Large GLUT4 Vesicles Are Stationary While Locally and Reversibly Depleted During Transient Insulin Stimulation of Skeletal Muscle of Living Mice: Imaging Analysis of GLUT4–Enhanced Green Fluorescent Protein Vesicle Dynamics

The construction of GLUT4-EGFP (16) has been described previously. EGFP was attached to the intracellular COOH-terminal part of GLUT4. In imaging experiments, the quadriceps muscles of 7- to 8-week-old male C57BL/6 mice (Charles River, Sulzfeld, Germany) were transfected with gene gun bombardment with 1 μg GLUT4-EGFP/0.5 mg gold as previously described (16). For GLUT4-EGFP expression, confirmation in vivo electroporation was carried out: either GLUT4-EGFP or EGFP plasmid was directly injected into mouse tibialis anterior muscles using a modification (17) of the protocol originally described by Aihara and Miyazaki (18). Mice were anesthetized with an intraperitoneal injection of pentobarbital sodium (90 mg/kg), and 25 μl 4 μg/μl DNA (either pEGFP or pGLUT4-EGFP) or 25 μl saline was injected longitudinally into the muscle with a 29-gauge needle. Stainless steel electrode needles fixed 4 mm apart were then inserted into the muscle belly, and square wave electrical pulses (200 V/cm) were applied eight times at a rate of 1 pulse/s (duration 20 ms) using a Grass S88 pulse generator (Grass Instruments, Quincy, MA).

Western blot tissue processing and immunoblotting: Frozen muscles were homogenized as described previously (20). Protein concentrations were determined by Bradford assay. Skeletal muscle proteins (25 μg) were separated by SDS-PAGE (21) for Western blot analysis (22). Specific primary antibodies against EGFP (8362-1; Clonetech) and GLUT4 (sc-7938; Santa Cruz Biotechnology) were used. Antibody-bound proteins were visualized using enhanced chemiluminescence (Amersham Biosciences). Protein bands were scanned by ImageScanner (Amersham Biosciences). Western blot showed a band corresponding to the molecular weight of GLUT4-EGFP (∼73 kDa) and no band corresponding to EGFP (∼27 kDa) alone, confirming that EGFP was attached to GLUT4 (Fig. 1A). Tibialis anterior muscles electroporated with pEGFP showed a band corresponding to the weight of EGFP (∼27 KDa). The fact that GLUT4-EGFP was intact also in in vivo gene gun–transfected muscle appeared from confocal imaging of GLUT4-immunostained GLUT4-EGFP–transfected flexor digitorum brevis muscle fibers (treated as previously described in 16), in which EGFP fluorescence always was located together with GLUT4 (Fig. 1B).

