Mobility of proteins associated with the plasma membrane by interaction with inositol lipids

The interaction of a range of protein domains with inositol lipids is now recognized as a central part of many aspects of cellular function (Cullen et al., 2001; McLaughlin et al., 2002). Some proteins are integral to the membrane and have binding domains for inositol lipids such as phosphatidylinositol-(4,5)-bisphosphate [PtdIns(4,5)P₂], which might regulate their activity (e.g. many ion channels) (Gamper et al., 2004; Runnels et al., 2002; Zhang et al., 1999). Other proteins are anchored to the membrane by their interaction with an inositol lipid, which can again be PtdIns(4,5)P₂ as is the case for a Ras-GAP, GAP1^IP4BP (Cozier et al., 2000). Moreover, the acute recruitment of cytosolic proteins to membranes caused by the generation of 3-phosphorylated inositol lipids is an important aspect of signal transduction. A classic example is the dual recruitment by PtdIns(3,4,5)P₃ of two interacting proteins, PDK-1 and PKB, the former of which activates the latter (Mora et al., 2004). Another group of proteins recruited to the plasma membrane by PtdIns(3,4,5)P₃ are those that interact with monomeric G proteins, such as the ARNO, Cytohesin, Grp-1 family (Cullen and Chardin, 2000) and GAP1^m (Cozier et al., 2000).

In some instances the interaction of the protein with the lipid regulates the activity of the protein, for example, the activation of PDK-1 by PtdIns(3,4,5)P₃ (Mora et al., 2004). However, in other examples, there is no such obvious activation, and it is generally accepted that the recruitment of the protein to the membrane in itself causes an effective activation by virtue of the increased concentration of the protein relative to its (membrane-bound) effectors because the recruited protein is now confined to two dimensions (for review, see McLaughlin et al., 2002). This indirect activation could take two forms. It might be that a local concentration of two proteins is engendered by local concentrations of their lipid partners, as might be expected from lipid rafts (Simons and Vaz, 2004) or other factors localizing PtdIns(4,5)P₂ (Huang et al., 2004; McLaughlin et al., 2002). Alternatively, the simple act of confining a freely diffusing protein to two dimensions instead of three might be sufficient to increase its concentration greatly relative to proteins confined to the same membrane (McLaughlin et al., 2002). This latter concept, although widely accepted, has not actually been tested and it is assumed that a protein anchored to a membrane by a lipid will have a diffusion speed similar to that of the soluble protein (and, indeed, similar to the lipid to which it is anchored) rather than that of an integral membrane protein, whose diffusion is intrinsically very much slower (Snapp et al., 2003).

Here, we have addressed some of these issues, using FRAP of proteins tagged with green fluorescent protein (GFP). Our original aim was to compare two similar proteins, GAP1^m and GAP1^IP4BP, which are reported to interact with the plasma membrane via PtdIns(3,4,5)P₃ (Lockyer et al., 1999) and PtdIns(4,5)P₂ (Cozier et al., 2000), respectively. We found marked differences between their mobilities when they were bound to the plasma membrane and so we extended the study to the same proteins in their soluble (cytosolic) form, and also to another PtdIns(4,5)P₂-binding protein, the PH domain of PI-PLCδ₁ (Cifuentes et al., 1994; Stauffer et al., 1998). We have also studied a transmembrane protein, ICAM, to complete the picture. Finally, because GAP1^IP4BP is constitutively bound to the plasma membrane and remains there when production of its putative regulator, Ins(1,3,4,5)P₄, is stimulated (Cozier et al., 2000), we investigated whether generation of Ins(1,3,4,5)P₄ or its lipid equivalent, PtdIns(3,4,5)P₃, would alter its mobility.

GFP-labelled GAP1^m and GAP1^IP4BP (Lockyer et al., 1997), live and dead Ins(1,4,5)P₃-3-kinase labelled with DsRedC1 and targeted to the cytosol by removing the N-terminal F-actin binding site (Nash et al., 2002), and the GFP-labelled pleckstrin homology (PH) domain of PI-PLCδ₁ [GFP-PH(PLC-δ)] (Stauffer et al., 1998) constructs were gifts from M. J. Schell (Department of Pharmacology, University of Cambridge, UK). GFP-ICAM was donated by M. Davies (Howard Hughes Medical Institute). Dulbecco's modified Eagle's medium (DMEM), foetal bovine serum (FBS), L-glutamine and penicillin/streptomycin were supplied by Gibco. FuGene6 transfection reagent was purchased from Roche Molecular Biochemicals. All other reagents were supplied by Sigma. Carbachol, where indicated, was used at 100 μM and atropine was used at 10 μM; insulin and wortmannin were both used at 100 nM. All stocks were prepared in imaging buffer (Bootman and Berridge, 1996) and drugs were applied as a bolus to the imaging chamber.

