Glucose transport in adipose cells is regulated by changing the distribution of glucose transporter 4 (GLUT4) between the cell interior and the plasma membrane (PM). Insulin shifts this distribution by augmenting the rate of exocytosis of specialized GLUT4 vesicles. We applied time-lapse total internal reflection fluorescence microscopy to dissect intermediates of this GLUT4 translocation in rat adipose cells in primary culture. Without insulin, GLUT4 vesicles rapidly moved along a microtubule network covering the entire PM, periodically stopping, most often just briefly, by loosely tethering to the PM. Insulin halted this traffic by tightly tethering vesicles to the PM where they formed clusters and slowly fused to the PM. This slow release of GLUT4 determined the overall increase of the PM GLUT4. Thus, insulin initially recruits GLUT4 sequestered in mobile vesicles near the PM. It is likely that the primary mechanism of insulin action in GLUT4 translocation is to stimulate tethering and fusion of trafficking vesicles to specific fusion sites in the PM.
Insulin regulates glucose transport in muscle and adipose cells through the intracellular redistribution of the glucose transporter 4 (GLUT4; Cushman and Wardzala, 1980; Suzuki and Kono, 1980). GLUT4 content in the plasma membrane (PM) of adipose cells is determined by a dynamic equilibrium between its exocytosis and internalization. In basal adipose cells, the content of GLUT4 in the PM remains low (<5%) as GLUT4 internalizes 10 times faster than it is delivered to the PM (Slot et al., 1991; Satoh et al., 1993; Holman et al., 1994; Malide et al., 2000; for review see Bryant et al., 2002). Insulin considerably stimulates the rate of GLUT4 exocytosis with relatively little effect on the rate of internalization (Satoh et al., 1993; Karylowski et al., 2004). Consequently, ∼50% of intracellular GLUT4 is translocated to the PM upon insulin activation, providing a >10-fold increase in the amount of transporter on the cell surface (Malide et al., 2000; for reviews see Pessin et al., 1999; Bryant et al., 2002).
GLUT4 is carried to the PM by specialized tubulo-vesicular compartments (referred to here as GLUT4 vesicles) tightly packed with the transporter (Rea and James, 1997). GLUT4 vesicles belong to specialized exocytic compartments actively excluding other recycling proteins (Martin et al., 1996; Malide et al., 1997; Hashiramoto and James, 2000; Lampson et al., 2001). In the basal state, the GLUT4 vesicles sequester up to 95% of the total GLUT4 in adipose cells (Martin et al., 1996; Malide et al., 1997; Lee et al., 1999). The vesicles slowly exchange GLUT4 with the rest of the intracellular pool and with the PM (Holman et al., 1994; Karylowski et al., 2004; Watson et al., 2004). The dynamics of basal GLUT4 recycling indicate that a fraction of the GLUT4 vesicles is mobile (Bryant et al., 2002). It is increasingly thought that insulin regulates the mobility of GLUT4 vesicles, yet the details and mechanisms of insulin's action remain controversial.
Live microscopy studies performed on 3T3-L1 adipocytes (3T3 fibroblast cell line that has been differentiated to adipose-like cells) demonstrate that in the basal state most of the GLUT4 vesicles remain relatively static (Oatey et al., 1997; Patki et al., 2001), although long-range movements of GLUT4 vesicles along microtubules were occasionally observed (Fletcher et al., 2000; Semiz et al., 2003). The microtubule network supports the GLUT4 translocation induced by insulin (Patki et al., 2001; Semiz et al., 2003; Karylowski et al., 2004). The disruption of the microtubular network prevents long-range movements of GLUT4 vesicles and substantially inhibits GLUT4 translocation to the PM (Fletcher et al., 2000; Patki et al., 2001). Hence, it has been suggested that insulin stimulates movement of GLUT4 vesicles from the cell interior toward the PM along microtubule tracks (Patki et al., 2001; Semiz et al., 2003). Translocation of the new vesicles to the PM would boost GLUT4 exocytosis because the vesicles carry essential parts of the docking/fusion machinery (Martin et al., 1996; Timmers et al., 1996; Grusovin and Macaulay, 2003).
