1. Introduction

Fluorescence microscopy (FM), including standard epi-fluorescence microscopy and a range of more specialized techniques, can be regarded as the most important characterization technique within histology, cell biology, molecular biology and related fields [1]. A relatively recent addition to the range of FM methods is total internal reflection fluorescence microscopy (TIR-FM), introduced by D. Axelrod around 1990 [2]. Commercially available TIR-FM systems are typically based on laser excitation through a high numerical aperture immersion objective (NA > 1) on an inverted microscope where the excitation light undergoes total internal reflection at the substrate-sample (typically glass-water) interface. An exponentially decaying (evanescent) electric field associated with the total internal reflection extends into the sample (see Fig. 1(a)) with an exponential decay length that depends on the incidence angle, wavelength and the refractive indices of the two media [3]. In evanescent-wave fluorescence microscopy, the excitation is confined to the surface, providing an excellent tool for studying, e.g., events at or within the cell membrane, cell morphology, cell motility, and focal adhesions. The technique gives a superior signal-to-noise (i.e. signal-to-background) ratio due to the absence of fluorescence excited outside the focal plane.

Evanescent-wave microscopy can also be accomplished using planar waveguide structures, where the sample (typically in an aqueous solution) forms at least a part of the cladding around an adjacent waveguide core. The penetration depth of the evanescent field into the sample in this case is defined by the relationship between the refractive indices of the sample, substrate and waveguide core materials, as well as the waveguide core thickness and the wavelength and polarization of the guided wave. Waveguide-excitation fluorescence microscopy (WExFM) has been previously demonstrated using a high-index single-mode waveguide layer (e.g., Si_xTi_1-xO₂ or Ta₂O₅) on top of a glass substrate [4,5], giving penetration depths in the range 100-200 nm (Fig. 1(b)). In this geometry, the waveguide mode was excited through a submicron-period grating patterned into the waveguide layer which requires the excitation light to be incident on the grating at a specific wavelength-dependent angle. The image quality obtained using this method was limited by uneven illumination resulting from imperfections in the in-coupling grating [4]. As an effort to extend the penetration depth of the evanescent field further into the sample, waveguide-based sensors having a “reversed” cladding geometry (i.e. where the substrate refractive index is lower than the sample index) have been reported, using special porous glass substrates (n=1.2) coated with a polystyrene waveguide film with imprinted gratings for excitation [6,7], see Fig. 1(c). The potential of using such sensors for monitoring cell growth was demonstrated [7], but there are no reports of reversed-geometry structures being used for waveguide-excitation fluorescence microscopy.

In the present work, we demonstrate a new principle of waveguide-excitation fluorescence microscopy using a symmetric waveguide structure (SWExFM), where the cladding material is index-matched to the sample solution (Fig. 1(d)). The symmetric waveguide structure has a number of important advantages over the “conventional” or “reversed” asymmetric configurations discussed above, including a large selection of possible polymer waveguide materials, possibility of tuning the penetration depth over a wide range by varying the structure geometry and/or materials, highly efficient end-fire excitation from optical fibers or on-chip light sources (eliminating the need for patterning of gratings in the waveguide film) and multiple-wavelength excitation through the same optical path. The SWExFM technology can provide continuous surface-confined illumination of areas up to macroscopic dimensions, it can deliver TIR-FM performance on standard (normal or inverted) light microscopes and it is compatible with laser sources as well as single-mode fiber-coupled white-light (e.g., supercontinuum) sources. The technique places no special requirements on the detection optics. The optical chip fabrication involves only standard cleanroom technologies such as spin-coating, photolithography and dry etching.

 

Fig. 1. Methods of evanescent-wave excitation. (a) Total internal reflection and (b-d) waveguide-excitation using a substrate with a refractive index that is (b) higher, (c) lower and (d) approximately equal to the sample solution. Arrows indicate the preferred method of coupling of excitation light to the evanescent mode, being either incident from within the substrate, onto a surface grating or end-fire coupled directly into the waveguide.

Download Full Size | PDF

2. Materials and methods

In order to fabricate a symmetric waveguide chip, the refractive index of the bottom cladding material is required to be closely matched to that of the sample solution. In order to match the refractive index of water, however, very few materials are available. In this work, we used an amorphous perfluorinated optical polymer (Cytop) with n ≈ 1.34, high optical transmission in the 200-1600 nm wavelength range, and excellent chemical resistance. As a waveguide core material, we selected polymethylmethacrylate (PMMA) due to its relatively high refractive index n ≈ 1.49 and low autofluorescence [8].