HEK-293 cells were passaged and maintained in DMEM (containing 10% FBS, 5 mM glutamine, 50 U ml^–1 penicillin and 50 U ml^–1 streptomycin) at 37°C, 5% CO₂. Two days before experimentation, the cells were seeded onto 25 mm glass coverslips (Menzel-Glaser, Germany) coated with poly-D-lysine. Cells were seeded at a density of 0.5×10⁶ ml^–1 and incubated overnight (37°C, 5% CO₂). One hour before transfection, the cells were washed with fresh medium. Vector containing the DNA encoding the GFP and DsRedC1 fusion constructs were transfected into cells using FuGene6 transfection reagent. For all transfections 2 μg DNA were used and, for experiments requiring co-transfections, 1 μg each construct was used. All experiments were conducted 16-24 hours after transfection.

Coated glass coverslips with transfected HEK-293 cells were mounted into an imaging chamber and immersed with 1 ml imaging buffer. The imaging buffer used is the same as described elsewhere (Bootman and Berridge, 1996) and contained the following: 121 mM NaCl, 5.4 mM KCl, 0.8 mM MgCl₂, 1.8 mM CaCl₂, 6 mM NaHCO₃, 5.5 mM glucose, 25 mM HEPES, pH 7.3. The imaging chamber was placed onto the heated stage of the Zeiss LSM 510 microscope. Transfected cells were selected using the fluorescence microscope. Before acquisition of data, a region of the cell was selected for photobleaching [the region of interest (ROI)]. In all experiments, a strip 3 μm wide spanning the width of the cell was selected (Fig. 2). The 488 nm line of the argon laser was used to excite GFP, with fluorescence emission collected through a 505-530 nm filter. Whole-cell and ROI images were collected every 196 milliseconds (128×128 pixels, optical slices were ∼3 μm thick). The sample was scanned at low laser intensity 50 times before the first bleach of 50 iterations, lasting ∼1.0-1.5 seconds. Photobleaching of the GFP required 100% of the argon laser power. Fluorescence-recovery data were then acquired using low-intensity laser scanning to minimize further photobleaching.

For experiments in which PtdIns(3,4,5)P₃ levels were manipulated, transfected HEK-293 cells were incubated either in serum-free medium for 2 hours before incubation with the PI3-kinase inhibitor wortmannin for 1 hour, or incubated in serum-containing medium for 2 hours before incubation with the PtdIns(3,4,5)P₃-generating agonist insulin for 1 hour. The imaging buffer was also supplemented with either wortmannin or insulin. These conditions either reduced or increased the intracellular levels of PtdIns(3,4,5)P₃, respectively.

For the triple-bleach protocol, the same ROI was subjected to three sequential photobleaches. Following the first bleach (B1), carbachol was added to the cells. After 5 minutes, the ROI was bleached for a second time (B2). After fluorescence recovery, atropine was added and, after 2 minutes, the cells were bleached for a third time (B3). In control experiments, the triple-bleach protocol was carried out in the absence of carbachol (between B1 and B2) and atropine (between B2 and B3), or just in the absence of atropine (between B2 and B3).

Raw ROI fluorescence data were normalized in Microsoft Excel. The initial fluorescence data point after bleaching was assigned value zero at time zero. Fluorescence recovery over time was normalized to this point. Non-linear regression analysis was applied to determine the kinetics of fluorescence recovery, with fluorescence recovery half-times (t_½) used to compare relative mobility's of fluorescent species. This analysis was achieved with Prisim3 software (Graph Pad).

Diffusion coefficients (D) for the GFP-fusion proteins were calculated using the Siggia simulation (Siggia et al., 2000). This simulates the diffusion of unbleached proteins into the ROI and compares experimental to simulated recovery plots to determine D. D reflects the mean-squared displacement of a protein undergoing Brownian motion over time, and is expressed as μm² second^–1 (Snapp et al., 2003).

Confocal images of live HEK-293 cells were taken using the Zeiss LSM 510 confocal microscope using a 63× oil-immersion lens. GFP fluorescence was excited with the 488 nm line of the argon laser, with an optical slice depth of 3 μm. All data are means±s.e.m.

