The Rockefeller University Press $30.00
J. Cell Biol. Vol. 193 No. 4 643–653
J.S. Bogan and D.K. Toomre contributed equally to this paper.
Correspondence to Derek K. Toomre: email@example.com; or Jonathan S.
Abbreviations used in this paper: FIPI, 5-fluoro-2-indolyl des-chlorohalopemide;
FWHM, full-width at half-maximum; GSV, GLUT4 storage vesicle; PLD, phospho-
lipase D; PM, plasma membrane; SDCM, spinning disc confocal microscopy;
shRNA, short hairpin RNA; TfR, transferrin receptor; TIRFM, total internal reflec-
tion fluorescence microscopy.
In adipocytes and myocytes, insulin causes translocation of
GLUT4 storage vesicles (GSVs) to the cell surface. Insertion
of GLUT4 glucose transporters at the plasma membrane (PM)
enhances glucose uptake and controls mammalian glucose
homeostasis (Watson et al., 2004; Huang and Czech, 2007).
Despite extensive study, where insulin acts to modulate GLUT4
trafficking is controversial (Dugani and Klip, 2005; Leney and
Tavaré, 2009). One model posits that the main effect of insulin is
to release GSVs from a deep, intracellular location (Bogan et al.,
2003; Govers et al., 2004; Muretta et al., 2008). A second, more
recent model challenges this view, and argues that fusion of
vesicles at the PM is the major rate-controlling step (Huang et al.,
2005; Koumanov et al., 2005; Bai et al., 2007).
Resolving where insulin acts has been hindered by sev-
eral challenges. First, GLUT4 resides in several intracellular
compartments, including the TGN, endosomes, and GSVs
(Bryant et al., 2002). How GLUT4 traffics among these com-
partments is a matter of debate. Both “static retention” and
“dynamic equilibrium” models have been proposed (Leney and
Tavaré, 2009; Muretta and Mastick, 2009). A key difference is
that in a “dynamic” model, insulin action at the PM can fully ac-
count for translocation. In contrast, the “static” model requires
release from an internal, sequestered pool. Second, biochemical
assays cannot easily quantify GLUT4 flux at individual traffick-
ing steps, nor distinguish action on exocytic and endocytic path-
ways. Insulin may act at multiple sites, and increased PM fusion
may actually result from release of an earlier rate-limiting step
(Chen and Saltiel, 2007; Yu et al., 2007; Fujita et al., 2010).
Finally, even with ultrasensitive imaging methods such as total
internal reflection fluorescence microscopy (TIRFM), the small
size of GSVs (60 nm) may preclude observation of unequivo-
cal fusion events. Detection of larger, brighter endosomes may
skew results. Indeed, using TIRFM, bona fide GLUT4-GFP
quantify the size, dynamics, and frequency of single ves-
icle exocytosis in 3T3-L1 adipocytes. We use a new GSV
reporter, VAMP2-pHluorin, and bypass insulin signaling
by disrupting the GLUT4-retention protein TUG. Remark-
ably, in unstimulated TUG-depleted cells, the exocytic
rate is similar to that in insulin-stimulated control cells.
In TUG-depleted cells, insulin triggers a transient, two-
fold burst of exocytosis. Surprisingly, insulin promotes
nsulin stimulates translocation of GLUT4 storage
vesicles (GSVs) to the surface of adipocytes, but pre-
cisely where insulin acts is controversial. Here we
fusion pore expansion, blocked by acute perturbation of
phospholipase D, which reflects both properties intrinsic
to the mobilized vesicles and a novel regulatory site at
the fusion pore itself. Prolonged stimulation causes cargo
to switch from 60 nm GSVs to larger exocytic vesicles
characteristic of endosomes. Our results support a model
whereby insulin promotes exocytic flux primarily by re-
leasing an intracellular brake, but also by accelerating
plasma membrane fusion and switching vesicle traffic
between two distinct circuits.
Dual-mode of insulin action controls GLUT4
Yingke Xu,1 Bradley R. Rubin,1,2 Charisse M. Orme,1,2 Alexander Karpikov,3 Chenfei Yu,2 Jonathan S. Bogan,1,2
and Derek K. Toomre1
1Department of Cell Biology; 2Section of Endocrinology and Metabolism, Department of Internal Medicine; and 3Department of Diagnostic Radiology, Yale University
School of Medicine, New Haven, CT 06520
© 2011 Xu et al. This article is distributed under the terms of an Attribution–Noncommercial–
Share Alike–No Mirror Sites license for the first six months after the publication date (see
http://www.rupress.org/terms). After six months it is available under a Creative Commons
License (Attribution–Noncommercial–Share Alike 3.0 Unported license, as described at
T H E J O U R N A L O F C E L L B I O L O G Y
JCB • VOLUME 193 • NUMBER 4 • 2011 644
much more fluorescent when transferred from an acidic to a
neutral environment (Miesenböck et al., 1998). We predicted
that VAMP2-pHluorin would be a more sensitive reporter of
GSV exocytosis compared with GLUT4-GFP.
Several criteria validated VAMP2-pHluorin as a probe to
image GSVs. First, VAMP2-GFP colocalized extensively with
GLUT4-DsRed (74%; Fig. 2 A). Second, insulin increased
the surface targeting of VAMP2-pHluorin (Fig. 2, B–D). Third,
VAMP2-pHluorin and GLUT4-DsRed colocalized in fusing
vesicles, as assessed using dual-color TIRFM (Fig. 2, E and F).
Of note, kiss-and-run fusion events were observed using
VAMP2-pHluorin, but not GLUT4-DsRed (Fig. 2 F). Finally,
TUG disruption redistributed both GLUT4 (Yu et al., 2007)
and VAMP2-pHluorin (Fig. S2, B–D) to the PM of unstimu-
Insulin regulates the stability of the
The pH sensitivity of pHluorin (pKa = 7.1), and even EGFP
(pKa = 6.0), has been exploited to visualize opening of the
fusion pore, which causes a sudden rise in fluorescence (Barg
et al., 2002; Bowser and Khakh, 2007). Using VAMP2-pHluorin,
we confirmed that GSV fusion pore formation could be identi-
fied seconds before full vesicle collapse (Fig. 3, A and B; and
Videos 3–5; Jiang et al., 2008). We define this operationally as
a “transition time” (or fusion pore duration).
