Probing events with single molecule sensitivity in zebrafish and Drosophila embryos by fluorescence correlation spectroscopy

Abstract

Zebrafish and Drosophila are animal models widely used in developmental biology. High-resolution microscopy and live imaging techniques have allowed the investigation of biological processes down to the cellular level in these models. Here, using fluorescence correlation spectroscopy (FCS), we show that even processes on a molecular level can be studied in these embryos. The two animal models provide different advantages and challenges. We first characterize their autofluorescence pattern and determine usable penetration depth for FCS especially in the case of zebrafish, where tissue thickness is an issue. Next, the applicability of FCS to study molecular processes is shown by the determination of blood flow velocities with high spatial resolution and the determination of diffusion coefficients of cytosolic and membrane-bound enhanced green fluorescent protein–labeled proteins in different cell types. This work provides an approach to study molecular processes in vivo and opens up the possibility to relate these molecular processes to developmental biology questions. Developmental Dynamics 238:3156–3167, 2009. © 2009 Wiley-Liss, Inc.

INTRODUCTION

To understand cell functions in vivo requires the study of fundamental molecular processes in vivo. Currently, most protein dynamics studies are performed in Petri dish–based cell culture systems. Cell cultures are engineered as isolated individual cells that can be artificially cultivated. The physiological relevance of findings made in vitro remains not clear because many questions of developmental biology cannot be addressed in these simplified models. Thus, techniques that can directly monitor biomolecular behaviors within a living organism are needed. Recent advances in microscopic techniques allow monitoring of single cells within thick tissues of living organisms. Confocal laser scanning microscopy generates high-resolution images of up to 100 microns depth within tissue and two-photon excitation can further increase the penetration depth to several hundreds of microns (Helmchen and Denk,2005). As a result, some animal models, such as nematode (Caenorhabditis elegans), fruit fly (Drosophila melanogaster), zebrafish (Danio rerio), and medaka (Oryzias latipes) can be directly studied using imaging based cell biological methods due to their small size and/or semitransparent body tissues (Beis and Stainier,2006; Lis,2007; Garcia-Lecea et al.,2008; Korzh et al.,2008). However, imaging based techniques have limited spatial resolution and events that happen on the protein level could not be seen directly. For this reason, indirect biophysical approaches, such as fluorescence correlation spectroscopy (FCS), were introduced to probe molecular dynamics in live cells (Kim et al.,2007). FCS analyzes fluorescence signals from a very small observation volume (∼10⁻¹⁵ L) and determines local concentrations and diffusion coefficients of fluorescently labeled molecules. The combination of imaging techniques with FCS thus allows quantitative analysis of protein dynamics within subcellular compartments (Terry et al.,1995; Brock and Jovin,1998; Pan et al.,2007a). Up to now, FCS application in living animals is still limited due to the thick tissue induced light scattering. In one example, Nagao and others reported diffusion coefficient measurements of green fluorescent protein (GFP) -labeled granules in medaka primordial germ cells using FCS and fluorescence recovery after photobleaching (FRAP) (Nagao et al.,2008). To avoid the deep tissue penetration, the medaka embryos were dissected and cells of interest were revealed. In contrast, working with much smaller animals, Petrasek and others applied scanning FCS to study the localization and redistribution of GFP-labeled NMY-2 and PAR-2 proteins during the asymmetric first division of C. elegans embryos (Petrasek et al.,2008). Working with transparent animals helps to alleviate this problem too. Recently, it has been shown that FCS can be used efficiently to measure blood flow velocities in live zebrafish embryos (Pan et al.,2007b; Korzh et al.,2008) and even to quantify protein–protein interactions (Shi et al.,2009).

Zebrafish and Drosophila can be used as models to study different biological questions. In Drosophila, a wide repertoire of genetic techniques can be implemented. Genetic manipulation by means of gain- or loss-of-function, and targeted expression of a desired gene product can be introduced in nearly any tissue or group of cells by means of the Gal4-UAS system (Brand and Perrimon,1993). It is possible by these means to genetically modify lipid metabolism or other conditions such as susceptibility to oxidative stress, which may alter membrane properties in vivo. Zebrafish, as a model of vertebrate development, has been widely used only relatively recently, but it is fast catching up as many methodologies developed in Drosophila are transferred to zebrafish research. As a model species, it is more complex, evolutionarily closer to humans, and amenable to standard genetic and molecular tools. Numerous human diseases, both genetic and acquired, can be introduced and studied in zebrafish, which made it a model vertebrate of choice for drug discovery and large-scale studies of genetics, development, and regeneration (Strahle and Korzh,2004; Korzh,2007; Lieschke and Currie,2007; Fetcho et al.,2008). The optical clarity of zebrafish embryos also allows investigation of cells deep within the tissue. In this study, we demonstrate several application of FCS in live zebrafish and Drosophila embryos showing that studies with single molecule sensitivity within an organism are possible. The example applications are general in nature and can be easily extended to other small organisms.