However, in nonstimulated primary adipose cells, a fraction of all GLUT4 vesicles is already located near the PM. Even in 3T3-L1 adipocytes, where GLUT4 is mainly concentrated in the perinuclear region (Patki et al., 2001), GLUT4-containing vesicles were also reportedly found near the PM (Oatey et al., 1997). These peripheral vesicles were suggested to be an initial source of GLUT4 translocated to the PM upon insulin stimulation (Fletcher et al., 2000; Semiz et al., 2003), although it has never been directly demonstrated. In primary adipose cells, the disperse cytoplasmic GLUT4 vesicles already make up a majority of the GLUT4-containing compartments (Lee et al., 1999; Malide et al., 2000). In freshly isolated primary adipose cells, the GLUT4 vesicles are uniformly spread over the whole cytoplasm with many of the vesicles located near the PM (Slot et al., 1991; Malide et al., 1997, 2000). Such spatial distribution of GLUT4 in basal primary adipose cells immediately suggests that GLUT4 vesicles proximal to the PM play an essential role in the insulin response in these cells. However, the dynamics of these GLUT4 vesicles in primary adipose cells remain unknown.
Recently, insulin has been shown to stimulate translocation of effector molecules providing specific targeting of GLUT4 vesicles to the PM (Grusovin and Macaulay, 2003; Inoue et al., 2003; Saltiel and Pessin, 2003). Thus, insulin may be involved in regulating tethering of GLUT4 vesicles located near the PM. To test this hypothesis, we designed an approach to visualize the dynamic behavior of GLUT4 vesicles located near the PM in primary adipose cells using total internal reflection fluorescence (TIRF) microscopy (TIRFM). Here, we report the resolution of the movements, docking, and fusion of single GLUT4 vesicles, in both the basal and insulin-stimulated states. We find that without insulin GLUT4 vesicles rapidly traffic near the PM along stationary trajectories formed by a microtubule network covering the entire PM, although GLUT4 vesicles transiently stop and loosely tether to the PM. Insulin stops this traffic, tightly tethering vesicles at the PM where they form clusters. After fusion, the slow diffusion of GLUT4 from clustered vesicles determines the overall increase of PM GLUT4. We conclude that one of the primary modes of insulin action on adipose cells is to halt a hitherto unrecognized form of GLUT4 trafficking that scans the cytoplasmic surface of a cell membrane, and then stimulates tethering and fusion of vesicles bearing glucose transporters into the cell membrane.
In isolated adipose cells, GLUT4 vesicles are scattered near the PM
When isolated from tissue, white adipose cells are round and their interior is filled with a large droplet of stored triglyceride (Fig. 1, A and B), leaving a thin (1–3-μm) layer of cytoplasm between the lipid droplet and the PM (Cushman, 1970; Malide et al., 2000). In our experiments, these cells, which ordinarily float in their medium, were slightly pressed against the coverslip such that a small flattened part of the PM and nearby cytoplasm (hereafter termed the “TIRF zone”) was illuminated by the evanescent wave (Fig. 1 B and Fig. S1 A, available). The three-dimensional distribution and dynamics of GLUT4 vesicles in the cytoplasm of basal cells was characterized by confocal and TIRFM.
The HA-GLUT4-GFP construct expressed in these cells (Dawson et al., 2001) was seen in distinct bright fluorescent spots (Fig. 1, C and D). This punctate pattern of GLUT4 distribution is consistent with the previous observations that up to 95% of intracellular GLUT4 is stored in vesicles (Lee et al., 1999; Malide et al., 2000). Confocal microscopy showed that for most of the transfected cells cultured overnight the punctate patterns of GLUT4 distribution (Fig. 1 C) resemble those of freshly isolated cells; i.e., GLUT4 predominantly resides in vesicles scattered over the whole cytoplasm (Malide et al., 1997). However, with increasing time in culture (>24 h), the bulk of the fluorescence starts to relocate to the perinuclear region (unpublished data). We used cells that were incubated overnight, but within 24 h after isolation; we selected those cells with substantial amounts of fluorescent vesicles scattered throughout the cytoplasm.