Four days after transfection, the overnight fasted mice were anesthetized subcutaneously with Hypnorm-Dormicum (0.8–1 ml/100 g body wt), and the skin covering the quadriceps muscle was opened to expose the quadriceps. The mice were put to rest on the side, and effects of respiration on the z-position were dampened by mounting and stabilizing with Colténe President cement (Colténe/Whaledent, Mahwah, NJ) embedding. Blood flow did not influence the z-position. The exposed muscle was covered by a coverslip, and the mice were mounted on the Leica TCS SP2 confocal microscope stage. A coverslip holder maintaining the coverslip in a position perpendicular to the objective was used. Confocal image series consisted of a single confocal image for each time point taken 15 μm from the surface of the muscle fiber. Confocal images were taken every 15 s in all experiments except long-term re-internalization studies. In long-term re-internalization studies, images were collected every 5 min from 30 to 100 min after insulin administration and every 10 min from 100 to 260 min after insulin. The 512 × 512 pixel images were collected with a ×63, 1.2 numerical aperture (NA) Leica HCX PL APO water immersion objective using 400-Hz scan speed and 2× line averaging. The theoretical depth of images was 0.413 μm. However, because of imaging in the in vivo environment, changes in refractive index and dispersion probably increase the depth of images to 0.5–0.6 μm. The 488-nm laser line with 40% of minimal laser power was used for excitation of EGFP, whereas green emission light was collected between 500 and 530 nm using the standard fluorescein isothiocyanate settings of the Leica software. Selective photobleaching of a region of interest (ROI) was performed by scanning the ROI 72 times in 2 min at 100% of maximal 488-nm laser power. Gain and offset settings in the Leica confocal software (V. 2.61) were adjusted so that images of nontransfected fibers appeared essentially black. Imaging of insulin-mediated GLUT4-EGFP translocation for 30 min was preceded by imaging of the same site of the same fiber during 30 min after saline injection. However, in photobleaching experiments because of the destructive nature of the technique, different sites in the same fiber were imaged after saline and insulin injection, respectively. One hour elapsed between the two injections. In experiments in which insulin-mediated GLUT4-EGFP translocation and subsequent re-internalization were studied, control experiments with saline injection were performed in different mice, because of potential relocation caused by animal movements during long experiments. For saline injection, at t = 0 min, a 20-μl saline bolus (0.9% saline and 0.1% BSA) was injected in a tail vein. For insulin stimulation, at t = 0 min, a 16.8-μl insulin bolus (656 mU in 20% [wt/vol] glucose and 0.1% BSA; Actrapid; Novo Nordisk) was injected in a tail vein, and this was followed by a glucose infusion, which in pilot studies had been shown to maintain euglycemia. The fact that translocation of GLUT4-EGFP was completely intact on a second stimulation with insulin after 5 h of recovery ensured that the procedures did not harm the preparation (data not shown).

The TIF images obtained with the Leica confocal software were imported into Metamorf Software (V. 6.1; Universal Imaging). Image stacks were created, and pixel values were corrected for noncellular fluorescence. Brightness and contrast were adjusted. For quantifications and visualization of GLUT4-EGFP images, a Fourier noise filter was applied. The amount of GLUT4-EGFP at membrane surfaces was quantified by drawing thin ROIs including sarcolemma or regions of t-tubules and excluding vesicular structures. For sarcolemma, measurements from both sarcolemma edges (excluding perinuclear areas) of the confocal picture were averaged. For t-tubules, measurements from four areas were averaged. Because of the resolution limit of light microscopy the cell surface membrane GLUT4-EGFP may be overestimated by quantification in the basal state and underestimated on translocation. Differences in fluorescence between images at identical time points reflect variation in level of GLUT4-EGFP expression within and between individual fibers. In image quantifications, values were expressed relative to values at t = 0 (f/f₀) in the same fiber to compensate for such variation. Thus, arbitrary units shown are the actual gray value of ROI fluorescence divided by ROI fluorescence value at t = 0 (f/f₀). Uncompressed QuickTime movies were produced from the image stacks and subsequently compressed with Sorenson 3 codec in QuickTime Pro 6.

Gene gun–transfected flexor digitorum brevis muscles were fixed, and fibers were isolated and immunostained as previously described (16). Confocal images were collected with the Leica TCS SP2 system. EGFP was excitated using the 488-nm laser line, and emission was collected in the fluorescein isothiocyanate channel. Alexa 594 was excitated using the 543-nm laser line, and emission was collected in the tetramethylrhodamine B isothiocyanate channel. Images were collected sequentially from the two channels through a ×63 water immersion objective (NA 1.2).

Tail vein blood (30 μl) was added to 1 μl 4°C Trasylol-EDTA (60 mg EDTA in 20,000 KIU/ml Trasylol) and centrifuged for 10 s in an Eppendorf 5415C centrifuge (Eppendorf, Hamburg, Germany). The plasma fraction was immediately transferred to −20°C. Plasma insulin was measured using Linco insulin Elisa kit (Linco Research, St. Charles, MO).

To study the dynamics of GLUT4 vesicle trafficking, skeletal muscle fibers of living mice were transfected with GLUT4-EGFP cDNA by gene gun bombardment, and 96 h later, the transfected fibers were imaged in their natural surroundings. We have previously shown that under these circumstances, GLUT4-EGFP is ∼2–3 times endogenous GLUT4 (16). Furthermore, Western blots of in vivo–transfected muscles confirmed that GLUT4-EGFP was expressed as the intact fusion protein with no free EGFP being present (Fig. 1A). Correspondingly, colocalization analysis of gene gun–transfected fibers showed that the green EGFP signal was always associated with GLUT4 protein (Fig. 1B).

