Multiscale diffusion in the mitotic Drosophila melanogaster syncytial blastoderm

Abstract

Despite the fundamental importance of diffusion for embryonic morphogen gradient formation in the early Drosophila melanogaster embryo, there remains controversy regarding both the extent and the rate of diffusion of well-characterized morphogens. Furthermore, the recent observation of diffusional “compartmentalization” has suggested that diffusion may in fact be nonideal and mediated by an as-yet-unidentified mechanism. Here, we characterize the effects of the geometry of the early syncytial Drosophila embryo on the effective diffusivity of cytoplasmic proteins. Our results demonstrate that the presence of transient mitotic membrane furrows results in a multiscale diffusion effect that has a significant impact on effective diffusion rates across the embryo. Using a combination of live-cell experiments and computational modeling, we characterize these effects and relate effective bulk diffusion rates to instantaneous diffusion coefficients throughout the syncytial blastoderm nuclear cycle phase of the early embryo. This multiscale effect may be related to the effect of interphase nuclei on effective diffusion, and thus we propose that an as-yet-unidentified role of syncytial membrane furrows is to temporally regulate bulk embryonic diffusion rates to balance the multiscale effect of interphase nuclei, which ultimately stabilizes the shapes of various morphogen gradients.

Positional information is a general mechanism for pattern formation in developing organisms: Cell fate specification is determined in a dosage-dependent manner by embryonic morphogen gradients. In the Drosophila syncytial blastoderm, maternal factors establish the axes and set up a system of positional information on which further patterning is built. There is a cascade of gene activity that leads both to the development of periodic structures, the segments, and to their acquiring a unique identity. This cascade involves the binding of transcription factors to regulatory regions of genes to produce sharp thresholds. There are striking similarities in the mechanisms for specifying and recording positional identity in Drosophila and vertebrates.

It has been shown that nuclei and cytoplasm within the Drosophila syncytial blastoderm become organized into independent nuclear/cytoplasmic “protoplasmic islands” known as “energids” (1). It has also been observed that the cortical plasma membrane and various cytoplasmic components—including secretory machinery, Golgi, endoplasmic reticulum, and even cytoplasmic proteins—become “compartmentalized” such that they effectively belong to a single unique nucleus before cellularization (2–4). Despite the concept that the syncytium is essentially a shared sheet of cortical cytoplasm, photobleaching has revealed that diffusion between energids takes place much more slowly than diffusion within individual energids during mitosis (2). This diffusive “compartmentalization” has clear implications for the way in which diffusion-based gradients are formed within syncytial embryos and has begun to be incorporated into morphogen gradient modeling (5, 6). However, despite the fundamental importance of this phenomenon, it remains poorly characterized and so far unexplained. Here, using a combination of mathematical/computational modeling and live-cell imaging, we characterize the effects of dynamic syncytial geometry on the effective diffusivity of cytoplasmic proteins in the Drosophila syncytial blastoderm.

Results

Theoretical Multiscale Diffusion During Mitosis.