The transition time varied widely, from less than a sec-
ond to several seconds (Fig. 3 B). Analysis of basal and insulin-
stimulated full fusion events yielded probability distributions in
Fig. 3 (C and D), with peaks at 1 and 3 s. A dual Gaussian
fit showed that most (56%) events in basal cells were of lon-
ger duration; this fraction dropped to 22% in insulin-treated
cells. Similar results were obtained using an arbitrary threshold
(Figs. 3, C and D) or cumulative probability analysis (Fig. S3 A;
P = 0.00035). Control experiments showed that insulin did not
affect the transition time of VAMP2-pHluorin in preadipocytes
(Fig. S3 B) or TfR-pHluorin in adipocytes (Fig. S3 C). Thus,
insulin may act specifically on GSVs to accelerate the transition
from pore formation to full expansion. Previous 2–5 s “dock-
ing” times assigned using GLUT4-GFP (Huang et al., 2005; Bai
et al., 2007) may be overestimated, as pore formation (a post-
docking step) was not detected. In primary adipocytes, GLUT4
release was slow and constrained, prompting speculation that
fusion pore opening may be regulated (Lizunov et al., 2005;
Stenkula et al., 2010).
We wondered if a prolonged transition time in un-
stimulated cells might increase the frequency of kiss-and-run
events, identified by a rapid fluorescence increase but no
spread (Fig. 3 E; see Materials and methods). Strikingly,
kiss-and-run events (defined operationally), in which the
pore opened, closed, and the vesicle stayed in the region and
slowly reacidified or moved axially, were common (37%)
in unstimulated cells, but were reduced to 5% of events
after insulin treatment (P < 0.01; Fig. 3 F). As chronic deple-
tion of phospholipase D1 (PLD1) caused vesicles to “dock” at
the PM, but not fuse (Huang et al., 2005), we tested if acute
perturbation of PLD with a general antagonist (1-butanol)
full fusion events were rare or difficult to detect (Lizunov et al.,
2005; Huang et al., 2007; Jiang et al., 2008).
To circumvent these issues, we used TUG knockdown
to release retained vesicles and VAMP2-pHluorin as a new
sensitive reporter for GSV fusion. Our data show that insulin
switches traffic between two circuits and acts in a dual mode:
the major activity is to release a brake, which retains GSVs
intracellularly in unstimulated cells. A second activity is to
accelerate fusion at the PM. Remarkably, insulin acts at the
exocytic fusion pore, and promotes dilation of the pore and full
Results and discussion
Differentiation of 3T3-L1 cells induces a
change in the size of GLUT4-containing
Single GLUT4 exocytic events in living cells may be caused by
fusion of endosomes or GSVs at the PM. Distinguishing these
vesicles is particularly important in 3T3-L1 adipocytes, which
are heterogeneous and have significant GLUT4 in endosomes
(Loo et al., 2009; Rubin and Bogan, 2009). We reasoned that
endosomes and GSVs may have distinct TIRFM exocytic sig-
natures (Fig. S1), as they differ considerably in size (100–250
vs. 50–70 nm diameters, respectively; Rodnick et al., 1992;
Kandror et al., 1995). We found that GLUT4-GFP produced a
bright vesicle fusion “flash” in 3T3-L1 preadipocytes, but fu-
sion was difficult to detect in mature adipocytes (Fig. 1, A–D;
Videos 1 and 2). This is consistent with the finding that GLUT4
traffics in endosomes in preadipocytes, but is targeted to GSVs
in mature adipocytes (Shi and Kandror, 2005).
As a novel means to estimate a vesicle’s size, we modeled
how its intensity varies during docking (Id) and fusion (If) in an
exponentially decaying evanescent field (Fig. 1 E, black curve;
see Materials and methods). The Id/If ratio is independent of the
vesicle’s absolute intensity and axial position, assuming that
the fluorophore is uniformly distributed on the surface of a
spherical vesicle. When docked, a larger vesicle will have a
smaller fraction of fluorophores in the evanescent field, and
Id/If will be smaller.
Our data show that GLUT4-containing exocytic vesicles
had a diameter of 153 ± 42 nm in preadipocytes, which is con-
sistent with GLUT4 recycling in endosomes, but that 56 ± 27 nm
vesicles were the main exocytic carriers in mature adipocytes
(Fig. 1 E). These smaller vesicles are identical in size to purified
GSVs, measured using electron microscopy (Kandror et al.,
1995). In addition, 92% of GLUT4 colocalized with transfer-
rin receptor (TfR) in TIRFM images of preadipocytes, but only
57% colocalized in adipocytes (Fig. S2 A). These data support
the finding that GSVs fuse directly with the PM.
Validation of VAMP2-pHluorin as a new
probe to visualize GSVs
To better identify GSV exocytosis, we used VAMP2-pHluorin.
VAMP2 is an established component of GSVs (Ramm et al.,
2000; Williams and Pessin, 2008; Bogan and Kandror, 2010),
and pHluorin is a pH-sensitive fluorescent protein that becomes
Dual-mode control of GLUT4 exocytosis • Xu et al.
Figure 1. 3T3-L1 adipocyte differentiation induces a change in the size of GLUT4 vesicles. (A–D) Single vesicle fusion events were visualized by TIRFM
imaging of 3T3-L1 cells stably expressing GLUT4-GFP (see Videos 1 and 2). Maximum projection time-lapse (10 Hz) images of preadipocytes (A) and mature
adipocytes (B) are shown, with galleries of single fusion events (asterisks) underneath (C and D). (E) Theoretical ratio of the fluorescence intensities of docked
(Id) to fully fused (If) membrane-labeled vesicles in an evanescent field of measured penetration depth, Dp (98 nm), were modeled as a function of vesicle size
(black curve, see Materials and methods). Id/If were measured for GLUT4 vesicles from preadipocytes and mature adipocytes and fitted to Gaussian distribu-
tions (inset), and the mean vesicle diameters were calculated using the black look-up curve (arrows). Bars: (A and B) 10 µm; (C and D) 1 µm.