In the basal state, GLUT4-EGFP was distributed as reported earlier, i.e., in larger and smaller vesicular structures and in the perinuclear area, and with no continuous presence at the sarcolemma and no regular continuous striated pattern (3,16,22) (Fig. 2B). No fluorescent inclusion bodies were seen, the distribution of GLUT4-EGFP being similar to that shown for endogenous GLUT4, as was also found previously (16). Ten minutes after insulin administration, a clear GLUT4-EGFP staining of the sarcolemma was seen, indicating translocation from the intracellular compartments (Fig. 2B). Maximal translocation to sarcolemma and t-tubuli were gradually reached within the next 10–20 min. Accompanying the translocation, a gradual depletion of intracellular GLUT4-EGFP–containing structures took place. The insulin-mediated reduction in larger perinulear and smaller vesicular GLUT4 structures seen in Fig. 2B–H is in line with previous immunocytochemical studies (3). However, in these studies, the methodology did not allow an account of the dynamics of the reduction (3). Fig. 2D shows how the GLUT4-EGFP–containing depots in the perinuclear area are gradually depleted in response to insulin stimulation. Similarly, the vesicle structures (>1 μm in size) at the sarcolemma did not move to or along the sarcolemma (Fig. 2F) and neither did the vesicular structures at the inner t-tubules move to or along the t-tubules (Fig. 2H). So, in the present study of muscle fibers, almost all GLUT4 vesicular storage structures remained stationary while being gradually depleted (Fig. 2B–H). This is in contrast to findings in 3T3-L1 adipocytes and fibroblasts, in which vivid movements of smaller and larger GLUT4 vesicular structures are striking (8)

To further analyze how the stationary vesicles are depleted, time-lapse image series (ΔT = 15 s) were analyzed frame by frame. In Fig. 3A, an enlarged part of the perinuclear area is shown. To clearly visualize the GLUT4-EGFP release from this compartment, the area outside the perinuclear region was scanned with high-intensity laser light, thus leaving all GLUT4-EGFP in the vicinity of the perinuclear area bleached and only the perinuclear GLUT4-EGFP fluorescent. Immediately after bleaching, saline or insulin was injected at t = 0, and images are shown from t = 0 to 16 min. In the studied mice, a vesicle outflow on the left side of the perinuclear area was present at t = 8–10 min after saline injection (Fig. 3A). However, this only resulted in weak and short-lived staining of the sarcolemma, which was not present any more at t = 16 min. In contrast, after insulin injection (Fig. 3B), at approximately t = 6 min, GLUT4-EGFP from the perinuclear area was sent out along the sarcolemma on both sides in a microvesicular fog, giving rise to strong and prolonged sarcolemma staining accumulating to a maximum at t = 16 min. The random, unilateral, and transient vesicle outflows after saline injection in individual mice are of course not reflected in the quantitation given in Fig. 3C based on data from four mice. Little exchange between the perinuclear GLUT4-EGFP compartment and the interior of the muscle fiber was seen, indicating that the perinuclear GLUT4-EGFP depot primarily interacts with the sarcolemma membrane surface. Quantification of ROIs confirmed that restaining of sarcolemma was fast and above prebleached levels only on insulin stimulation and not after saline injection (Fig. 3C).

In a similar fashion, the GLUT4-EGFP storage vesicles located more deeply within the muscle fiber were gradually depleted and spread out along the t-tubules in a microvesicular fog as is shown in Fig. 3D (horizontal arrowheads and vertical arrowhead, respectively). Note the fluorescent arm to the right sending a microvesicular fog of GLUT4-EGFP out along the t-tubules. GLUT4-EGFP storage vesicles (just <1 μm in size) (Fig. 3E) located at the inner t-tubules are followed from t = 0 to 6 min with a time resolution of 15 s. The small vesicles remained stationary and from approximately t = 4 min started to diminish in size being almost totally depleted at t = 6 min and 45 s after insulin (Fig. 3E). Only a few vesicles were moving, and the direction then was always along the length of the muscle fiber (data not shown). Imaging with higher time resolutions (ΔT = 1–1.6 s) did not reveal vesicle movements differing from those described above (data not shown).

