan Publishers Limited. All rights reserved.
ARTICLES

TSG101
a b n=3

CHMP4A/B/C

Percentage of CHMP4B
Syntenin
SDC1CTF Syntenin CD63
n = 4 87

(relative to control)
RNAi 234 325 668 334 325 100

CHMP4C
CHMP4B
CHMP4A
nt

TSG101

(relative to control)
RNAi

VPS22

nt
TSG101 100 80

Percentage
β-actin RNAi 80 Syntenin 60

nt

nt
CHMP4B
SDC1CTF 60 CHMP4B 40
40 20
Syntenin
20 CD63
nt

nin
CD63

nt
CHMP4B RNAi
TSG101 VPS22 A B C A/B/C

nte
Flotillin-1 RNAi CHMP4

Sy
lys exo

Percentage (relative to control)


d RNAi RNAi n = 333 333 333 879 333 33 3 344

VPS4AK173Q
c RNAi ALIX
SDC1CTF

CHMP2A

VPS4A/B
I212D

n=3 140

CHMP6

CHMP3
Syntenin

VPS4B
VPS4A

VPS4A
WT

SDC1CTF
Percentage (relative

eYFP–mALIX CD63

200 Syntenin 100


SDC1CTF

nt

nt
CD63
to control)

Syntenin 150 SDC1CTF


exo 100 60
Syntenin
CD63 50
20
WT I212D CD63

Q
CHMP6 CHMP2A CHMP3 VPS4A VPS4B VPS4A/B
lys ALIX eYFP-tagged

4A K173
eYFP–mALIX
Endogenous
RNAi RNAi

VPS
Figure 5 The biogenesis of syntenin exosomes requires ESCRTs . (a) Western formation; the mutant form fails. (d) ESCRT-III protein requirements.
blots (left) and histogram (right), illustrating the effectiveness of TSG101 Western blots of HSC exosomes (left) and histogram (right), illustrating the
and CHMP4B RNAi, and the effects of TSG101 (ESCRT-I), VPS22 effect of dominant-negative VPS4AK173Q , versus that of wild-type VPS4A,
(ESCRT-II) and CHMP4 (ESCRT-III) RNAi (relative to that of control, nt and the effects of RNAi targeting CHMP6, CHMP3, CHMP2A or VPS4A/B,
RNAi) on the levels of SDC1CTF, syntenin and CD63 in HSC exosomes. versus that of nt RNAi, on syntenin exosome formation. Note VPS4 and
(b) Effect of syntenin RNAi on CHMP4B levels in HSC exosomes. CHMP2A requirements, contrasting with a lack of dominant-negative
(c) ALIX–CHMP4 interaction requirement. MCF-7 cells were transfected, to VPS4A effects. Similarly, VPS4AE228Q (defective in ATP hydrolysis)46 did
express low levels of eYFP (-) or eYFP fused to wild-type (WT) mALIX or to also not affect the recovery of exosomal syndecan, CD63 and syntenin (data
mALIXI212D , deficient in CHMP4 binding, and were treated with ALIX RNAi not shown). All histograms represent means (and s.d.) from n independent
to suppress endogenous ALIX (left, western blots; right, quantifications). experiments (as indicated). Uncropped images of blots/gels are shown in
Note that for similar levels of expression, wild-type ALIX supports exosome Supplementary Fig. S6.

syntenin might lead to a compound failure of (a specific subset of) ILVs the formation of syndecan–syntenin exosomes, but RNAi targeting
to form or persist, and of endosomes to mature. CHMP3 and CHMP6 had no effect (Fig. 5d). These results are
reminiscent of the ESCRT-III protein requirements for HIV-1 release47 ,
Mechanisms of exosome biogenesis and might indicate that syntenin/LYPX(n)L viral late domains recruit
The activities of ALIX in membrane trafficking probably depend on its CHMP4 through ALIX, circumventing the need for CHMP6 (whereas
interactions with multiple proteins, including CHMP4 and TSG101. in canonical pathways, VPS22 in ESCRT-II recruits CHMP6, and
The knockdown of CHMP4A/B/C or TSG101 significantly reduced CHMP6 recruits CHMP4). Besides recycling ESCRT-III components,
the exosomal release of syndecan-CTF, CD63 and syntenin (Fig. 5a). VPS4 may also play a direct role in the process of membrane
TSG101 downregulation has strong effects on general endosomal abscission11,48 . Consistent with this, and in contrast to dominant-
trafficking44,45 , arguing against a specific role for TSG101 in the negative VPS4, RNAi targeting VPS4A/B markedly reduced the level
biogenesis of syndecan–syntenin exosomes. Of note, gain (Fig. 2a) or of exosomal release, supporting the involvement of ESCRTs (Fig. 5d).
loss (Fig. 3a) of syntenin left the TSG101 signals in exosomes unaffected. Interestingly, whereas HIV-1 release does not depend on ESCRT-II (for
In contrast, loss of syntenin significantly decreased the recovery recruiting ESCRT-III; ref. 20), RNAi targeting VPS22 (ESCRT-II) also
of CHMP4B from exosomes (Fig. 5b). Low levels of eYFP-tagged significantly reduced the production of syndecan–syntenin exosomes
murine ALIX (mALIX) (approximating the levels of endogenous (Fig. 5a). Thus, the processes of membrane bending and abscission
ALIX in control cells) rescued the exosomal yields of syndecan, that direct syndecan and syntenin to ILVs and exosomes are likely
CD63 and syntenin in cells treated with ALIX RNAi; in contrast, to involve ESCRTs and/or to occur in endosomal compartments that
eYFP–mALIXI212D , deficient in CHMP4 binding, failed to rescue require ESCRTs for their architectural and functional integrity.
exosome production (Fig. 5c). Thus, a role for ESCRTs, in particular Recent reports have revealed roles for higher-order oligomerization
ESCRT-III in the biogenesis of syntenin–ALIX exosomes is likely. A of vesicular cargo49 . We therefore surmised that heparan sulphate,
requirement for the ATPase VPS4 is considered as a defining feature and heparan sulphate cargo impacting on syndecan clustering and
of the ESCRT pathway11–13 . Surprisingly, but as observed by others oligomerization25–27 , might be required for exosome formation.
for CD63-containing exosomes40 , expression of dominant-negative Consistently, disrupting the heparan sulphate structure, by the
VPS4A (VPS4AK173Q , defective in ATP binding46 ), did not affect the knockdown of EXT1/2 (heparan sulphate polymerase), the knockdown
recovery of exosomal syndecan, CD63 and syntenin (Fig. 5d). As this of NDST1/2 (an enzyme required for the creation of sulphated,
might indicate that the ESCRT-III components whose availability ligand-binding domains in heparan sulphate) or growing the cells
critically depends on VPS4-mediated recycling are not required for in culture media supplemented with heparitinase, markedly reduced
the biogenesis of syntenin exosomes (and, conversely, the required the level of exosomal release (Fig. 6a). Anti-syndecan ectodomain
ESCRT components not limited by recycling), we also tested other monoclonal antibodies reduced exosomal syntenin release, possibly
ESCRT-III components. RNAi targeting CHMP2A markedly reduced because the ensuing clusters include syndecan that is bound to the

