Both MHC Class II and its GPI-Anchored Form Undergo Hop Diffusion as Observed by Single-Molecule Tracking

Cell culture

For the details of culturing CHO cells expressing TM-I-E^k or GPI-I-E^k, see Vrljic et al. (1). Cells expressing these proteins were plated on 18 × 18-mm coverslips (IWAKI, Chiba, Japan) for SPT observations or on 12 mm-φ glass-bottom dishes (IWAKI) for SFMT observations, and were grown for a day before the microscope experiment. The glass surface was coated with 50 μg/ml fibronectin from human plasma (CalBiochem, San Diego, CA), as described previously (1).

Preparation of colloidal gold probes

The TM-I-E^k and GPI-I-E^k expressed on the CHO cell surface were first labeled with anti-I-E^k-Fab, and then the cells were further incubated with anti-mouse IgG-Fab-coated 40 nm colloidal gold particles. The anti-I-E^k-Fab was prepared from anti-I-E^k-IgG (14-4-4s) purified from the hybridoma (HB32, ATCC, Manassas, VA) supernatant. The minimal protecting amount (MPA, defined as the minimum concentration of the protein needed to stabilize colloidal gold particles in suspension) of anti-mouse IgG-antibody Fab (Cappel, Organon Teknica, Durham, NC) was determined to be 3.3 μg/ml (40,41). Colloidal-gold probes coated with the MPA of Fab were prepared by mixing 50 μl of 36 μg/ml anti-mouse IgG Fab in 2 mM borate buffer (pH 9.2) and a 500 μl suspension of colloidal gold (BBInternational, Cardiff, UK) on a slowly tumbling shaker for 15 min at room temperature. The gold probe was stabilized by the further addition of bovine serum albumin (BSA) at a final concentration of 1% (w/v). After sedimentation, the gold probe was resuspended in RPMI medium supplemented with 10% fetal bovine serum and 1% BSA, sterilized by filtration through a 0.22-μm filter (Millipore, Bedford, MA), and then used within 5 h.

SFMT of fluorophore-labeled TM-I-E^k, GPI-I-E^k, and DOPE

See Fig. 2 (left). TM-I-E^k and GPI-I-E^k were labeled by using either anti-I-E^k Fab conjugated with Alexa594 (Molecular Probes, Eugene, OR); the final molar ratio of fluorochrome to protein was 0.2–0.3:1) or Moth Cytochrome c peptide, MCC 95–103 (IAYLKQATK, custom-ordered from the Peptide Institute, Osaka, Japan) labeled at the N-terminus with Cy3 mono functional dye (Amersham-Pharmacia, Uppsala, Sweden; precisely one dye molecule per peptide chain). Cells were incubated with 4–40 nM Alexa594-anti-I-E^k Fab or with 30–300 nM (0.05–0.5 μg/ml) Cy3-MCC 95–103 peptide for 15–30 min at 37°C, in RPMI 1640 medium supplemented with 10% FBS and 1% BSA without phenol red (the same medium was used for microscope observations). For SFMT of DOPE, first Alexa594-DOPE (custom-ordered from Molecular Probes) in methanol (14 μM) was mixed with RPMI 1640 medium without 10% FBS and phenol red by vigorous vortexing (70 nM final concentration), and then this solution was added to the cells cultured on a glass-bottom dish at 37°C (final concentration 7 nM).

All observations of the cells were carried out at 37°C for up to 20 min. No efforts for deoxygenation, such as the procedures employed by Vrljic et al. (1,2) and Nishimura et al. (3), were made. Individual Alexa594 molecules were monitored on the upper cell membrane at video rate (30 Hz), using the oblique illumination mode of a homebuilt objective lens-type total internal reflection fluorescence microscope (42–44). Briefly, a 594.1-nm laser beam (He-Ne laser; Melles Griot, Carlsbad, CA) and a 532-nm laser beam (the second harmonic of the Nd:YAG laser beam, Model 4501-050; Uniphase, San Jose, CA) were attenuated with neutral density filters, circularly polarized, and then steered into the edge of a high numerical aperture (NA) objective lens (PlanApo 100×, NA = 1.4; Olympus, Tokyo, Japan) with a focus at the back-focal plane of the objective lens on an Olympus inverted microscope (IX-70).

The precision of the position determination was estimated from the standard deviation of the coordinates of Alexa594-Fab and Cy3-peptide adsorbed to a poly-L-lysine-coated coverslip, overlaid by a 15% polyacrylamide gel (5% cross-linker) (16,45), and was ≈20 nm at a time resolution of 33 ms.

SPT of TM-I-E^k, GPI-I-E^k, and DOPE, tagged by 40-nm-φ colloidal gold particles

See Fig. 2 (right). For the TM-I-E^k and GPI-I-E^k observations, these membrane molecules were labeled with anti-I-E^k Fab fragments and then labeled with gold probes coated with anti-mouse IgG antibody Fab fragments (gold particles could not be directly coated with anti-I-E^k Fab fragments, due to the inability of these Fab fragments to protect the gold surface). First, the anti-I-E^k Fab (8 μg/ml final concentration) was incubated with CHO cells expressing TM-I-E^k or GPI-I-E^k, which were cultured on coverslips, and then after washing the cells, gold probes conjugated with anti-mouse IgG Fab were applied to the cells. All observations of the gold probes were carried out at 37°C within 20 min after the application of the gold probes to the cells, which was very effective for reducing the number of gold particles exhibiting slow diffusion or immobilization.