RESULTS AND DISCUSSION

Autofluorescence Study

One advantage of FCS is the extreme sensitivity down to the single molecule level (Rigler et al.,1993). This requires high signal to noise ratio of the sample and minimal background interference. However, autofluorescence in live cells and tissues is inevitable and it is generally considered as one of the major problems encountered in intracellular FCS applications. The contribution of autofluorescence to the signal to noise ratio of FCS measurement has been previously discussed (Brock et al.,1998; Schwille et al.,1999). Here, we first examined the autofluorescence in live zebrafish embryo, the distribution in the embryo body and its emission spectrum, aiming to minimize the autofluorescence interference. We focused on embryos around 3 days postfertilization (dpf) and subsequent FCS measurements were performed at the same stage. An LSM 510 Meta (Carl Zeiss, Singapore) confocal microscope was used for all autofluorescence studies. To obtain an overview of the autofluorescence distribution, a ×10 objective (Plan-Neofluar/NA=0.3, Carl Zeiss) and long-pass 505 emission filter were chosen to record images of a 3 days postfertilization (dpf) wild-type embryo excited by 488-nm laser line (100 μW) as shown in Figure 1A. The autofluorescence was elevated in the eye (arrow 1) and yolk (arrow 2), whereas at the trunk region (arrow 3) autofluorescence is much lower (Fig. 1C). FCS works better with low probe concentration and an upper limit of probe concentration is usually 100 nM in intracellular applications. In our experiments, the EGFP fluorescence intensity is controlled within 200 kHz with 30 μW 488-nm laser excitation. However, autofluroescence intensity at zebrafish yolk region can be several times higher than 200 kHz at the same experiment condition and it can mask the EGFP signals, which makes FCS analysis difficult. To minimize the influence of autofluorescence, the subsequent FCS measurements were conducted in the trunk region posterior to midbrain, where the main tissues are the epidermis, muscle, vascular and neuronal system. We further studied the autofluorescence spectrum of muscle fibers in the trunk region using the LSM 510 Meta Lambda Mode. A ×40 objective (Plan-Neofluar/NA=0.75, Carl Zeiss) was used for this purpose and the spectrum was recorded from 505 nm to 719 nm in steps of 10.7 nm. Figure 1D shows the normalized autofluorescence spectrum together with the EGFP spectrum and the 510AF23 emission filter (Omega Optical, Brattleboro, VT) used for FCS measurements. The autofluorescence peaked at 570 nm and 630 nm, which is distinct from EGFP (maximum at 509 nm). Thus, the 510AF23 emission filter can efficiently block most of the autofluorescence signal and pass the EGFP signal.

Figure 1

Zebrafish embryo autofluorescence study. A: Overview of autofluorescence expression in lateral view 3 days postfertilization (dpf) AB wild-type embryo, excited with 100 μW 488-nm laser, long-pass 505 as emission filter. Overlay of transmission and fluorescence channel. The embryo was treated with PTU at 18 hpf to prevent pigmentation. B: Enlarged view of the square region in A. C: Fluorescence intensity trace of the red line in A. D: Normalized autofluorescence spectrum taken in the trunk region (yellow cross in B), shown together with enhanced green fluorescent protein (EGFP) fluorescence spectrum and 510AF23 emission filter. Scale bar = 200 μm.

We also examined the autofluorescence in Drosophila embryos. The Drosophila embryo is relatively opaque (Huisken et al.,2004). To measure cells within tissue, the embryo was dissected. Our confocal images suggest that the autofluorescence in the Drosophila embryo is mostly concentrated in the gut. Here, we performed FCS measurements on the aCC/RP2 motor neurons in the nerve cord (Goodman and Doe,1993) aiming to investigate membrane fluidity. To avoid interference by strong autofluorescence and light scattering and distortion of the confocal volume by extraneous tissue, before measurement the gut was removed to reveal nerve cord during sample preparation (see the Experimental Procedures section). The aCC/RP2 motor neurons are on the dorsal most surface of the ventral nerve cord. After sample preparation, the aCC/RP2 motor neurons within the nerve cord lie less than 5 μm from the glass surface. Control experiments using a live/dead fluorescence stain were performed to ensure that the dissected embryos were still alive at the time of FCS measurements (Supp. Fig. S1, which is available online).

We finally performed FCS measurements with autofluorescence alone in both zebrafish and Drosophila embryos. The signal to noise ratio of FCS is determined by the count rate (number of photons detected) per molecule per second (cps) of the fluorophore (Koppel,1974). The observation volume was positioned in non-EGFP expressing muscle fibers and motor neurons in zebrafish embryos as well as non-EGFP expressing motor neurons in Drosophila embryos, and no autocorrelation was observed with autofluorescence alone (Supp. Fig. S2). In contrast, FCS measurements in EGFP expressing cells in both zebrafish and Drosophila embryos generate robust autocorrelations with cps values around 6 kHz at the same laser excitation power of 30 μW. Autofluorescence did not generate any autocorrelations and thus contributes only as background noise. This background signal leads to an overestimation of the concentration of the target molecules, because it influences the amplitude of the autocorrelation function (ACF), but it does not affect the width of the ACF and, thus, the temporal information. In addition, the autofluorescence intensity (defined as count rate per second) recorded by the avalanche photodiode detector is 0.6–1 kHz in the zebrafish embryo trunk region (excluding skin and vascular system) and around 2 kHz in the Drosophila nerve cord, compared with 0.2–0.6 kHz in buffer solutions. Autofluorescence contributes less than 5% to the overall intensity in EGFP-expressing cells in zebrafish embryo and around 10% in Drosophila embryos.

In terms of autofluorescence interference, the optically transparent zebrafish embryo is advantageous. We noticed that all strong autofluorescence observed in the zebrafish embryo arose from the less transparent regions, including eyes and yolk. By choosing for measurements the more transparent trunk region, the influence of autofluorescence was minimized and experiments can be done in vivo leaving the embryo intact at the end of measurements. Therefore working with larvae of fish strains where pigmentation is suppressed genetically (e.g., albino) or by chemicals (PTU), such measurements could be repeated during development of the same animal several times. In Drosophila, the removal of the gut removes the main source of autofluorescence. Furthermore, the autofluorescence in zebrafish and Drosophila embryos does not increase over the 3 to 4 hr of measurement time, contrary to that in cell cultures (Brock et al.,1998).

Penetration Depth Study

For Drosophila embryos, the dissected embryo with exposed nerve cord is mounted onto the cover glass, which places the aCC/RP2 motor neurons within a depth of 5 μm from the glass surface. For zebrafish embryos, although the optical transparency aids FCS applications, the thick tissue limits the working distance within the embryo at which the measurements could be performed efficiently due to the strong and multiple light scattering. We used an Islet-1-EGFP transgenic line to study the maximum penetration depth of FCS measurements. In this transgenic line, the neural-specific Islet-1 promoter/enhancer drives EGFP expression in primary neurons, including a subset of sensory neurons and motor neurons (Higashijima et al.,2000; Fig. 2A). The confocal image showed that fluorescence intensity of EGFP expressed vagal motor neurons drops to only 15% as the image plane progresses from a surface 20 μm to 120 μm within the embryo body (Fig. 2B). This demonstrates the decrease in photon collection efficiency of a confocal setup with penetration depth. To examine the working distance of FCS, we measured the diffusion time τ_d of cytosolic EGPF in zebrafish motor neurons at different cell depths and compared the measurement accuracy. The cell depth is determined by a z-scan from the FCS measurement point toward the bottom skin of the embryo indicated by a fluorescence intensity change when the focus shifts from embryo skin to agarose. Due to the high EGFP expression level, the motor neurons were prephotobleached for 10 sec with 1 mW 488-nm laser. To confirm that photobleaching does not change the diffusion time, we photobleached the same cell for 5, 10, and 15 sec and obtained the same diffusion time (0.26 ± 0.03 ms). It should be noted that FCS measurements should avoid regions close to the heart as the body movement induced by the heartbeat introduces an additional periodic time trace into the ACF curve making analysis difficult (Supp. Fig. S3). We measured 30 cells at different depths and the acquired τ_d values were then plotted against its depth as shown in Figure 2C. Each point on the graph represents the mean of five individual measurements within the same cell. Generally, if the cell is located within 50 μm inside the embryo, the τ_d values are constant within the margins of error and measurements have a small standard deviation (SD). When the cell is located between 50 to 80 μm, the τ_d has a slightly increased average value but with a noticeably larger SD. If the cell is located beyond 80 μm, the autocorrelation is lost in almost all measurements and no parameters can be obtained (of five cells with cell depth over 80 μm measured, only one could be analyzed). Based on these results, we conclude that FCS can be applied for accurate measurements as deep as 50 μm inside the embryo using one-photon excitation (OPE). FCS can still provide meaningful data with penetration depth between 50 and 80 μm and sufficient number of measurements should be conducted in this case for statistical analysis. In addition, the obtained cps values were also plotted against cell depth as shown in Figure 2D. The average cps of FCS measurements within 50 μm is ∼ 6 kHz, while beyond 50 μm it drops to only ∼2 kHz. This again suggests that accurate FCS measurements were limited within 50 μm depth. The limited working distance is restricted by strong and multiple light scattering and spherical aberrations induced by the heterogeneous refractive index of tissues (Hell et al.,1993), as cells comprise different subcompartments and macromolecules, which make the cell optically nonhomogeneous. Furthermore, light scattering particularly affects the signal to noise ratio in confocal illumination, which achieves optical sectioning with a detection pinhole that rejects out-of-focus light. In scattering tissues, the fluorescence light of interest will be scattered and thus will be partially blocked by the pinhole itself. It has been suggested that confocal microscopy can work at most 100 μm within biological tissue using OPE (Denk et al.,1990). So the range we explored here of up to 80 μm is probably a realistic estimate of the accessible range with FCS using OPE. A slight extension of penetration might be possible with better fluorophores.

Figure 2

Penetration depth study using one-photon excitation. A: Dorsal view of vagal motor neurons of 3 dpf embryo of Islet-1-EGFP transgenic line as indicated in the dashed ellipse. B: Fluorescence intensity measurements of EGFP-expressing motor neurons at different depths. C: Diffusion time measurements by fluorescence correlation spectroscopy (FCS) at different cell depths. D: cps of FCS measurements at different cell depths. Scale bar = 50 μm.

Two-photon excitation (TPE) has been reported to reduce light scattering, autofluorescence and phototoxicity for intracellular applications (Denk et al.,1990). It has been rapidly adapted for FCS and successfully applied for measurements made on tobacco leaf epidermal tissues with thick cell wall (Schwille et al.,1999). We decided to evaluate whether TPE could be used for FCS measurements in zebrafish embryo at working distances deeper than 80 μm. We used a Ti:sapphire laser (100-fs pulse) in our FCS setup. The system was first calibrated and optimized using a standard fluorescence dye, tetramethylrhodamine (TMR). A cps of 14 kHz was obtained for TMR, which is similar to other reports (Schwille et al.,1999; Mutze et al.,2007). The excitation wavelengths were optimized to 840 nm for TMR and 890 nm for EGFP (Supp. Fig. S4). We repeated the measurement of cytosolic EGFP diffusion time versus cell depth using TPE (Fig. 3A). However, TPE did not improve the penetration depth in the case of EGFP in our setup. The determination of diffusion times worsened beyond 80 μm. It seems that TPE-FCS did not improve the working distance in living embryos in this case partially owing to two reasons. The first reason is the use of a de-scanned detection scheme. Under de-scanned detection, emission light will be reflected back from the scanning mirrors and pass several lenses and the pinhole before detection. Photon collection efficiency is low compared with non–de-scanned detection schemes so the potential of TPE was not fully explored (Le Grand et al.,2008). The second and more important reason is the inefficiency of EGFP as a TPE dye. The low fluorescence yield of EGFP using TPE dramatically decreased the signal to noise ratio of FCS measurements. The average cps of EGFP obtained in embryos using TPE is only ∼0.5 kHz compared with ∼6.0 kHz using OPE. The low molecular brightness of EGFP using TPE has so far been attributed to photobleaching (Dittrich and Schwille,2001; Petrasek and Schwille,2008) and saturation (Berland and Shen,2003). In an attempt to improve in this regard, we injected TMR into the embryo's blood circulation and accurately determined the blood flow speed up to ∼200 μm within tissue (Fig. 3B), confirming that TPE provides better penetration depths when using appropriate dyes. It should be noted that the penetration depth is dependent on the opacity of the embryo tissue, which itself is variable at different embryonic stages. Using more optically transparent strains of fish such as albino, nacre, or casper mutants (Henion et al.,1996; Lister et al.,1999) could improve penetration depth due to reduced pigmentation. Using fluorophores with higher molecular brightness and photostability can also increase the signal to noise ratio of FCS measurements and subsequently lead to deeper penetration depth.