682 NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012


© 2012 Macmillan Publishers Limited. All rights reserved.
ARTICLES
a SDC1CTF Syntenin CD63 b

Percentage (relative to control)


n= 6 3 3 6 6 5 4 4 3 3 2 2 n=3
RNAi EXT1/2 – +
RNAi nt NDST1/2
100 120

Percentage of syntenin
80

Percentage of
nd

total eluted

(relative to control)
SDC1 60 80 100
d
40 60 80
SDC1CTF
20 40 60
β-actin
20 40
0.2 0.4 0.6
20
NaCl (M) RNAi EXT1/2 NDST1/2 – NDST1/2
Hpt – – + + Anti-SDC mAb + – +
c d RNAi nt EXT1/2
CTF
WT

SDC3 PSI – +
SDC1CTF SDC1CTF
RNAi SDC1/4 + + + Syntenin RNAi SDC1/4 + + Syntenin
n=3 e Cath. I – + – +
SDC1/3CTF CD63 CD63
SDC1CTF

Percentage (relative to control)


RNAi SDC1/4 – – + +
n= 8 8 9
(relative to control)

Syntenin 400 Syntenin


exo

Syntenin

exo

exo
Percentage

300 200 Flotillin-1


CD63 CD63
150 SDC1CTF
200

lys
100 ALIX
SDC1/3CTF 100 SDC1
50 n=3
lys

Percentage of syntenin
Syntenin 150
lys
SDC1CTF

(relative to control)
ALIX SDC3 WT CTF
Syntenin PSI
100
f g 1h 2h 4h 1h 2h 4h
FGF2 – + – + n= 2 3 3 2 3 3 2 3 3 + + – – –
50
FGF2 + + + + + + + +
RNAi SDC1/4 – – + + RNAi SDC1/4 – + – + + – + + – + + – +
300 SDC1CTF
(relative to control)

Syntenin Syntenin OE + + + – + + – + + – + + + Cath. I + – +


CD63 250
Percentage

CD63 HA–FGFR OE – + + + + + + + + + + + +
RNAi nt SDC1/4
exo

200
FGFR
exo

Syntenin 150
SDC1CTF 100
FGFR
ALIX 50
SDC1
lys

FGF2 + – + p-ERKs
lys

SDC4 RNAi nt SDC1/4


Syntenin
ERKs

Figure 6 Syntenin exosomes require syndecan oligomerization and cleavage. as 100%; left, western blots; right, quantifications). (e) The cell-permeable
(a) Heparan sulphate requirement. MCF-7 cells were treated with nt (-), cysteine protease/cathepsin-inhibitor Z-Phe-Gly-NHO-Bz (Cath. I, 2 µM)
EXT1/2 or NDST1/2 RNAi, or grown in the presence of heparitinase reduces syndecan-CTF in lysates, and when syndecans are near limiting
(Hpt). Left: cell lysates digested (d) or not digested (nd) with heparitinase. (syndecan RNAi), strongly reduces exosomal syntenin (top, western blots;
In EXT -RNAi-treated cells, syndecan-1 is detected without digestion, bottom, quantifications). (f) MCF-7 cells were treated with nt RNAi, or with
demonstrating reduced heparan sulphate polymerization. Middle: cell RNAi targeting endogenous syndecans, and serum-starved, overnight, in the
lysates were loaded on DEAE-Sepharose, and eluted by step increments presence or absence of 2 ng ml−1 of FGF2. The histogram (right) indicates
in salt concentration. In NDST -RNAi-treated cells, SDC1 elutes at lower exosomal mean signal intensities (relative to nt RNAi, no FGF2, taken as
salt concentrations, demonstrating reduced sulphation/negative charge of 100%) as detected in western blotting (left). Note that FGF enhances the
the heparan sulphate on syndecan. Right: exosomal releases. (b) Syndecan production of syndecan–syntenin exosomes, pending syndecan support.
clustering. RNAi-treated cells incubated with or without anti-syndecan (g) MCF-7 cells expressing HA-tagged FGFR were stimulated with FGF2
(-1 and -4) ectodomain monoclonal antibodies (mAbs); HSC exosomes (2 ng ml−1 ) for different lengths of time, or left untreated. The cells
were tested for syntenin contents. (c) Cells transfected with empty vector expressed syntenin at endogenous levels or overexpressed (OE) syntenin,
(-), wild-type SDC3 (WT) or N-truncated SDC3 (CTF) and treated with and were treated with control or syndecan RNAi. Syntenin-stimulated
syndecan RNAi (left, western blots; right, quantifications). Note that CTF exosomal FGFR release depends both on syndecan and on FGF, but not
fails to sustain exosome formation (taking the release by sham-transfected on FGF signalling (pERK activation) per se (which persists in the absence
cells as 100%). (d) Presenilin inhibitor (PSI, 0.1 µM), raising CTF levels of syndecan). All histograms represent means (and s.d.) from n independent
in cells with a partial knockdown of syndecans, fails to stimulate exosomal experiments (as indicated). Uncropped images of blots/gels are shown in
syntenin and CD63 release (taking the release in the absence of inhibitor Supplementary Fig. S6.

extracellular matrix (through heparan sulphate), subtracting syndecan syndecan, an N-truncated syndecan (mimicking syndecan-CTF;
from internalization routes. In contrast, in cells treated with EXT1/2 ref. 50) failed to stimulate the exosomal accumulation of syntenin
RNAi, anti-syndecan antibody rescued exosomal release (Fig. 6b). and CD63 in syndecan-depleted cells (Fig. 6c). Syndecan-CTFs are
Together, these data suggest that heparan sulphate-mediated syndecan substrates for presenilin, which releases the syndecan-ICD for turnover
clustering matters for syntenin exosome formation. by the proteasome50 . Thus, presenilin activity might prevent CTF
As syndecans accumulate in exosomes as CTFs, we also assessed accumulation in cells and compromise rescue attempts by CTF
the significance of CTF production. Compared with full-length overexpression. Presenilin inhibition, raising the levels of CTF in