Analysis of TIRFM images showed a similar punctate pattern of GLUT4 distribution (Fig. 1 D), demonstrating that the spots mostly corresponded to single GLUT4 vesicles (see Material and methods). In the basal state, we detected 15.7 ± 3.2 vesicles per a 100-μm2 area of the PM (65 cells, SD) in the TIRF zone. The proximity of many GLUT4 vesicles to the PM is consistent with earlier papers on GLUT4 distribution in adipose cells in primary culture (Lee et al., 1999; Malide et al., 2000).
GLUT4 vesicles near the PM are highly mobile
Time-lapse videos made using TIRFM demonstrated a high basal trafficking of GLUT4 vesicles in the vicinity of the PM (Video 1, available). To quantify this traffic, we used a projection algorithm producing distinct traces (Fig. 2 A) for all moving vesicles in the TIRF zone. In basal cells, we detected on average 10.0 ± 1.4 traces/100 μm2/min (22 cells, SD) representing trajectories of moving vesicles. The vesicles often underwent long-range (>10 μm) lateral movements, presumably approaching the PM along the way (as deduced from periodic increases in vesicle fluorescence). Vesicles tended to stop for a period of time (varying from a fraction of a second to 100 s) at dedicated places, and sometimes two or more vesicles stopped at the same location. Ultimately, vesicles exited the TIRF zone laterally or in a direction perpendicular to the coverslip (for a detailed description, see the section Kinetic analysis of GLUT4 recycling in primary adipose cells in the online supplemental material).
Moving vesicles took predefined trajectories. Often, vesicles followed exactly the same trajectories one after another (Fig. 2 B and Video 2, available at l). In adipose cells transfected with fluorescent tubulin, we observed an extensive microtubular network (Fig. S1 B; Malide and Cushman, 1997). Long microtubular pathways (>30 μm) that cover the entire cytoplasm (Fig. S1 B) could very well account for the observed movements of GLUT4 vesicles. Indeed, the GLUT4 vesicles closely follow the paths marked by fluorescent tubulin (Video 3, available; Semiz et al., 2003). The distribution of vesicle velocity exhibits two distinct peaks at 0.6 and ∼2 μm/s (Fig. S1 C), both falling into the range of velocities characteristic for fast kinesin or dynein motors (Higuchi and Endow, 2002; Ma and Chisholm, 2002).
Insulin stimulates tethering of mobile GLUT4 vesicles to the PM
In stimulated cells we observed a drastic reduction in the traffic (Video 4, available). Only 1.4 ± 1.4 traces/100 m2/min (18 cells, SD) were detected 10 min after the insulin treatment, approximately eightfold less compared with the basal state (Fig. 2, C and D). Our analysis indicates that the reduction of the traffic is due to immobilization of GLUT4 vesicles at the PM. First, upon insulin stimulation, GLUT4 vesicles attach to the PM much tighter. Fig. 3 A shows two GLUT4 vesicles that attached to the PM shortly after appearance in the TIRF zone. After attachment, the vesicles do not “wiggle” near the point of attachment (Li et al., 2004) as much in the presence of insulin as they do in the basal state; the amplitude of wiggling has two peaks (42 ± 11 and 88 ± 26 nm, n = 15, SD) in the basal state, whereas in the insulin-stimulated state, only the first peak (49 ± 20 nm, n = 15, SD, statistically indistinguishable from the first “basal” peak) is present (Fig. 3, B and C). Thus, insulin stimulation almost completely eliminates the long-range “wiggling” of the attached vesicles. Moreover, these vesicles rarely move again and so the mobile vesicles all remain in the TIRF zone. Second, insulin diminished the distance traveled by the vesicles in the TIRF zone (from 15 ± 6 to 5 ± 2 μm, n = 70, SD; Fig. 3 D). Thus, insulin stimulates tighter and quicker attachment of mobile GLUT4 vesicles to the PM of adipose cells (see Fig. S4, available, and the section Kinetic analysis of GLUT4 recycling in primary adipose cells in the online supplemental material).