JCB • VOLUME 193 • NUMBER 4 • 2011 646
Figure 2. Validation of VAMP2-pHluorin as a new probe to visualize GSVs. 3T3-L1 adipocytes were electroporated with VAMP2 and/or GLUT4 and
imaged by SDCM and TIRFM. (A) VAMP2-GFP and GLUT4-DsRed colocalize extensively (arrows). (B–D) Insulin-stimulated VAMP2-pHluorin translocation.
3T3-L1 adipocytes electroporated with VAMP2-pHluorin were imaged by 3D time-lapse SDCM. (B) Maximum z-projection images were taken before (0 min)
and after (5–30 min) 100 nM insulin addition. (C) Fluorescence intensity line profiles (yellow line in B) were plotted for images before (green) and 30 min
after (red) insulin. (D) The ratio of edge-to-middle intensities from line profiles of 10 cells was plotted (error bars indicate mean ± SEM; ***, p < 0.001).
(E and F) VAMP2-pHluorin colocalizes with GLUT4-DsRed and allows detection of full fusion (E) and kiss-and-run (F) fusion events using dual-color TIRFM.
Bars: (A and B) 10 µm; (E and F) 1 µm.
647Dual-mode control of GLUT4 exocytosis • Xu et al.
or a new inhibitor (5-fluoro-2-indolyl des-chlorohalopemide
[FIPI]; Su et al., 2009) affected the frequency of kiss-and-run
events. Both drugs increased the frequency of kiss-and-run
events in basal and insulin-stimulated cells (Figs. 3, G and H;
62 ± 5% for 1-butanol after insulin stimulation); as a control,
2-butanol had no effect (Fig. 3 G). These data implicate PLD
in the regulation of fusion pore dynamics. Together, the data
suggest a working model for GSV fusion (Fig. 3 I), whereby
insulin lowers the barrier to full vesicle fusion and thus reduces
the kiss-and-run frequency.
Figure 3. Insulin regulates the stability of the vesicle fusion pore. (A) pHluorin’s fluorescence brightens (arrowhead) upon fusion pore formation, and then
spreads laterally as the vesicle collapses into the PM (asterisk; see Fig. S1 [C and D] for fusion criteria). (B) Three examples of VAMP2-pHluorin fusion events
with different durations of pore opening. (C and D) Vesicle “transition times” (arrowhead to asterisk) were measured in basal (C, n = 163) and insulin-
stimulated (D, n = 894) cells, plotted as histograms, and fitted with a dual Gaussian distribution. (E) A representative kiss-and-run fusion event, whereby
the signal slowly dimmed but did not spread (see Fig. S1 D). Bars, 1 µm. (F–H) Percentage of kiss-and-run events in basal and insulin-stimulated cells
(F, n = 734 from three cells; **, P < 0.01), after addition of 1-butanol (G, n = 759 from three cells) or 2-butanol (n = 733 from three cells; ***, P < 0.001),
and in FIPI-treated cells, before and after insulin (H, n = 150 from two cells). Error bars indicate mean ± SEM. (I) Working model of the energy landscape
of vesicle fusion. Insulin reduces the barrier to full fusion after formation of the fusion pore.
JCB • VOLUME 193 • NUMBER 4 • 2011 648
Figure 4. TUG disruption increases GSV exocytic flux in unstimulated 3T3-L1 adipocytes. (A) Immunoblots of PM fractions from control, TUG shRNA, and
rescued (shRNA+TUG) cells. The ratio of VAMP2 to TfR abundance is indicated for each sample, normalized to unstimulated control cells. (B) PM VAMP2:TfR
was measured in three experiments by fractionation and immunoblotting. Mean ± SEM is plotted (error bars). *, P < 0.05; and ***, P < 0.001 relative to
basal control. (C) VAMP2-pHluorin was expressed transiently in control and TUG shRNA cells, and fusion events (crosses) were marked over a 3-min window
649Dual-mode control of GLUT4 exocytosis • Xu et al.
the idea that in addition to mobilizing GSVs, insulin acts at the
fusion pore itself.
As TUG depletion and insulin stimulation caused similar
rates of exocytosis, we tested if insulin signaling is activated
by TUG knockdown. Insulin-stimulated phosphorylation of
Akt is proposed to be sufficient to induce GLUT4 translocation
(Ng et al., 2008), yet data are equivocal on this point (Gonzalez
and McGraw, 2006; Chang et al., 2007). TUG depletion did not
cause Akt phosphorylation in unstimulated 3T3-L1 adipocytes,
and insulin-stimulated Akt phosphorylation was similar in con-
trol and TUG shRNA cells (Fig. 4 G). Thus, the enhanced exo-
cytic rate in basal TUG shRNA cells is independent of Akt.
Small GSVs are mobilized by acute
To analyze the size of insulin-stimulated exocytic vesicles, we
extended our method based on the ratio of fluorescence intensi-
ties in docked and fused states. We reasoned that ratios of
VAMP2-pHluorin intensities after fusion pore opening and after
full fusion would permit us to assess vesicle size, and to resolve
if insulin regulates distinct exocytic carriers. In basal adipocytes,
vesicles containing VAMP2-pHluorin were the same size as
those containing TfR-pHluorin, which supports the idea that they
originate from endosomes (Fig. 5 A). After brief (3–6 min) insu-
lin stimulation, small vesicles, identical in size to GLUT4-GFP
vesicles in mature adipocytes, were the predominant exocytic
carrier (Fig. 5, A and B). Similarly sized exocytic vesicles were
observed in unstimulated TUG shRNA cells, which supports the
idea that TUG regulates GSVs (Fig. 5 B). We hypothesized that
GSV cargo might not engage the TUG retention mechanism dur-
ing continued insulin exposure. Supporting this idea, only larger-
sized vesicles fused at the surface of cells after prolonged insulin
stimulation (Fig. 5, A and B), which is consistent with the idea
that the GSV compartment is bypassed.
Intriguingly, the frequency of kiss-and-run events after
prolonged (>15 min) insulin exposure was 25.5 ± 4.6% (n = 230
from two cells), much higher than the 5% observed after brief
insulin stimulation. Thus, the effect of insulin to reduce kiss-
and-run frequency at shorter time points (3–6 min) may in part
be secondary to its mobilization of small vesicles. Yet, insulin
markedly reduced the kiss-and-run frequency in TUG shRNA
cells, as noted earlier, in which small vesicles are the predomi-
nant exocytic carrier. A parsimonious interpretation of these
data is that GSVs have intrinsic properties that allow them to
respond to an insulin signal at the fusion pore, which then spe-
cifically promotes full fusion of these vesicles.