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012 683


© 2012 Macmillan Publishers Limited. All rights reserved.
ARTICLES

lysates and exosomes of cells with partial syndecan knockdown, did cleavage potentially also disconnects endocytosed cargo from budding
not stimulate the exosomal accumulation of CD63 and syntenin machineries, once ESCRT-III proteins have assembled on the necks
(Fig. 6d), further documenting that syndecan-CTFs are not sufficient of budding membranes11 , syndecan ectodomains, with their heparan
for exosome formation. Treating cells with cathepsin inhibitor, sulphate, would become unnecessary for sequestering cargo to budding
reducing the accumulation of syndecan-CTF in cell lysates by ∼40% membranes. Syndecan ligation (for example, by FGF2) causing
(without significant effects on the levels of intact syndecan), had syndecan to become associated with detergent-resistant membrane
no significant effect on exosomal syntenin (Fig. 6e). Yet, cathepsin domains51 , the lateral segregation of lipids in endosomal membranes
inhibitor and syndecan RNAi synergized in suppressing exosomal may also be a stabilizing factor. Possibly this pertains to the importance
syntenin releases (Fig. 6e), suggesting that, although non-sufficient, of ceramide for exosome formation40 . Clearly, further investigations
syndecan-CTF production might be required for exosome formation. are required to clarify the spatiotemporal and mechanistic relationships
Thus, the syntenin–ALIX axis is likely to be driven by cargo-induced between syndecan clustering, syntenin recruitment, syndecan cleavage
oligomerization of syndecan, with also a major role for the cleavage and consolidated cargo sequestration in membrane buds.
of endosomal syndecan. We found that syntenin–ALIX also controls exosome formation
The heparan sulphate-assisted docking of FGF to FGFR and ensuing in HeLa cells (M.F.B., unpublished results). Yet, the specificity and
dimerization of the receptor represents one example of physiological regulation of this pathway remain to be explored. Reportedly, syntenin
cargo that can induce heparan sulphate clustering22 . This implies has multiple interaction partners, and clusters of membrane and
that FGF–FGFR signalling potentially approximates syndecans to membrane-associated proteins that represent alternative bait for
each other, recruiting syntenin. Consistent with that notion, FGF2 the PDZ domains of syntenin might thus function along with or
stimulation enhanced exosome production (that is, exosomal CD63 independently of syndecan (for example, ephrin-B1/2; refs 52,53
and syntenin), and syndecan downregulation, not interfering with and Fz signalling54 ). The functional property of heparan sulphate
FGF signalling per se (as indicated by pERK levels; see Fig. 6g), depends on its fine structure, which is highly plastic and is set
attenuated this FGF effect (Fig. 6f). By extension, we examined by modifying enzymes that are cell-type- and differentiation-stage-
whether syntenin, adapting ALIX to syndecan, might direct associated specific23,24 . This might select what cargo is trafficking through
FGF:FGFR cargo to exosomes (Fig. 6g). Syntenin overexpression syndecan–syntenin vesicles. As the syndecan cytoplasmic domain25–27 ,
stimulated exosomal release of FGFR. This stimulation depended on syntenin30,31 and ALIX55 are subject to modifications (phosphorylation,
syndecan, as it was prevented by syndecan RNAi. Consistent with ‘open’ and ‘closed’ configurations) that might regulate their mutual
the ligand-induced approximation model, exosomal accumulation engagement, the model outlined here in MCF-7 cells might be
of FGFR depended on FGF. modified by cellular context.
Taken together, the above findings identify syndecan–syntenin–ALIX
DISCUSSION as an important regulator of membrane trafficking and heparan
Our study establishes a direct mechanistic link between syndecans sulphate-assisted signalling. By extension, activation or deregulation of
and ILV/exosome formation, identifying an as yet unsuspected this axis might also influence heparan sulphate-associated pathological
role for heparan sulphate proteoglycans in vesicular trafficking and processes, including cancer, the propagation of prions, inflammation,
endocytic controls on signalling processes. The cytosolic adaptor amyloid deposition and neurodegenerative disease. 
syntenin connects syndecans to ALIX, an auxiliary component of
the ESCRT machinery that supports endosomal membrane budding METHODS
and abscission. This connection not only segregates syndecans and Methods and any associated references are available in the online
syndecan cargo, for example FGF:FGFR complexes assembled on version of the paper at www.nature.com/naturecellbiology
the heparan sulphate chains of syndecan, to budding endosomal
Note: Supplementary Information is available on the Nature Cell Biology website
membranes and exosomes; it also supports budding processes and
the production of exosomes. Thus, syndecans emerge as key regulators ACKNOWLEDGEMENTS
of post-engagement receptor trafficking26 . The sheer abundance of the We are grateful to H. Ceulemans, G. Reekmans and J. Grootjans for technical
support, and to P. Baatsen for access to the electron microscopy facility at
syndecans (up to 106 copies on the surface of an adherent cell) and the the Department for Human Genetics, VIB-KULeuven. We thank C. Dotti and
functional versatility of their heparan sulphate chains imply that the P. Courtoy for critical feedback on the manuscript. This work was supported by
signalling and trafficking of a large number and variety of receptors the Fund for Scientific Research-Flanders (FWO), the Concerted Actions Program
of the K.U. Leuven, the VIB, the Belgian Federation against Cancer (STK), the
might be regulated in that way. Interuniversity Attraction Poles of the Prime Ministers Services (IUAP) and the
Together our data suggest that syndecan cargo creates syndecan EMBO young investigator programme (P.Z.). A.M. was supported by a Marie Curie
assemblies that can recruit syntenin–ALIX and support membrane postdoctoral fellowship of the EU (FP7); Y.I. by an EMBO-long term postdoctoral
fellowship; and A.G. by a PhD-student fellowship of the Agency for Innovation by
budding. Although syntenin can be detected at the plasma Science and Technology (IWT), Flanders.
membrane32,33 , the formation or the stability of these assemblies
AUTHOR CONTRIBUTIONS
might require the concentration of syndecan in endosomal domains,
The work is an equal contribution of two laboratories. M.F.B., Z.Z. and A.M.
where syndecan cleavage occurs. This cleavage, removing the highly performed the cellular experiments; E.M. and Y.I. did the Biacore, A.G. the
negatively charged and auto-repulsive syndecan ectodomains, might fluorescence spectroscopy, G. Degeest the electron microscopy, and F.D. the
allow syndecan-ICD to remain clustered and syntenin bound to co-immunoprecipitation and the Y2H analyses; E.V. made the various expression
constructs and C.C. assisted with cell culture and biochemical analyses; P.Z. and
syndecan (as CTF), stabilizing the recruitments and sustaining the G. David designed experiments, analysed data and wrote the manuscript. All authors
membrane budding process. Of note, although endosomal syndecan discussed results and commented on the manuscript.