Three-layer planar waveguide structures with sample wells in the top cladding for introducing aqueous samples were fabricated as follows: The lower cladding was formed by spin-coating a 4” polished silicon wafer with a 4-μm layer of Cytop CTX-809A (Asahi Glass Co.) after priming the wafer with AP3000 adhesion promoter (DOW Chemical Co.). A 25 nm layer of aluminum was deposited on the Cytop surface and subsequently removed by etching in NaOH solution. The fluoropolymer has carboxyl end groups that bind to the aluminum layer and remain at the surface after etching, reducing the hydrophobicity of the surface and improving the adhesion of subsequent layers. The Cytop surface was coated with a layer of polymethylmethacrylate obtained by spin coating 950PMMA dissolved in anisole (MicroChem Corp.). We deposited Al on the PMMA layer in order to protect it from the photoresist solvent and subsequently coated the wafer with positive photoresist (ma-P 1210, Micro resist technology GmbH). The photoresist was patterned using standard UV lithography to provide an etch stop layer covering the sample well regions and the underlying aluminum was removed elsewhere. The top cladding was formed by spin-coating a 4-μm layer of Cytop, followed by Al deposition and removal as above. The wafer was then coated with a layer of negative photoresist (ma-N 1410, Micro resist technology GmbH). The negative photoresist was used as an etch mask, patterned to expose the areas of the sample wells. The sample wells were created by reactive ion etching in O₂:CHF₃ (20:60 sccm) plasma at 100W RF power and 15 mtorr. The wafers were diced into individual chips, leaving the positive and negative resist layers in place to prevent contamination during dicing. A typical chip size was 10 mm × 10 mm with a 2 mm × 2 mm sample well in the center of the chip. Finally, the remaining photoresist and aluminum were removed in NaOH solution, leaving a clean exposed area of PMMA in the sample wells. No additional surface treatment was performed on the PMMA layer prior to cell growth.

MCF7 breast cancer cell line [9] was plated directly on the polymer-coated chips and cultured in DMEM/F12 culture medium (GIBCO) containing 5% fetal calf serum (GIBCO). After a 2-day culture period, cells were fixed in 3.5% Formaldehyde, permeabilized using 0.1% Triton-X-100 and prepared for immunocytochemistry. Cells were immunostained with monoclonal antibody against the transmembrane adhesion protein E-cadherin (HECD-1, Zymed) and Alexa Fluor 546 Goat Anti Mouse IgG1 fluorescent secondary antibody (Invitrogen).

 

Fig. 2. Schematic illustration of the experimental configuration. Excitation light was provided by a supercontinuum source (SuperK) and filtered using an acousto-optic tunable filter (AOTF). The light was coupled through a single-mode optical fiber into the planar waveguide. Fluorescence excited by the evanescent tail of the bound mode was picked up by the microscope objective. Transmission through the waveguide was monitored using a second objective to ensure optimum coupling.

Download Full Size | PDF

Our experimental setup is shown schematically in Fig. 2. Imaging of cells on waveguide chips was carried out using an upright microscope (Zeiss Axiotech Vario 100). The excitation light was provided by a SuperK Versa quasi-CW (80 MHz) supercontinuum source fitted with an acousto-optic tunable filter (Koheras A/S), providing single-mode output tunable from 450 nm to 700 nm. For the experiments described here, a central excitation wavelength of 551 nm with a FWHM bandwidth of 3.2 nm was used. The output from the SuperK was coupled into a single-mode polarization-maintaining optical fiber (HB450, Fibercore Ltd.) with an efficiency of about 30%, giving 150 μW of power for coupling into the waveguide. The output end of the fiber was manually aligned to the waveguide chip using a 3-axis translation stage for end-fire coupling to the waveguide. Waveguide chips with fixed cells immersed in water or buffer solution in the sample well, were covered with a 170-μm thick glass slide on top of a polydimethylsiloxane (PDMS, Dow Corning Sylgard® 184) sealing ring. Carbon powder was mixed into the PDMS before curing for absorbing scattered excitation light. The purpose of the sample well is to protect the chip facet from the sample solution during measurement and to avoid stray light from the in-coupling region from entering the sample.

High-resolution SWExFM images were obtained using a Zeiss C-Apochromat 63× water-immersion objective (NA=1.2) and a QuantEM:512SC electron-multiplying CCD camera (Photometrics). Scattered excitation light was blocked using a 570 nm long-pass filter (Schott glass). For comparison, we collected images of the same cells using a conventional fluorescence microscope (Nikon Eclipse E800, 100× NA=1.3 oil-immersion objective, 100W halogen lamp with a G-2A filter block, Nikon DXM1200F digital camera) and a confocal laser scanning microscope (Zeiss LSM 5 Pascal, 63× NA=1.4 oil immersion objective, 543 nm HeNe-line excitation, 560-615 nm bandpass filter for detection) adjusted for large depth of field (7 μm).