For SPT of DOPE, gold probes coated with anti-fluorescein antibody Fab fragments were bound to fluorescein-DOPE (fluorescein was simply used as a tag, rather than a fluorescent probe), which was preincorporated in the cell membrane, as described in Fujiwara et al. (5) and Murase et al. (6). Briefly, after fluorescein-DOPE was incorporated in the cell membrane by the addition of 2 μg/ml (final concentration) of fluorescein-DOPE, gold probes conjugated with the anti-fluorescein antibodies' Fab fragments were applied to cells cultured on 18 × 18-mm coverslips. For the observations with enhanced frame rates, a digital high-speed camera with a C-MOS sensor was used (Fastcam-APX RS; Photron, Tokyo, Japan (5,6,14,19)). For high-speed video microscopy, bright-field optical microscopy, rather than Nomarski microscopy, was employed. We used an Alpha-Plan-Fluar 100× oil immersion objective lens with a numerical aperture of 1.45 (Carl Zeiss, Oberköchen, Germany). The sequence of images was replayed at the video rate (30 Hz) with analog and digital enhancements by an image processor (DVS-3000, Hamamatsu Photonics, Hamamatsu, Japan), and was recorded on a digital video tape recorder (DSR-20, Sony, Tokyo, Japan).

The precision of the position determination was estimated by the same method employed in SFMT using 40-nm-φ gold particles, and was 16 nm at a time resolution of 20 μs.

Obtaining the trajectories of membrane molecules and the plots of mean-square displacement (MSD) versus time

The positions (x and y coordinates) of the selected gold particles or fluorescent molecules were determined by a computer that employs the method developed by Gelles et al. (46). For each trajectory, the mean-square displacement (MSD), 〈Δr(Δt)²〉, for every time interval (Eq. 1) was calculated according to the formula (16,47,48)

(1)

where δt is the video frame time and x(jδt + nδt), y(jδt + nδt) describes the particle position after a time interval Δt_n = nδt after starting at position (x(jδt), y(jδt)), N is the total number of frames in the video recording sequence, n and j are positive integers, and n determines the time increment.

To keep the statistical spread in the MSD within reasonable levels for the duration of 0–1 s, TM-I-E^k, GPI-I-E^k, and DOPE trajectories of 3 s (90 video frames) were used for the quantitative analysis ((49–51); see Figs. 3–5).

Classification of the mode of diffusion, calculation of the diffusion coefficient, and analysis of the high-speed SPT trajectories

For a detailed description of the data analysis methods, see Fujiwara et al. (5) and Suzuki et al. (19).

We designate the effective diffusion coefficients (with an indication of the midpoint of the linear fitting) as D^eff(time resolution)_midpoint. For example, D^eff(33ms)_100ms (in this case, 33 ms is the time resolution employed for the observation) and D^eff(20μs)_60μs (observation rate of 20 μs/frame) correspond to D_2–4, defined in Kusumi et al. (16) and described in Suzuki et al. (19).

A statistical method for classifying each trajectory into confined-hop diffusion, simple Brownian diffusion, simple-Brownian diffusion with a drift, or immobile modes, based on the MSD-Δt plot, was described by Kusumi et al. (16). Briefly, all of the trajectories were first classified into mobile and immobile ones (see Fig. 5 and its related text in Results), and the mode-of-motion classification was carried out only for the trajectories that were classified as mobile (see Fig. 4 a and its related text in Results). The ensemble-averaged RD is ≪, ≈, or ≫ 1, when the molecules are undergoing confined-hop diffusion, simple Brownian diffusion, or simple-Brownian diffusion with a drift (directed diffusion mode), respectively. Fig. 4 a shows the theoretical curves for

1. Simple Brownian diffusion, in which MSD(Δt) = 4DΔt.

2. Directed diffusion mode, in which a molecule moves in a direction at a constant drift velocity (v_x, v_y), with superimposed random diffusion, MSD(Δt) = 4DΔt + v² (Δt)², where

3. Confined diffusion, in which a molecule undergoes Brownian diffusion while totally confined within a limited area (compartment; 0 ≤ x ≤ L_x, 0 ≤ y ≤ L_y) during the observation period.