684 NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012


© 2012 Macmillan Publishers Limited. All rights reserved.
ARTICLES

COMPETING FINANCIAL INTERESTS 28. Grootjans, J. J. et al. Syntenin, a PDZ protein that binds syndecan cytoplasmic
The authors declare no competing financial interests. domains. Proc. Natl Acad. Sci. USA 94, 13683–13688 (1997).
29. Beekman, J. M. & Coffer, P. J. The ins and outs of syntenin, a multifunctional
intracellular adaptor protein. J. Cell Sci. 121, 1349–1355 (2008).
Published online at www.nature.com/naturecellbiology
30. Grootjans, J. J., Reekmans, G., Ceulemans, H. & David, G. Syntenin-syndecan
Reprints and permissions information is available online at www.nature.com/reprints binding requires syndecan-synteny and the co-operation of both PDZ domains of
syntenin. J. Biol. Chem. 275, 19933–19941 (2000).
1. Bobrie, A., Colombo, M., Raposo, G. & Théry, C. Exosome secretion: molecular 31. Zimmermann, P. et al. Characterization of syntenin, a syndecan-binding PDZ protein,
mechanisms and roles in immune responses. Traffic 12, 1659–1668 (2011). as a component of cell adhesion sites and microfilaments. Mol. Biol. Cell 12,
2. Fevrier, B. & Raposo, G. Exosomes: endosomal-derived vesicles shipping extracellular 339–350 (2001).
messages. Curr. Opin. Cell Biol. 16, 415–421 (2004). 32. Zimmermann, P. et al. PIP(2)-PDZ domain binding controls the association of
3. Couzin, J. Cell biology: the ins and outs of exosomes. Science 308, syntenin with the plasma membrane. Mol. Cell 9, 1215–1225 (2002).
1862–1863 (2005). 33. Zimmermann, P. et al. Syndecan recycling [corrected] is controlled by syntenin-PIP2
4. Liegeois, S., Benedetto, A., Garnier, J. M., Schwab, Y. & Labouesse, M. The interaction and Arf6. Dev. Cell 9, 377–388 (2005).
V0-ATPase mediates apical secretion of exosomes containing Hedgehog-related 34. Gottlinger, H. G. How HIV-1 hijacks ALIX. Nat. Struct. Mol. Biol. 14,
proteins in Caenorhabditis elegans. J. Cell Biol. 173, 949–961 (2006). 254–256 (2007).
5. Al-Nedawi, K. et al. Intercellular transfer of the oncogenic receptor EGFRvIII by 35. Lee, S., Joshi, A., Nagashima, K., Freed, E. O. & Hurley, J. H. Structural basis for
microvesicles derived from tumour cells. Nat. Cell Biol. 10, 619–624 (2008). viral late-domain binding to Alix. Nat. Struct. Mol. Biol. 14, 194–199 (2007).
6. Valadi, H. et al. Exosome-mediated transfer of mRNAs and microRNAs is a novel 36. Zhai, Q. et al. Structural and functional studies of ALIX interactions with YPX(n)L
mechanism of genetic exchange between cells. Nat. Cell Biol. 9, 654–659 (2007). late domains of HIV-1 and EIAV. Nat. Struct. Mol. Biol. 15, 43–49 (2008).
7. Thery, C., Zitvogel, L. & Amigorena, S. Exosomes: composition, biogenesis and 37. Thery, C. et al. Molecular characterization of dendritic cell-derived exosomes.
function. Nat. Rev. Immunol. 2, 569–579 (2002). Selective accumulation of the heat shock protein hsc73. J. Cell Biol. 147,
8. Nabhan, J. F., Hu, R., Oh, R. S., Cohen, S. N. & Lu, Q. Formation and release of 599–610 (1999).
arrestin domain-containing protein 1-mediated microvesicles (ARMMs) at plasma 38. Mathivanan, S. & Simpson, R. J. ExoCarta: a compendium of exosomal proteins and
membrane by recruitment of TSG101 protein. Proc. Natl Acad. Sci. USA 109, RNA. Proteomics 9, 4997–5000 (2009).
4146–4151 (2012). 39. Latysheva, N. et al. Syntenin-1 is a new component of tetraspanin-enriched
9. Wehman, A. M., Poggioli, C., Schweinsberg, P., Grant, B. D. & Nance, J. The microdomains: mechanisms and consequences of the interaction of syntenin-1 with
P4-ATPase TAT-5 inhibits the budding of extracellular vesicles in C. elegans embryos. CD63. Mol. Cell Biol. 26, 7707–7718 (2006).
Curr. Biol. 21, 1951–1959 (2011). 40. Trajkovic, K. et al. Ceramide triggers budding of exosome vesicles into multivesicular
10. Raiborg, C. & Stenmark, H. The ESCRT machinery in endosomal sorting of endosomes. Science 319, 1244–1247 (2008).
ubiquitylated membrane proteins. Nature 458, 445–452 (2009). 41. Stenmark, H. et al. Inhibition of rab5 GTPase activity stimulates membrane fusion
11. Henne, W. M., Buchkovich, N. J. & Emr, S. D. The ESCRT pathway. Dev. Cell 21, in endocytosis. EMBO J. 13, 1287–1296 (1994).
77–91 (2011). 42. Rink, J., Ghigo, E., Kalaidzidis, Y. & Zerial, M. Rab conversion as a mechanism of
12. Wollert, T. & Hurley, J. Molecular mechanism of multivesicular body biogenesis by progression from early to late endosomes. Cell 122, 735–749 (2005).
ESCRT complexes. Nature 464, 864–869 (2010). 43. Vanlandingham, P. A. & Ceresa, B. P. Rab7 regulates late endocytic trafficking
13. Hurley, J. & Hanson, P. Membrane budding and scission by the ESCRT machinery: downstream of multivesicular body biogenesis and cargo sequestration. J. Biol.
it’s all in the neck. Nat. Rev. Mol. Cell Biol. 11, 556–566 (2010). Chem. 284, 12110–12124 (2009).
14. Babst, M. MVB vesicle formation: ESCRT-dependent, ESCRT-independent and 44. Doyotte, A., Russell, M., Hopkins, C. & Woodman, P. Depletion of TSG101 forms a
everything in between. Curr. Opin. Cell Biol. 23, 452–457 (2011). mammalian ‘Class E’ compartment: a multicisternal early endosome with multiple
15. Stuffers, S., Sem Wegner, C., Stenmark, H. & Brech, A. Multivesicular endosome sorting defects. J. Cell Sci. 118, 3003–3017 (2005).
biogenesis in the absence of ESCRTs. Traffic 10, 925–937 (2009). 45. Razi, M. & Futter, C. E. Distinct roles for Tsg101 and Hrs in multivesicular body
16. Theos, A. C. et al. A lumenal domain-dependent pathway for sorting to intralumenal formation and inward vesiculation. Mol. Biol. Cell 17, 3469–3483 (2006).
vesicles of multivesicular endosomes involved in organelle morphogenesis. Dev. Cell 46. Bishop, N. & Woodman, P. ATPase-defective mammalian VPS4 localizes to
10, 343–354 (2006). aberrant endosomes and impairs cholesterol trafficking. Mol. Biol. Cell 11,
17. Odorizzi, G. The multiple personalities of Alix. J. Cell Sci. 119, 3025–3032 (2006). 227–239 (2000).
18. Matsuo, H. et al. Role of LBPA and Alix in multivesicular liposome formation and 47. Morita, E. et al. ESCRT-III protein requirements for HIV-1 budding. Cell Host Microbe.
endosome organization. 303, 531-534 (2004). 9, 235–242 (2011).
19. Fujii, K., Hurley, J. H. & Freed, E. O. Beyond Tsg101: the role of Alix in ’ESCRTing’ 48. Jouvenet, N., Zhadina, M., Bieniasz, P. D. & Simon, S. M. Dynamics of ESCRT
HIV-1. Nat. Rev. Microbiol. 5, 912–916 (2007). protein recruitment during retroviral assembly. Nat. Cell Biol. 13, 394–401 (2011).
20. Martin-Serrano, J. & Neil, S. J. Host factors involved in retroviral budding and release. 49. Fang, Y. et al. Higher-order oligomerization targets plasma membrane proteins and
Nat. Rev. Microbiol. 9, 519–531 (2011). HIV gag to exosomes. PLoS Biol. 5, e158 (2007).
21. Rapraeger, A. C., Krufka, A. & Olwin, B. B. Requirement of heparan sulfate 50. Schulz, J. et al. Syndecan 3 intramembrane proteolysis is presenilin/gamma-
for bFGF-mediated fibroblast growth and myoblast differentiation. Science 252, secretase-dependent and modulates cytosolic signaling. J. Biol. Chem. 278,
1705–1708 (1991). 48651–48657 (2003).
22. Schlessinger, J. et al. Crystal structure of a ternary FGF–FGFR-heparin complex 51. Tkachenko, E. & Simons, M. Clustering induces redistribution of syndecan-4 core
reveals a dual role for heparin in FGFR binding and dimerization. Mol. Cell 6, protein into raft membrane domains. J. Biol. Chem. 277, 19946–19951 (2002).
743–750 (2000). 52. McClelland, A. C., Sheffler-Collins, S. I., Kayser, M. S. & Dalva, M. B. Ephrin-B1
23. Bishop, J. R., Schuksz, M. & Esko, J. D. Heparan sulphate proteoglycans fine-tune and ephrin-B2 mediate EphB-dependent presynaptic development via syntenin-1.
mammalian physiology. Nature 446, 1030–1037 (2007). Proc. Natl Acad. Sci. USA 106, 20487–20492 (2009).
24. Hacker, U., Nybakken, K. & Perrimon, N. Heparan sulphate proteoglycans: the sweet 53. Xu, N. J., Sun, S., Gibson, J. R. & Henkemeyer, M. A dual shaping mechanism
side of development. Nat. Rev. Mol. Cell Biol. 6, 530–541 (2005). for postsynaptic ephrin-B3 as a receptor that sculpts dendrites and synapses. Nat.
25. Couchman, J. Transmembrane signaling proteoglycans. Annu. Rev. Cell Dev. Biol. Neurosci. 14, 1421–1429 (2011).
26, 89–114 (2010). 54. Luyten, A. et al. The postsynaptic density 95/disc-large/zona occludens protein
26. Lambaerts, K., Wilcox-Adelman, S. A. & Zimmermann, P. The signaling syntenin directly interacts with frizzled 7 and supports noncanonical Wnt signaling.
mechanisms of syndecan heparan sulfate proteoglycans. Curr. Opin. Cell Biol. 21, Mol. Biol. Cell 19, 1594–1604 (2008).
662–669 (2009). 55. Zhou, X., Si, J., Corvera, J., Gallick, G. E. & Kuang, J. Decoding the intrinsic
27. Morgan, M. R., Humphries, M. J. & Bass, M. D. Synergistic control of cell adhesion mechanism that prohibits ALIX interaction with ESCRT and viral proteins. Biochem.
by integrins and syndecans. Nat. Rev. Mol. Cell Biol. 8, 957–969 (2007). J. 432, 525–534 (2010).