We began our study by exploring whether PtdIns(3,4,5)P₃-induced recruitment to a membrane has any effect on a protein's mobility, using GAP1^m as a classic example of such a protein (Lockyer et al., 1999). We also reasoned that an interesting control would be GAP1^IP4BP, which is a very similar protein (Cullen et al., 1995) but is constitutively membrane bound owing to an apparent interaction with PtdIns(4,5)P₂ (Cozier et al., 2000). We thus set up protocols to study the mobility of both proteins in the cytosol and plasma membrane.

HEK-293 cells transiently expressing GFP, GFP-GAP1^IP4BP or GFP-GAP1^m were used directly in serum-free medium or (a) incubated in serum-free medium for 2 hours and then with the PI3-kinase inhibitor wortmannin (100 nM, 1 hour) or (b) incubated in serum-containing medium for 2 hours followed by incubation with the PtdIns(3,4,5)P₃-generating agonist insulin (100 nM, 1 hour). These latter two conditions were designed, respectively, either to deplete or to enhance the intracellular levels of PtdIns(3,4,5)P₃, so that we could directly compare the mobility of cytosolic and membrane-associated GFP-GAP1^IP4BP and GFP-GAP1^m. The subcellular distribution of GFP was identical under both PtdIns(3,4,5)P₃-reduced and PtdIns(3,4,5)P₃-increased conditions, being ubiquitous throughout the cell (Fig. 1A). The behaviour of GAP1^IP4BP and GAP1^m in our experiments was identical to that reported by others. That is, GAP1^m is cytosolic but translocates to the plasma membrane when PtdIns(3,4,5)P₃ is generated by insulin (Fig. 1C) (Lockyer et al., 1999), whereas GAP1^IP4BP is localized to the plasma membrane (Cozier et al., 2000; El-Daher et al., 2000) until prolonged treatment with wortmannin depletes PtdIns(4,5)P₂, which causes a shift to a partially cytosolic localization (Fig. 1B) (Cozier et al., 2000).

The horizontal white bar on each confocal image in Fig. 2 shows the section of the cell that was analysed to generate the adjacent fluorescence profile. We first analysed the mobile fraction for the principal constructs used in this study, because this parameter is representative of the total number of molecules free to diffuse into the bleached area (Siggia et al., 2000) and could impinge on our subsequent quantitative analysis of the mobility of the proteins. Also, any protein anchored to PtdIns(4,5)P₂ might reveal evidence of pools of PtdIns(4,5)P₂ with restricted mobility, and any such sequestering might be detected as any immobile fraction in the FRAP experiments. So, we analysed the M_f values of GFP, GFP-GAP1^IP4BP and GFP-GAP1^m in cells treated with either insulin or wortmannin to ensure that cytosolic or plasma-membrane localization of GAP1^IP4BP or GAP1^m made no significant difference. For GFP, the M_f values in the presence of insulin or wortmannin were 85.03±2.06% (n=7) and 76.02±3.20% (n=8), respectively. The corresponding values for GFP-GAP1^IP4BP were 81.21±2.83% (n=8) and 86.94±2.43% (n=7), and for GAP1^m 86.91±2.87% (n=7) and 89.60±1.16% (n=7). These data show that the different treatments discussed above do not significantly alter the proportions of molecules free to diffuse into the bleached region.

HEK-293 cells transiently expressing GFP, GFP-GAP1^IP4BP and GFP-GAP1^m were photobleached, and the kinetics of fluorescence recovery analysed. The ROI was 3 μm wide and spanned the width of the cell, encompassing both membrane and cytosolic compartments (Fig. 2A). An example of this process is shown in Fig. 2 for a GFP-GAP1^IP4BP-expressing HEK-293 cell. The trace of raw fluorescence data (Fig. 2B) records fluorescence changes in the ROI during the experiment, with pre-bleaching (i), immediate-post-bleaching (ii) and fluorescence recovery (iii) highlighted. Data obtained from the fluorescence recovery curves were normalized and fitted to a one-phase exponential-association equation. The protein mobilities are expressed as the half-time of fluorescence recovery (t_½) although, in some experiments (as discussed below), we also calculated the diffusion coefficient (D). In control HEK-293 cells, we found that the diffusion coefficient of GFP is 23.00±7.61 μm².second^–1 (n=4), which is similar to the values obtained by others (Dayel et al., 1999).