The intensity of all images displayed in the paper was normalized by linearly mapping 99.5% of the intensity histogram of the raw image data onto a 256-level gray scale (zero level unchanged). The pixel resolution of (5× oversampled) epi-fluorescence images was reduced by half using bicubic downsampling. No additional image processing was performed.

 

Fig. 3. Calculated parameters for waveguide chips. Effective index of the waveguide modes (bottom) and the corresponding penetration depths into pure water (top) for common fluorophore excitation wavelength. The limit for live-cell imaging and the onset of multimode behavior for the Cytop-PMMA chip structure are indicated.

Download Full Size | PDF

3. Results

3.1 Waveguide simulations

The penetration depth of the bound waveguide mode into the sample (defined as the distance where the electric field amplitude has dropped to 1/e of its value at the surface) can be written as [3]:

where n_eff is the effective index of the waveguide mode, determined by numerically solving Maxwell’s equations using the boundary conditions imposed by the layer structure. Using a one-dimensional mode solver, we determined the mode index for different core thicknesses, taking into account the material dispersion of the core, cladding and sample (water). Refractive index values used in the calculation were based on data provided by the polymer suppliers and the IAPWS [10]. Figure 3 shows effective indices and penetration depths calculated for excitation wavelengths corresponding to commonly used fluorophores. For clarity, we only show results for p-polarized modes that exhibit slightly larger penetration depths and better excitation efficiency than s-polarized modes [11]. Single-mode operation is observed up to core thickness of approximately 350 nm (blue) to 500 nm (red), while for thicker layers of PMMA, the structure also supports higher order modes. Multimode structures could potentially be used for multiple-depth imaging by selective mode excitation, as suggested by Horvath et al. [7].

 

Fig. 4. Comparison of waveguide-excitation and epi-fluorescence imaging. a) Cluster of MFC7 breast cancer cells labeled with antibody against the transmembrane protein E-cadherin imaged using symmetric-waveguide-excitation fluorescence microscopy. The estimated penetration depth of the excitation light is about 185 nm into the cells. Arrows indicate dark bands caused by scattering from particles in the waveguide film. b) Same cluster imaged using a standard fluorescence microscope (dashed line indicates field of view). c) Detail of the evanescent-wave image (pixel size corresponds to 200 nm = 0.7r). d) Detail of the epi-fluorescence image (pixel size corresponds to 50 nm = 0.18r after downsampling). Scale bar in all images represents 10 μm.

Download Full Size | PDF

Samples with variations in refractive index will experience a penetration depth depending on the local index. The refractive index immediately inside the membrane of live cells is larger than that of water (around 1.36, for example, in a HeLa cell at λ=633 nm [12]) giving an increased penetration depth of the evanescent field compared to water. In order for the light to be confined to the waveguide in the presence of index variations in the sample, the effective index of the guided mode must exceed the maximum value of the refractive index of the sample. For the currently investigated structures, this puts a lower limit on the practical waveguide core thickness for live cell imaging of approximately 100 nm (blue) to 150 nm (red), as indicated in Fig. 3. Larger penetration depths are associated with more weakly bound waveguide modes and in practice, therefore, light is more easily scattered out of the waveguide mode by sudden changes in the refractive index of the sample.

In our case, the waveguide core covered the entire chip area but the core layer can also be patterned to form channel waveguides for in-plane-localized excitation, improved stray light rejection, transfer of the excitation signal to several sample wells and even (if combined with, e.g., thermo-optic control) on-chip modulation of excitation intensity for time-lapse imaging, switching between sample wells, switching between excitation wavelengths, phase modulation for structured illumination [13], etc.

 

Fig. 5. Comparison of images obtained using waveguide-excitation and laser scanning microscopy. (a) Symmetric waveguide excitation with a calculated penetration depth of 185 nm for the excitation light. (b) Same cell cluster imaged using a laser scanning microscope with a large field of view. In the former case, the signal in the near-surface region is significantly enhanced and imaging time is reduced by an order of magnitude. Scale bars represent 10 μm in both images.

Download Full Size | PDF

4.2 Fluorescence imaging

As a validation study, we investigated clusters of cancer cells from a well characterized standard cell line (MCF7). The cells were labeled with an antibody against the trans-membrane adhesion protein E-cadherin to emphasize the difference between surface-bound excitation and epi-illumination. MCF-7 cells were seeded onto preformed chips and subsequently placed into the culture medium. Cell growth was observed on the exposed PMMA surface in the sample well as well as on the Cytop surface outside the well. The cells were allowed to multiply to form clusters of 5–10 cells before imaging. Fluorescence imaging was carried out on different instruments, using high-magnification oil or water-immersion objectives.