Dual-mode working model of insulin action
Our data indicate that GLUT4 traffics to the PM by two cir-
cuits, controlled by a dual brake/accelerator mode of insulin
GSV translocation to the PM in the
absence of TUG function
We next sought to bypass insulin signaling by manipulating TUG,
a protein that retains GLUT4 intracellularly in unstimulated cells
and that releases GLUT4 upon insulin stimulation (Bogan et al.,
2003; Yu et al., 2007). We hypothesized that TUG disruption
mobilizes GSV components (GLUT4 and VAMP2) to the PM.
Supporting this idea, biochemical fractionation of 3T3-L1
adipocytes showed that VAMP2 was enriched in PM fractions
from unstimulated cells containing a TUG short hairpin RNA
(shRNA; Fig. 4, A and B). This effect was similar to that ob-
served for GLUT4 (Yu et al., 2007). TfR distribution was
unaffected, which is consistent with the idea that TUG regulates
GSVs and not endosomes (Rubin and Bogan, 2009). Moreover,
confocal microscopy revealed high PM targeting of VAMP2-
pHluorin in unstimulated TUG shRNA cells (Figs. S2, B and D).
To test if TUG regulates GSV exocytosis, we imaged
VAMP2-pHluorin in control and TUG shRNA adipocytes using
TIRFM (Videos 3–5). Insulin greatly increased the exocytic rate
in control cells, yet this rate was high before insulin addition in
TUG shRNA cells (Fig. 4 C). Quantitative analysis (Fig. 4 D)
revealed that insulin increased the VAMP2-pHluorin exocytosis
rate by about fourfold in control 3T3-L1 adipocytes, similar to
previous data (Huang et al., 2007; Lopez et al., 2009). As a con-
trol, the rate of TfR-pHluorin fusion was unchanged. Strikingly, in
TUG shRNA cells, the basal rate of VAMP2-pHluorin exocytosis
was similar to that in insulin-stimulated control cells (Fig. 4 D,
broken red line), and was rescued by shRNA-resistant TUG
(Fig. 4 D, broken green line). In the TUG shRNA cells, insulin
promoted a transient (1–2 min), twofold increase in the fusion
rate. Accordingly, we tested if more vesicles are in close proximity
(e.g., tethered) to the PM in these cells. We found that the density
of GLUT4-GFP vesicles in the evanescent field in unstimulated
TUG shRNA cells was increased 1.5-fold compared with control
cells (Fig. 4 E; P < 0.001), and was rescued by shRNA-resistant
TUG. These data support the idea that a TUG-regulated step is
a major site at which insulin controls GSV exocytosis, and that
insulin also stimulates GSV fusion with the PM.
We next tested if insulin decreased kiss-and-run events in
TUG shRNA cells, which would support a direct action of insu-
lin to regulate the fusion pore. In basal TUG shRNA cells, the
kiss-and-run frequency was 26.2 ± 3.2%, and this rate decreased
to 6.8 ± 0.6% (approximately fourfold) after insulin stimula-
tion. The slightly lower basal rate was rescued by shRNA-
resistant TUG (to 33.6 ± 2.4%). These data support a direct
action of insulin on fusion pore dynamics. TUG knockdown
cells had very short transition times (t1/2 = 0.5 s), which sug-
gests an intrinsic property of the GSVs to promote fusion. None-
theless, insulin slightly, but significantly (P = 0.001), shortened
fusion pore duration (Fig. S3, D–F). Together, the data support
before (green) and after (red) insulin stimulation (see Videos 3 and 4). Bars, 10 µm. (D) Frequencies of VAMP2-pHluorin and TfR-pHluorin fusion events in
control (n = 1,070 from five cells) and TUG shRNA 3T3-L1 adipocytes (n = 895 from five cells). VAMP2-pHluorin events are also plotted for “rescued” TUG
shRNA+TUG cells (containing shRNA-resistant TUG, n = 262 from three cells). Data are mean ± SEM (error bars). (E) GLUT4-GFP was transfected in control,
TUG shRNA, and rescued (TUG shRNA+TUG) cells and imaged by TIRFM. Density of single vesicles in the evanescent field of unstimulated cells (n = 20
shRNA, 12 control, and 12 rescued cells; Dp = 98 nm; n = 10; ***, P < 0.001). (F) Percentage of kiss-and-run events in TUG shRNA cells (n = 676 from
three cells; **, P < 0.01). Error bars indicate mean ± SEM. (G) Immunoblots were performed as indicated on control and TUG shRNA adipocytes.
JCB • VOLUME 193 • NUMBER 4 • 2011 650
action (Fig. 5 C). First, insulin releases GLUT4 from TUG,
which serves as a brake to prevent GSV translocation in unstim-
ulated cells. Remarkably, the rate of GSV fusion was similar
in unstimulated TUG knockdown cells and insulin-stimulated
control cells. Thus, a major rate-limiting step for GSV trans-
location is upstream of vesicle fusion, which is counter to asser-
tions based in part on cell-free assays (Koumanov et al., 2005;
Leney and Tavaré, 2009) but consistent with a large immobile
GLUT4 pool in basal adipocytes (Fujita et al., 2010). Second,
insulin accelerates exocytosis of a fusion-ready pool of vesicles,
as observed in TUG-depleted cells. This dual mode can account
for the transition from basal to insulin-stimulated states.
Insulin also promoted a switch in the size of exocytic
vesicles (Fig. 5). In preadipocytes and unstimulated adipocytes,
the exocytic vesicles were identical in size to those containing
TfR, whereas after acute (3–6 min) insulin stimulation, the pre-
dominant carriers were smaller (60 nm) vesicles (GSVs). The
finding that GSV cargo is translocated in two distinct types of
vesicles is not compatible with models in which insulin only in-
creases the kinetics of exocytosis (Martin et al., 2006), but sup-
ports the idea that insulin enlarges the recycling pool (Govers
et al., 2004; Muretta et al., 2008). Together with the result that
larger vesicles are translocated during prolonged insulin expo-
sure, our data unite the static retention and dynamic equilibrium
models, as both participate in distinct phases of insulin action.