Insulin immobilizes GLUT4 vesicles into clusters on the PM
Although the initial distribution of GLUT4 vesicles around the cell surface is relatively uniform (Fig. 1, C and D), we observed that insulin stimulates formation of fluorescent domains alternating with dark regions (Fig. 4, A and C). To characterize the degree of nonuniformity of the spatial distribution of GLUT4 vesicles near the PM, we applied a pair-distance correlation analysis. Insulin-stimulated cells showed a distinct peak on the pair-distance correlation curve (Fig. 4 B), demonstrating that a significant percentage of vesicles are clustered. The width of the main peak gives an estimate of average cluster size (∼5 μm). Pair-distance correlations obtained with cells in the basal state exhibited no significant deviation from a randomly scattered point distribution (all points lie inside the 99% confidence interval), as tested by simulation. These observations are consistent with previous studies that insulin targets GLUT4 vesicles to specific places of the PM of 3T3-L1 adipocytes (Patki et al., 2001; Semiz et al., 2003). Moreover, insulin signaling in adipose cells has been proposed to be associated with specific domains of the PM (Saltiel and Pessin, 2003). If insulin promotes tethering of GLUT4 vesicles to the PM, then to complete the insulin response docked vesicles must fuse to release GLUT4 to the PM. We further confirmed that insulin stimulates tethering and fusion of GLUT4 vesicles to the PM within the characteristic time of the appearance of GLUT4 in the PM reported previously (Dawson et al., 2001; Tengholm and Meyer, 2002).
GLUT4 vesicles fuse with the PM shortly after docking in clusters
Formation of clusters was already seen after 2 min of insulin treatment (Fig. 4 C, top). Shortly after, we saw an increase of GLUT4 fluorescence between the clusters, where no vesicles were detected (Fig. 4 C, bottom). The time lag between the increase of fluorescence integrated over the cluster region (Fig. 4 D, red curve) and over the dark region (Fig. 4 D, black curve) was several minutes. That fluorescence redistributes from clusters to the dark regions of the PM strongly indicates that vesicles in the clusters fuse to the PM. We confirmed the appearance of GLUT4 on the PM by antibody labeling; moreover, wortmannin, known to interfere with insulin signaling, ablated the translocation of GLUT4 to the PM (Fig. S2, A and B, available; Dawson et al., 2001). Next, we analyzed the dynamics of GLUT4 release from individual vesicles tethered to the PM.
In insulin-stimulated cells, GLUT4 vesicles lost their fluorescence shortly after attachment; the vesicles approached the PM and stopped, and then the vesicle fluorescence dimmed. Fig. 5 A demonstrates the fluorescence spreading from the same GLUT4 vesicle indicated by the green arrow in Fig. 3 A to the PM; the corresponding radially symmetric Gaussian fits of the fluorescence intensity are placed below each frame (Video 5, available). The fits demonstrate that while the peak fluorescence decayed (Fig. 5 B, red curve), the profile of vesicle fluorescence widened (Fig. 5 B, black curve) but the total amount of fluorescence in the circular region (4 μm in diameter) surrounding the vesicle remained constant (Fig. 5 B, green curve), verifying that the vesicle underwent fusion rather than moved away from the PM (Loerke et al., 2002).