The MSD-Δt plot levels off and asymptotically approaches a constant value, as expressed by

For the analysis of the trajectories obtained by using high-speed SPT with 20-μs resolutions and classified into the confined diffusion mode (under the analysis conditions employed therein), the MSD-Δt plots in the x or y directions were fitted with a homemade program based on the hop diffusion theory of Powles et al. (52), in which a particle undergoes diffusion in the presence of semipermeable barriers placed at an equal distance (termed “hop fit” in this article; see also (5,6,19)). In the analysis, the time evolution of the probability distribution depends on three parameters: the distance between barriers, L, the true diffusion coefficient in the absence of barriers, D_micro, and the permeability of the barriers, P. Powles et al. (52) also derived a relationship between the permeability and the long-term diffusion coefficient, D_MACRO, D_MACRO/D_micro = [1 + (PL)⁻¹]⁻¹.

The correct hop rate (or the residency time within a compartment) was evaluated from the macroscopic diffusion coefficient, determined by SFMT with a fluorescent probe, and the compartment size, determined by SPT with a gold probe. Individual compartments for each trajectory were automatically identified by the computer program (5,6,19).

Drug treatments

Latrunculin A (Molecular Probes) and cytochalasin D (Sigma, St. Louis, MO) were dissolved in methanol. The treatment of cells with latrunculin A was done by incubating the cells in the RPMI medium supplemented with 1% BSA (Sigma), containing 1 μM latrunculin A (0.01% methanol) under the microscope observation at 37°C, and the observation was carried out between 5 and 20 min after the drug addition. Likewise, with 1 μM and 10 μM cytochalasin D (0.006% and 0.06% methanol), the cells were treated for 1 min at 37°C, and then the observation was continued for up to 5 or 12 min. Control cells were treated with an equivalent amount of methanol alone.

Typical SFMT trajectories for TM-I-E^k, GPI-I-E^k, and Alexa594-DOPE obtained at video rate

All experiments were carried out at 37°C. Vrljic et al. (1,2) and Nishimura et al. (3) observed the diffusion of individual TM-I-E^k and GPI-I-E^k using a Cy5-peptide (MCC 95–103 peptide, with one Cy5 dye attached per peptide at its N-terminus) that binds to a specific site on MHC class II molecules with high specificity. In this study, we used Alexa594-conjugated anti-I-E^k Fab (the final molar ratio of fluorochrome to protein was 0.2–0.3:1) in addition to the peptide (conjugated with Cy3) (Fig. 2). They were observed at a 33-ms resolution for a period of 3 s in the upper cell membrane of CHO cells by SFMT, using the oblique illumination mode of a homebuilt objective lens-type total internal reflection fluorescence microscope. Each fluorescent spot of the Alexa594-conjugated anti-I-E^k Fab and Cy3-peptide in the microscope image was found to represent a single molecule, based on the single-step photobleaching (1,5,42).

Fig. 3 a shows typical trajectories of Alexa594-conjugated anti-I-E^k-Fab bound to TM-I-E^k and to GPI-I-E^k, and Alexa594-DOPE, a DOPE molecule tagged with Alexa594 in the headgroup region and incorporated in the cell membrane of CHO cells. Fig. 3 b shows ensemble-averaged MSD-Δt plots for TM-I-E^k and GPI-I-E^k observed with Alexa594-Fab (red) and Cy3-peptide (cyan), as well as Alexa594-DOPE (right) on the timescale of 1 s. These plots were fitted by straight lines in the range between 0 and 1 s, suggesting that TM-I-E^k, GPI-I-E^k, and DOPE on average underwent apparent simple Brownian diffusion on the 1-s scale. The averaged D^eff(33ms)_500ms values of TM-I-E^k and GPI-I-E^k labeled with Alexa594-Fab and of Alexa-DOPE were 0.13, 0.25, and 0.12 μm²/s, respectively (designated as D^eff(time resolution)_midpoint; see Materials and Methods). The diffusion coefficients of TM-I-E^k and GPI-I-E^k with the Cy3-peptide probes were virtually the same as those with the Alexa594-Fab probes.

Classification of single-molecule trajectories into different diffusion modes

We next performed a statistical analysis of each single-molecule trajectory to classify each trajectory into simple Brownian, directed, confined-hop, or immobile modes, based on the MSD-Δt plot (16), rather than characterizing the collective diffusion properties. Fig. 4 a shows representative theoretical curves of MSD-Δt plots for simple Brownian diffusion, directed diffusion, and confined-hop diffusion, with the same microscopic diffusion coefficient.

Firstly, any fluorescent spot exhibiting a diffusion coefficient in a 100-ms time-window (D^eff(33ms)_100ms) <0.007 μm²/s was classified into the immobile mode in this experiment, in the sense that it cannot be distinguished from the immobilized probe on the coverslip. The value of 0.007 μm²/s was selected as the lowest detectable limit for D^eff(33ms)_100ms, because it was a spot exhibiting the top 2.5 percentile value in the distribution of the nominal D^eff(33ms)_100ms in the range below 0.009 μm²/s, which the Alexa594-Fab or Cy3-peptide attached to the cover glass exhibited (Fig. 5 a, top). The fraction classified as the immobile mode was in the range of 3–9% for all of the molecules studied here. These immobile probes might be the probes that are bound to the cell surface nonspecifically.