The plasma membrane of the Drosophila syncytial blastoderm forms dynamic furrows between neighboring energids as the embryo undergoes nuclear division cycles in the absence of cell division (7). These furrows may potentially affect diffusion through geometry alone. To predict the effect of these furrows on diffusion, we measured the 1D spread of simulated pseudo-Brownian particles within systems mimicking that of the Drosophila syncytial blastoderm (Fig. 1A). Particles were allowed to diffuse in a simulated sheet of cytoplasm 15 μm thick. For our characterization, we varied the spacing between furrows (ω = 5 μm, 10 μm, and 20 μm) and the length of the furrows (L = 0–90% of the system height in increments of 10%). To initiate the simulations, all particles were randomly placed at the same lateral position (x = 0, 0 < y < h) and subsequently allowed to diffuse by pseudo-Brownian motion, and their distribution was characterized as a function of elapsed time. In the case of ideal diffusion, particles adopt a Gaussian distribution (Fig. 1B, Left) with a variance that increases linearly with time (8), the slope of which is proportional to the effective diffusion coefficient according to Eq. 4 (Materials and Methods). Indeed, our simulations revealed that the Gaussian distribution of particles in systems with no furrows (L = 0%) exhibited time-dependent variances where D_C = D_eff (Fig. 1C; L = 0%). When furrows were introduced into the system, particle distribution appeared non-Gaussian (Fig. 1B, Right; here L = 90%), and the temporal increase in the variance of the distribution exhibited three distinct regimes (Fig. 1C). At early times, when the width of the particle distribution is less than the spacing between furrows, the log-log slope of the σ² vs. t was approximately unity, and the intercept corresponded to the cytoplasmic diffusion coefficient, D_C. At intermediate times, the slope became less than unity, and the distribution became largely non-Gaussian (Fig. 1B, Right). At long times, when the spacing between furrows became much smaller than the width of the particle distribution, particles readopted a Gaussian distribution, and the slope of the variance once again became unity on the log-log plot, albeit shifted downward, with the intercept proportional to the effective “long-range” diffusion coefficient, D_eff. This multiscale effect was found to have a strong dependence on furrow length (Fig. 1D). Linear plots show the drastic effect of the furrows on the slope of σ² vs. t, which is directly proportional to the effective diffusion coefficient (Fig. 1E). Thus, our simulations predict that the presence of furrows within the syncytial blastoderm causes a multiscale effect below the percolation threshold (9) where the effective diffusion coefficient of a molecule depends on the time scale and length scale over which diffusion takes place. There is a quasi-ideal early time regime, an anomalous transition time regime, and a quasi-ideal long-range time regime with an effective diffusion coefficient, D_eff, that is less than the instantaneous cytoplasmic diffusion coefficient, D_C. A similar analysis involving tracking individual particles and using their mean squared displacements (MSDs) to determine the effect of furrows on diffusion yielded equivalent results (SI Materials and Methods).

Fig. 1.

Multiscale diffusion in the mitotic syncytial blastoderm. (A) Simulations were initiated with 100,000 particles randomly positioned along the y axis (t = 0, x = 0, 0 < y < h) that were allowed to diffuse by pseudo-Brownian motion at t > 0. (B) For ideal diffusion (no furrows), particles adopted a Gaussian distribution with a variance that increased linearly with elapsed time. For obstructed diffusion (furrows present, L = 90% shown here), particles exhibited deviations from a Gaussian distribution, which spread more slowly with time. (C) The dependence of σ² on time exhibits a slope of 1 on a log-log plot for ideal diffusion (black). The multiscale effect (pink) exhibits pseudoideal diffusion (slope ∼ 1 on log-log plot) for short and long time scales. The downward shift at long time scales reflects a decrease in the effective diffusion coefficient. (D) Simulation results show the effects of furrow length and furrow spacing on the spreading of particles. (E) The slopes of σ² vs. t (apparent on linear plots) are directly proportional the effective diffusion coefficient of the particles. (F) σ²/2t, which indicates the effective diffusion as a function of time, illustrates the decrease in effective long-range diffusion attributable to the presence of furrows. (G) Effective diffusion coefficients, D_eff, exhibit an inverse dependence on furrow length, the effect of which is more severe for closely spaced furrows. Simulation results are shown as data points, and the phenomenological model is shown as solid lines.

The effective diffusion coefficient, as determined by Eq. 6 (Fig. 1F), plateaus over long time scales, indicating that the effect of the furrows is below the percolation threshold and for all system geometries (9). The effective diffusion coefficient, D_eff, was quantified by linearly fitting our simulation data to Eq. 4. For the variable system geometries considered in our simulations, we found the effect of furrow length and spacing on D_eff to approximately obey the phenomenological equation

where L is the furrow length (%), ω is the furrow spacing (μm), and ω* is a characteristic length scale of 1 μm. Thus, D_eff exhibits an inverse dependence on furrow length, the effect of which is more severe for closely spaced furrows (Fig. 1G).

Fluorescence Recovery After Photobleaching (FRAP) Simulations.

To assess the predicted apparent diffusion coefficient, D_app, we simulated FRAP experiments with varying furrow spacing and furrow width (as above). We distributed “bleached” molecules randomly between two furrows and measured the number of these bleached molecules between the same two furrows as a function of time (Fig. 2A). As expected, the number of bleached particles decreased with time due to lateral diffusion in all cases (Fig. 2B). Because the total concentration of bleached and unbleached molecules is not affected by photobleaching and is therefore spatiotemporally constant during the FRAP simulation (10, 11), we inferred the number of visible molecules as the difference between total and bleached molecules according to Eq. 14 (SI Materials and Methods). Assuming the “intensity” within the original energid is proportional to the number of visible molecules it contains, we thus used Eq. 15 to generate simulated normalized FRAP curves (Fig. 2C).