The coefficient of lateral diffusion of GLUT4 from such immediate fusion events can be estimated as 2 × 10−9 cm2/s from the kinetics of the fluorescence profile widening (Fig. 5 B, black curve; Schmoranzer et al., 2000). Using FRAP, we obtained a similar estimate of the diffusion coefficient of GLUT4 in the PM, 1.4 × 10−9 cm2/s (Fig. S3 A, available), thus pointing out that GLUT4 vesicles can quickly release the transporter to PM. However, the quick unrestricted release of GLUT4 corresponding to detectable widening of the fluorescence profile, as in Fig. 5 A, was detected only rarely. We estimated the characteristic times of GLUT4 release from individual fusing vesicles by an exponential fit of the decay of vesicle fluorescence (as in Fig. 5 B, red curve). Fig. 5 C demonstrates that most of the vesicles release GLUT4 much slower than 1 s, the estimated characteristic time of unrestricted release calculated from the above diffusion constant (Fig. 5 B). Yet, Fig. 5 C shows that the decay of the vesicle fluorescence was indeed due to insulin stimulation: insulin-stimulated vesicles lose fluorescence much faster (in 20 ± 12 s) than vesicles in the basal state (in 77 ± 13 s). A decay time slower than 70 s was statistically indistinguishable from the characteristic time of vesicle bleaching under constant laser illumination (80 ± 10 s). In control experiments performed on cells in the basal state using intermittent illumination, no fluorescence decay was detected for most of the vesicles, confirming that bleaching was the reason for the slow decay (τ > 70 s). Because fluorescence is not lost but redistributes in times substantially faster than 70 s, these data further confirm that insulin-stimulated tethering of GLUT4 vesicles in clusters is followed by fusion of the vesicles with the PM. Fig. 5 C also shows that the fluorescence of a small fraction of vesicles in nonstimulated cells (Fig. 5 C, blue bars) decayed quickly, indicating fusion events in the basal state. These rare exocytic events can account for slow basal recycling of GLUT4 (Holman et al., 1994; Karylowski et al., 2004).
Insulin changes both the kinetic behavior and the spatial distribution of GLUT4 vesicles in the TIRF zone. Within 15 min of insulin treatment, the vesicles moved into clusters (Fig. 4 A) and became closer to the PM (Fig. 5 D). Thus, insulin initially mobilized vesicles in the TIRF zone, relocating them to fusion sites on the PM. New GLUT4 vesicles that are being constantly formed to replenish the vesicle population provide GLUT4 to the PM (Bryant et al., 2002). Fig. 2 D demonstrates that vesicle movements are infrequently detected 10 min after insulin stimulation, when the new nontrafficking distribution of GLUT4 is established (Holman et al., 1994). We suggest that these short movements of GLUT4 vesicles in the steady-state after insulin treatment represent the same pathways that vesicles take in the basal state. However, the average vesicle displacement is much shorter because the probability of irreversible attachment to the PM, mostly leading to fusion (Fig. S4 and the section Kinetic analysis of GLUT4 recycling in primary adipose cells in the online supplemental material), is dramatically augmented by insulin.
Our kinetic model of GLUT4 recycling (see the online supplemental material) indicates that insulin augments both the probability of tethering at specific fusion sites and the subsequent priming of the vesicles leading to their fusion to the PM. In insulin-stimulated cells, the GLUT4 vesicles are expected to fuse at the first encountered fusion site as typical cargo vesicles (Schmoranzer and Simon, 2003). The probability of fusion is much less in nonstimulated cells; thus, mobile GLUT4 vesicles bypass many fusion sites, remaining on the microtubule network for a long time. Fig. 6 summarizes our model for the insulin stimulation of GLUT4 exocytosis. Upon formation, the GLUT4 vesicles move on microtubules until they fuse, like vesicles delivering constitutive post-Golgi cargo (Schmoranzer and Simon, 2003). The GLUT4 vesicles fuse to the PM at special fusion sites, with the probability of tethering and fusion being regulated by insulin.