Figure 3

Penetration depth study using two-photon excitation. A: Diffusion time measurements at different cell depths. B: TPE‒ fluorescence correlation spectroscopy (FCS) flow measurements of injected TMR in the blood vessel. This measurement was conducted at cell depth ∼ 200 μm. At the bottom is the intensity trace of the measurement.

Blood Flow Measurement in Live Zebrafish Embryos

Blood flow measurements play an important role for studies of vascular development. Mechanical signals such as shear stress induced by blood flow is essential for normal development of organs involved in circulation (Hove et al.,2003; Korzh et al.,2008). Currently, several approaches have been implemented to measure flow velocities in small animal vessels, e.g. fast confocal laser scanning microscopy (Lucitti et al.,2007), optical coherence tomography (Larina et al.,2009), and laser speckle imaging (Cheng et al.,2003). However, those imaging based methods provided data characterized by good temporal but low spatial resolution, and a high level of spatial resolution is a key factor to understand the local changes of blood flow in the vessel. FCS could be an alternative choice for blood flow velocity measurement. We have previously shown that FCS can measure flow velocities with a spatial resolution of 0.5 μm and determine flow directions in two and three dimensions (Pan et al.,2007b,2009). As an example, we characterize the spatial flow profiles in the trunk blood vessels in 3 dpf zebrafish embryos. For this purpose, an embryo was mounted in a lateral view, and FCS measurements were performed in the dorsal aorta and cardinal vein in the trunk region (Fig. 4A). We found that the autofluorescence present in the serum generates sufficient cps for FCS analysis thus no additional dye injection into the circulation is required. Because this can be accomplished independently of erythrocytes simply based on autofluorescence of serum, FCS can be used to measure blood flow in vessels during embryogenesis when red blood cells are not present. A typical ACF curve obtained in the cardinal vein is shown in Figure 4B. For data fitting, a simple one-flow model is not adequate to describe blood flow in animal as there is alternating fast and slow blood flow in the cardiac cycle stemming from systolic (heart contraction, peak pressure occurs during systole period) and diastolic (heart relaxation, lowest pressure occurs during diastole period) pressure. Therefore, the ACF curves were fitted with a three-dimensional two-flow model (Eq. 8) and the fast and slow portions represent average systolic and diastolic flows, respectively (Pan et al.,2007b). Due to the fast changes in blood flow velocities a better resolution of the flow in full profiles is not possible with the current data and only average velocities for systole and diastole are given. FCS measurements were performed across the whole diameter of a vessel in steps of 1.5 μm. The flow velocities were then plotted against positions inside the vessel. The dorsal aorta is around 15 μm in width, and the cardinal vein is around 19 μm. A comparison of systolic flow velocities in both dorsal aorta and cardinal vein is shown in Figure 4C and the diastolic flow velocities in Figure 4D. As demonstrated, the flow velocity has a maximum in the center of the vessel for both diastolic and systolic flow and the flow velocities are close to zero near the vessel walls. The parabolic profile fits the prediction of the theoretical calculation of flow velocity in a cylindrical tube (Chien et al.,1977) and the values obtained here are comparable to other reports (Malone et al.,2007). The figures also suggest that the velocities of both systolic and diastolic flow in the dorsal aorta are faster than those in the cardinal vein. This is expected due to the different levels of cardiac-mediated blood pressure in these vessels and according to their widths.

Figure 4

Blood flow measurements in live zebrafish embryos. A: Confocal image of the dorsal aorta and cardinal vein, 3 days postfertilization (dpf) AB, lateral view. B: A typical FCS flow measurement in the cardinal vein, showing both experimental curve (red) and fitting curve (black). At the bottom are the residuals of the fit. Experimental curves are fitted with a two flow model representing systolic and diastolic flow. C,D: Blood flow velocity profiles of systolic and diastolic flow, respectively. Results show measurements taken in the dorsal aorta (black) and cardinal vein (red).

Diffusion Coefficient Measurements in Live Zebrafish Embryos

Energy-independent random diffusive motion is the most common mechanism for the translocation of intracellular proteins. Thus accurate measurements of protein diffusion coefficients are important for our understanding of protein function. Here, we demonstrated that FCS can accurately measure diffusion coefficients of cytosolic EGFP and EGFP tagged membrane proteins in live Drosophila and zebrafish embryos.

In zebrafish embryos, we injected EGFP plasmid into a single blastomere at the 16-cell stage first and observed the fluorescence expression at both muscle fibers (Fig. 5A) and motor neurons (Fig. 5B). The mosaic distribution of cell descendants of injected blastomere provides a range of cell types with different EGFP expression levels to choose from. FCS measurements were performed in both muscle fibers and motor neurons that are within 50 μm from the body surface. The ACF curves suggested slightly different diffusion behavior (Fig. 5D). The average diffusion time τ_d obtained was 0.29 ± 0.06 ms in motor neurons and 0.36 ± 0.04 ms in muscle fibers. Measurement of the reference dye fluorescein in aqueous solution, with a diffusion coefficient of 300 μm²s⁻¹ (Wohland et al.,1999), yields a τ_d of 43.7 μs, and the corresponding EGFP diffusion coefficients can be estimated to be 45.3 ± 9.4 μm²s⁻¹ in motor neurons and 36.5 ± 4.1 μm²s⁻¹ in muscle fibers (Eq. 6). Note that the diffusion coefficients for EGFP inside cells represent lower limits because the focal volume is slightly increased due to the refractive index mismatch between water and tissue. However, the values are very close to typical values found in two-dimensional cell cultures. According to the Stokes-Einstein relation (Eq. 9), the diffusion coefficient of a molecule is inversely proportional to the viscosity of the medium. Thus, the result suggested a higher viscosity of the cytoplasm of muscle fibers compared with motor neurons. Since it has been shown that cancerous cells have increased cellular elasticity and are considerably “softer” than normal cells (Beil et al.,2003; Darling et al.,2007), possibly the value of cytoplasm viscosity could be used as a parameter for characterization of cell types and states in mixed cell populations.

Figure 5

Diffusion coefficient measurements in zebrafish embryos. A,B: Confocal images of enhanced green fluorescent protein (EGFP) expressed muscle fibers (at posterior trunk region) and motor neurons (at hindbrain/spinal cord junction region), respectively. Lateral view. C: Confocal image of Cxcr4b-EGFP–expressing muscle fiber, membrane expression pattern was observed. D: Normalized autocorrelation curves of cytosolic EGFPs in muscle fibers and motor neurons. E: Autocorrelation and fitting curves of Cxcr4b-EGFP on membrane of muscle fibers, together with fitting residuals and fluorescence intensity trace. The inset is a schematic drawing of the construct. Scale bars = 40 μm in A, 10 μm in B, 20 μm in C.

Cxcr4b is a chemokine receptor of the G-Protein Coupled Receptor (GPCR) gene family (Chong et al.,2001,2007,2009). These receptors play an important role in defining directionality of cell migration of germ cells, lateral line, and somites during embryogenesis (Doitsidou et al.,2002; Gilmour et al.,2004; Hollway et al.,2007). We further injected Cxcr4b-EGFP plasmid into a single blastomere at 16-cell stage and observed distribution of EGFP-tagged Cxcr4b in labeled cells. As shown by confocal microscopy the EGFP signal in muscle fibers was found along the membrane (Fig. 5C). In following experiments, Cxcr4b-EGFP plasmid was injected in lower concentration (40 ng/μl) as it has been shown that overexpression of Cxcr4b leads to developmental abnormalities (Doitsidou et al.,2002). The EGFP-tagged Cxcr4b was found to be uniformly distributed along the muscle fiber plasma membrane. The FCS observation volume was placed on the bottom membrane and the z-position was adjusted in steps of 0.1 μm until the maximum fluorescence intensity was observed. A typical ACF curve is shown in Figure 5E. The fluorescence intensity trace shows no intensity bursts, suggesting no higher order aggregation of Cxcr4b on the membrane. Due to the internalized EGFP in cytoplasm, the experimental curves were fitted with a two-components model (Eq. 7) and the short diffusion time, representing cytosolic fluorescent proteins, was fixed at 0.5 ms. Because membrane dynamics could influence the FCS curves, we tested for the presence of a third component. An addition of a third parameter did not improve the fitting results. In two-thirds of the measurements, the second and third diffusion time are the same. In the other one-third of the measurements, one somewhat longer diffusion time can be found. However, this time is not consistent and more importantly the χ_μ², as a measure of the goodness of the fit (Meseth et al.,1999), is not significantly improved, as determined by an F-test, compared with the two-particle fit. Therefore, membrane movement did not significantly contribute to our measurements. Nevertheless, it should be noted that this is not necessarily the case for all cell types and developmental stages and one has to control for membrane movements. Therefore, the two-component model with the faster one fixed at 0.5 ms was used to fit the Cxcr4b-EGFP curve. The fitting yields a diffusion time of 22.1 ± 7.1 ms for the Cxcr4b-EGFP on the membrane. The corresponding diffusion coefficient is 0.60 ± 0.19 μm²s⁻¹, in agreement with previously published results obtained in cell culture (Barak et al.,1997).

Diffusion Coefficient Measurements in Live Drosophila Embryos

In live, filleted Drosophila embryos, we measured the diffusion behavior of fluorescent fusion proteins expressed in individual cells by means of the Gal4/UAS expression system (Brand and Perrimon,1993). For this purpose, the evenskipped GAL4 driver was used to drive expression of the different EGFP reporters in two identified motor neurons in the central nervous system. In control experiments, filleted embryos were labeled with the live/dead stain Sytox Green after 1, 2, 3, or 4 hr to ensure that the fillets were still alive at the time FCS measurements were done, between 1 and 2 hr after filleting (Supp. Fig. S1). The temperature sensitivity of the Gal4/UAS system allows modulation of protein expression levels by raising or lowering the temperature at which the embryos develop. The expression level can be further adjusted by varying the chromosomal copy number of the driver and reporter, depending on the reporter being used. By using a single copy of the driver and reporter and aging the embryos at 18°C, the average particle number of fluorescent proteins passing through the observation volume was kept below 10. An evenskipped promoter-Gal4 driver designated RRaPP3 was used to drive marker protein expression in dorsal-lying aCC/RP2 motor neurons (Goodman and Doe,1993; Fujioka et al.,2003) and two proteins, cytosolic EGFP and membrane-bound mCD8-EGFP (Lee and Luo,1999), were expressed in the embryos to determine whether differently localized fusion proteins (free or membrane bound) would give different diffusion rates. Images of typical motor neurons used for FCS measurements are shown in the insets of Figure 6A,B. We first measured cytosolic EGFP diffusion in the aCC/RP2 motor neurons. FCS measurements were completed within 2 hr after each embryo dissection and a typical ACF curve is shown in Figure 6A. A total of 30 measurements (10 neurons × 3 measurements) gave a diffusion time of 0.23 ± 0.03 ms, similar to results obtained on zebrafish embryos (Table 1). In comparison, a total of 72 FCS measurements (24 neurons × 3 measurements) were done on membrane-bound mCD8-EGFP in aCC/RP2 motor neurons. Excluding two extreme outliers with diffusion times of 169.0 and 266.2 ms, 70 measurements gave an average diffusion time of 24.3 ± 14.6 ms (see Supp. Fig. S5 for distribution histogram). A typical ACF curve is shown in Figure 6B. To explain these outliers, several mechanisms could be invoked, such as transient binding of the diffusing molecules to other proteins, or sporadic protein aggregation. In addition to its expression at the plasma membrane, mCD8-EGFP labels the nuclear membrane when expressed in neurons (Lee et al.,2000). Hence, fluorescence signal from both the plasma membrane and nuclear membranes could in principle contribute to the ACF curves.

Figure 6

Diffusion coefficients measurements in Drosophila embryos. A: Autocorrelation and fit curves of cytosolic enhanced green fluorescent protein (EGFP) in motor neurons, shown together with residuals and intensity traces. Experimental curves were fitted with a three-dimensional one-particle one-triplet model (Eq. 4), representing cytosolic tagged protein. Scale bar = 10 μm. B: Autocorrelation and fit curves of membrane-bound mCD8-EGFP in motor neurons. Experimental curves were fitted with a two-dimensional two-particle one-triplet model (Eq. 7), representing cytosolic and membrane bound fluorescently tagged protein. Scale bar = 10 μm. Inserts are corresponding confocal images.

Table 1. Diffusion Coefficients Measured in Zebrafish and Drosophila Embryos

Sample	τd ± SD (ms)	D ± SD (μm^2 s^−1)	Sample size
EGFP in motor neuron (zebrafish)	0.29 ± 0.06	45.3 ± 9.4	60
EGFP in muscle fiber (zebrafish)	0.36 ± 0.04	36.5 ± 4.1	80
Cxcr4b-EGFP in muscle fiber (zebrafish)	22.1 ± 7.1	0.60 ± 0.19	20
EGFP in motor neuron (Drosophila)	0.23 ± 0.03	57.1 ± 7.4	30
mCD8-EGFP in motor neuron (Drosophila)	24.3 ± 14.6	0.54 ± 0.32	70

• EGFP, enhanced green fluorescent protein.

In this study, we showed that protein dynamics can be studied directly on a molecular level in live zebrafish and Drosophila embryos using FCS. For this purpose, we first examined the autofluorescence expression patterns and determined the usable penetration depth. We showed in zebrafish that the blood flow velocities can be measured with high spatial resolution even in the absence of red blood cells and external dye injections potentially allowing the quantitative study of cardiovascular related diseases in a zebrafish model. On a molecular level, we accurately measured the diffusion coefficients of EGFP or EGFP labeled proteins in the cytoplasm and on the membrane of cells at various depths up to at least 50 μm within living zebrafish embryo. The penetration depth and the possibility to freely orient the embryo during measurements are sufficient to allow access to most tissues and organs within the small embryos and larvae studied here. The viscosity of the cytoplasm can be estimated from EGFP diffusion coefficients and the values can be used to differentiate between cell types or possibly between different cell states.

In Drosophila embryos, we showed that the in vivo membrane fluidity of cells can be quantified by measuring the diffusion behavior of a membrane marker, and that a membrane bound marker (mCD8-EGFP) could be distinguished based on its diffusion characteristics from a cytoplasmically localized marker (cytosolic EGFP). This opens the way for similar studies to be carried out in a variety of genetic backgrounds that may impinge upon membrane composition and dynamics. Overall, the introduction of FCS, a technique with single molecule sensitivity, into small model animals allows measurement of a wide range of molecular processes and raises the prospect of observing molecular causes of system-wide events.

EXPERIMENTAL PROCEDURES

FCS Theory

FCS measurements collect temporal fluorescence signals from a confined observation volume (∼1 fL). In a system of thermodynamic equilibrium, temporal fluorescence fluctuations contain information about the dynamic processes that govern these fluctuations. In FCS the fluorescence intensity F(t) at time t is autocorrelated with itself at a time t+τ later F(t+τ). The normalized ACF is given by:

(1)

where < > denotes the time average. Experimental ACF curves were fitted with predetermined models. For a fluorescent particle freely diffusing in three dimensions the ACF, including a triplet state, is given by:

(2)

with:

(3)

and

(4)

N is the number of particles in the confocal volume; F_trip is the fraction of the particles that have entered the triplet state; τ_trip is the triplet state relaxation time; τ_d is the average time the fluorescent molecule requires to diffuse through the confocal volume; ω₀ and z₀ are the radial and axial distances where the excitation intensity reaches 1/e² of its value at the center of the confocal volume; G_∞ is the convergence value of G(τ) for long time and was close to 1 in all measured cases. The diffusion coefficient (D) is inversely proportional to τ_d:

(5)

To obtain an absolute value for the diffusion coefficient, a standard dye fluorescein (Flu) with a known diffusion coefficient of 300 μm²s⁻¹ was measured for calibration (Wohland et al.,1999). The diffusion coefficient of an unknown sample can then be calculated by comparing their τ_d:

(6)

If two different size particles diffuse in a confined 2 dimensional plane, considering one triplet state, the ACF is given by:

(7)

F₂ is the fraction of the second particle; τ_d1 and τ_d2 are the diffusion times of the first and second particle, respectively.