How insulin switches cargo to the two circuits is unclear, but
may involve AS160 modulation of Rab activity (Sakamoto and
Unexpectedly, insulin regulated fusion pore stability, re-
ducing the fraction of kiss-and-run events by more than sevenfold
and promoting full vesicle fusion. In part, this may reflect the mo-
bilization of distinct carriers (GSVs) with increased propensity to
fuse at the PM. Yet, the data also support an independent effect
of insulin to regulate the fusion pore directly, likely through PLD.
Thus, our data reveal the fusion pore as a novel site of insulin ac-
tion, which may shed light on how insulin action is impaired in
type 2 diabetes. In addition to PLD, Munc18c may mediate insu-
lin effects at this site (Huang et al., 2005; Smithers et al., 2008).
Finally, we speculate that fusion pore stability may be a control
point in other regulated exocytic systems, such as neurons or pan-
Materials and methods
Cell culture and biochemical methods
3T3-L1 cells were cultured in DME containing 10% fetal bovine serum, and
differentiated in media supplemented with 500 µM isomethylbutylxanthine,
0.25 µM dexamethasone, and 160 nM insulin as described previously
(Yu et al., 2007). After 8–10 d of differentiation, cells were electroporated
Figure 5. Insulin regulates two distinct pools of GLUT4-containing vesicles. 3T3-L1 adipocytes were transfected with plasmids encoding TfR-pHluorin or
VAMP2-pHluorin, and imaged by TIRFM before and after 100 nM insulin stimulation. (A) Ratios of the fluorescence intensities of fusion pore open and fully
fused states, Id/If, were measured using the indicated reporters. Histograms were plotted and fitted with Gaussian distributions, with basal VAMP2-pHluorin
is shown as a reference (blue). TfR-pHluorin data were similar for basal and insulin-stimulated cells; insulin data are shown. (B) Cumulative probability
plots of data. Significance was assessed using a Kolmogorov-Smirnov test with basal VAMP2-pHluorin data as a reference set. (C) Dual brake–accelerator
model of GSV translocation. Insulin acts both to release an intracellular brake (1) and to accelerate docking (2) and full fusion (3) at the PM, and switches
vesicle traffic between two cycles (see text).
651Dual-mode control of GLUT4 exocytosis • Xu et al.
depth (decay to 1/e or 37%) of the evanescent field was calculated to
be 98 nm for the TIRFM experiments.
Vesicle size calculation
To calculate the hypothetical size-dependent change in TIRFM fluorescence
intensity caused by geometric collapse of a “docked” spherical vesicle in
an exponentially decaying evanescent field (see Fig. 1 E), we performed
a numerical simulation. Specifically, to calculate the theoretical change in
fluorescence intensity from a docked (Id) state (or fusion pore open state,
Fig. 5) that contains a surface-labeled (2 nm shell) sphere of arbitrary
radius, R (in nm), to a fused “pancake” state (If) of 4 nm thickness and
radius 2R (to mimic collapse into the cell PM), we computed the intensity
in both states by numerical integration using the formula:
IC x y z ( , , )exp(z d dxdydz/ )
where C is unity inside of the “shell” or “pancake” and zero outside. The
numerical integration was performed using MATLAB (MathWorks), using a
penetration depth of 98 nm. Of note, this ratiometric calculation does not
depend on the real distance of the PM to the glass surface and the intensity
of the evanescent field I0 at the glass surface.
Stacks of time-lapse images were processed and analyzed using algo-
rithms implemented in ImageJ 1.42 (National Institutes of Health) and
custom-written MATLAB programs. Images were prepared with Photoshop
(Adobe); for display purposes, highly magnified images of single vesicles
were “bicubic” smoothed. Videos were compressed with QuickTime Pro
(Apple) using a JPEG 2000 algorithm. No nonlinear algorithms were used
to alter fluorescent signals. Colocalizations were quantified using ImageJ
and the “colocalization” plugin JACoP (provided by F. Cordelieres, Institut
Curie, Orsay, France). The degree of overlap of markers within individual
cells was determined using Pearson’s correlation coefficient. Single vesicles
in the TIRFM field were identified by ImageJ with the “particle detector”
plugin (provided by I. F. Sbalzarini, Eidgenössische Technische Hochschule
Zürich, Zürich, Switzerland).
For analysis of single fusion events, each acquired image sequence
was manually reviewed multiple times to visually identify putative vesicle
fusion events. The coordinates of the fusion events were marked and a
small region of interest around each fusion was used for further analysis.
Circular regions with diameters of 10 pixels (1.8 µm) were used to cal-
culate the intensity of single vesicles. The local background, the average in-
tensity of a ring just outside this circle, was subtracted. The fusion profiles
were analyzed with the implementation of a Gaussian fit. The integrated in-
tensity, peak intensity and full-width at half-maximum (FWHM) of a Gaussian
distribution were used to validate and distinguish between full fusion and
kiss-and-run fusion events (Fig. S1, C and D). The peak intensity drops
with a concurrent increase in FWHM when a vesicle undergoes full fusion
(Fig. S1 C), whereas a kiss-and-run fusion event was defined operationally as
having only a decrease in peak intensity without an increase of the FWHM
(Fig. S1 D). For GLUT4-GFP fusion events, the intensity of docked vesicles, Id,
was monitored in the penultimate frames before full fusion events of intensity
If. For VAMP2-pHluorin or TfR-pHluorin, Id was assigned to the mean intensity
of a vesicle after the opening of the fusion pore, but before full fusion.
Binned counts of events were normalized such that bin areas
summed to unity: each normalized bin height represents the probability per
unit bin of observing the values in that bin. These observed probability
densities were fitted with Gaussian probability densities using a maximum
likelihood estimation in the Statistics Toolbox of MATLAB. In some cases, a
mixture of two Gaussian probability densities was used:
whereby is fractional contribution of the first Gaussian, and 1 and 2
are the standard deviations of the two Gaussians. Unless otherwise indi-
cated, data are presented as the mean ± SEM, and were analyzed using
a Student’s t test.