The mobility as judged by the t_½ of recovery (Fig. 3) of GFP was the same in the cytosol of HEK-293 cells when PtdIns(3,4,5)P₃ levels were increased or reduced (t_½ 0.95±0.29 seconds (n=6) or 0.86±0.24 seconds, respectively (n=6)). The mobility of GFP-GAP1^m was also similar under conditions of reduced PtdIns(3,4,5)P₃ levels, when it is cytosolic (t_g 1.71±0.17 seconds (n=6)), or increased PtdIns(3,4,5)P₃ levels, when it is localized to the plasma membrane [t_½ 1.44±0.14 seconds (n=6)]. The diffusion coefficients for cytosolic versus plasma-membrane GAP1^m were 5.56±0.52 (n=5) and 3.83±0.67 (n=6) μm² second^–1, respectively.

The calculation of diffusion coefficients involves assumptions about the diffusion of a soluble protein that essentially ignores the third dimension (Snapp et al., 2003). There can therefore be some apparent variation when comparing the t_½ values with D for the proteins analysed here. This is true in the case of soluble versus membrane-bound GAP1^m (above); also see below, for example, where GFP-PH(PLC-δ) and GFP-GAP1^m have different t_½ values [3.62±1.83 seconds (n=6) and 1.77±0.64 seconds (n=6), respectively], yet they have similar D values [3.79±2.22 μm² second^–1 (n=6) and 3.55±1.44 μm² second^–1 (n=6), respectively]. D is the more reliable measure of mobility because it excludes many of the variabilities introduced by the experimental protocol. The calculation of D simulates the diffusion of unbleached proteins into the ROI and compares experimental with simulated recovery plots to extract D. It thus includes a component of the fluorescence of the whole cells and so allows for the difference in size of the ROI (relative to the cell) between different cells, which can cause differences in t_½ even if the actual diffusion is the same.

The most interesting conclusion that emerges from these data is that the diffusion coefficient of membrane-bound GAP1^m is not significantly different from that of the soluble protein (P>0.05, unpaired Student's t-test), suggesting that GAP1^m is freely diffusible in the plane of the plasma membrane. Thus, subject to the caveat that we are using transfected proteins (which might supersaturate some physiological protein-protein interactions), we suggest that the data provide the first quantitative evidence that binding an inositol lipid in a membrane has little effect on the free diffusion of a previously soluble protein (here, GAP1^m).

GAP1^IP4BP makes a very interesting contrast with GAP1^m. It is usually found on the plasma membrane (Cozier et al., 2000; El-Daher et al., 2000), probably because of its interaction with PtdIns(4,5)P₂ (Cozier et al., 2000). When mostly cytosolic [under conditions of depleted PtdIns(3,4,5)P₃ and probably also depleted PtdIns(4,5)P₂], its t_½ is 1.59±0.17 seconds (n=6) (Fig. 3A). This signal is a mixture of the cytosolic and a small amount of remaining plasma-membrane-localized GFP-GAP1^IP4BP (Fig. 2Ai), so we also independently analysed the cytosolic component. This gave a value of t_½ 1.18±0.15 seconds (n=11), which is similar to the total signal (Fig. 3A) and is not significantly different from GAP1^m (P>0.05, unpaired Student's t-test). This is not surprising given the close similarity of the two proteins but is nevertheless an important control to confirm that GAP1^IP4BP has no intrinsically unusual properties in its mobility.

However, when attached to the plasma membrane in insulin-treated cells, the mobility of GAP1^IP4BP was markedly different, with a t_½ of 4.18±0.27 seconds (n=6) (Fig. 3) and a diffusion coefficient of 0.92±0.23 μm² second^–1. This is clearly different from GAP1^m, and indeed from GAP1^IP4BP itself when cytosolic, and this slower diffusion points to either to an interaction with another cellular component or to a much more restricted mobility of PtdIns(4,5)P₂, the lipid with which GAP1^IP4BP is probably interacting (Cozier et al., 2000).

PtdIns(4,5)P₂ has been suggested to have some limits to its free diffusion, either because it is in lipid rafts (McLaughlin et al., 2002; Munro, 2003; Simons and Vaz, 2004; van Rheenen and Jalink, 2002) or in caveolae (Galbiati et al., 2001) or some other localized region (Huang et al., 2004). PtdIns(4,5)P₂-binding proteins might therefore also show a restricted mobility, with GAP1^IP4BP being a case in point. This has important ramifications for function, so we expressed the GFP-tagged PH domain of PI-PLCδ₁, which specifically binds PtdIns(4,5)P₂ (Cifuentes et al., 1994). This has been used as a PtdIns(4,5)P₂-specific probe by many labs (Cullen et al., 2001; Stauffer et al., 1998) and, because it is a single domain (as opposed to a multidomain protein), is arguably more likely to be free of other interactions. We compared GFP-PH(PLC-δ₁) directly with the GFP-GAP1 proteins using a more restricted ROI and focused on the plasma membrane only (Table 1). For the GAP1 proteins, we used conditions under which each is plasma-membrane localized. The t_½ was calculated for each protein and we again calculated the diffusion coefficients (D) using the Siggia simulation (Siggia et al., 2000).