Next, the mode-of-motion classification was carried out, but only for trajectories that were classified into the mobile mode. This method employs the parameter RD(N, n), which describes the long-term deviation of the actual MSD(N, n) at the time nδt (N = the full length in the number of image frames in a trajectory; n = the number of frames used for the analysis in the MSD-Δt plot, δt = duration of each frame) from the expected MSD based on the initial slope of the MSD-Δt plot, 4D_2-4nδt, i.e., RD(N, n) = MSD(N, n)/[4D_2-4nδt] (Fig. 4 a and see Materials and Methods). In the case of molecules undergoing simple Brownian diffusion, the average value of RD(N, n) is 1, although the value for each individual trajectory of Brownian molecules would show a statistical spread at ∼1. Using Monte Carlo simulation, we generated 5000 90-step trajectories (here N is 90), and the ideal statistical spread of the RD(90, 30) was obtained, as shown in Fig. 4 b (top). For the classification of the trajectories into different diffusion modes, RD(90, 30) values that gave the 2.5 percentile of particles from both ends of the distribution, referred to as RD_min(90, 30) and RD_max(90, 30), were determined, as described by Kusumi et al. (16) (shown in Fig. 4 b by vertical red and cyan lines, respectively). When the trajectory of an experimental molecule shows an RD value between RD_min and RD_max, it is classified into the simple Brownian diffusion mode, and when RD > RD_max or RD < RD_min, it is classified into the directed or confined-hop diffusion mode, respectively. Greater than 74% of the TM-I-E^k, GPI-I-E^k, and DOPE trajectories were classified into the simple Brownian diffusion mode, irrespective of the probes employed here (Fig. 4 b). These results were consistent with the results reported by Vrljic et al. (1,2) and Nishimura et al. (3), but not with those reported by Schütz et al. (36). None of the trajectories exhibited confinement within φ ∼ 80 nm domains, as reported by Lenne et al. (38) and Wenger et al. (39), but the video-rate (a frame rate at once every 33 ms) might be too slow to detect transient confinement for several 10s of milliseconds.

The Alexa594-Fab and Cy3-peptide probes display similar diffusion behaviors

Since the majority of the mobile molecules exhibited apparent simple Brownian diffusion at video rate (Fig. 4 b), the motion of each molecule observed at this frame rate can be characterized by a single effective diffusion coefficient. The distributions of the effective diffusion coefficients, D^eff(33ms)_100ms, for the median values TM-I-E^k and GPI-I-E^k labeled with Alexa594-Fab or Cy3-peptide, as well as the distributions for these probes bound to the glass surface (control for immobile molecules), are shown in Fig. 5 a. The median values for TM-I-E^k and GPI-I-E^k with Alexa-Fab were 0.14 and 0.23 μm²/s, respectively, and those with the Cy3-peptide were 0.15 and 0.23 μm²/s, respectively (excluding the immobile fraction, defined as the spots with D^eff(33ms)_100ms <0.007 μm²/s; see the cyan line in Fig. 5 a). There were no significant differences in these diffusion coefficients of TM-I-E^k and GPI-I-E^k between the two different labels (Alexa594-Fab and Cy3-peptide). These results, along with the results of the ensemble-averaged MSD-Δt plots for the Alexa594-Fab and Cy3-peptide probes (Fig. 3 b), indicate that both of these probes can be employed to study the dynamics of TM-I-E^k and GPI-I-E^k. Since the labeling efficiency of Alexa594-Fab is much better than that of Cy3-peptide, we performed all of the following experiments using Alexa594-Fab probes.

Video-rate SPT observations of gold-tagged TM-I-E^k, GPI-I-E^k, and DOPE

Next, we intended to observe the diffusion of these molecules at a much higher frame rate to examine the possibility of anomalous diffusion, such as hop diffusion or confinement within ∼80-nm domains for tens of milliseconds (5,6,14,19). Since achieving higher time resolutions with fluorescent probes is difficult due to the problem of low signal/noise ratios, we carried out single-particle tracking (SPT) with 40-nm-φ colloidal gold probes (see Fig. 2 and Materials and Methods).

We first observed gold-labeled molecules at video rate, to compare their diffusions with those of fluorescently-labeled molecules. Fig. 3 a (bottom) shows typical trajectories of gold-labeled TM-I-E^k, GPI-I-E^k, and DOPE, observed at a 33-ms resolution using SPT. No gold-tagged molecule was classified into the immobile mode. A statistical analysis classified almost all of the trajectories of the gold-tagged molecules, as well as those of the fluorescently-labeled molecules, into the simple Brownian diffusion mode (Fig. 4 b, bottom). The distributions of D^eff(33ms)_100ms for these molecules are shown in Fig. 5 b (black open bars; compare those with gray bars, representing data with the Alexa-Fab probe). The median values of the diffusion coefficients of gold-labeled TM-I-E^k, GPI-I-E^k, and DOPE in a 100-ms time-window (D^eff(33ms)_100ms) were 0.044, 0.032, and 0.061 μm²/s, respectively, which are 3–7-fold smaller than those of the Alexa594-labeled molecules. These results suggest that the diffusion of gold-labeled molecules may be slowed, due to steric hindrance and/or the crosslinking effect of gold probes attached to these molecules.