Figure 4 shows results from the fluorescence imaging on the waveguide chips, compared to conventional epi-fluorescence imaging carried out on a standard fluorescence microscope. The waveguide chip had a core thickness of around 450 nm which is close to the maximum thickness for single-mode operation at the excitation wavelength used λ_peak=551 nm), giving a calculated penetration depth of approximately 160 nm in water. Experimentally determining the actual penetration depth is not trivial, especially within the cells, and we have not carried out such measurements as part of this work.

In principle, filtering of the excitation signal is not required since it propagates perpendicular to the optical axis but in practice it is necessary in order to discriminate between the fluorescence signal and scattering from occasional particles present in the waveguide layer. In general, the SWExFM method gave clear images with a uniform fluorescence signal from cell membranes in contact with the waveguide surface and an increased signal in the cell-cell contact regions, as expected. No visible bleaching of the fluorescence signal was observed even during extended observation times of several hours, presumably due to a combination of efficient narrow-bandwidth excitation and optimum cut-off filter selection.

SWExFM imaging yields an essentially two-dimensional section of the sample and reveals details of the cell membrane adjacent to the waveguide surface that are masked by bulk signals in epi-fluorescence images (compare Figs. 4(a) and 4(b)). Furthermore, it gives a more clear definition of the cell morphology (compare details in 4(c) and 4(d)). In these images, the lateral resolution (defined as r=0.6λ/NA) is similar in both microscopes (approximately 300 nm), but the SWExFM images appear more grainy due to the pixel size of the camera, which corresponds to around 0.7×r, while in the epi-fluorescence images the pixel size is 0.09×r.

In some cases, we observe a shadowing effect in the SWExFM images due to localized scattering of the excitation light, which can typically be traced to particles in the waveguide film. This results in slightly darker bands appearing parallel to the direction of illumination, as seen in the lower half of Figs. 4(a) and indicated by arrows. This problem can be reduced or eliminated by improved filtering of the PMMA solution, by preventing contamination during processing and/or by illuminating the observation area from several directions simultaneously. Many observation areas on our chips were completely free of such dark bands (see Figs. 5(a)) which represents a significant improvement over previously reported work on WExFM imaging [4]. Shadowing can also arise from abrupt variations in the refractive index of the sample, especially for chips with larger penetration depths as noted above, or from imperfections at the incoupling facet which can be reduced by using index-matching liquid (n=1.34) between the fiber and the chip, by polishing the facets or by introducing a single-mode channel waveguide section at the edge of the chip.

We also compared the SWExFM technique to imaging with a laser-scanning microscope set to a large depth of field as shown in Fig. 5. The lateral resolution of the laser scanning system is given approximately by 0.4λ/NA = 160 nm and the lateral scan step was 130 nm. Again, the SWExFM yields detailed information from the sample region adjacent to the chip surface while the laser scanning microscope provides 3D information about the cell cluster which masks details of the near-surface region. The confocal principle of a laser scanning system can, of course, be used for imaging a thin section (down to about 0.5 μm in thickness) of the cell cluster close to the interface in order to eliminate bulk fluorescence. However, when signals originating from the near-surface region are of interest, the low photobleaching, simplicity and speed of the SWExFM method make it a more suitable technique, in particular for imaging fast processes in live cells such as protein-protein interactions. For comparison, the image in 5(a) was obtained in 125 ms (camera set to 1/3 of maximum gain) while the total scan time for image 5(b) was close to 10 s for a similar field of view.

5. Conclusions

In summary, we have demonstrated a new evanescent-wave fluorescence imaging technique based on waveguide excitation. By using waveguides with a symmetric cladding environment, in-coupling of excitation light is greatly simplified and image quality is improved, compared to previously reported waveguide-excitation methods. The SWExFM technique is the only reported waveguide-excitation method that delivers sufficiently uniform illumination for high-quality fluorescence imaging. While total internal reflection fluorescence microscopy is better suited for confining the excitation field to within 100 nm of the surface, the SWExFM method is more flexible with respect to substrate types, range of penetration depths, field of view, and decoupling of the excitation and imaging optics. Furthermore, SWExFM can, in principle, be carried out on any light microscope and it provides a number of possibilities for controlling the illumination on chip using planar waveguide circuits, e.g. for replacing electro-mechanical shutters in time-lapse imaging. The polymer platform used for fabricating the waveguide chips is also compatible with microfluidics or other lab-on-a-chip functionalities, e.g. for regulating the local environment during live cell imaging.

The SWExFM technique is a significant contribution to current microscopy techniques applied to measure and visualize biological samples such as cell cultures. The fact that the SWExFM technique can be used with conventional microscopes and can exhibit minimum photobleaching effects makes it an attractive choice for researchers focusing on various aspects of cellular processes.