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 7 | JULY 2012 685


© 2012 Macmillan Publishers Limited. All rights reserved.
METHODS DOI: 10.1038/ncb2502

METHODS exosome preparations were adjusted accordingly. Except for RNAi targeting nSMase,
Expression vectors and reagents. Expression vectors for GST (pGEX-5X, none of the treatments (chemicals, enzymes), transfections or RNAi significantly
Amersham Pharmacia Biotech) and fluorescent protein (pmCherry-C and peYFP-C, affected cell survival or growth, requiring major (>10%) adjustments of the
Clontech) fusion proteins, and for yeast two-hybrid experiments (pGAD10 and samples.
pAS2, Clontech), were constructed from syntenin complementary DNAs (described
previously28,30,31 ) and from full-length cDNA clones encoding mouse (ID 4038684) Exosome fractionation. For further fractionation by buoyant density, HSC
and human (ID 2455231) ALIX (obtained from the IMAGE consortium, through exosomes were suspended in 2 ml 60% Optiprep (Axis-Shield PoC), and a
Invitrogen). Mutant forms were made by QuickChange site-directed mutagenesis 6–40% gradient of Optiprep in DMEM/F12 was carefully deposited on top. After
PCR (Stratagene). For gain-of-function studies in cells, cDNA encoding full-length centrifugation for 16 h at 140,000g , at 4 ◦ C, fractions were collected from the top.
non-tagged syntenin was cloned in pcDNA3.1/Zeo(+) (Invitrogen). Expression The density of each fraction was extrapolated from its attenuance, measured at
vector for mGFP–CD63 was a gift from J. Klumperman (Utrecht University, 340 nm. Then, the different fractions were diluted with DMEM/F12 and further
The Netherlands), cer–RAB5Q79L was received from W. Annaert (K.U. Leuven, spun for 3 h at 140,000g , at 4 ◦ C, to collect the exosomes. For fractionation by CD63
Belgium; ref. 56) and constructs for expressing GFP–VPS4A were kindly provided content, HSC exosomes were suspended in 1 ml of PBS-0.1%BSA and incubated,
by M. Simons (University of Göttingen/Max-Planck-Institute for Experimental for 30 min at 4 ◦ C, under gentle agitation, with 1.5 µg of anti-CD63 monoclonal
Medicine, Germany). All other expression vectors were described previously28,30–33 . antibody or monoclonal antibody 3G10 (specific for a neo-epitope, generated by
All constructs were verified by sequencing. Except for RNAi targeting human bacterial enzyme digestion, not occurring in non-enzyme-treated mammalian cells,
nSMase-2 (Flexitube short interfering RNA (siRNA), Qiagen), all RNAi targeting and used as a control). After spinning (140,000g , 1 h, 4 ◦ C) to remove unbound
human sequences (ON-Target plus Smartpools) and the non-targeting control antibody, pellets were suspended in 1 ml of PBS-0.1%BSA, supplemented with
RNAi (siSTABLE) were purchased from Dharmacon. Anti-syntenin, and anti-SDC 20 µl of magnetic beads with covalently bound anti-mouse IgG (Dynal) and further
(-1/3 ICD and -4 ectodomain) antibodies were as described previously31,32,57,58 . incubated for 30 min at 4 ◦ C, under gentle agitation. Collected beads were washed
GST–ALIX (full length, mouse) fusion protein and GST were used to generate twice with PBS-0.1%BSA, once with PBS, and then resuspended in lysis buffer.
affinity-purified ALIX-specific rabbit polyclonal antisera (Eurogentec). Other
antibodies were from commercial sources and were used as recommended by the Western blotting. To allow detection of the syndecans, the samples were
manufacturers. Antibodies directed against TSG101 (C-2), HSP70 (W27), RAB7 digested with (0.006 U ml−1 ) heparitinase and (0.166 U ml−1 ) chondroitinase ABC
(H-50), CHMP4B (C-12) and β-actin (AC-15) were from Santa Cruz; antibody (Seikagaku Corporation) for 3 h at 37 ◦ C in heparitinase buffer (0.1% Triton, 0.1 M
against SDC1 ectodomain (BB4) was from Serotec; antibody against SDC4 ICD NaCl, 1 mM CaCl2 , 50 mM 6-aminohexanoic acid and 50 mM HEPES, at pH 7).
was from Abnova; antibodies against CD63 (MEM-259) and Calnexin (ab10286) Samples were fractionated in 4–12% gradient gels (NuPAGE Novex Bis-Tris gels,
were from Abcam; antibody against HA (16B12) was from Eurogentec; anti-GM130 Invitrogen), electro-transferred to Protran nitrocellulose membrane (0.2 µm), and
(clone 35) and anti-flotillin-1 were from BD Biosciences; anti-LAMP-2 (H4B4) was incubated with the indicated antibodies. Signals were visualized with enhanced
from the Developmental Studies Hybridoma Bank of the University of Iowa, USA; chemiluminescence detection reagent (Amersham Pharmacia Biotech) and were
anti-phospho-p44/42 MAPK (9101) was from Cell Signaling. Cathepsin inhibitor quantified by densitometric scanning, using Quantity One software (Biorad).
I (Z-FG-NHO-Bz) and InSolution γ-Secretase Inhibitor X were from Calbiochem.
See Supplementary Table S1 for more information on antibodies and siRNA. Cell-surface-biotinylation experiments. MCF-7 cells were biotinylated with
reducible sulpho-NHS-SS-biotin (Pierce) at 4 ◦ C, as described previously33 .
Recombinant proteins, overlay and surface plasmon resonance experiments. Biotinylated cells were allowed to resume their membrane trafficking for 16 h.
Recombinant proteins for in vitro binding studies were prepared as GST fusions, Exosomes and cells were then collected as described above. Biotinylated proteins