To analyze the recycling route of GLUT4-EGFP during maximal translocation, repeated photobleaching was used. By bleaching a ROI covering a part of the sarcolemma and t-tubules, but not the perinuclear area (Fig. 4A), it should be possible to follow the GLUT4-EGFP exchange between the perinuclear area and the t-tubules during maximal GLUT4-EGFP translocation. Fig. 4A and B shows a part of a maximally stimulated muscle fiber bleached 41 and 78 min after insulin administration. The bleached region is dark compared with the strongly fluorescent perinuclear GLUT4-EGFP compartment and the surrounding GLUT4 containing t-tubule area. Within a few minutes, GLUT4-EGFP staining is seen “running” from the perinuclear depot and rapidly staining the sarcolemma (Fig. 4A). In contrast, GLUT4-EGFP striations emerge only slowly and weakly in the bleached area, not reaching the prebleach level. This indicates that the perinuclear GLUT4-EGFP depot is highly dynamically recycling GLUT4-EGFP and exchanging GLUT4-EGFP primarily with the sarcolemma surface. This impression was also confirmed by image quantification (Fig. 4C and D). Furthermore, the weak and slow restaining of the t-tubules indicates that here, GLUT4-EGFP is either not recycling or locally recycling at the t-tubules. This is further supported by the finding that the fluorescence intensity just outside (2 μm) the bleached area does not change (Fig. 4C and D). If there were any GLUT4-EGFP recycling exchange with adjacent or distant t-tubule areas, fluorescence would occur in the bleached area, and the fluorescence outside the bleached ROI should gradually be reduced because of dilution with bleached nonfluorescent GLUT4-EGFP. From quantifications (Fig. 4C and D), it is clear that GLUT4-EGFP fluorescence further away (20 μm) also did not change. Thus GLUT4-EGFP within the t-tubules is recycling locally or not at all during maximal GLUT4 translocation.

We have previously described that on insulin stimulation, recruitment and translocation are 10 min delayed in t-tubules compared with sarcolemma (15). In addition, as described above, recycling dynamics differ between the two surfaces during maximal stimulation. Thus, it was of interest to analyze the re-internalization of GLUT4-EGFP from sarcolemma and t-tubules. So, we followed the re-internalization of GLUT4-EGFP by confocal in vivo analysis after insulin and saline bolus injection. After t = 80 min, the maximal staining was maintained until t = 140 min after insulin. From 140 min onward, the distinct coloring of sarcolemma and t-tubules gradually disappeared (Fig. 5A). This reflected waning insulin action as plasma insulin levels decreased below maximally stimulating concentrations (Fig. 6D). In accordance with cessation of significant insulin action, blood glucose concentration remained constant, when glucose infusion was stopped at 195 min (Fig. 6E). Direct inspection (Fig. 5) and quantification analysis (Fig. 6B and C) showed that GLUT4-EGFP re-internalization was delayed in t-tubules compared with sarcolemma. Detailed time-lapse analysis showed a gradual refilling with GLUT4-EGFP of the various depleted vesicle compartments (Fig. 7). Fig. 7A shows refilling of vesicle structures close to and in the sarcolemma. Fig. 7B and C shows the gradual refilling of the GLUT4-EGFP storage compartments in the perinuclear area and in the inner t-tubule area, respectively. Apart from a slight general photobleaching (∼10%), no changes were observed in transfected muscle fibers in saline-injected mice studied for 4.5 h (Fig. 5B).