Online supplemental material
Fig. S1 shows how single vesicle fusion events were analyzed, and in-
cludes examples of GLUT4-GFP in 3T3-L1 preadipocytes and mature
adipocytes, and of VAMP2-pHluorin full fusion and kiss-and-run events in
mature 3T3-L1 adipocytes. Fig. S2 shows that differentiation of 3T3-L1
(Lonza) according to the manufacturer’s instructions using Transfection solu-
tion “L” and cultured on glass-bottomed 35 mm dishes (MaTek Corporation)
for microscopy (see Live cell imaging). Plasmids encoding superecliptic
VAMP2-pHluorin, TfR-pHluorin, TfR-mCherry, and myc-GLUT4-DsRed were
gifts of J. Rothman (Yale School of Medicine, New Haven, CT), C. Merrifield
(Medical Research Council Laboratory of Molecular Biology, Cambridge,
England, UK), M. Ehlers (Duke University Medical Center, Durham, NC),
and M. Cormont (Centre Méditerranéen de Médecine Moléculaire, Nice,
France), respectively. For stable expression of exogenous proteins or TUG
shRNA, 3T3-L1 cells were infected with retroviruses as described previ-
ously (Bogan et al., 2001; Yu et al., 2007).
Subcellular fractionation and immunoblotting were performed as de-
scribed previously (Yu et al., 2007). In brief, four to six 10-cm plates of adi-
pocytes were used per condition. Cells were stimulated for 8–10 min using
480 nM insulin, washed in cold PBS, and scraped in ice-cold 250 mM
sucrose, 10 mM Tris, pH 7.4, and 0.5 mM EDTA (TES) containing protease
inhibitors (Complete; Roche). All subsequent steps were done at 4°C
or on ice. Cells were homogenized in a Dounce-type Teflon tissue grinder
(Kontes no. 22; VWR International), then centrifuged at 12,000 rpm for
15 min in an SS-34 rotor (Sorvall). The pellet was resuspended in TES and
centrifuged again at 12,000 rpm for 20 min in an SS-34 rotor. To isolate
PM fractions, the pellet was resuspended in 1 ml TES, layered on top of
a 1.12 M sucrose cushion (made in TES) in a 2 ml centrifuge tube, and
centrifuged at 36,000 rpm for 20 min in a TLS-55 rotor (Beckman Coulter).
The interface was removed using a syringe, diluted in TES, and centrifuged
at 37,000 rpm for 9 min in a TLA-120.2 rotor (Beckman Coulter). The pel-
let was resuspended and centrifuged again under identical conditions. The
pellet was resuspended in SDS-PAGE sample buffer, and equal amounts
of protein (by EZ-Q assays) in each sample were analyzed by SDS-PAGE
and immunoblotting. NuPAGE gels and buffers were used (Invitrogen),
and proteins were transferred to nitrocellulose membranes using a semidry
apparatus. Detection was done both on film, using enhanced chemilumi-
nescence (Thermo Fisher Scientific), and with an infrared imaging system
(LI-COR Biosciences). For images acquired on film, densitometry was done
on exposures in the linear range, which were digitized using a trans-
illuminating flatbed scanner. Antibodies used were: mouse anti-myc (9E10;
Covance), anti-Akt and phospho-Akt antibodies (Cell Signaling Technol-
ogy), anti–-actin (Abcam), anti-VAMP2 (a gift of J. Rothman), anti-TfR
(a gift of P. De Camilli, Yale School of Medicine, New Haven, CT), and
anti-TUG (Bogan et al., 2003). Antibodies were typically used at dilu-
tions of 1:1,000. Secondary antibodies were conjugated to peroxidase
(Bio-Rad Laboratories) for detection on film, or to IRDye700 or IRDye800
(LI-COR Biosciences) for digital imaging.
Live cell imaging
Electroporated 3T3-L1 adipocytes were imaged 15–20 h after transfec-
tion. Cells grown in MatTek chambers were serum starved for 2 h, then
imaged in KRH buffer, pH 7.4, containing 125 mM NaCl, 5 mM KCl,
1.3 mM CaCl2, 1.2 mM MgSO4, 20 mM d-glucose, 25 mM Hepes, and
0.2% bovine serum albumin. Insulin, FIPI, and 1- and 2-butanol were ob-
tained from Sigma-Aldrich, and insulin was used at a final concentration
of 100 nM. Cells were kept in an Air-therm (WPI) temperature-regulated
environmental box at 37°C throughout the experiments.
The TIRFM setup was based on an IX-70 inverted microscope
(Olympus), equipped with argon (488 nm) and argon/krypton (568 nm)
laser lines (Melles Griot), a 60× 1.45 NA oil immersion objective lens
(Plan-ApoN; Olympus), and a TIRFM condenser. Cells were imaged in one
channel at 5–10 Hz or two channels by sequential excitation at 3–5 Hz,
without binning, and detected with a back-illuminated Andor iXon887
EMCCD camera (512 × 512, 0.18 µm per pixel, 16 bits; Andor Technolo-
gies) with a 1.5× expansion lens. The TIRFM system was controlled by
Andor iQ software, and the calculated depth of the evanescent field was
98 nm (see TIRFM calibration).
A Yokagawa-type spinning disc confocal microscope system was
used for fast 4D imaging (PerkinElmer). The system was mounted onto an
inverted microscope (IX-71; Olympus) quipped with a 1 Kb × 1 Kb EM
charge-coupled device camera (Hamamatsu Photonics). The spinning disc
confocal microscopy (SDCM) system was controlled by Volocity software,
cells were imaged by using a 100× 1.4 NA oil objective lens, and expo-
sure times were 0.1–0.25 s.
Silica beads (Lot GK1529943W; Kisker-Biotech) with a 20-µm diameter
were used for TIRFM calibration, as described previously (Mattheyses and
Axelrod, 2006). The bead diameter was determined by taking a z-stack
image using PIFOC piezo device (Physik Instrumente). The penetration
JCB • VOLUME 193 • NUMBER 4 • 2011 652
Huang, S., and M.P. Czech. 2007. The GLUT4 glucose transporter. Cell Metab.
Huang, P., Y.M. Altshuller, J.C. Hou, J.E. Pessin, and M.A. Frohman. 2005.