6xHis-tagged or from self-splicing intein fusion protein (non-tagged) and used in were detected with streptavidin–HRP and signals were quantified as described above.
overlay and surface plasmon resonance experiments, as described previously30,54,59 .
Surface plasmon resonance was measured using a Biacore T200 instrument for Microscopy. For microscopic analysis, cells were plated on non-coated eight-well
Kd determination and a Biacore 2000 for the other experiments. For the triple chamber slides (Nalge Nunc International). The day after transfection, cells were
binding experiments a total of 1,000 response units (RU) of biotinylated SDC2 immunostained as described previously31 . Images were recorded with a Leica
cytoplasmic domain peptide, and for Kd determinations 100 RU of biotinylated AS MDW microscopy workstation (Leica Microsystems) or a confocal Olympus
peptides, were immobilized on streptavidin biosensor chips. For the ALIX–syntenin Fluoview 1000 instrument. The filling of RAB5Q79L endosomes was quantified with
binding experiments, a total of 9,000 RU of GST–syntenin full length, GST–syntenin ImageJ (National Institutes of Health, USA), and analysed for statistical significance
peptides corresponding to amino acids 1–30 or 31–60, or GST as a negative control, using Graphpad Prism 5.
were immobilized on a CM-5 biosensor chip.
Electron microscopy. For assessing MVB morphology, cells were incubated for
Immunoprecipitation from microsome extracts. Preparation of microsomal 30 min at 37 ◦ C with anti-CD63 antibody (Abcam) labelled with HRP (Z 25054
membranes and immunoprecipitations were as described previously30,31 . Molecular Probes). Cells were then washed with PBS and fixed for 1 h in 3%
glutaraldehyde in 0.1 M sodium cacodylate (NaCaC) buffer. After several washes
Cells and transfections. MCF-7 cells were obtained from ATCC. Cells were in 0.1 M NaCaC, cells were incubated with DAB (Sigma) in 0.05 M Tris–HCl, at
routinely grown in DMEM/F12 (1:1) medium (Life Technologies) supplemented pH 7.6. The enzymatic reaction was started by adding H2 O2 to a final concentration
with 10% fetal bovine serum (Hyclone). Serum used for experiments was depleted of of 0.01%, and allowed to proceed for 30 min. After washing in Tris–HCl buffer,
exosomes, by prior overnight centrifugation at 140,000g . For transient expressions, cells were postfixed for 1 h with 1% osmium tetroxide in 0.1 M phosphate buffer,
the cells were transfected the day after plating using Fugene 6 reagent (Roche). For washed, scraped, re-suspended in 2% agar and pelleted. Sliced agar blocks were
RNAi experiments, cells were transfected at 50% confluency with 100 nM siRNA dehydrated in a series of ethanol, and embedded in Agar 100 resin polymerized
using Oligofectamine (Invitrogen); cells were analysed 48 h after RNAi treatment. for 48 h at 60 ◦ C. Sections (60 nm; Ultracut E Leica) were counterstained with
5% uranyl acetate and lead citrate (Reynold’s), and were examined with a Jeol
Exosomes and total cell lysates. For comparative analyses, in gain- and loss- (JEM-2100) instrument. MVE/MVB sectional areas were measured using ImageJ
of-function studies, exosomes were collected from equivalent amounts of culture software (National Institutes of Health, USA) and data analysed using Graphpad
medium, conditioned by equivalent amounts of cells in triplicate cultures. When Prism 5.
reaching 80% of confluency, the cell layers were rinsed with DMEM/F12 and Exosomes purified on a continuous gradient as described above were re-
refreshed with DMEM/F12 containing 10% exosome-depleted FBS. The cell- suspended in PBS containing 4% paraformaldehyde, and negatively stained with 5%
conditioned media were collected 16 h later, pooling media from corresponding uranyl acetate.
triplicate cultures. Exosomes were isolated from these media by three sequential
centrifugation steps at 4 ◦ C: 15 min at 500g , to remove cells; 30 min at 10,000g , Fluorescence spectroscopy and two-colour fluorescence coincidence
to remove cell debris; and 3 h at 140,000g , to pellet exosomes, followed by detection. Freshly isolated HSC exosomes were diluted in PBS. Single-colour
one rinse (suspension in PBS/centrifugation at 140,000g ), to remove soluble fluorescence detection, and two-colour fluorescence coincidence detection mea-
serum and secreted proteins. Exosomal pellets prepared by differential HSC (HSC surements on these samples were performed at room temperature (∼21 ◦ C), using
exosomes) were then re-suspended in PBS or lysis buffer (PBS supplemented an LSM 510 Confocor-II combination system equipped with argon ion (488 nm)
with 1% NP40, 1 mM EDTA, 5 µg ml−1 leupeptin, 1 µg ml−1 pepstatin and 1 mM and HeNe (543 nm) lasers. The excitation light was focused into the sample by
phenylmethylsulphonyl fluoride). Corresponding cell layers were washed twice a Zeiss 40x/1.2 Plan-Apochromat water-immersion objective and a 488/543 main
in cold PBS, and scraped on ice in lysis buffer. Lysates from corresponding beam splitter. The detected light was split over two channels using a 570 nm dichroic
triplicate cultures were pooled, and cleared by centrifugation at 16,000g , for mirror. To narrow the emission spectrum, 600–650 nm and 505–530 nm band-pass
10 min at 4 ◦ C. Cell lysates were normalized by protein content, and corresponding filters were set before channel 1 and 2 respectively. Fluorescence fluctuations