The data in Table 1 clearly demonstrate that the PH domain of PI-PLCδ₁ has a mobility similar to plasma-membrane-localized GAP1^m. The molecular weights of GFP-PH(PLC-δ₁) and GFP-GAP1^m are different (55 kDa and 124 kDa, respectively) but an expected difference in D caused by this is unlikely to be detectable in these experiments. The diffusion coefficient of PtdIns(4,5)P₂ in membranes has not been measured directly to our knowledge, but studies of other lipids (e.g. phosphatidylcholine in sarcoplasmic reticulum) have reported a figure of around 6 μm² second^–1 (Scandella et al., 1972). Thus, it is clear that proteins such as GFP-PH(PLC-δ₁) and GFP-GAP1^m, which are attached to a membrane solely by interaction of their PH domain with a lipid [an interaction whose structural details are known in other proteins for PtdIns(4,5)P₂ (Ferguson et al., 1995) and PtdIns(3,4,5)P₃ (Thomas et al., 2002)] diffuse only slightly more slowly than the lipid itself. We should note that there are reports (Murase et al., 2004) that lipids in the outer layer of the plasma membrane diffuse more slowly than those in the sarcoplasmic reticulum (Scandella et al., 1972), and our data on GFP-PH(PLC-δ₁) and GFP-GAP1^m might be pointing to the possibility that different parameters apply for the inner layer of the plasma membrane.

We can draw two further conclusions from these data. First, the difference between GAP1^IP4BP and GAP1^m (Table 1) is unlikely to be caused by their probably binding different inositol lipids [PtdIns(4,5)P₂ and PtdIns(3,4,5)P₃, respectively]. Second, the similarity between PH(PLC-δ₁) and GAP1^m (Table 1) implies that the mobilities of PtdIns(4,5)P₂ and PtdIns(3,4,5)P₃ are similar, and thus within the limits of this approach there is no evidence for compartmentalization of PtdIns(4,5)P₂ (Huang et al., 2004; Munro, 2003; Simons and Vaz, 2004; van Rheenen and Jalink, 2002).

The comparative low mobility of GAP1^IP4BP (Table 1) is therefore probably due to its interaction with another cellular component. Because the levels of all the GFP-tagged proteins in this study are high compared with endogenous protein, such a component is probably an abundant one in the cell. An example might be the structural proteins in the sub-plasma-membrane skeleton, such as a spectrin, which itself interestingly has a PtdIns(4,5)P₂-binding domain (Viel and Branton, 1996), although presently we have no data that address this possibility.

Another possibility is that GAP1^IP4BP might be interacting with a plasma-membrane protein (that is, one that is embedded in or spanning the membrane). We therefore finally compared the mobility of ICAM, which is a single-span transmembrane protein in the plasma membrane. Its mobility in direct comparison with the other proteins studied here should also be of intrinsic interest in the wider context of plasma-membrane function. The diffusion coefficient of ICAM in our experiments (Table 1) is similar to that found for other transmembrane single-span proteins (for review, see Snapp et al., 2003), which reinforces our analysis that proteins targeted to the membrane by inositol lipids are indeed more mobile.

Both GAP1^m and GAP1^IP4BP can bind Ins(1,3,4,5)P₄ specifically in vitro, and several characteristics of these proteins suggest that the likely upstream regulator of GAP1^m is PtdIns(3,4,5)P₃, whereas GAP1^IP4BP is more likely to be responsive to Ins(1,3,4,5)P₄ (Cozier et al., 2000; Cullen et al., 1995). Generating Ins(1,3,4,5)P₄ does not displace GAP1^IP4BP from the plasma membrane, even though it will displace it from an artificial PtdIns(4,5)P₂-containing bilayer (Cozier et al., 2000), and so, given the probable interaction of GAP1^IP4BP with some other cellular component, we considered it useful to ask whether Ins(1,3,4,5)P₄ might alter its mobility.