Insulin-stimulated plasma membrane fusion of Glut4 glucose transporter-
containing vesicles is regulated by phospholipase D1. Mol. Biol. Cell.
Huang, S., L.M. Lifshitz, C. Jones, K.D. Bellve, C. Standley, S. Fonseca, S. Corvera,
K.E. Fogarty, and M.P. Czech. 2007. Insulin stimulates membrane fusion
and GLUT4 accumulation in clathrin coats on adipocyte plasma membranes.
Mol. Cell. Biol. 27:3456–3469. doi:10.1128/MCB.01719-06
Jiang, L., J. Fan, L. Bai, Y. Wang, Y. Chen, L. Yang, L. Chen, and T. Xu. 2008.
Direct quantification of fusion rate reveals a distal role for AS160 in insu-
lin-stimulated fusion of GLUT4 storage vesicles. J. Biol. Chem.
Kandror, K.V., L. Coderre, A.V. Pushkin, and P.F. Pilch. 1995. Comparison
of glucose-transporter-containing vesicles from rat fat and muscle
tissues: evidence for a unique endosomal compartment. Biochem. J.
Koumanov, F., B. Jin, J. Yang, and G.D. Holman. 2005. Insulin signaling meets
vesicle traffic of GLUT4 at a plasma-membrane-activated fusion step.
Cell Metab. 2:179–189. doi:10.1016/j.cmet.2005.08.007
Leney, S.E., and J.M. Tavaré. 2009. The molecular basis of insulin-stimulated
glucose uptake: signalling, trafficking and potential drug targets.
J. Endocrinol. 203:1–18. doi:10.1677/JOE-09-0037
Lizunov, V.A., H. Matsumoto, J. Zimmerberg, S.W. Cushman, and V.A. Frolov.
2005. Insulin stimulates the halting, tethering, and fusion of mobile
GLUT4 vesicles in rat adipose cells. J. Cell Biol. 169:481–489.
Loo, L.H., H.J. Lin, D.K. Singh, K.M. Lyons, S.J. Altschuler, and L.F. Wu. 2009.
Heterogeneity in the physiological states and pharmacological responses
of differentiating 3T3-L1 preadipocytes. J. Cell Biol. 187:375–384.
Lopez, J.A., J.G. Burchfield, D.H. Blair, K. Mele, Y. Ng, P. Vallotton, D.E. James,
and W.E. Hughes. 2009. Identification of a distal GLUT4 trafficking
event controlled by actin polymerization. Mol. Biol. Cell. 20:3918–3929.
Martin, O.J., A. Lee, and T.E. McGraw. 2006. GLUT4 distribution between the
plasma membrane and the intracellular compartments is maintained by an
insulin-modulated bipartite dynamic mechanism. J. Biol. Chem. 281:
Mattheyses, A.L., and D. Axelrod. 2006. Direct measurement of the evanescent
field profile produced by objective-based total internal reflection fluores-
cence. J. Biomed. Opt. 11:014006. doi:10.1117/1.2161018
Miesenböck, G., D.A. De Angelis, and J.E. Rothman. 1998. Visualizing secretion
and synaptic transmission with pH-sensitive green fluorescent proteins.
Nature. 394:192–195. doi:10.1038/28190
Muretta, J.M., and C.C. Mastick. 2009. How insulin regulates glucose trans-
port in adipocytes. Vitam. Horm. 80:245–286. doi:10.1016/S0083-
Muretta, J.M., I. Romenskaia, and C.C. Mastick. 2008. Insulin releases Glut4
from static storage compartments into cycling endosomes and increases
the rate constant for Glut4 exocytosis. J. Biol. Chem. 283:311–323.
Ng, Y., G. Ramm, J.A. Lopez, and D.E. James. 2008. Rapid activation of Akt2
is sufficient to stimulate GLUT4 translocation in 3T3-L1 adipocytes.
Cell Metab. 7:348–356. doi:10.1016/j.cmet.2008.02.008
Ramm, G., J.W. Slot, D.E. James, and W. Stoorvogel. 2000. Insulin recruits
GLUT4 from specialized VAMP2-carrying vesicles as well as from the
dynamic endosomal/trans-Golgi network in rat adipocytes. Mol. Biol.
Rodnick, K.J., J.W. Slot, D.R. Studelska, D.E. Hanpeter, L.J. Robinson, H.J.
Geuze, and D.E. James. 1992. Immunocytochemical and biochemical
studies of GLUT4 in rat skeletal muscle. J. Biol. Chem. 267:6278–6285.
Rubin, B.R., and J.S. Bogan. 2009. Intracellular retention and insulin-stimulated
mobilization of GLUT4 glucose transporters. Vitam. Horm. 80:155–192.
Sakamoto, K., and G.D. Holman. 2008. Emerging role for AS160/TBC1D4 and
TBC1D1 in the regulation of GLUT4 traffic. Am. J. Physiol. Endocrinol.
Metab. 295:E29–E37. doi:10.1152/ajpendo.90331.2008
Shi, J., and K.V. Kandror. 2005. Sortilin is essential and sufficient for the forma-
tion of Glut4 storage vesicles in 3T3-L1 adipocytes. Dev. Cell. 9:99–108.
Smithers, N.P., C.P. Hodgkinson, M. Cuttle, and G.J. Sale. 2008. Insulin-triggered
repositioning of munc18c on syntaxin-4 in GLUT4 signalling. Biochem.
J. 410:255–260. doi:10.1042/BJ20070802
Stenkula, K.G., V.A. Lizunov, S.W. Cushman, and J. Zimmerberg. 2010. Insulin
controls the spatial distribution of GLUT4 on the cell surface through
adipocytes causes the targeting of GLUT4 to TfR-negative vesicles, as im-
aged by TIRFM. SDCM shows that VAMP2-pHluorin is distributed to the
cell surface in both basal and insulin-stimulated TUG knockdown cells.