Previously, in the NRK cell membrane using the same Cy3-DOPE and gold-tagged DOPE, Fujiwara et al. (5) found that these probes gave the same diffusion coefficients, as long as the time-window for evaluating the diffusion coefficient was <100 ms. This difference is probably due to the very small compartment size in the CHO cell membrane (40 vs. 230 nm in NRK cells), as described below. Gold-labeled molecules collide with the compartment boundaries ∼30-fold ([230/40]²) more often in CHO cells than in NRK cells, which is likely to make the D^eff(33ms)_100ms of these gold-labeled molecules very sensitive to low levels of gold-induced crosslinking.

High-speed SPT with gold-tags revealed the hop diffusion of TM-I-E^k, GPI-I-E^k, and DOPE

The movements of gold-tagged TM-I-E^k, GPI-I-E^k, and DOPE were examined at a time resolution of 20 μs (at 50,000 frames/s), an enhancement by a factor of 1667 from the normal video rate (once every 33 ms). Their typical trajectories, shown in Fig. 6 a, suggest that all of these molecules undergo hop diffusion in the plasma membrane. In Fig. 6 a, each color indicates plausible confinement within a compartment. Individual plausible compartments were identified by software developed in our laboratory as well as by visual examination (5,6,19). The residency time for each compartment is indicated in the same color.

Fig. 6 b shows the ensemble-averaged MSD-Δt plots of the gold-tagged molecules, on a timescale of 5 ms (out of 100-ms trajectory or 250 frames out of 5000 frames). In this display, it is clear that the MSD-Δt curve is not linear, particularly in the range of 0–0.5 ms, indicating that these molecules undergo anomalous diffusion. These plots should be compared with those in Fig. 3 b, where the first point is at 66 ms (the first point in the MSD-Δt curve is removed, because all of the high-frequency noise components are accumulated in the first point in the MSD-Δt plot). Note that in the MSD-Δt plots shown in Fig. 6 b, the main curvature is seen between 0 and 0.5 ms, and the MSD value at 0.5 ms is only 0.0004–0.001 μm².

These ensemble-averaged MSD-Δt plots were fitted with a theoretical equation representing hop diffusion over equally spaced, semipermeable barriers (52). The fit parameters included L, D_micro, and D_MACRO (due to the lack of time resolution or insufficient frame rate, even at 50,000 frames per second for the compartment size of ∼40 nm, the correct D_micro cannot be obtained). For further details of the analysis, see Classification of the Mode of Diffusion, Calculation of the Diffusion Coefficient, and Analysis of High-Speed SPT Trajectories. The following hop parameters were obtained from this fitting. For TM-I-E^k: L = 52 nm and D_MACRO = 0.12 μm²/s. For GPI-I-E^k: L = 36 nm and D_MACRO = 0.086 μm²/s. For DOPE, L = 50 nm and D_MACRO = 0.13 μm²/s. These should be compared with the data shown later in Fig. 8 and Table 2. It is interesting to find that D_MACRO for GPI-I-E^k are smaller than that for TM-I-E^k or DOPE. Since SFMT with fluorescent probes, which do not induce aggregation of the target molecules, exhibited more or less similar D_MACRO for all of the three molecules (Figs. 3 and 5; also see Table 2), this suggests the readiness of GPI-I-E^k for being crosslinked with gold particles. Therefore, although D_MACRO for GPI-I-E^k is slightly greater than that for TM-I-E^k or DOPE, GPI-I-E^k molecules might have greater tendency to associate with each other.

A statistical classification of each individual trajectory into the categories of simple Brownian, confined-hop, and directed diffusion was carried out, as described in Fig. 4 and the related text, and shown in Fig. 7. Here all of the analyzed trajectories are 5000 frames' long (100 ms), and for each trajectory, MSD-Δt was calculated from all of the possible combinations of any two points in the trajectories, to produce MSD-Δt for the full timescales of 2, 5, and 8 ms (100, 250, and 400 points). These results indicated that 75–89% of the TM-I-E^k, GPI-I-E^k, and DOPE molecules undergo confined-hop diffusion, which was totally missed in the observations made at video rate.

The MSD-Δt plot for each trajectory (between 0 and 5 ms) was fitted with a theoretical equation representing hop diffusion over equally spaced, semipermeable barriers (52), as described above. The compartment size L and the macroscopic diffusion coefficient were the fit parameters, and the average residency time over a single trajectory was calculated from L and D_MACRO (L²/4D_MACRO).

Fig. 8 shows the distributions of the compartment sizes and the residency times within each compartment, determined for each trajectory. There were no significant differences in the compartment sizes sensed by the three molecules (medians, ∼40 nm). The median residency times were 4–7 ms, but note that these values are for gold-tagged molecules, which might be crosslinked by the gold probes. The correct values for the ensemble-averaged residency times of these three molecules were estimated in the following manner (the correct values for each individual molecule could not be determined).

Recently, Morone et al. (27) succeeded in directly determining the mesh size of the membrane skeleton on the cytoplasmic surface of the plasma membrane, by determining the three-dimensional structure of the membrane skeleton using electron tomography. They found that the compartment sizes determined by the high-speed diffusion measurements of gold-tagged molecules (crosslinked at various degrees, depending on the molecule under investigation) agreed well with the mesh size of the membrane skeleton on the cytoplasmic surface of the plasma membrane. From this agreement, it can be deduced that high-speed SPT of gold probes provides the correct compartment size. However, SPT may not give the correct hop rate, due to the crosslinking by gold particles. In contrast, Alexa594-labeled molecules would give more correct macroscopic diffusion rates over many compartments, whereas the low time resolution of SFMT does not allow direct observations of the hop events and the compartment size. Therefore, the correct hop rate (averaged value) can be estimated using the median D^eff(33ms)_100ms of Alexa594-labeled molecules and the median compartment sizes obtained by using gold-labeled molecules. The median residency times are 3.2 ms for TM-I-E^k ([0.042 μm]²/4 × 0.14 μm²/s), 1.4 ms for GPI-I-E^k, and 2.9 ms for DOPE (indicated in parentheses in Fig. 8 and listed in Table 2). In summary, all of the molecules we examined here exhibited rapid hop diffusion between ∼40-nm compartments, with a dwell time of 1–3 ms in each compartment on average. Confinement within φ ∼ 80 nm domains for several 10s of milliseconds was never observed for any single molecule inspected in this study.

A useful parameter, which can be calculated from D_MACRO, D_micro, and the compartment size, as described by Powles et al. (52) and Murase et al. (6), is the probability of passing a barrier when the membrane molecules are already at the boundary (PP). This shows how easily a molecule in the plasma membrane at a compartment boundary can pass the barrier. As D_micro, we used 8 μm²/s for DOPE (5,6) and GPI-I-E^k (assuming that its microscopic diffusion coefficient is the same as that for the phospholipid DOPE), and 6 μm²/s for TM-I-E^k (based on 4.5–6 μm²/s by (19) for G-protein-coupled receptor, which contains seven transmembrane domains, and employing the upper-limit value for TM-I-E^k, because it has only two transmembrane domains in the dimeric structure). The PP value of GPI-I-E^k was approximately twofold larger than those for DOPE and TM-I-E^k (Table 2). In other words, GPI-I-E^k tends to pass through the barrier two-times more readily than TM-I-E^k and DOPE. The mechanism underlying this greater hop probability for the GPI-anchored protein is unknown.

The actin-based membrane skeleton is responsible for cell membrane compartmentalization

We have examined the effects of drugs that affect actin organization on the diffusion of TM-I-E^k, GPI-I-E^k, and DOPE in the plasma membrane. CHO cells were incubated in the observation medium containing latrunculin A (final 1 μM) on the microscope stage at 37°C for 5 min. To avoid the secondary and/or large-scale drug effects, microscope observations were completed within 20 min after the drug addition. The latrunculin A treatment induced slight decreases in the rhodamine-stainable actin filaments visible by fluorescence microscopy (see our Supplementary Material, Data S1, Fig. S1, row b). Fig. 9 (red open bars) shows the distributions of the compartment sizes and the apparent residency times within each compartment, determined for each high-speed SPT trajectory of TM-I-E^k, GPI-I-E^k, and DOPE after latrunculin A treatment. Under these conditions, larger compartments appeared, with the median diameter increased by a factor of ∼1.5, or the area by a factor of ∼2.3. Meanwhile, the apparent residency times (called “apparent” due to the use of gold particle probes for high-speed SPT, which would crosslink the target molecules, prolonging the residency time) were not affected at a statistically meaningful level.

Under the same conditions, the macroscopic diffusion coefficients for fluorescently-tagged molecules were determined by SFMT at video rate (Fig. 10). D^eff(33ms)_100ms (macroscopic diffusion coefficient) was increased by 30–70% (p < 0.05; Fig. 10, red open bars). From these observations, the averages of the corrected residency times and PP (probability of passing the barrier) were calculated, as summarized in Table 2. The residency times increased, but the PP remained about the same, suggesting that the latrunculin-induced increases of the macroscopic diffusion are due to the expansion of the compartments, rather than the increased probability of passing the compartment boundaries (the increases of the residency times would also probably be due to the expanded compartments, which would decrease the frequency of the molecule reaching the compartment boundaries, rather than the increase in the probability of passing the fence).

The effect of cytochalasin D on the diffusion of TM-I-E^k, GPI-I-E^k, and DOPE in CHO cells

We next examined the effects of another drug, cytochalasin D. Its mode of action on actin organization is different from that of latrunculin A. While latrunculin A binds to actin monomers and blocks their polymerization, resulting in the amount of actin filaments (53,54), cytochalasin D at micromolar concentrations binds to both the barbed and pointed ends of actin filaments and prevents their further polymerization, as well as their interactions with other barbed-end binding proteins (55).