Fig. 2.

Simulated FRAP. (A) Simulations were initiated with 100,000 bleached particles randomly positioned within an energid (−ω/2 < x < ω/2, 0 < y < h), and particles were allowed to diffuse by pseudo-Brownian motion at t = 0. (B) The number of bleached particles in the energid decreased as a function of elapsed time. (C) The number of visible molecules in the energid can be inferred by assuming a constant total average concentration (bleached plus visible molecules) of molecules within the energid. This number was then used directly to represent visible intensity within the energid as a function of time. (D) Apparent diffusion coefficients, D_app, from simulated FRAP fitting exhibit an inverse dependence on furrow length, the effect of which is more severe for closely spaced furrows.

Qualitatively, it is clear that the presence of furrows markedly decreases both the apparent kinetics of recovery and the apparent immobile fraction. To quantify the predicted effect of the furrows on FRAP recovery, we fit the simulated recovery curves to Eq. 15 for recovery within a rectangular photobleached region (12). Notably, the implicit assumptions of Eq. 15 [in particular, complete photobleaching, no immobile fraction, uniform regions of interest (ROIs), ω << system length] are fully satisfied in our simulations. Given the fact that most of these assumptions cannot be satisfied experimentally (13), our analysis provides a conservative estimate of the effect of furrows on actual FRAP experiments. Similar to the results of the particle-spreading simulations, the apparent diffusion coefficient, D_app, exhibits an inverse dependence on furrow length, the effect of which is more severe for closely spaced furrows (Fig. 2D). It is also worth noting that, despite the fact that there is no immobile fraction in these simulations, the apparent extent of recovery in our simulations appears significantly reduced because of the presence of furrow (Fig. 2C). Thus, aside from actually affecting effective diffusion rates, the furrows may also lead to the (incorrect) observation of an immobile fraction for a fully mobile molecule, especially in the case of larger photobleached areas.

Membrane Furrow Dynamics.

With a predictive understanding of how furrow geometry theoretically affects diffusion within the syncytium, we sought to directly measure these parameters by using fluorescence microscopy. We first characterized the spatiotemporal geometry of the system by recording time-lapse movies of the fluorescent membrane reporter protein Spider::GFP (Fig. 3A). The plasma membrane undergoes drastic reorganization upon the arrival of nuclei to the cortex. Membrane furrows become visible between nuclei around nuclear cycle 10 and extend most prominently into the cytoplasm during mitosis, with the greatest precellularization extension length found during mitosis of cycle 13. After the interphase of nuclear cycle 14, the membrane furrows extend deeply into the cytoplasm and cellularize the syncytium. The thickness of the cortical layer that is accessible to diffusion, as assessed by the depth of cytosolic protein, was ∼15 μm and did not vary significantly between nuclear cycles (SI Materials and Methods). Furrow lengths were determined by measuring the distance from the basement vitelline membrane to the tips of the furrows (Fig. 3B). The average spacing between furrows, which decreases with time during the syncytial blastoderm phase in concert with the decrease in average nuclear dimensions, was also measured during mitosis in each nuclear cycle (Fig. 3C).

Fig. 3.

Membrane furrow dynamics. (A) Cross-sections of early embryos expressing Spider::GFP reveal drastic reorganization of the plasma membrane through nuclear cycles 10–14. Transient membrane furrows are most prominent during mitosis (M) and significantly recede during interphase (I). (B) Furrow length, measured as the average minimum distances between the cortical membrane and furrow tips, oscillates according to the cell cycle, exhibiting a maximum during nuclear cycle 13. The increase after interphase 14 represents cellularization. (C) The spacing between furrows decreases as the density of nuclei increases with each progressive nuclear cycle. Furrow spacing was calculated by taking the average minimum distances between adjacent furrow tips. (D) The overall effect of furrow length and furrow spacing was determined as a function of development time. We denoted the maximum furrow length during mitosis 13 to be t = 0. (Scale bar: 20 μm.)