It has long been known that in primary adipose cells GLUT4 is predominantly sequestered in specialized vesicles uniformly scattered over the cytoplasm (Lee et al., 1999; Malide et al., 2000). Such a uniform distribution was shown to be characteristic of freshly isolated cells; with increasing time in culture, GLUT4 vesicles tend to relocate to the perinuclear region (Malide et al., 1997). However, it has remained unclear whether these scattered GLUT4 vesicles, especially those situated relatively close to the PM (Lee et al., 1999; Malide et al., 2000), represent a static pool, awaiting insulin stimulation to trigger fusion of the vesicles to the PM, or a dynamic pool constantly exchanging GLUT4. We demonstrate here that in freshly isolated adipose cells, GLUT4 vesicles located near the PM are mobile. The vesicles move along microtubules, thus scanning the extensive microtubule network that densely underlies the entire PM (Malide and Cushman, 1997). Accordingly, the mean velocities of the mobile vesicles correspond well to those reported for microtubule-based transport. The first characteristic velocity of ∼0.6 μm/s corresponds to that reported for GLUT4 vesicles driven by conventional kinesin on microtubules in 3T3-L1 cells (Semiz et al., 2003). The second one, ∼2 μm/s, which represents rapid, long-range movements of GLUT4 vesicles, falls into the range of velocities characteristic for fast kinesin or dynein motors (Higuchi and Endow, 2002; Ma and Chisholm, 2002). The combination of fast and slow movements, alternating with periods of idling, results in a seemingly random distribution of GLUT4 vesicles over the microtubule network (Fig. 1, C and D). Hence, white adipose cells appear to use microtubules extensively to support trafficking of GLUT4 vesicles, which is consistent with the general idea of the microtubular network supporting transport of post-Golgi vesicles to the PM (Schmoranzer and Simon, 2003).
Importantly, our data indicate that the network architecture plays an important role in the regulation of GLUT4 translocation to the PM (Fletcher et al., 2000; Semiz et al., 2003). In the basal state, GLUT4 vesicles often move close to the PM where the microtubules do (Schmoranzer and Simon, 2003). Although the vesicles stop at such sites, they are weakly tethered (Fig. 3) and ultimately escape, only rarely fusing with the PM. However, upon insulin stimulation, the tethering becomes much tighter (Fig. 3) and the vesicles accumulate in clusters (Fig. 4) in very close proximity to the PM (Fig. 5 D). The clustering likely takes place at the point of microtubule attachment to the PM as GLUT4 vesicles appear to move on the microtubules until they tether and fuse (Schmoranzer and Simon, 2003). Furthermore, clustering may also correspond to the insulin signal initiation sites proposed by Saltiel and Pessin (2003), to which various effector molecules, including molecular tethers, could be recruited. Transformations of the actin network in the periphery of adipose cells could also play a role in localization of vesicle fusion at specific sites (Malide and Cushman, 1997; Saltiel and Pessin, 2003).
In insulin-stimulated adipose cells, GLUT4 vesicles reaching the PM are quickly recruited and the long-range vesicle movements are thus halted. Contrary to this finding, in 3T3-L1 cells, insulin stimulates long-range movements of GLUT4 vesicles toward the PM, rather than inhibiting trafficking (Patki et al., 2001; Semiz et al., 2003). The discrepancy could be due to different basal distributions of GLUT4. If the distribution is shifted to the perinuclear region, as is characteristic of 3T3-L1 cells (Patki et al., 2001), then more long-range trafficking would be expected to bring these vesicles to the PM. If a substantial amount of GLUT4 vesicles is already near the PM (Slot et al., 1991; Malide et al., 2000), as in adipose cells in primary culture (Fig. 1 D), then the GLUT4 vesicles need only move slightly toward the PM to augment the rate of GLUT4 exocytosis.