Fig. S3 shows that insulin regulates VAMP2-pHluorin fusion pore dynamics
in mature 3T3-L1 adipocytes, and reduces the transition time from pore
opening to full vesicle fusion. Control experiments demonstrate that there is
no effect of insulin on fusion pore dynamics in preadipocytes, or in mature
adipocytes containing TfR-pHluorin. The effect of insulin is also observed
in TUG shRNA cells. The supplemental videos present data that are sum-
marized and quantified in the main text and figures. Video 1 shows TIRFM
imaging of exocytosis of GLUT4-GFP in a 3T3-L1 preadipocyte. Video 2
presents TIRFM of GLUT4-GFP in a mature insulin-stimulated 3T3-L1 adi-
pocyte. Video 3 shows TIRFM of VAMP2-pHluorin translocation in a wild-
type 3T3-L1 adipocyte. Video 4 shows TIRFM of VAMP2-pHluorin in a
TUG knockdown 3T3-L1 adipocyte. Video 5 shows a high-magnification
cropped view of VAMP2-pHluorin exocytosis after insulin stimulation in a
3T3-L1 adipocyte. Online supplemental material is available at http://
We thank Drs. J. Rothman, P. De Camilli, C. Merrifield, M. Ehlers, M. Cormont,
and E. Karatekin for reagents and advice.
This work was supported by fellowships to Y. Xu (China Postdoctoral
Science Foundation 20080441255 and NSFC 30770596), B.R. Rubin
(American Heart Association), and C.M. Orme (National Institutes of Health
[NIH] F30DK086109), as well as by grants to J.S. Bogan (W. M. Keck Foun-
dation and NIH R01DK075772) and D.K. Toomre (NIH DP2OD002980),
and the Cell Biology Core of the Yale Diabetes Endocrinology Research
Submitted: 24 August 2010
Accepted: 11 April 2011
Bai, L., Y. Wang, J. Fan, Y. Chen, W. Ji, A. Qu, P. Xu, D.E. James, and T. Xu. 2007.
Dissecting multiple steps of GLUT4 trafficking and identifying the sites
of insulin action. Cell Metab. 5:47–57. doi:10.1016/j.cmet.2006.11.013
Barg, S., C.S. Olofsson, J. Schriever-Abeln, A. Wendt, S. Gebre-Medhin, E.
Renström, and P. Rorsman. 2002. Delay between fusion pore opening and
peptide release from large dense-core vesicles in neuroendocrine cells.
Neuron. 33:287–299. doi:10.1016/S0896-6273(02)00563-9
Bogan, J.S., and K.V. Kandror. 2010. Biogenesis and regulation of insulin-
responsive vesicles containing GLUT4. Curr. Opin. Cell Biol. 22:
Bogan, J.S., A.E. McKee, and H.F. Lodish. 2001. Insulin-responsive compart-
ments containing GLUT4 in 3T3-L1 and CHO cells: regulation by amino
acid concentrations. Mol. Cell. Biol. 21:4785–4806. doi:10.1128/MCB
Bogan, J.S., N. Hendon, A.E. McKee, T.S. Tsao, and H.F. Lodish. 2003.
Functional cloning of TUG as a regulator of GLUT4 glucose transporter
trafficking. Nature. 425:727–733. doi:10.1038/nature01989
Bowser, D.N., and B.S. Khakh. 2007. Two forms of single-vesicle astrocyte
exocytosis imaged with total internal reflection fluorescence microscopy.
Proc. Natl. Acad. Sci. USA. 104:4212–4217. doi:10.1073/pnas.0607625104
Bryant, N.J., R. Govers, and D.E. James. 2002. Regulated transport of the
glucose transporter GLUT4. Nat. Rev. Mol. Cell Biol. 3:267–277. doi:
Chang, L., S.H. Chiang, and A.R. Saltiel. 2007. TC10alpha is required for insulin-
stimulated glucose uptake in adipocytes. Endocrinology. 148:27–33.
Chen, X.W., and A.R. Saltiel. 2007. TIRFing out studies on Glut4 trafficking.
Dev. Cell. 12:4–5. doi:10.1016/j.devcel.2006.12.008
Dugani, C.B., and A. Klip. 2005. Glucose transporter 4: cycling, compartments and
controversies. EMBO Rep. 6:1137–1142. doi:10.1038/sj.embor.7400584
Fujita, H., H. Hatakeyama, T.M. Watanabe, M. Sato, H. Higuchi, and M. Kanzaki.
2010. Identification of three distinct functional sites of insulin-mediated
GLUT4 trafficking in adipocytes using quantitative single molecule imaging.
Mol. Biol. Cell. 21:2721–2731. doi:10.1091/mbc.E10-01-0029
Gonzalez, E., and T.E. McGraw. 2006. Insulin signaling diverges into Akt-
dependent and -independent signals to regulate the recruitment/docking
and the fusion of GLUT4 vesicles to the plasma membrane. Mol. Biol.
Cell. 17:4484–4493. doi:10.1091/mbc.E06-07-0585
Govers, R., A.C. Coster, and D.E. James. 2004. Insulin increases cell sur-
face GLUT4 levels by dose dependently discharging GLUT4 into a
cell surface recycling pathway. Mol. Cell. Biol. 24:6456–6466. doi:10
653Dual-mode control of GLUT4 exocytosis • Xu et al.
regulation of its postfusion dispersal. Cell Metab. 12:250–259. doi:10
Su, W., O. Yeku, S. Olepu, A. Genna, J.S. Park, H. Ren, G. Du, M.H. Gelb, A.J.
Morris, and M.A. Frohman. 2009. 5-Fluoro-2-indolyl des-chlorohalopemide
(FIPI), a phospholipase D pharmacological inhibitor that alters cell
spreading and inhibits chemotaxis. Mol. Pharmacol. 75:437–446. doi:
Watson, R.T., M. Kanzaki, and J.E. Pessin. 2004. Regulated membrane traffick-
ing of the insulin-responsive glucose transporter 4 in adipocytes. Endocr.
Rev. 25:177–204. doi:10.1210/er.2003-0011
Williams, D., and J.E. Pessin. 2008. Mapping of R-SNARE function at distinct
intracellular GLUT4 trafficking steps in adipocytes. J. Cell Biol. 180:375–
Yu, C., J. Cresswell, M.G. Löffler, and J.S. Bogan. 2007. The glucose transporter
4-regulating protein TUG is essential for highly insulin-responsive glu-
cose uptake in 3T3-L1 adipocytes. J. Biol. Chem. 282:7710–7722. doi: