1. Introduction

Membrane activities play a crucial role in cell dynamics such as motility, morphological variations, cell-cell adhesion, endocytosis, exocytosis, etc [1]. The membrane topography and motion are also closely related to the configuration and structure of the underlying cortex, which is mainly composed of the actin filaments. Conventionally, phase-contrast and fluorescence microscopy techniques are employed to investigate the membrane activities as well as the actin dynamics [2–4]. However, these optical observations only provide information from the boundaries of the cell or from the dye-labeled cytoskeletons. It is very difficult to learn the membrane topography on the cell surface by using conventional optical techniques. Low-coherence reflectometry is also a useful technique for studying cell membrane activities because of its nanometer depth resolution and fast response [5]. But to acquire a three-dimensional surface plot with this technique requires two-dimensional scanning of the light spot on the cell surface. On the other hand, atomic force microscopy (AFM) has been a powerful tool for studying cell mechanics [6–8]. Nevertheless the scanning speed of the AFM tip is usually too slow to profile a sufficiently large area for dynamic assessments on a living cell. Meanwhile, because of the nano-newton tip force, the interpretation of AFM measurements on cell membranes needs careful modeling. For example, the fractions of cellular viscoelasticity contributed by the membrane and the cytoskeletons are difficult to determine from the data obtained by AFM.

In order to clearly and accurately observe the membrane activities of a living cell, an instrument should match several requirements related to the resolution, imaging speed, and field of view. The size of common mammalian cells is tens of micrometers, and hence the feature sizes of membrane activities could be from tens of nanometers to a few micrometers. The moving speed of a motile cell such as a fibroblast could be several tens of micrometers per hour [3, 4], such that the time scale of the membrane activities would range from 100 ms to several minutes. Imaging or profiling technologies that provide the above spatial and temporal resolutions are suitable for the studies of cell membrane movements. For the convenience of overall observations, the field of view should be equal to or larger than a single cell. In addition, owing to the high flexibility of the membranes, non-contact measurement is favorable. Finally, this instrument should be able to work with fluorescence microscopy for identifying some specific membrane proteins as well as the membrane-associated cytoskeletons.

In this paper we present the use of a novel optical technique, non-interferometric widefield optical profilometry (NIWOP) [9], to measure the membrane topography of a living cell. The NIWOP technique directly obtains height variations from processed images that are captured by a CCD camera, eliminating the scanning mechanism for acquiring an image. The probe is a water-immersion objective of which the working distance is in the millimeter range. Therefore the membrane activities as well as the cell motion are not perturbed during the measurement. We observe the propagation of membrane ripples from the edge toward the center while the cell has adhered to the bottom of the culture dish. In order to verify the driving mechanism of the ripple propagation, we use cytochalasin D to dissolve the cortex actin networks while recording the membrane motion. The results clearly show that actin filaments are closely related to the motion of the membrane ripples.

 

Fig. 1. Setup of the non-interferometric widefield optical profilometer. The system is constructed on a microscope with epi-illumination.

Download Full Size | PDF

2. Method and material

2.1 Setup and operation of the NIWOP

Figure 1 shows the setup of our NIWOP system [9]. The setup uses a conventional optical microscope (Nikon Eclipse L150) with epi-illumination as the mainframe. We employ a power-regulated tungsten-halogen lamp as the light source, of which the power fluctuations are measured to be smaller than 0.1%. A bandpass filter (CVI Laser BG-39) selects the illumination wavelength of 350–610 nm. Such large a bandwidth prevents the interference caused by the reflections from the cell surface and the bottom of the culture dish. An 8 cycles/mm grid pattern made by coating chromium stripes on a fused-silica window is projected onto the sample as the synthesized aperture to produce optical sectioning [10]. The spatial frequency of this grid pattern projected onto the cell surface is 2.4 μm^-1. Light reflected from the cell membrane is collected by a 40× water-immersion objective with a 0.75 numerical aperture (NA). The culture dish containing the cells is placed on a PZT-driven vertical stage (Physik Instrumente P-762.ZL), which has a 10-nm smallest step size and 0.1% linearity in closed-loop operation. This PZT stage is used to precisely control the average membrane position relative to the focal plane and to calibrate the height measurement of the NIWOP. A 14-bit CCD camera cooled at -40°C is used to capture the images. The field of view of this NIWOP system is 146 × 97 μm² and the pixel size of the images is 190 nm. For the images used in this paper, the exposure time is 500 ms/frame.