Our data indicate that adipose cells might also have a means of regulating GLUT4 release at the level of the individual GLUT4 vesicle: GLUT4 is not freely released from the fusing vesicles. The release may be constrained by the trapping of GLUT4 by TUG (tether, containing a UBX domain, for GLUT4; Bogan et al., 2003) or other factors involved in packing vesicles with the transporter might interfere with GLUT4 release. Alternatively, GLUT4 may be constrained by the fusion pore, as seen for lipid dyes in small fusion pores induced by influenza HA (Zimmerberg et al., 1994; Frolov et al., 2000). We cannot determine the delay in fusion after tethering, but this gradual release of GLUT4 obviously must come after a fusion pore forms between the vesicle and the PM. If the fusion pore were sufficiently large, extracellular glucose could access the GLUT4 constrained in the fusing vesicles. These clustered transporters would contribute to the cell's glucose uptake, which is consistent with the fact that insulin stimulates glucose uptake in adipose cells maximally in 5–10 min (Vega and Kono, 1979; Satoh et al., 1993). Importantly, GLUT4 trapped in a slowly fusing vesicle cannot be retrieved from the PM via normal endocytosis. Thus, the constrained fusion may explain the overshoot of the PM GLUT4 seen after insulin application (Bogan et al., 2001; Zeigerer et al., 2002). The constrained release of cargo proteins from fused vesicles, also observed for VAMP, CDC63, and Dopamine-β-hydroxylase, could be the general mechanism of regulated redistribution of granule membrane proteins to the PM (Allersma et al., 2004).
In summary, we propose that GLUT4 vesicles follow common pathways of constitutive exocytosis, exploiting microtubule tracks on their way to the PM and revealing constrained release of membrane cargo. However, the probability of tethering and fusion of these vesicles to the PM is specifically sensitive to insulin. Insulin is known to stimulate constitutive exocytosis in general (Bryant et al., 2002), though to a lesser extent than GLUT4 exocytosis. Molecular mechanisms providing the specificity of insulin action on the GLUT4 vesicles remain to be established.
Materials And Methods
White adipose cells were isolated from the epididymal fat pads of 180–250-g male rats (CD strain; Charles River Laboratories), and then transfected with 8 mg/ml HA-GLUT4-GFP plasmid, and in some experiments supplemented with 0.25 mg/ml tubulin-GFP plasmid as described previously (Malide et al., 1997; Dawson et al., 2001). Tubulin-GFP plasmid was provided by A. Tsvetkov (University of Illinois College of Medicine, Chicago, IL). Coverslips coated with poly-l-lysine were placed on the surface of the medium containing the cells shortly after the transfection. A portion of floating adipose cells spontaneously adhered to the coverslip after overnight incubation at 37°C with 5% CO2 in DME containing 3.5% BSA (fraction V; Intergen) as described in Malide et al. (1997). Cells were stimulated with 67 nM (10 mU/ml) insulin added to the extracellular medium.
Confocal microscopy and TIRFM
A coverslip with adhered cells was placed into Delta-TPG dish (Bioptechs) in Krebs-Ringer bicarbonate Hepes buffer, pH 7.4, containing 200 nM adenosine and 5% BSA (Al-Hasani et al., 1998), supplemented with 2% Ficoll to avoid cell damaging. The cells were slightly squashed between the coverslip and the bottom of the dish (Fig. 1, A and B). All experiments were performed at 37°C maintained by Delta T4 temperature controller (Bioptechs). Confocal microscopy was performed on a microscope (model LSM510; Carl Zeiss MicroImaging, Inc.) using a 100× 1.4 NA oil-immersion objective. Stacks of confocal images were collected from each cell and three-dimensional reconstructions were performed by Imaris software (Bitplane). The Prism-less TIRFM setup (Schmoranzer et al., 2000) was based on a microscope (model IX-70; Olympus) equipped with a 60× 1.45 NA objective (Olympus), argon laser (488 nm; Spectra Physics), and intensified CCD camera (VE1000SIT; MTI). Images were digitized by acquisition board (Flashbus; Integral Technologies) controlled by MetaMorph (Universal Imaging Corp.). Time-lapse images were acquired at high rates (10–30 frames/s) to monitor vesicle trafficking and fusion and at slow rates (<0.5 frames/s) to monitor changes of total fluorescence from an area of the PM. For the slow acquisition, a mechanical shutter (UniBlitz D122; Vincent Associates) synchronized with the MetaMorph software was used to block laser illumination between frames. Penetration depth d of the evanescent field was measured to be 180 ± 20 nm by a calibration procedure with 40-nm fluorescent beads attached to the piezo-driven micropipette.
