1. Introduction

Live-cell microscopy plays an important part in many areas of life science and has contributed to an increased knowledge of cell structure and function [1,2]. Two crucial factors need to be carefully considered with all live-cell microscopy techniques; the cell environment and the invasiveness of the imaging method used. The former involves factors such as temperature stability, CO₂ concentration, humidity, and biocompatibility of the cell substrate, while the latter relates to the perturbation brought on by the imaging method itself.

A common way to study living and fixed cells is through fluorescence microscopy, where exogenous fluorescent molecules or genetically encoded fluorescent proteins are excited by visible light and the emitted fluorescence signal is detected with an imaging system [3]. However, even though low levels of white light do not show any obvious adverse effects on cells, extensive fluorescence excitation within a given wavelength band can result in photobleaching and/or phototoxicity in the imaged specimen [4]. Phototoxicity and photobleaching are the two major limitations in live-cell fluorescence microscopy. They are caused by fluorophores in an excited singlet or triplet state that generate singlet oxygen and other reactive oxygen species [5]. Irreversible alteration of the fluorophore causes loss of fluorescence signal and may therefore also result in toxicity effects within the cells. The rate of photobleaching increases with excitation intensity, meaning that high numerical aperture objectives used to maximize spatial resolution simultaneously accelerate photobleaching and toxicity effects [6]. Hence, a balance must be reached between attaining highly detailed images on one hand, and obtaining meaningful quantitative information of a living specimen without disrupting or inducing damage to the specimen on the other [1,2]. Efficient collection of fluorescence emission, together with minimum excitation time and power are thus crucial in for imaging of live cells. Consequently, speed of acquisition is important, especially were different types of fluorophores are being excited using multiple wavelengths, where mechanical switching between filters or light sources may be both time consuming and require continuous refocusing of the imaging optics due to chromatic aberrations [1,2].

Many interesting biochemical events take place in close vicinity to the cell membrane. To observe such events with a high signal-to-background ratio, evanescent-wave microscopy techniques have proven a valuable tool. Currently, the most commonly used evanescent-wave imaging method is total internal reflection fluorescence (TIRF) microscopy [7–9]. Evanescent-wave fluorescence microscopy is a convenient option in live-cell fluorescence imaging, since localized evanescent-wave excitation minimizes the number of fluorophores excited within the cell and thus reduces photobleaching and phototoxicity effects [10]. Time-lapse imaging, where excitation is provided only within the volume of interest for a minimum amount of time (as defined by the sensitivity and signal-to-noise ratio of the imaging system) at selected intervals, further reduces phototoxicity effects and is thus important for live-cell imaging where the required observation time can extend over minutes, hours or even days.

Evanescent-wave excitation can be realized using planar dielectric waveguides, where the sample under investigation is in contact with the waveguide core layer. Application of such devices for imaging of fixed cells and optical screening of live-cells has been reported previously [11–15]. Recently, we demonstrated a novel polymer waveguide platform for evanescent-wave excitation, enabling the study of biological objects with high specificity and resolution, as demonstrated by fluorescence imaging of fixed cells [16,17]. The cladding index of the polymer waveguide is closely matched to that of water, making it ideal for constructing lightwave circuits for optical interfacing with biological samples. The planar nature of the illumination allows for direct imaging of large specimen areas, as opposed to the focused excitation typically used in TIRF microscopy. Due to the inherently large refractive index contrast of the waveguide core and cladding materials used, the waveguide platform also allows for the realization of highly integrated waveguide devices for compact on-chip routing and filtering of the excitation light [18].

In order to extend the applicability of the previously reported waveguide platform, we have added a thermo-optically driven modulation/attenuation functionality, reported in the present paper. This was achieved by fabricating waveguide-based Mach-Zehnder interferometers, where a phase shift is introduced by heating of one of the interferometer arms. Through the combination of active and passive functionalities, the platform can be readily used for multicolor time-lapse fluorescence imaging of live cells with full on-chip control over excitation wavelengths and illumination times, coupled with an electronic triggering of the image acquisition system, eliminating the need for external acousto-optic or mechanical switching of filters or light sources. The compatibility of the platform with culturing and imaging of live cells expressing green fluorescent protein (GFP) was experimentally confirmed, as described in more detail in Sec. 4.

Mach-Zehnder interferometer (MZI) structures provide an efficient method of thermo-optic control of light on the chip. Ideally, the MZI should be compact, have low power consumption and a sufficiently short response time. A large refractive index contrast between the waveguide core and cladding layers means that higher mode confinement is possible, resulting in lower bend loss and more compact optical devices. In the present case, the waveguiding structure is composed of two polymers with a relatively large refractive index contrast of ≈0.2 [18] and thermo-optic coefficient of the fundamental waveguide mode around 1 × 10⁻⁴/K [19].

A simple MZI (see Fig. 1 ) is constructed by splitting an incoming optical signal evenly into two arms using a Y-splitter junction and then recombining the signal at a second Y-junction. A perturbation of the optical signal in one arm, i.e. through external heating, causes a phase-shift compared to the signal traveling in the unperturbed arm. Assuming that the phase shift is introduced by external heating, i.e. in a heated arm of length L which has undergone a temperature change ΔT, the resulting output signal can be expressed in the following form [20–22]:

 

Fig. 1 Optical microscope image showing a fully fabricated MZI structure with gold contact pads and heater over one arm, with 532-nm TM-polarized light coupled into the structure from the right by end-fire coupling using a lensed fiber.

Download Full Size | PDF

2. Chip fabrication

A detailed description of the fabrication process and optical properties of the polymer waveguide platform, as well as several passive integrated optical components, can be found in [16–18]. The waveguide structure is composed of three highly transparent polymer layers. The core layer consists of polymethylmethacrylate (PMMA 950k, Microchem Corp.) with a refractive index of n≈1.49 while the cladding layers are made of the fluorinated polymer Cytop (Asahi Glass Co.) with refractive index n≈1.34. PMMA is a well-studied high-resolution electron-beam lithography resist [24], facilitating patterning of buried channel waveguides and optical devices by direct e-beam writing [18]. For the present study, MZIs of different lengths were constructed, with input- and output waveguides displaced using circular 90° bends (see Fig. 1) to prevent stray light from the input fiber from affecting the output signal. The channel waveguides were approximately 440-nm high and 500-nm wide, resulting in a highly confined fundamental mode over the range of typical fluorescence excitation wavelengths (488-633 nm). The thickness of the top and bottom Cytop cladding layers was 1.9 µm and 4.0 µm, respectively.

A microscope image of a finished MZI device with light propagating through the waveguides, is shown in Fig. 1. Contact pads and heaters were fabricated by photolithography and deposition of 2 nm of chromium and 120 nm of gold, using thermal evaporation. The thin chromium layer serves as an adhesion layer between the gold and the underlying Cytop polymer. The heaters (narrow wire segments in Fig. 1) were 10-μm wide and 600-μm long (R≈120 Ω). The straight segments of the MZI shown in Fig. 1 are 400-μm long. This length provided a suitable extinction ratio with low driving power, as compared to shorter devices. From Eq. (1), the ΔT of the waveguide cores required to achieve extinction can be estimated as ≈7°C for this device length. In cases where excitation over large areas is desired, the sensing well must be separated from the MZI by a sufficient distance to allow the light exiting from the MZI to diverge in a planar slab waveguide. This simultaneously ensures that heating of the MZI does not affect the temperature of the investigated sample. Even though light scattered from the waveguide is sufficiently strong to be visualized, propagation loss is too low to be accurately determined with the waveguide lengths used here. This is consistent with the previously reported propagation loss of 0.3 dB/mm in similarly prepared waveguides [18].

3. MZI characterization

A tapered optical fiber (Nanonics Inc.) was used to couple linearly polarized light from a frequency-doubled Nd:YAG laser (532 nm) into the MZI device. The spot size of the lensed fiber closely matched the mode size of the polymer waveguide [18] and the position of the input fiber was fine-tuned for maximum transmission using a piezo-electric xyz-translation stage. The output signal was collected at the opposite facet using a 100× objective and focused either onto a CCD camera (Thorlabs) for mode profile characterization, or onto an amplified Si photodetector (Thorlabs, 17MHz bandwidth) connected to a digital oscilloscope. Output images were collected both in the on-state (maximum transmission, no heating power) and the off-state (minimum transmission, heating power applied). Cross-sectional profiles through such images are shown in Fig. 2(a) . The results confirm the strong confinement of the waveguide mode (the measured mode field diameter of ≈0.8 µm is limited by the resolution of the 0.95NA microscope objective [18]), low level of scattered light in the cladding, and an excellent extinction ratio of about 25 dB (99.7%). The value of the extinction ratio serves as an indicator for the high quality of the fabrication process, including the performance of the y-splitters and consistently low propagation loss in the waveguides.