Because any such change might be small and, given the inherent errors in comparing mobilities quantitatively between different cells, we designed a protocol that should give the maximum chance of detecting any Ins(1,3,4,5)P₄-induced change. In all cells, a selected ROI in the membrane was bleached three times in total (B1, B2 and B3). After B1, the cell was allowed sufficient time to recover before the addition of carbachol (100 μM) or of vehicle. Carbachol stimulates Ins(1,4,5)P₃ production and so Ins(1,3,4,5)P₄ should be produced; we confirmed in separate experiments with Fluo-3-loaded cells that more than 95% of the cells responded to carbachol with a clear increase in intracellular calcium (not shown). In some experiments, we transfected cells with Ins(1,4,5)P₃-3-kinase to maximize Ins(1,3,4,5)P₄ production, as Cozier et al. did in their experiments (Cozier et al., 2000). We used the cytosolic construct of the A isoform of Ins(1,4,5)P₃-3-kinase (Schell et al., 2001), and in some further experiments transfected with a kinase-dead construct (Nash et al., 2002) as an additional control. For all these experiments, 5 minutes following the addition of carbachol or vehicle, the cells were bleached a second time (B2), in the same ROI (Fig. 4). The cells were allowed to recover once again before the addition of atropine (10 μM) or vehicle. 5 minutes later, the cells were bleached for a third time (B3), in the same ROI (Fig. 4).

In a typical experiment with only vehicle added, t_½ at B1, B2 and B3, respectively, were 4.22±0.48 seconds (n=12), 5.15±0.48 seconds (n=14) and 6.62±0.58 seconds (n=14), and the corresponding diffusion coefficients were 0.94±0.35 μm² second^–1, 0.79±0.14 μm² second^–1 and 0.97±0.12 μm² second^–1; these are not significantly different from one another. We should note that the corresponding M_f values for this series of experiments are 79.43±2.59 (n=12), 77.11±4.2 (n=14) and 80.85±2.86 (n=14); similarly constant M_f values for B1, B2 and B3 were obtained for all the other experiments described below (Table 2) and, taken together, these data suggest that the triple bleaching of one region of a cell does not cause significant laser-induced damage. For cells treated with first carbachol and then atropine, the t_½ values for B1-B3 were 5.39±0.69 seconds (n=13), 5.08±0.56 seconds (n=11) and 5.77±0.598 seconds (n=5), respectively, so the generation of Ins(1,3,4,5)P₄ has had no detectable effect on mobility. If atropine was omitted before B3, the corresponding t_½ in the above series of experiment for B3 was 6.08±0.67 seconds (n=7). In a series of experiments on cells that were co-transfected with Ins(1,4,5)P₃-3-kinase and treated with carbachol and then atropine, the t_½ values for B1-B3 were 4.56±0.51 seconds (n=7), 4.53±0.45 seconds (n=7) and 4.68±0.52 seconds (n=6), respectively. With the kinase-dead construct, the equivalent t_½ values were 3.18±0.40 (n=6), 3.29±0.23 (n=6) and 3.71±0.24 (n=6), respectively. Thus, within the limits of these experiments, we can conclude that production of Ins(1,3,4,5)P₄ has no detectable effect on the mobility of GAP1^IP4BP. The principal caveat to this is that GAP1^IP4BP produced at such high levels might not respond to Ins(1,3,4,5)P₄ and, for a definitive answer, the endogenous protein would have to be analysed.

Finally, it is relevant to ask in the same context whether PtdIns(3,4,5)P₃ generation alters GAP1^IP4BP mobility, given that has not yet been ruled out that PtdIns(3,4,5)P₃ could be an additional, or alternative, regulator of this protein. The diffusion coefficient for GAP1^IP4BP in control cells was 0.71±0.44 μm² second^–1 (n=10) (Table 1), which is very similar to the value of 0.92±0.23 μm² second^–1 in insulin-treated cells (Fig. 3), in which PtdIns(3,4,5)P₃ has been generated. We cannot be sure how much PtdIns(3,4,5)P₃ is present in the cells that have received neither wortmannin nor insulin but, because it must be at a level that is low enough to be insufficient to translocate GAP1^m, we conclude that PtdIns(3,4,5)P₃ generation does not detectably alter GAP1^IP4BP mobility.

We are most grateful to E. Snapp and J. Lippencott-Schwartz for providing us with the programmes to analyse FRAP data and to generate diffusion coefficients. We also thank the members of the Irvine lab for many helpful discussions. F.B. was supported by an MRC studentship, D.B. by a Programme Grant from the Wellcome Trust and R.F.I. by the Royal Society.