Considering a system with two flow components coupled with free diffusion, the ACF is given by (Pan et al.,2007b):

(8)

F_f1 is the fraction of the first flow component, i.e., the fraction of the time the system exhibits flow rate 1 compared with the time the system spends in flow rate 2. τ_f1 and τ_f2 are the first and second characteristic flow times for a molecule through the focal volume (τ_f = ω₀/V_f, where V_f is the flow velocity).

In addition, assuming the fluorescent molecule is spherical with radius of r, the viscosity η of the medium can be determined by the Stokes-Einstein relation:

(9)

where k is Boltzmann's constant and T is the absolute temperature.

FCS Instrument Setup

FCS measurements were performed on a modified commercial confocal laser scanning microscope (FV300, Olympus, Singapore) as described previously (Pan et al.,2007a). Briefly, the laser beam (30 μW, 488 nm, Melles Griot, Singapore) is reflected by a long-pass excitation dichroic mirror (488/543/633, Olympus, Singapore) into the scanning unit (G120DT, GSI Lumonics). A water immersion objective (×60, 1.2 NA, Olympus, Singapore) then focuses the laser beam into the sample and collects the emitted fluorescence emission. The Fluorescence light is de-scanned and pass through the same dichroic mirror. The light is then spatially filtered by a confocal pinhole (150 μm) which rejects out-of-focus light. A custom-built slider then directs the light to either the FV300 photomultipliers for imaging, or to the avalanche photodiode (SPCM-AQR-14, Berkshire, UK) for FCS analysis. The single pinhole for both imaging and spectroscopy guarantees the accurate positioning of the FCS observation volume after confocal image acquisition (Pan et al.,2007a). In the FCS part, the fluorescence light is focused by a lens (f = 60 mm, Achromats, Linos, Goettingen, Germany) and pass through a 510AF23 emission filter to the photodiode. Signals are then processed online by a hardware correlator (Flex02-01D, correlator.com, Zhejiang, China). FCS curve fitting is performed by a self-written program in Igor Pro (Wavemetrics, Tigard, OR).

For two-photon FCS measurements, a Ti:sapphire laser (Chameleon-XR, Coherent, Santa Clara, CA) is aligned into the same optical pathway. A continuously variable neutral density filter (NDC-25C-2, Thorlabs, Newton, NJ) is used to adjust the laser power. The laser power is attenuated to around 20 mW to avoid severe photobleaching for EGFP measurements. A short-pass excitation dichroic mirror (675dcsx, Chroma, Rockingham, VT) and an IR blocking filter (e700sp-2p, Chroma) were used instead of the long pass dichroic mirror and a maximum pinhole size (300 μm) was chosen rather than 150 μm for better photon collection efficiency. The size of the observation volumes were calculated by measuring dyes with known diffusion coefficients (fluorescein with a diffusion coefficient of 300 μm²s⁻¹ for 488 nm OPE, TMR with a diffusion coefficient of 480 μm²s⁻¹ for 840 nm TPE; Culbertson et al.,2002). The size of the observation volume with 488 nm (OPE) and 840 nm (TPE) laser excitation is 0.36 ± 0.04 fL and 0.41 ± 0.07 fL, respectively.

Zebrafish Embryo Preparation

Zebrafish is maintained according to the Zebrafish book (Westerfield,2000), the IACUC regulations and rules of the IMCB zebrafish facility. One-cell stage wild-type AB zebrafish embryos are collected, dechorionated and transferred to a Petri Dish with molded agarose injection holder. A total of 100 pL of designated DNA plasmid (100 ng/μl) is injected into a single blastomere at 16-cell stage. The embryos are then incubated in egg water at 28.5°C for optical development, and PTU (0.003% 1-phenyl-2-thiourea in 10% Hank's saline, Invitrogen, Singapore) is added at 18 hr postfertilization (hpf) to prevent pigmentation. Fluorescence expression is examined at 3 dpf under a UV dissecting microscope and normal embryos with low expression level in proper cell lines were selected for FCS measurements. Selected embryos are anesthetized using Tricaine (ethylm-aminoboenzoate, Sigma, Singapore) and mounted in 0.5% low-melting-temperature agarose (Invitrogen) in a glass bottom Petri dish (GW-3512, WillCo-Wells, Amsterdam, Netherlands). The Petri Dish containing the zebrafish embryo is then mounted in the fluorescence correlation microscope for confocal imaging acquisition first. Cells with proper EGFP expression level are selected for following FCS analysis (see flow chart in Fig. 7). A single spot is chosen inside the cell for point scanning and the custom-built slider then switches the system for FCS measurements. Each measurement takes 30 sec. For blood flow measurements, a 3 dpf embryo is imaged under the transmitted light channel and the image is used to guide FCS positioning.