 

Fig. 2 (a) Mode profile measured at the MZI output facet (inset) and cross-sections along the lateral direction with the device in the on-state (black symbols) and the off-state (red symbols). (b) Modulation of output intensity with power applied to the heating element. (c) Sub-millisecond MZI response time, measured with a square-pulse driving signal.

Download Full Size | PDF

Figure 2(b) shows the transmission through the MZI device as a function of applied heating power. The response was determined while driving the device with a slow (approx. 1 Hz) sawtooth function. Heating power was determined by recording the current through the device as a function of driving voltage. Maximum extinction occurs at approximately 10 mW driving power which compares favorably to many reported thermo-optic devices [25]. The solid curve in Fig. 2(b) shows a cosine fit of the data, confirming the expected behavior of the output with increasing temperature (Eq. (1)). A slight hysteresis is visible in Fig. 2(b), corresponding to heating and cooling cycles of the device, respectively. Temperature-induced hysteresis in optical response has been observed in other polymer thermo-optic devices where maximum heating powers and distance from heating element to the primary heat sink (Si substrate) were similar to the present case [26].

The device response curve for a 200-Hz square-pulse driving signal can be seen in Fig. 2(c). The turn-off time (heating) is faster than the turn-on time (cooling), with the average 10%-to-90% change being around 0.4 ms and 0.7 ms, respectively. Most polymer-based modulators have typical rise times on the ms-scale, while SOI-based thermo-optic devices with response times down to the μs-range have been reported [25]. In order to realize faster response times in our devices (at the expense of increased driving power due to more efficient heat sinking) cladding layer thicknesses must be reduced considerably. Because of the strong light confinement in the structure, this can be readily achieved without introducing additional propagation loss due to mode leakage.

4. Live cell imaging

For live-cell imaging, stable dSH2-GFP-expressing [27] LLC PK1 cells were plated directly on waveguide chips. Sensing areas were prepared by etching through the top cladding, as described in [16]. Because of the near-surface confinement of the excitation light (approximately 200-nm penetration depth into the cytoplasm), the waveguide-excitation method is optimal for studying cytosolic and cell membrane compounds while eliminating, e.g., proteins in the nucleus from being excited. We therefore chose to use the platform to study the movements of phosphotyrosine-binding Src homology 2 (SH2) domain of pp60^Src (dSH2) [27], expressed in small focal adhesions that play a central role in cell migration. Focal adhesions are the most common integrin-mediated cell-matrix adhesions in cultured cells. Tyrosine phosphorylation affects focal adhesion formation and stability and provides docking sites for SH2 [28].

The chip was placed on a microscope stage (maintained at 37°C) and embedded in buffered phenol-red-free Dulbecco's Modified Eagle Medium:Nutrient Mixture (DMEM/F12), supplemented with 15mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) and 2.5 mM L-glutamine (pH 7.36), to ensure that the cells would not show any stress symptoms due to lack of CO₂. In order to trigger focal adhesion movements, the cells were stimulated with an epidermal growth factor (EGF) at a concentration of 100 ng/mL just before imaging [29]. The cells were studied with a 50× objective using laser excitation (473 nm). The GFP fluorescence (emission maximum at 509 nm wavelength) was imaged through a long-pass filter blocking scattered excitation light. Images were captured using an EM-CCD camera (Photometrics QuantEM512C) every 120-150s (50 ms integration time) for a total of 20 minutes, as shown in Fig. 3 . In the time-lapse sequence, EGF-stimulated movement of focal adhesions is clearly observed at t=6 min. (see Media 1).

 

Fig. 3 Movements of focal adhesions in living cells, imaged with a 50× objective at the indicated times (minutes). The arrow in the t=0 frame indicates the position of a moving focal adhesion (see Media 1).

Download Full Size | PDF

5. Conclusion

We have demonstrated the applicability of a novel polymer waveguide platform for evanescent-wave live-cell imaging by observing minor movements of focal adhesions in living cells. By integrating thermo-optic light modulators into the waveguide platform, we have shown that on-chip control of illumination with sub-millisecond switching times and large extinction ratios is possible. Combined with passive functions such as routing and filtering, this offers the possibility of designing chips for multi-wavelength time-lapse imaging or other microscopy or optical biosensing applications.