Our simulation data (Fig. 1G, data points) was well represented by Eq. 1, which allowed us to use experimental measurements of furrow length (Fig. 3B) and furrow spacing (Fig. 3C) to estimate the overall predicted effect of membrane furrows on bulk diffusion as a function of development time (Fig. 3D). We predict an oscillatory reduction in effective diffusion with a large effect during mitosis, which becomes increasingly severe with each successive nuclear cycle, having a significant effect during the mitosis of nuclear cycle 13.

Transient Compartmentalization.

Because the effect of membrane furrows on diffusion is expected to be greatest during mitosis, we predict that diffusive compartmentalization between energids should correlate similarly. To assess this prediction, we performed repeated fluorescence loss in photobleaching (FLIP) of circular ROIs on the cortical surface of the syncytial blastoderm expressing fluorescently tagged cytoplasmic proteins and assessed compartmentalization. As expected for a compartmentalized system, FLIP resulted in jagged boundaries (Fig. 4A) and steep fluorescent discontinuities at energid–energid interfaces (Fig. 4C) during mitosis, presumably due to the rapid depletion of fluorescence intensity from energids that overlapped with the bleach ROI. However, as the cell cycle progressed out of mitosis, the fluorescence intensity profile from the bleach ROI appeared symmetrical and smooth, as expected for unobstructed diffusion (Fig. 4B), and the steep fluorescent transitions at energid boundaries were lost (Fig. 4D). Thus, we only found evidence of diffusive compartmentalization between energids during metaphase. Although the asymmetric distribution of Dorsal within the blastoderm complicates precise quantitation of its diffusion rate with FLIP, equivalent results were also found for uniformly distributed fluorescently labeled elongation factor (ELF::GFP).

Fig. 4.

Diffusive compartmentalization is transient and depends on the cell cycle (FLIP). (A and B) Continued photobleaching within a circular ROI (dashed line) on the cortex of an early embryo expressing Dorsal::GFP results in the formation of jagged boundaries at energid interfaces, indicative of diffusive compartmentalization during mitosis (A) and smooth, symmetric concentration profiles during interphase (B). (C and D) Intensity profiles across the ROI reveal steep transitions in fluorescence at energid interfaces (arrows) during mitosis (C) and smooth transitions in fluorescent during interphase (D). The photobleached ROI is shown in red in the intensity profiles. (Scale bar: 20 μm.)

Discussion

Particle Simulations and Transient Compartmentalization.

It has previously been shown that the cytoplasmic morphogen Dorsal partitions into nuclear/cytoplasmic protoplasmic islands (1) commonly referred to as energids. Photobleaching experiments show that the protein exhibits rapid diffusion within an individual energid but only slow diffusion between energids, suggesting some type of diffusive compartmentalization within a system of nuclei sharing a common cytoplasm. Photobleaching experiments further show that the lateral mobility of the morphogen appears more restricted adjacent to the cortex of the early embryo and more freely mobile toward the deep cytoplasm, which is compatible with cortically anchored membrane furrows restricting lateral diffusion (2). The effect should also be expected to appear more pronounced in confocal photobleaching close to the cortex of the embryo (at or below the depth of the nuclei), which may also explain why this compartmentalization phenomenon has not been observed in the deep cytoplasm.

Using computational simulations based on experimentally measured parameters, we predicted from first principles the effect of membrane furrows on effective diffusion with the early embryo. We placed an ensemble of particles at a unique lateral position (x = 0) within a periodic system mimicking the early embryo, with varying furrow length, L, and furrow spacing, ω (Fig. 1), and allowed the particles to “diffuse” by simulated Brownian motion. The distribution of the particles was fit to a Gaussian distribution according to Eq. 3, and the variance, σ², was measured as a function of elapsed time. Eq. 6 was ultimately used to infer the effective diffusion coefficient as a function of elapsed time. We found that, unlike ideal diffusion, σ² did not exhibit a linear dependence on elapsed time when furrows were present in the system (Fig. 2D). Consequently, the inferred diffusion coefficient depends on the time scale (and, accordingly, the length scale) over which diffusion takes place, and it was found to plateau at a long-range effective diffusion coefficient smaller than the cytoplasmic diffusion coefficient (Fig. 1F). Similar results were found by assessing the predicted apparent diffusion coefficient, D_app, in simulated FRAP experiments (Fig. 2).