The present study provides new information regarding the GLUT4 translocation process by directly monitoring the dynamic trafficking of GLUT4 depots during early translocation, steady-state recycling, and re-internalization of GLUT4 in skeletal muscle in situ in a living animal on insulin injection. We have previously studied the early translocation phase (15), but the present study adds considerably to the former by monitoring at higher magnification and higher time resolution. A major new finding is that the majority of insulin-mediated GLUT4 translocation does not involve any high degree of movement of GLUT4 storage vesicles (>1 μm) over longer or even shorter distances, e.g., from the cell interior to the sarcolemma. Surprisingly, the storage vesicles, whether located in the perinuclear region, at the sarcolemma, or at the t-tubules, remain stationary and “melt” away as they are depleted locally without any significant movement away from their original position (Figs. 2 and 3). Another major finding is that in the perinuclear zone, photobleaching experiments showed that during initial translocation and steady-state maximal translocation, microvesicles (<1 μm) are released from the storage vesicles and sent out along the sarcolemma (Figs. 3A and 4A and B). Similarly, during initial translocation, storage vesicles at the t-tubules send out GLUT4 microvesicles (Fig. 3B). The local GLUT4 storage vesicle depletion without gross movement fits well with previous findings that also insulin signaling in skeletal muscle is elicited locally, whether at the sarcolemma or within the t-tubules (15).

In agreement with the present findings, in GLUT4-EGFP studies in 3T3-L1 fibroblasts and adipocytes, the majority of the vesicles remained stationary in response to insulin stimulation (8). However, it was found that an intracellular pool of GLUT4-EGFP vesicles moved briskly over longer distances (8). The finding of moving vesicles in these not fully differentiated cells may reflect their lack of development. Still, also in fully differentiated adipocytes, GLUT4-EGFP vesicle movement is seen (9). These vesicular movements observed by total internal reflection fluorescence microscopy (TIRFM) may be considered analogs to the microvesicular recruitment from perinuclear storage sites in our study. However, the vesicle movements seen in adipocytes were halted rather than enhanced by insulin as seen in the present study. This halting effect on GLUT4 vesicles due to an increased fusion frequency induced by insulin has also been observed by other TIRFM studies in 3T3-L1 adipocytes (23–25). Whereas in fully differentiated adipocytes, the lipid droplet restricts GLUT4 movement in the center of the cells, in muscle, GLUT4 vesicles located in the interior of the fiber at the t-tubules may be “locked” into a matrix of abundant quantities of contractile proteins, preventing them from moving. In studies in which microtubules were depolymerized to inhibit vesicular trafficking in skeletal muscle, insulin-mediated glucose uptake was intact (26). These previous studies together with our present findings indicate that no major vesicle movement takes place during translocation. On the other hand, the actin cytoskeleton might play a role, because actin depolymerization reduces GLUT4 translocation in 3T3-L1 adipocytes and L6 myotubes (27,28). However, other findings in 3T3-L1 adipocytes indicate no effect of actin depolymerization on the movement of GLUT4-EGFP vesicles (8). Thus, actin may primarily be involved in mediating contact between upstream signals (28) and/or between the plasma membrane and the GLUT4-containing vesicles.

Despite the fact that we in the basal state see some GLUT4 vesicles being sent out along the sarcolemma from the perinuclear GLUT4 depot (Fig. 3A), this does not lead to any maintained accumulation as seen during insulin stimulation (Fig. 3B). This difference may reflect insulin-mediated changes in GLUT4 re-internalization, because several TIRFM-based studies in 3T3-L1 adipocytes have indicated that one of the effects of insulin is to slow GLUT4 internalization from the plasma membrane (23,25).

We also, for the first time in skeletal muscle, carried out detailed studies of GLUT4-EGFP vesicle dynamics during steady-state recycling. The view that GLUT4 movement in the interior of the cell is restricted is supported by the photobleaching experiments (Fig. 4). We found that GLUT4 recycling within the t-tubule area either does not take place or is very local, because no exchange with bleached GLUT4-EGFP was observed outside the bleached area even in areas close (2 μm) to the bleached region. Apparently, microvesicles from the perinuclear depot travel much longer (10–20 μm) along the sarcolemma during early translocation and steady-state recycling (Figs. 3A and 4). It could be argued that the difference might reflect that increases in GLUT4-EGFP staining may be easier to detect in sarcolemma than in t-tubules, because the former in contrast to the latter is continuously present through the z-plane of the observed images. However, we find this possibility unlikely because restaining of the bleached t-tubule area, if at all occurring, was very slow, and, accordingly, compatible with possible drifting of the specimen during imaging. Apart from less hindrance by tissue architecture to microvesicle movement at the surface compared with the interior of the cell, it could be that microtubules are involved in microvesicle movement at the sarcolemma but not at the t-tubules, because microtubules are more abundant at the sarcolemma than at the t-tubules (29). If so, the fact that depolymerization of microtubules does not impair glucose transport in muscle (26) would be compatible with the view that t-tubules are the main site of glucose transport.