While the treatment with 1 μM latrunculin A for 5–20 min (all of the observations were conducted within 20 min after latrunculin addition) strongly affected the molecular motion, the treatment with 1 and 10 μM cytochalasin D for 2–12 min influenced neither the compartment size nor the apparent residency time, under the conditions where the cells did not round up, but the number of stress fibers was decreased and actin clusters were formed (Fig. 9, cyan open bars, Data S1, Fig. S1 rows c and d, and Table 2). These results, obtained at a 20-μs resolution, are consistent with the report by Vrljic et al. (1), showing no influence of cytochalasin D (for 30 min or 60 min at 37°C with 1, 10, and 40 μM) on the movement of TM-I-E^k and GPI-I-E^k at a 100-ms resolution.

SFMT observations at low frame rates

We found that the average D^eff(33ms)_100ms values for fluorescently-labeled TM-I-E^k, GPI-I-E^k, and DOPE at 37°C are 0.15, 0.33, and 0.15 μm²/s, respectively (the medians are 0.14, 0.23, and 0.15 μm²/s, respectively). Vrljic et al. (1) reported that the mean diffusion coefficients of TM-I-E^k and GPI-I-E^k were 0.18 and 0.22 μm²/s, respectively, at a 100-ms resolution at 22°C. Nishimura et al. (3) reported them to be 0.59 and 1.1 μm²/s, respectively, at a 15.4-ms resolution at 37°C. The diffusion coefficients of MHC class II proteins and GPI-anchored proteins reported previously are summarized in Table 1. The diffusion coefficients of transmembrane MHC class II proteins exhibited broad distributions of 0.13 × 10⁻⁴–0.59 μm²/s. In addition, the reported diffusion coefficients of GPI-linked proteins are also broadly distributed (0.0035–1.1 μm²/s). The recent results reported by Lenne et al. ((38); ∼1.1 μm²/s) and Nishimura et al. ((3); 1.1 μm²/s) are the largest in Table 1, whereas the values found in this study lie in the middle of these distributions. The FCS used by Lenne et al. (38) might be quite insensitive to slowly-diffusing or immobile molecules, and thus the reported data might be skewed toward fast-diffusing species (D > 0.1 μm²/s; for a review, see (56)). Meanwhile, Vrljic et al. (1) and Nishimura et al. (3) measured D^eff(100ms)_50–250ms and D^eff(15.4ms)_15.4ms, respectively, which might be sensitive to high-frequency noise components (see the Supplementary Material text 1 of (19)). In addition, Vrljic et al. (2) noted that the diffusion coefficients slightly depended on the fetal bovine serum used to culture the CHO cells (see the Supplementary Material Fig. S1 of (2)), which might partially explain these differences. Accordingly, it is concluded that the effective diffusion coefficients of TM-I-E^k and GPI-I-E^k found here are in general agreement with the data presented by Vrljic et al. (1) and Nishimura et al. (3).

We found that GPI-I-E^k diffused 1.6-fold faster than TM-I-E^k (p < 0.05, in the Wilcoxon test; Fig. 5 a), in agreement with the previous observations (1,3,30–34). Meanwhile, Shvartsman et al. (37) found the diffusion coefficient measured by FRAP for the GPI-linked modified form of influenza hemagglutinin was slightly smaller than that for the native transmembrane form. However, a simple comparison of our results with the data reported by Shvartsman et al. (37) cannot be done, because the native transmembrane hemagglutinin is in a trimeric form, with raft-associating properties.

In the following, we address the three key questions we posed in the introduction of this report.

Do GPI-anchored proteins actually undergo hop diffusion if they are observed at higher frame rates?

High-speed SPT at a 20-μs resolution, with the aid of SFMT at video rate, revealed that all three of the molecules examined here, TM-I-E^k, GPI-I-E^k, and DOPE, undergo rapid hop diffusion between 40-nm compartments, with an average dwell time of 1–3 ms in each compartment in CHO cells. However, these results are at variance with the data published by Wieser et al. (35). They observed a GPI-anchored protein, CD59, labeled with Alexa647-Fab in T24 cells using SFMT. Their MSD-Δt plots contained only six points on the timescale of 0–6 ms, whereas we have 250 points on the timescale of 0–5 ms (Fig. 6 b). They stated that the MSD-Δt plot for CD59, extrapolated to time 0 from the linear region, gave the y intercept of ∼0, and concluded that CD59 intrinsically undergoes slow simple Brownian diffusion. The data in Fig. 6 b suggest that it would be very difficult to find the correct y intercept with only six points, given the large single-molecule tracking errors (see the error bars in Fig. 6 b).

Does GPI-I-E^k and not TM-I-E^k exhibit confinement within 80-nm or 700-nm domains for several 10s of milliseconds or longer?

Previously, Schütz et al. (36), as well as Lenne et al. (38) and Wenger et al. (39), suggested that raft-associating molecules exhibit confinement within 700-nm (36) and 80-nm domains (38,39) for several 10s of milliseconds or longer. Therefore, in this study, we paid special attention to whether we could find confinement with GPI-I-E^k, but not with TM-I-E^k. If such confinement occurs, then it should be easily detectable with our time resolution of 20 μs, because the trajectory should contain stationary periods over 1000 image frames, and the precision of our determination of the particle's coordinates is sufficiently high (16 nm). However, we have never seen such a trajectory in our SPT observations. Furthermore, we failed to see any immobilized particles (throughout the observation period of 100 ms).

Instead, we found that virtually all of the molecules we observed undergo hop diffusion, indicating that the entire plasma membrane of the CHO cell is parceled up into apposed domains of ∼40 nm, with regard to the translational diffusion of GPI-I-E^k, TM-I-E^k, and DOPE. These molecules undergo short-term confined diffusion within these 40-nm compartments for 1–3 ms on average, and long-term hop diffusion between these compartments, and thus undergo macroscopic diffusion. Therefore, such macroscopic diffusion was detected in observations with slow frame rates as slow simple Brownian diffusion, characterized by an effective macroscopic diffusion coefficient of 0.14–0.23 μm²/s.

The residency times, compartment sizes, and probabilities of passing barrier (PP) of GPI-anchored proteins are similar to those found for transmembrane proteins and DOPE (summarized in Table 2). Given that ∼35–50% of GPI-I-E^k, ∼5–25% of TM-I-E^k, and nearly 0% of DOPE were found in the Triton X-100-resistant fractions (1,57,58), these results suggest that either the raft association affects the diffusion of membrane molecules only slightly or the raft domain might be very small and/or short-lived.

Does the treatment of cells with drugs that affect actin polymerization, such as cytochalasin D and latrunculin A, induce changes in the diffusion of GPI-anchored proteins?

Vrljic et al. (1,2) and Nishimura et al. (3) previously failed to detect any changes in the diffusion characteristics, after the cells were treated with cytochalasin D. Furthermore, Frick et al. (59) and Schmidt and Nichols (60) found that the disruption of the actin cytoskeleton by treating the cells with latrunculin A, gelsolin, or siRNA of spectrin had no detectable effect on the diffusion of membrane molecules by FRAP.

Therefore, in this investigation, the effects of both latrunculin A and cytochalasin D on the translational diffusion of GPI-I-E^k, TM-I-E^k, and DOPE were examined. Consistent with the results by Vrljic et al. (1) and Nishimura et al. (3), cytochalasin D treatments, under the conditions where the global cell morphology is not seriously affected, did not induce any changes in the diffusion of the three molecules examined here. In contrast, in the high-speed SPT experiments by Murase et al. (6), cytochalasin D greatly affected the compartment size in the fetal rat skin keratinocyte cells in culture.

The effect of cytochalasin D on the diffusion of membrane molecules is complex, probably because it depends on the level of actin expression as well as the feedback reactions of the cell (summarized in the On-line Supporting Information of (20,44), and Supplementary Materials of (19)). The cell could react to the loss of the links between the plasma membrane and the actin filaments, due to the barbed-end capping by cytochalasin D, by generating more actin filaments. This might compensate for the initial effect of cytochalasin D. Such compensation effects should always be considered in any study modulating actin filaments (59,60). Therefore, these results indicate that due caution is required for interpreting the pharmacological data.

The cell treatment with another actin-depolymerizing drug, latrunculin A, increased the compartment size for all three molecules examined in this study, without affecting the probability of passing the barrier when the membrane molecule was already in the boundary area between the compartments. This suggested that the actin-based membrane skeleton is responsible for the partitioning of the plasma membrane for the translational diffusion of both transmembrane and GPI-anchored proteins. This explains the ∼50% increase in the median macroscopic diffusion coefficient after the latrunculin treatment for all three molecules. Furthermore, it should be emphasized that such changes are observable only under well-controlled conditions: 1 μM latrunculin A and only during 5–20 min after the drug addition. Milder treatments had much milder effects, whereas harsher treatments would elicit compensating reactions of the cell or harm the cells.

Small changes in the macroscopic diffusion coefficient, such as a ∼50% increase in its median value, would be difficult to detect with normal SFMT at low frame rates, with a few observed points in the MSD-Δt plot, or with techniques observing ensemble-averaged values. In this study, SFMT at video rate for 1 s (30 observed points in the MSD-Δt plot, see Fig. 3 b) was combined with high-speed SPT. This combination allows sensitive detection of the changes in the diffusion of membrane molecules, upon drug addition.

In conclusion, all three of the molecules, GPI-I-E^k, TM-I-E^k, and DOPE, undergo hop diffusion in the plasma membrane, which is sensitive to changes in the actin-based membrane skeleton. The compartment size for these molecules is the same (∼40 nm). This result supports the fence and picket models. Meanwhile, the residency time for GPI-I-E^k is 1.4 ms (median), which is approximately twofold shorter than those for TM-I-E^k (3.2 ms) and DOPE (2.9 ms). Namely, GPI-I-E^k hops faster than TM-I-E^k or DOPE. The reason for the faster hop rate for GPI-I-E^k is still unknown.