The effect of furrow length, L, and furrow spacing, ω, on the effective diffusion coefficient was found to approximately obey Eq. 1. Using experimentally measured values for L (Fig. 3B) and ω (Fig. 3C), we predicted the furrows to have a nuclear cycle-dependent, oscillatory, and decreasing effect on effective diffusion shown (Fig. 3D). To check our prediction that the degree of diffusive compartmentalization should accordingly correlate with the phase of the nuclear cycle, we repeatedly photobleached a lateral ROI of the cortical cytoplasm and monitored the extent to which we could observe compartmentalized behavior. Indeed, we observed transient diffusive compartmentalization (jagged energid boundaries and steep fluorescence discontinuities), with a pronounced effect during mitosis and no visible effect during interphase (Fig. 4). Thus, our data demonstrate that the previously described diffusive compartmentalization phenomenon is nuclear cycle-dependent and support the prediction that furrows temporally regulate effective rates of the diffusion in the syncytium.

Implications for the Bicoid Gradient.

A common general model for the establishment and maintenance of embryonic morphogen gradients relies on reaction and diffusion, often involving synthesis from a localized source, diffusion throughout the embryo, and degradation. Although this type of synthesis/diffusion/degradation (SDD) model has been proposed for the Bicoid gradient in Drosophila melanogaster decades ago (14), specifics regarding this model remain a matter of controversy. The length scale, λ, of the SDD model depends on the effective diffusivity, D_eff, and the degradation constant, α, of the protein according to the equation

Quantitative assessment of the actual diffusion coefficient of Bicoid has remained controversial, with some recent estimates differing by several orders of magnitude (5, 15–17). Previous FRAP experiments have suggested that the diffusion coefficient is too small for the simple SDD model to explain the Bicoid gradient (15). These experiments were performed during mitosis 13, when furrows are most prominent and also very closely spaced, therefore causing a drastic effect on effective diffusion (Fig. 3). Fluorescence correlation spectroscopy measurements, which essentially measure the cytoplasmic diffusion coefficient and should not be expected to be affected by membrane furrows, have suggested a high cytoplasmic diffusion coefficient for Bicoid (16). Our multiscale modeling provides an explanation for the discrepancy between the physiological diffusion coefficient of Bicoid measured by these different techniques. It is also worth noting that previous FRAP experiments did not account for the “halo effect,” which arises because the photobleaching step is not actually instantaneous (17, 18). The previous photobleaching duration of ∼5 s (15) most likely caused large systematic errors in the assessment of the actual diffusion coefficient by FRAP, which ultimately resulted in underestimates of the effective diffusion coefficient of Bicoid.

On the basis of our model, one may expect the predicted change in effective diffusivity throughout the syncytial blastoderm stage to cause a corresponding change in the length scale of the Bicoid gradient as well. However, the constant length scale of the Bicoid gradient (15) suggests that the syncytium is able to compensate for these nuclear cycle-dependent effects. It has recently been predicted (19) that the presence of nuclei should have a drastic effect on the effective bulk diffusion of morphogens because of geometric and nucleoplasmic/cytoplasmic shuttling effects, causing bulk diffusion rates to be reduced during interphase (2). Our results provide an explanation for how the syncytium is able to maintain constant length scales of diffusion-mediated gradients despite the multiscale reduction in effective diffusion introduced by interphase nuclei. The presence of mitotic membrane furrows temporally regulates (reduces) long-rage rates of diffusion that balance the effect of interphase nuclei. The predicted reduction in effective diffusion by both interphase nuclei and mitotic furrows becomes more severe during each nuclear cycle in the syncytial blastoderm, resulting in a stabilized change in effective diffusion over time, further suggesting a reduction in the degradation rate constant, α, with time in the syncytial blastoderm. Thus, we propose that an as-yet-unidentified role of syncytial membrane furrows may be to temporally regulate bulk embryonic diffusion rates to keep the shapes of various morphogen gradients constant. It may also be interesting to characterize the diffusion rates of recently reported nuclear import-deficient mutants of Bicoid that seem to indicate that nuclei do not shape the Bicoid gradient but rather function solely during its interpretation (20). Although the readout of the Bicoid gradient has been shown to occur early in the syncytial blastoderm (21), our analysis of the stability of the Bicoid gradient at later cycles in the syncytium has implications for any diffusion-based gradient.

Materials and Methods

Simulated Effective Diffusion.

All custom code was written for Monte Carlo simulations and curve fitting using MATLAB 7.8.0 (R2009a) (MathWorks).