The re-internalization of GLUT4-EGFP during waning of insulin effect was also studied for the first time in skeletal muscle (Figs. 5–7). The impression that GLUT4 was re-internalized from sarcolemma and t-tubules did not reflect bleaching of GLUT4-EGFP in these membranes due to repeated imaging. This is so because the decrease in fluorescence at the two membrane areas differed and, furthermore, was several hundred percent whereas bleaching in control muscles from saline injected mice was only ∼10% during a comparable period of laser light exposure. It has previously been found that on insulin administration, GLUT4-EGFP translocation is delayed between sarcolemma and t-tubules, probably reflecting the time needed for insulin diffusion into the t-tubules (15). Similarly, we found that 150 min after insulin administration, GLUT4-EGFP is gradually re-internalized with a delay in t-tubules compared with sarcolemma (Fig. 6A and B). In parallel, the same GLUT4-EGFP storage compartments as found in the basal state are repleted (Fig. 7). This finding and the fact that the delays in onset and decay, respectively, of insulin action in t-tubules compared with sarcolemma are comparable indicate that the delay in GLUT4-EGFP re-internalization reflects a delay in insulin disappearance from t-tubules, and that the re-internalization machinery is similar in the two plasma membrane surfaces. Correspondingly, it has been found that during initial translocation the GLUT4 recruitment machinery at sarcolemma and t-tubules, respectively, are also similar (15).

The fact that GLUT4-EGFP was gradually re-internalized in muscle after 150 min of exposure to recombinant human insulin agrees with the time course of the plasma insulin concentration (Fig. 6C). Although at 150 min, the insulin concentration was 18,000 μU/ml, subsequent concentrations probably were submaximal. This is so in mouse muscle, because human insulin most likely has lower receptor affinity than mouse insulin. Thus, mice have two types of insulin, both having three amino acids different from human insulin (30). Furthermore, mouse insulin has been shown to cross-react poorly with human insulin in immunoassays (31). In addition, we have previously found that in concentrations up to 10,000 μU/ml human insulin does not elicit GLUT4-EGFP translocation in our experimental set-up (H.P.M.M.L., unpublished observations). Low receptor affinity of human insulin is also in accordance with the fact that in the present study, the glucose infusion could be stopped at t = 195 min without any subsequent decline in plasma glucose concentration, even though plasma insulin concentration at that time was around 7,000 μU/ml (Fig. 6D).

In conclusion, on insulin stimulation of muscle, GLUT4 vesicular storage structures (>1 μm) remain stationary and are depleted locally near the sarcolemma and deeply inside the muscle cells at the t-tubules. During maximal translocation, recycling of GLUT4 in microvesicles over longer distances (10–20 μm) only takes place at the sarcolemma and not at the t-tubules. So, in muscle, an efficient glucose supply to all cell compartments is secured by local, depletable GLUT4 depots at the sarcolemma and along the t-tubules and by a widespread t-tubule system, which has a surface area two to three times that of the sarcolemma (32), and allows rapid distribution of insulin and glucose from plasma (15).

J.B. has received National Institutes of Health Grant T32DK07260. L.J.G. has received National Institutes of Health Grant AR45670. This work has received support from the Faculty of Health Sciences, University of Copenhagen; the Danish Diabetes Association; the Weimann Foundation; the Novo Nordic Research Foundation; the Beckett Foundation; and the Danish National Research Foundation.

We thank L. Kall for bioanalytical assistance and S. Lohmann for engineering assistance.