Figure 7

Zebrafish embryo preparation (i) one-cell stage embryos were collected and dechorionated. The embryos were aligned onto the agar holder in a Petri Dish and the orientation of the embryo was adjusted with a glass dropper that the animal pole was on top. The Petri dish was filled with egg water and the embryos were let to develop to the 16-cell stage. (ii) The DNA plasmid was microinjected into one of the four central blastomeres. The embryos were then incubated in egg water at 28.5°C to develop to 3 days postfertilization (dpf). (iii) Selected embryos were anaesthetized and mounted into 0.5% low-melting-temperature agar in a glass bottom Petri dish. The embryo body was pushed close to the surface of cover glass and the orientation of the embryo was adjusted with a needle. The specimen was then mounted in the microscope stage for subsequent confocal imaging and fluorescence correlation spectroscopy (FCS) measurements.

Drosophila Embryo Preparation

For measurements of membrane-bound EGFP, the membrane-targeted reporter UAS-mCD8-EGFP, a fusion protein between mouse lymphocyte marker CD8 and EGFP, was expressed in aCC/RP2 motor neurons by means of an evenskipped-GAL4 driver RRaPP3 (a kind gift of Miki Fujioka, Fujioka et al.,2003). The mCD8-EGFP construct consists of the full length α chain of CD8 with EGFP attached at the C-terminus (Lee and Luo,1999). Virgin females of the genotype RRa-Gal4, UAS-mCD8-EGFP were crossed to yellow,white males. The progeny (stage 16 embryos) of the cross with genotype RRa-Gal4/+, UAS-mCD8-EGFP/+ were used for FCS measurement. The embryos were first collected at 22°C for 2 hr and were then left to age at 18°C for 21 to 22 hr until stage 16 (stages according to Campos-Ortega and Hartenstein,1985), at which time formation of the embryonic nervous system is nearly complete. Under these temperature conditions, levels of EGFP expression were low enough to produce accurate FCS measurements. By using a single copy of the driver (RRa-Gal4) and reporter (UAS-mCD8-EGFP), and aging the embryos at 18°C, the average particle number of fluorescent proteins passing through the observation volume was ∼7. For cytosolic EGFP expression, virgin females carrying RRaPP3 were crossed to males carrying UAS-EGFP (Halfon et al.,2002). Embryos from this cross with genotype RRaPP3/UAS-EGFP produce freely diffusing EGFP in the cytoplasm of aCC/RP2 motor neurons, and were analyzed by FCS measurement at stage 16. The embryo collection and aging process were the same as aforementioned except that the embryos were left to age at 18°C for approximately 23 hr until stage 16. Stage 16 Drosophila embryos were dechorionated by hand after transferring them from collection plates onto double-sided sticky tape, and then were aligned on an agar block with ventral sides up. Dechorionated, aligned embryos were transferred to the dissection/imaging dish (FluoroDish, World Precision Instruments, Sarasota, FL) onto a small piece of double-sided sticky tape, and covered gently with hemolymph-like (HL3) saline solution (Kuromi and Kidokoro,1999) containing 0.1 units/ml insulin (BioGen Pte. Ltd., Singapore). Immersed embryos were dissected in HL3 at room temperature. Pulled 1.2-mm capillary needles were used for dissection (detailed in Fig. 8). The dechorionation and dissection procedures took approximately 35 min. Only stage 16 embryos were dissected for FCS measurement. Each measurement took 20 sec and 3 measurements were done successively in a single motor neuron.

Figure 8

Drosophila embryo preparation (i) Drosophila embryos were spread onto double-sided sticky tape and manually dechorionated. Dechorionated embryos were aligned on the agar with the ventral side facing up. a: The orientation of the embryos is shown in this enlarged picture. (ii) A strip of double-sided sticky tape was cut and placed on the cover glass bottom of the FluoroDish. The agar with the aligned embryos was placed on a cylindrical object. The embryos were transferred to the double-sided sticky tape on the FluoroDish as shown. (iii) The embryos were aligned with the dorsal side facing up. Before dissecting the embryos, the FluoroDish was filled with 4 ml of dissecting medium to cover the embryos. (iv) The embryos were dissected and the gut was removed with a sharp needle. Care was taken not to damage the nerve cord. (v) The dissected embryo was flipped over and stuck onto the cover glass beside the sticky tape, with the help of a second blunt needle. Embryos are not lifted above the surface of the dissecting medium, as this causes dispersion of the tissue. The dissected embryos are stuck firmly onto the glass to avoid drift with respect to the cover glass bottom. The nerve cord of the embryo now faces downward toward the cover glass bottom for FCS measurement. b: Enlarged schematic diagram showing the arrangement of the aCC/RP2 motor neurons within the central nervous system. FCS measurements were done on the upper membrane of aCC/RP2 motor neurons which were the most distant from the cover glass bottom.

Plasmids

A stop codon was removed from Cxcr4b gene previously cloned in a pPCR-Script vector (Stratagene, CA) (Chong et al.,2001). The Cxcr4b gene was digested out using SacII and XhoI and then subcloned into pEGFP-N1 (Clontech) vector digested with the same enzymes. The Cxcr4b-EGFP construct was sequenced and membrane expression in zebrafish embryo was confirmed by confocal image.

Acknowledgements

X.S., L.S.T., are NUS scholarship holders. X.P. was awarded a scholarship by the Graduate Program in Bioengineering. T.W. was funded by the Biomedical Research Council of Singapore. V.K. laboratory in the IMCB is supported by the institutional grant from the Agency for Science, Technology and Research of Singapore. Funding for this project was provided in part by the Institute of Bioengineering and Nanotechnology (Biomedical Research Council, Agency for Science, Technology and Research of Singapore).