The details of operation and calibration of the NIWOP technique have been described in our previous articles [9, 11]. In brief, we take three images with the grid pattern at spatial phases 0, 2π/3, 4π/3, and then use the homodyne detection principle to remove the grid pattern and obtain optically sectioned images [10]. In this work, the capture and processing time for one NIWOP image is about 6 sec. With a proper selection of the grid period and the NA of the objective, the slope at the linear region of the axial response curve can be very sharp, and hence the surface profile can be accurately determined as the membrane surface is placed slightly away from the focal plane (into the linear region). Figure 2 shows a typical response curve in the linear region obtained by the setup in Fig. 1 when scanning the bottom of a culture dish along the optical axis. At each axial position we record the total intensity on a small area (40 × 40 pixels) of the dish bottom for 20 times, and calculate the standard deviation of these 20 measurements. Among all the 10 axial positions, the maximum standard deviation is ±3.1%, and the average value of the 10 standard deviations is ±1.4%. Because the slope of the linear region is 0.68 μm^-1, while we take the average value of standard deviations as the depth resolution, it corresponds to 20 nm. With present setup the dynamic range is as large as 1.0 μm.

 

Fig. 2. Linear region of the axial response curve of our NIWOP system. Every dot and error bar represent the average value and standard deviation of intensity measured for 20 times at that axial position. The slope of the fitting line is 0.68 μm^-1. The operational dynamic range is 1.0 μm.

Download Full Size | PDF

2.2 Cell culture and sample preparation

We cultured HS-68 fibroblasts in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum and 1% antibiotic pen-strep-ampho. The cells were placed in a 35-mm plastic dish (FALCON 3001 BD) and cultured at 37°C in a 5% CO₂ atmosphere incubator. Before the observation, we added phosphate buffered saline into the culture medium to reduce the optical absorption of the culturing medium and to improve the contrast of images. No other treatment to the cells is required. During the observation the cell was kept at room temperature (23°C).

3. Results and discussion

3.1 Removing the effect of cell heterogeneity from NIWOP images

The major advantage of NIWOP is that it provides nanometer height measurement on a sample surface. However, the inhomogeneous reflectivity of a cell leads to incorrect results. Here we explain how to remove the surface heterogeneity such as reflectivity from the NIWOP profiling results. Figure 3(a) shows the bright field image of a living fibroblast, which is obtained as we remove the grid from the illumination optical path. The contrast of the bright field image is from the heterogeneity of refractive index on the cell surface as well as in the cytoplasm. Therefore, in the raw NIWOP image such as Fig. 3(b) the contrast results from both the topography and the cell heterogeneity, and the interpretation of this image is difficult. Comparing Figs. 3(a) and 3(b), we also note that the ripple-like contrast near the cell edges in Fig. 3(b) does not appear in the corresponding regions in Fig. 3(a). This observation excludes the possibility that these ripples result from interference because NIWOP does not enhance interferometric effects.

In the original NIWOP system, we had obtained the actual topography of surface with various reflectivities through the division of the raw NIWOP image by the bright field image [9]. However, this simple processing can only be applied on specimens composed of a single reflective surface. Because the absorption of the cell membrane and cytoplasm is small around the cell edge (the lamellipodia), we should now consider the cell as a dielectric layer with two reflective interfaces: one is the interface between the upper cell membrane and the culture medium, the other is the interface between the lower cell membrane and the bottom of the culture dish. Assuming that the absorption is negligible, we can solve the surface profile of this dielectric layer by the following method:

 

Fig. 3. (a) Bright field image of a living HS-68 fibroblast. (b) The raw NIWOP image of the same cell. The membrane ripples are visible, but overlapped with other optical contrast. (c) The NIWOP image after the processing described in the text. The gray scale in this image corresponds to height variations from 0 to 350 nm. Membrane ripples are much clearer.