NATURE CELL BIOLOGY


© 2012 Macmillan Publishers Limited. All rights reserved.
DOI: 10.1038/ncb2502 METHODS

were measured 50 times for 10 s, and each experiment was repeated at least three described above. Cells were lysed on ice, in lysis buffer supplemented with 1 mM
times, independently. Data were corrected for bleeding through, background and Na3 VO4 .
noise. The percentage of bleeding through in channel 1 (red) was calculated
using exosomes containing eGFP–syntenin. A signal of intensity fivefold above Statistical analysis. P values were calculated by two-tailed t test. For the data on
background fluorescence was defined as a peak; in addition, when a peak was MVE size, the Mann–Whitney test for non-normal distributions was used. All data
detected in channel 2 (green), any coincident signal in channel 1 (red) had to be were analysed using Graphpad Prism 5.
higher than 5 times the background plus the bleeding through level to be considered
as a bona fide (red) peak. 56. Sannerud, R. et al. ADP ribosylation factor 6 (ARF6) controls amyloid precursor
protein (APP) processing by mediating the endosomal sorting of BACE1. Proc. Natl
Syndecan clustering experiments. To reduce proteoglycan recruitment/cluster- Acad. Sci. USA 108, E559–E568 (2011).
ing, by preventing heparan sulphate engagements, the culture media were 57. David, G., van der Schueren, B., Marynen, P., Cassiman, J. J. & van den Berghe,
supplemented with bacterial heparitinase (0.006 U ml−1 ). For antibody-induced H. Molecular cloning of amphiglycan, a novel integral membrane heparan sulfate
clustering, the culture media were supplemented with the monoclonal anti- proteoglycan expressed by epithelial and fibroblastic cells. J. Cell Biol. 118,
syndecan antibodies BB4 (specific for syndecan 1; 0.1 µg ml−1 ) and 8G3 (specific for 961–969 (1992).
58. Lories, V., Cassiman, J. J., Van den Berghe, H. & David, G. Multiple distinct
syndecan 4; 1 µ g ml−1 ). Exosomes were collected after 16 h of medium conditioning.
membrane heparan sulfate proteoglycans in human lung fibroblasts. J. Biol. Chem.
264, 7009–7016 (1989).
Growth factor receptor trafficking. To test for FGFR transfer to exosomes 59. Ivarsson, Y. et al. Cooperative phosphoinositide and peptide binding by PSD-95/Discs
by western blotting, MCF-7 cells were transiently transfected with HA-tagged Large/ZO-1 (PDZ) domain of polychaetoid, Drosophila zonulin. J. Biol. Chem. 286,
FGFR-1 (ref. 60). When reaching near 80% of confluency, cells were serum- 44669–44678 (2011).
starved overnight in DMEM/F12 supplemented with 0.1% BSA. Starved cells were 60. Zhang, Z., Coomans, C. & David, G. Membrane heparan sulfate proteoglycan-
exposed to 2 ng ml−1 of FGF2 in 0.1%BSA-DMEM/F12, supplemented with 0.1 mM supported FGF2-FGFR1 signaling: evidence in support of the ‘cooperative end
Na3 VO4 , for the indicated lengths of time. Exosomes were prepared by HSC as structures’ model. J. Biol. Chem. 276, 41921–41929 (2001).

NATURE CELL BIOLOGY


© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

DOI: 10.1038/ncb2502

Figure S1 Mapping the syntenin-Alix interaction. (a) Syntenin-Alix by intra-molecular Bro1-PRD interactions) that prohibits Alix interactions
interaction, as detected by yeast two-hybrid experiments. Numbers indicate with ESCRT and viral proteins55. (b) Overlay experiment with recombinant
amino-acid residues. FL, full-length; m, mouse; h, human. Bait constructs GST-fusion proteins, showing that full-length Alix interacts with full-length
(left column) were cloned in pAS2, prey constructs (rows) in pGAD10. syntenin (FL), but not with the PDZ domains and C-terminus of syntenin
Positive clones were selected on triple minus plates (Leu-, Trp, His-) (DN-Ter). (c) Biacore experiments showing the direct interaction of GST-
containing 5 mM 3-aminotriazol. mAlix (383-869) corresponds to the prey Alix, used as analyte, with various immobilized GST-syntenin constructs,
isolated from a mouse whole embryo library in the original screen, using as indicated, but not with GST alone, used as a negative control (lower
full-length mouse syntenin as bait. Alix was the sole prey found to interact sensorgram). (d) Biacore experiments showing the interaction of non-tagged
with the N-Ter domain of syntenin in subsequent mapping experiments, recombinant syntenin with immobilized syndecan-2 (initial part of the
and also in separate unbiased screens using the N-Ter domain as bait. sensorgrams) and, subsequently, that of syndecan-2:syntenin complexes
Note that truncated forms of Alix containing the V domain interact with with GST-Alix (wild-type), but not with mutant GST-Alix F676D (later parts of
the two parts of the N-Ter domain of syntenin that contain LYPXL motifs, the sensorgrams). In both sensorgrams, perfusion of syntenin stops at 400
(1-30) and (31-60), and that mutation of the three LYPXL motifs (YP to sec; GST-Alix perfusions start at 1000 sec. All data together indicate that
AA) abolishes the interaction. Full-length Alix constructs do not interact syntenin-Alix interaction requires the LYPXL motifs in the N-terminal domain
with syntenin in this assay. Possible reasons might be trivial, but include an of syntenin and the V domain of Alix, with a critical role for residue F676 of
intrinsic auto-inhibitory mechanism (closed protein conformation, supported Alix. Uncropped data are shown in Fig. S6.

WWW.NATURE.COM/NATURECELLBIOLOGY 1
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S2 Syndecan, syntenin and CD63 release in exosomes. (a) Syndecan exosomes derived from MCF-7 cells that express, separately, either mCherry-
over-expression. Left panesl: Western blots of heparitinase-digested total cell syntenin or eGFP-syntenin (left; negative control), and for exosomes derived
lysates (lys) and HSC exosomes (exo) of MCF-7 cells, transiently transfected from MCF-7 cells co-expressing mCherry-syntenin and mGFP-CD63 (right).
with a construct encoding syndecan-1 or an empty vector (-), as indicated on
top of the lanes, showing the relative abundance of the intact (~65kDa) core
*
Note the extensive coincident detection ( ) of green and red peaks in the right
panel. (d) Similar fluorescence data were collected for exosomes isolated from
protein and its C-terminal fragment (CTF) (~10kDa). Upper right panels: intact MCF-7 cells, as indicated. For each experiment, the extent of co-appearance
forms and CTFs of over-expressed syndecan-1 co-fractionate during density- (number of peaks recorded in the green channel with co-incident peak in the red
gradient centrifugation, and fractionate as in Fig. 1g-h. Lower right panels: channel, expressed as percentage of the total number of spikes recorded in the
also in cells that over-express syndecan-1, the over-expression of wild-type green channel) was calculated from 50 datasets (50 recordings of 10 seconds).
syntenin stimulates the exosomal accumulation of syndecan-1 (intact and CTF). The mean percentages of co-appearance (± SD), calculated from independent
Over-expression of (triple LYP-LAA) mutant syntenin, deficient for Alix binding experiments are listed. Coincident fluorescent peaks were barely detectable in
(synteninDAlix), in contrast, has no or little effect. Compared to non-targeting exosome mixtures derived from cells separately expressing mCherry-syntenin
(nt) RNAi, syntenin RNAi and Alix RNAi decrease exosomal syndecan-1 (intact and mGFP-CD63, or in exosomes isolated from mixed spent culture media
and CTF). (b) Ceramide requirements. Western blots of HSC exosomes derived of such cells, excluding coincidence due to exosome compaction (as a result
from MCF-7 cells treated with non-targeting (nt) RNAi or with RNAi targeting of HSC) and incomplete resuspension. Adding detergent (octylglucoside),
neutral sphingomyelinase-2 (SMase2), assaying for syndecan-1 CTF, syntenin, during the isolation procedure or to the final exosome preparation, disrupted
CD63 and flotillin-1. Signals measured by densitometry were expressed as the fluorescent signals (data not shown). Such effect is consistent with the
percentage of the corresponding signals obtained in controls (nt RNAi); bars detection of fluorescence contained in individual vesicles, rather than individual
represent mean values obtained from 6 independent experiments. Error bars molecules. Of note, we found that, unlike endogenous flotillin, FP-labelled
represent standard deviations (SD). (c) Panels illustrating simultaneous green flotillin(s) did not incorporate in exosomes, excluding similar experiments for
and red fluorescence recordings over 10 seconds, obtained for a mixture of that exosomal marker. Uncropped data are shown in Fig. S6.