Stacks of time-lapse images were processed using projection algorithms implemented in ImageJ 1.32 (National Institutes of Health). Mean projection image of a stack was subtracted from its max projection image to obtain the traces of all moving vesicles. The number of traces (length >2 μm) within a square region of interest (ROI) of 100 μm2 was counted visually. The Gaussian fit of radial pixel intensity profile showed that for the moving vesicles 90% of fluorescence was confined within a circular vesicle ROI of 3 pixels. Thus, bright spots were counted as vesicles if (a) the pixel intensity profile had a local maximum and (b) the average intensity in the vesicle ROI positioned on the maximum was 20% higher than the background fluorescence. The thickness of the layer where we could detect single vesicles was ∼400 nm. To measure intensity of the vesicle fluorescence, images were first background subtracted using a rolling-ball algorithm (ImageJ) with a 10-pixel radius; and then the fluorescence intensity was calculated as a sum of all pixels values in the vesicle ROI.
The degree of vesicle clustering after application of insulin was estimated by pair-distance correlation analysis. Coordinates of the peak intensity of the vesicle ROI were taken as vesicle coordinates and transferred in ASCII format to Maple (Waterloo Maple Inc.). Custom written macros counted the number of vesicles within a certain distance r (running from 0.5 to 30 μm) from each vesicle and built the cumulative frequency distribution of all vesicle-vesicle distances:
where n is the total number of vesicles, A is the region area, and Ir is a counter variable that is set to one if the distance between vesicles i and j dij ≤ r, otherwise Ir = 0. For the analysis, circular regions of 30-μm-radius were selected from TIRF images of the cells. Because the cumulative functions were built for the same boundary conditions and were normalized over the number of vesicles, it allowed us to average it between different cells. To compare the vesicle distribution with a random spatial distribution, the random point patterns were generated by computer simulation. 30 simulations were performed using a region with the same boundary conditions and with the number of random points equal to the average amount of vesicles in the ROI defined on cells. The SD and confidence intervals were approximated from simulations of complete random spatial distributions.
Online supplemental material
Online supplemental material describes in detail the imaging of the microtubular network in adipose cells, immunofluorescence of GLUT4 exposure on cell surface, and FRAP analysis of GLUT4 mobility in the PM. Videos 1 and 2 show traffic of GLUT4 vesicles in adipose cells in the basal state, and Video 3 demonstrates the movement of the vesicles on fluorescent microtubule tracks. Video 4 shows the traffic reduced by insulin stimulation. Video 5 shows a docking-fusion event and release of GLUT4 to the PM. Fig. S1 shows the TIRFM image of the PM in the TIRF zone, the microtubular network, and a histogram of vesicle velocities. Fig. S2 shows colocalization of immunofluorescence to surface GLUT4 in the PM after application of insulin and in control experiments with wortmannin. Fig. S3 shows the analysis of the GLUT4 diffusion, whereas Fig. S4 and Table S1 describe the kinetic model of GLUT4 recycling in primary adipose cells.
The authors would like to thank Mary Jane Zarnowski for assistance in isolation and transfection of rat adipose cells.
V.A. Lizunov and H. Matsumoto contributed equally to this paper.
H. Matsumoto's present address is Dept. of Molecular Biology, Saitama Medical School, Saitama, 350-0495, Japan.
Abbreviations used in this paper: GLUT4, glucose transporter 4; PM, plasma membrane; ROI, region of interest; TIRF, total internal reflection fluorescence; TIRFM, TIRF microscopy.