Download Full Size | PDF

On the region occupied by the cell in Fig. 3(b), the actual membrane topography T_m (x, y) and the intensity of the raw NIWOP image I(x, y) can be expressed by a simple algebraic equation:

where I_m(x,y) and I_d(x,y) represents the intensity of reflection from the medium-membrane and membrane-dish interface respectively, α is the normalized slope of the linear response curve as the one shown in Fig. 2, and Z_d is the distance between the dish bottom and a reference plane corresponding to the origin of the x-axis in Fig. 2. Since we recorded the height of the PZT vertical stage during each exposure while obtaining Fig. 2, Z_d is actually a known number. Here we ignore the profile of the bottom membrane because in typical measurements we find I_m ~ 100I_d. It is known that the gap between the cell bottom and the substrate is about 50 nm, and the focal contacts reduce this distance to 10–15 nm [1]. The variations in I_d caused by the bottom membrane profile are thus not detectable using the current NIWOP system. In contrast, the cell region in the bright-field image, Fig. 3(a), can be expressed as:

without the topographic modulations.

In order to get the membrane profile T_m(x, y), we need to solve two unknowns in Eq. (1), I_m(x, y) and I_d(x, y). To obtain these two intensities from upper and bottom interfaces of the cell, we need to know the refractive indices of three materials: the culture medium (n_medium), the dish bottom (n_bottom), and the cytoplasm (n_c). Assuming n_medium = 1.33 (the refractive index of water), we can determine n_bottom from the reflectivity calibrated from the intensity measured on the region not occupied by the cell in Fig. 3(a). With n_medium and n_bottom known, we can solve n_c(x,y) using the reflection intensity I_B at each pixel. Then we obtain I_m(x,y) and I_d(x,y). Thereafter, by solving Eq. (1) we achieve the correct membrane topography T_m(x,y).

The above solution can only be applied on the regions occupied by the cell because the bottom of the culture dish is not a dielectric layer, and therefore the height calibration method of the no-cell region is different. With this constraint, we cannot compare the height of cell membrane relative to the bottom of the dish while the profile on the membrane is correct. One may employ a specific image processing algorithm that can identify the cell boundary, and then uses corresponding calibration methods to calibrate the profiles of cell membrane and dish bottom, respectively. However, to develop such a program is not necessary for the study of the membrane ripples.

This calibration procedure works for specific regions on a cell in which the height variation as well as the cell thickness are smaller than the dynamic range. For example, the thickness of lamellipodia at the leading edge of a fibroblast is hundreds of nanometers [6]. The NIWOP system used in this study is thus adequate for studying the membrane dynamics on lamellipodia. For thicker cells the dynamic range of NIWOP can be adjusted simply by using grid patterns of different spatial frequencies without changing other optics in the setup [9]. This feature makes a NIWOP system readily applicable on samples of various roughness scales. As to cells thicker than the width of an optical section, the original calibration technique for single-surface samples applies.

Figure 3(c) shows the membrane topography obtained by solving Eqs. (1) and (2). The gray scale represents surface height from 0 to 350 nm. Compared to Fig. 3(b) with mixed contrast, the membrane ripples are clearly seen in this processed image. Because the above calculation treats the cell as a dielectric layer and omits the absorption in cytoplasm, the height calibration in the middle region of the cell, especially above the cell nucleus, may still be incorrect. In the following sub-sections we discuss only the motion of membrane ripples near the edge (the lamellipodia) of a cell.

3.2 Membrane ripple dynamics

In Fig. 4 we show the propagation of the membrane ripples in a 13.3 × 20.9 μm² area on a HS-68 fibroblast near the cell edge (as indicated by the dashed rectangle in the whole-view image). For a better visualization we use an arrow to indicate one of the ripples. The height of this ripple is about 250 nm, and the average propagation speed is approximately 1.3 μm/hour away from the edge. (See the movie in Fig. 5 for a continuous motion of the ripple.) Because this cell has already flattened itself on the dish bottom, this centripetal membrane motion is quite intriguing. We suspect that it is related to the dynamic of the underlying cell cortex: the actin filaments of the cell cortex are also moving away from the cell edge. If this hypothesis is correct, it would be possible to stimulate cell contraction at this site.