2 WWW.NATURE.COM/NATURECELLBIOLOGY
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S3 Exosome production. (a) Cell surface-biotinylated MCF-7 cells were conditioning), corrected for the number of particles of similar size present in
treated with various RNAi, targeting syndecan, syntenin or Alix, as indicated media at the beginning of the experiment (i.e. recovered by a rinse of parallel
on top. Non-targeting RNAi (nt) was used as control. Total cell lysates (lys) and cultures), was considered as the number of exosomes produced. Means (and
HSC exosomes (exo) were analysed by Western blotting, testing for biotinylated SD) of three independent experiments; courtesy of NanoSight. (c) Confocal
protein, using streptavidin. Representative Western blots are shown on the fluorescence spectroscopy. MCF-7 cells were transfected with mGFP-CD63,
left. Right, bar graphs representing mean signal intensities (relative to signals and with or without syntenin expression constructs. Green fluorescence present
in controls), and standard deviations (SD, error bars), calculated from several in HSC exosomes was monitored over time (10 seconds, 50 times). The
(n) independent experiments, for each RNAi, as indicated at the bottom. (b) number of peaks, total fluorescence in peaks, and modal peak fluorescence
Nanoparticle tracking. HSC exosome preparations from MCF-7 cells treated intensity, were measured in three independent experiments. Bars represent
with control RNAi, or with RNAi targeting syntenin, Alix or syndecan, and means (and SD) of the fold stimulations in syntenin-transfected cells, relative
also from cells that over-express (OE) syntenin were analyzed by nanoparticle to values obtained in mock-transfected cells (set as one). Note that syntenin
tracking analysis (NanoSight), providing information on the concentration over-expression more than doubles the total amount of (CD63) fluorescence
and size of particles. The number of particles of exosomal size (30-110 released, by increasing the number of peaks (exosomes), but barely affects the
nm) recovered from culture media at the end of the experiment (16h of modal signal intensity (amplitude). Uncropped data are shown in Fig. S6.

WWW.NATURE.COM/NATURECELLBIOLOGY 3
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S4 Syndecan, syntenin and Alix co-localize in Rab5Q79L endosomes. Merged images represent the merged mCh-syntenin (red channel), Alix (green
Confocal sections of MCF-7 cells transfected with cerulean-Rab5Q79L, mCherry- channel) and syndecan-1 ICD (blue channel) signals. The cer-Rab5Q79L signal
syntenin, syndecan-1 and Alix. Syndecan-1 intracellular domain (ICD) and is only shown as a separate image. Bar, 20µm. Boxed areas show Rab5Q79L
Alix were detected with, respectively, cognate mouse and rabbit antibodies. endosomes at high magnification. Uncropped data are shown in Fig. S6.

4 WWW.NATURE.COM/NATURECELLBIOLOGY
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S5 Endosomal membrane budding. (a) Western blots of HSC is counter-intuitive, as a defective membrane budding process, consuming
exosomes derived from MCF-7 cells that were sham-transfected (-) or lesser amounts of the limiting endosomal membrane, would tend to yield
over-expressing syntenin. Cells were treated with control (nt) or Rab7 larger endosomes. Thus, syntenin depletion reduces endosome filling with
RNAi, as indicated on top. (b) Gallery of electron micrographs, illustrating vesicles, but may also affect the progression of membranes and cargo
the size, filling and morphology of the late endosomal compartment, as through the endosomal compartment, slowing down the maturation of this
revealed by peroxidase-conjugated anti-CD63 internalization (30 min) and compartment. (d) Electron microscopic measurements of MVB/endosome
staining with DAB, in MCF-7 cells treated with control (nt), syntenin or filling, in cells treated with Rab7 RNAi, or both Rab7 RNAi and syntenin
Rab7 RNAi, or a combination of both syntenin RNAi and Rab7 RNAi, as RNAi. For each individual peroxidase/DAB-marked MVB/endosome, the
indicated. Bar, 200nm. (c) MVB/endosome size. The plot documents the total sectional area occupied by the MVB/endosome is plotted versus the
distributions and mean values (bars) of the total sectional areas occupied by area occupied by the stain in that MVB/endosome. The right panel shows
endosomes that contain DAB stain (n, indicating the number of endosomes the data points enclosed by the insert, at higher resolution. Note reduced
examined). Note that in the absence of syntenin, residual MVB/endosomes endosome filling in syntenin-depleted cells. Uncropped data are shown in
are significantly smaller in size than in corresponding controls. This result Fig. S6.

WWW.NATURE.COM/NATURECELLBIOLOGY 5
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 Uncropped data. The uncropped illustrations used to prepare the main and supplementary figures of this manuscript are shown with reference to
their specific sub-figures indicated on each panel.

6 WWW.NATURE.COM/NATURECELLBIOLOGY
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

WWW.NATURE.COM/NATURECELLBIOLOGY 7
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

8 WWW.NATURE.COM/NATURECELLBIOLOGY
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

WWW.NATURE.COM/NATURECELLBIOLOGY 9
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

10  WWW.NATURE.COM/NATURECELLBIOLOGY
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

WWW.NATURE.COM/NATURECELLBIOLOGY 11
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

12  WWW.NATURE.COM/NATURECELLBIOLOGY
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

WWW.NATURE.COM/NATURECELLBIOLOGY 13
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

14  WWW.NATURE.COM/NATURECELLBIOLOGY
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 continued

WWW.NATURE.COM/NATURECELLBIOLOGY 15
© 2012 Macmillan Publishers Limited. All rights reserved.
S U P P L E M E N TA R Y I N F O R M AT I O N

Table S1 Additional information about antibodies and siRNA.

16  WWW.NATURE.COM/NATURECELLBIOLOGY
© 2012 Macmillan Publishers Limited. All rights reserved.

You might also like