Morphogen diffusion in the Drosophila early embryo takes place in a cortical sheet of cytoplasm ∼15 μm deep (SI Materials and Methods). To assess the effect of membrane furrows on long-range diffusion across the syncytial blastoderm, we approximated the system as diffusion between parallel plates (membranes) and considered diffusion in one lateral dimension, x (Fig. 1A). In our simulations, we allowed particles to independently and randomly “walk” (diffuse) in two dimensions to approximate Brownian motion and ultimately monitored their 1D lateral spread in the presence of furrows of varying length and spacing. In the absence of obstacles, particles adopt a Gaussian distribution that spreads with elapsed time (Fig. 1B), according to the normalized probability density function

where D_eff is the effective (bulk) diffusion coefficient of the particles and t is time (8). The variance of this distribution increases linearly with time according the equation

Taking the logarithm of both sides of Eq. 4, we obtain

and thus, for this ideal case, the slope of log(σ²) vs. log(t) will be unity. The intercept is determined by the diffusion coefficient, and therefore a downward shift on the log-log plot indicates a decrease in effective diffusivity (Fig. 1C).

We can directly assess effective diffusion by rearranging Eq. 4 to obtain

and thus

which should both be constant for ideal diffusion.

We modeled the volume accessible to diffusion in the syncytium as an infinite space along the x dimension consisting of periodic energids with a height h and individual widths ω (Fig. 1A). For each bulk diffusion simulation, particles were randomly placed along the y axis (x = 0; 0 < y < h) and subsequently allowed to diffuse by pseudo-Brownian motion beginning at t = 0. We simulated Brownian motion by displacing each particle a small, fixed prescribed length in a random direction at each iteration, with the simulation time step calibrated for the appropriate designated instantaneous diffusion coefficient according to the equation

where Δt is the real time interval associated with each iteration of the simulation, Δd is the fixed displacement length of a particle in each iteration, and D_C is the chosen instantaneous diffusion coefficient of the particles. In the case of pure ideal diffusion, the effective diffusion coefficient is equal to the instantaneous (i.e., small-scale) cytoplasmic diffusion coefficient, such that D_eff = D_C (Fig. 1C). Nonideal system geometries, such as that of the syncytium, will have no effect on the instantaneous diffusion coefficient of a molecule but may be expected to decrease its effective diffusivity.

For our simulation geometry, the height h of the system was set to 15 μm, the approximate constant thickness of the cortical cytoplasmic sheet in the Drosophila syncytial blastoderm (SI Materials and Methods). The two major system parameters that were varied in our simulations were the length of the membrane furrows (furrow length, L: varied from 0% to 90% of the system height in increments of 10%) and the spacing between furrows (furrow width, ω: 5 μm, 10 μm, and 20 μm). All membranes were modeled to be completely elastic, such that a particle “bounces” off of a membrane with no loss of kinetic energy. The displacement length of the particles was always at least an order of magnitude less than any system geometry characteristic lengths, most notably the “gap” between the ceiling membrane and the membrane furrows; the gap (L − h) had a minimum simulation value of 1.5 μm. All simulations contained 100,000 particles and used diffusion displacement of <0.1 μm (0.0894 μm), which corresponds to a simulation time step of 0.002 s for D_C = 1.0 μm²/s. Distribution parameters (i.e., σ) were obtained by using the embedded MATLAB (MathWorks) function normfit.

Fly Strains and Live Imaging.

Spider::GFP and Dorsal::GFP Drosophila strains were maintained as previously described (2, 4). Flies were allowed to lay embryos for 1 h at 25 °C on grape juice agar plates. Embryos were dechorionated and mounted on their posterior pole in Lab-Tek chambers [coated with silane (22) for end-on imaging], immersed in PBS, and imaged on an LSM 510 confocal microscope with a Zeiss 40× (N.A. 1.2) A-plan objective. For end-on imaging, embryos were mounted on their posterior pole and imaged as previously described (23). Embryos were imaged at ∼60 μm from the posterior pole, at a position where the diameter of the embryo was ∼140–160 μm. For photobleaching experiments (Fig. 4), circular ROIs of 40 μm in diameter were continuously photobleached with the embedded raster-scanning laser module, and images were acquired every 10 s for analysis. Quantification was performed with LSM Image Examiner software (Zeiss).