In order to verify the potential driving mechanism of the ripple propagation shown in Fig. 4, we add 20 μM cytochalasin D (Calbiochem Biochemicals) into the culture medium to dissolve the actin filaments of the cell cortex. After 10 minutes, we observe chaotic membrane ripples as shown in Fig. 6(a). The average amplitude is much higher than that of a normal cell because the cell cortex is softened by cytochalasin D, as we measured in a previous study [12]. These ripples do not have a specific propagation direction as those in Fig. 4. Moreover, the original long ripples disappear as a result of the dissolution of the actin-filament networks. In Fig. 6(b) the membrane flattens, and an obvious cell contraction occurs. The cell shape and membrane profile do not have further change in Fig. 6(c).

3.3 Comparison between the NIWOP and other optical techniques

The surface profiles measured in this work may also be observed by real-time low-coherence interferometry, such as that proposed by Dubois et al. [13]. Compared with interferometric techniques, the optical setup of NIWOP is only a desk-top microscope and hence simpler to operate and more insensitive to environmental perturbations, such as mechanical or thermal drifts. In addition, because of the simple configuration of NIWOP, it is easy to combine other contrast mechanisms of light microscopy, such as fluorescence, phase-contrast, polarization, etc., to obtain more physiological information accompanied the topographical measurements.

 

Fig. 4. Propagation of the membrane ripples on a fibroblast. Panels (a) to (e) are the zoom-in images of the region enclosed by the dashed rectangle in the whole-view image. An arrow indicates one of the ripples for a better visualization of the propagation. From (a) to (e) we see the ripples are moving away from the cell edge with an average speed about 1.3 μm/hour. The dynamics is more obvious in Fig. 5.

Download Full Size | PDF

 

Fig. 5. (2.1 MB) Movie of the membrane-ripple propagation. 20 μM cytochalasin D was added into the culture medium at 3:00.

Download Full Size | PDF

 

Fig. 6. Membrane profile of the same cell in Fig. 4 after the treatment of 20 μM cytochalasin D. The time interval between each image is 10 minutes. The dashed line indicates the boundary of the cell. (a) 10 minutes after the treatment. We see sudden arising of multiple peaks on the membrane. (b) The cell contracts and the membrane ripples become more random and smaller compared to those in (a). (c) 10 minutes later, the cell shape and membrane profile do not show significant variation. [Media 1]

Download Full Size | PDF

Compared to our previous study using optical force to uniformly deform the cell and differential confocal microscopy to detect the nanometer membrane motion [12], the NIWOP provides a whole-cell observation, and thus is more suitable for the development a model of cell-motility under external chemical or mechanical perturbations. Moreover, because no scanning mechanism is required, the optical alignment is simpler and the cost is much lower.

A drawback of the NIWOP technique is that the height signal originates directly from the reflected light, and therefore scattering debris in the culture medium results in noise of the data. Although the optical sectioning feature of this technique eliminates most of the scattering light above or below the focal plane, small debris near the cell surface still affects the height measurement. Hence, for accurate topographic information on a living cell, the culture medium has to be filtered for the use in the NIWOP measurement.

4. Conclusion

We demonstrate the application of the NIWOP technique on measuring the membrane ripples of a living cell. This system can non-intrusively and rapidly image the three-dimensional membrane topography in a physiological environment with nanometer depth resolution. The lateral resolution is the same as that provided by high NA objectives. If the accurate height measurement is not required but small topographical features on the membrane are of interest, the super-resolution NIWOP technique which has lateral resolution one-sixth of the wavelength on slowly varying surface can be employed [11]. The system can also combine the fluorescence microscopy to find out the correlation between membrane surface topography and the underlying organelles or proteins. We believe the NIWOP will be a powerful tool in cell research.