Surface plasmon resonance imaging of excitable cells

The custom-built multimodal imaging system we have developed combines SPR imaging with a number of microscopy sub-systems that include bright-field, epi-fluorescence, TIRM and SPRF. Figure 2 shows a schematic of the optical system and optical pathways. The system is also capable of performing electrophysiological measurements simultaneously using microelectrode technology. The rig was mounted on a floating optical table (Thorlabs, Newton, NJ, USA) to minimize mechanical perturbations, to which such sensitive methods are highly susceptible.

Zoom In Zoom Out Reset image size

A 640 nm light emitting diode (LED) was used in conjunction with Köhler illumination to illuminate the back focal plane (BFP) of the Nikon Oil Objective lens (CFI Apo TIRF 60×, NA  =  1.49, oil-immersion lens, Nikon Inc., Tokyo, Japan) with a uniform light intensity [32]. The BFP was monitored on a complementaty metal oxide semiconductor (CMOS) camera (Unibrain, 640  ×  480 pixels, 5.6 µm pixel size, Fire-i^™, Unibrain Inc., San Ramon, CA, USA), shown in figure 3(a).

Zoom In Zoom Out Reset image size

Reflection microscopy was performed with the 470 nm LED uniformly illuminating the BFP of a water-dipping objective (60×, NA  =  1.00, Nikon Inc, Tokyo, Japan). The reflected light passed through the same objective lens, and was imaged on to a CCD camera (Grasshopper3, 1920  ×  1440 pixels, 4.54 µm pixel size, S3-U3-28S5M-C Point Grey, Richmond, BC, Canada). Transmission microscopy was performed using the 470 nm LED and the transmitted light was imaged onto a second CMOS camera (640  ×  480 pixels, 9.9 µm pixel size, SV643M, EPIX Inc., Buffalo Grove, IL, USA). The 470 nm and 590 nm LEDs could also be used for exciting fluorescence, in combination with appropriate emission filters.

TIRM was achieved using the 640 nm LED, in conjunction with Köhler illumination to uniformly illuminate the BFP of the objective. The aperture diaphragm (D1) within the Köhler illumination was closed so that contrast was obtained in reflection by frustration of the critical angle (see supplementary information for information on the critical angle). The reflected light was imaged on the Unibrain camera.

When the objective pupil is filled with plane-polarized light from the 640 nm LED (figures 3(a) and (b)), in the sector where the azimuthal angle is 0° the polarization state is pure p-polarized light, while pure s-polarized light arrives at the sample where the azimuthal angle is  ±90°. The dark arcs visible in figure 3(a) show the angle of incidence where the light is coupled into SPs and is therefore no longer being reflected. There is no dark band in the vertical direction (±90°) due to the light being pure s-polarized, and therefore incapable of exciting SPs.

SPRF was performed by diffusional loading of Alexa Fluor^®, 680-dextran (3kDa, Invitrogen^™, D34681) into the cytoplasm of the cell of interest, using a micropipette attached to the cell soma. The fluorescent dye was then excited using the 680 nm superluminescent LED (SLED, 1 mW, Superlum Diodes Ltd, Cork, Ireland) which was tuned to excite plasmons beneath the cell of interest. The SPR fluorescence was imaged on the Point-grey Grasshopper3 camera. SPRF microscopy could be performed to confirm that the cells on the sample were in the evanescent field prior to functional SPR imaging.

The SPR imaging system was developed around a high NA (NA  =  1.49) objective lens and is able to exploit both angular and intensity modulation by either shifting the illumination angle (laterally translating the focal spot in the BFP) or holding this angle constant at the angle of maximum SPR gradient and monitoring the resulting intensity.

Angular modulation uses monochromatic light to excite the SPs, and the excitation can be seen as a dip in the angular spectrum. The detector senses a shift in the angle where the reflected light intensity is at a minimum. Angular modulation was achieved using the 680 nm SLED, generating light that is focused onto the objective BFP, resulting in an illuminating beam at the sample of an adjustable, narrow range of angles. A SLED was used over a laser or LED because they have a high spatial coherence but a low temporal coherence, allowing the light to be focused to a very tight spot [33, 34], without suffering from the effects of laser speckling [35]. Additionally, SLEDs have extremely low noise [36], which is advantageous in the high-speed acquisition of rapid signals such as electrical activity in cells.

The output of the fiber-coupled 680 nm SLED was collimated using an achromatic lens (f  =  100 mm) and passed through a polarizer so that only p-polarized light was incident on the sample. The polarized light was focused to a diffraction-limited spot on the BFP of the objective lens with a second achromat lens (f  =  75 mm). The diffraction-limited spot, along with the polarizer, reduces the level of background caused by the s-polarized light that is unable to excite SPs. The radial position of the illumination spot in the BFP dictates the angle of incidence at the sample. The light scattered off refractive index discontinuities in the sample when cells are cultured on the gold coverslips returns from the same elevation angle but all azimuthal angles. This illustrates that momentum is conserved during plasmon scattering, as shown in figure 3(d). To change the angle of incidence at the sample the focused SLED beam was scanned across the BFP of the sample using a stepper motor traveling at 0.1 mm s⁻¹. The reflected light was imaged on the EPIX camera at 24 fps, with an exposure time of 0.124 ms to provide reflection values as a function of the angle of incidence. Regions of interest were selected on the surfaces using Image-J and the z-axis profile was exported to MATLAB (figure 3(c)).

SPR sensing with intensity modulation (IM) works by fixing both the angle of incidence and wavelength and measuring the strength of the coupling between the light wave and the SP; detection is achieved by measuring the change in the intensity of the reflected light (figure 1(b)). Therefore, for IM detection the angle of incidence (θ) was fixed at the angle with the steepest gradient, ΔR/Δθ (around 30% of the minimum intensity recorded at the trough of the SPR dip, marked on figure 3(c)) and the variation of the reflected light intensity was monitored on the EPIX camera.

The glass coverslips (19 mm, Karl Hecht GmbH and Co KG, Sondheim, Germany) were treated with (3-mercaptopropyl) tri-methoxysilane (MPTS, Sigma Aldrich, 174617) to present thiol groups prior to thermal metal evaporation [37]. MPTS was used instead of the usual chromium or titanium adhesion layer to preserve the surface plasmon quality [38]. Silanisation was performed immediately after solvent and plasma cleaning. The plasma cleaned coverslips were immersed in a boiling solution of 50:1:1 isopropanol: MPTS: deionised water for 10 min, then rinsed with isopropanol, dried under nitrogen and transferred to an oven at 100 °C for 10 min. The boiling, rinsing and baking steps were repeated three times.

The MPTS glass coverslips were coated with ~50 nm thick gold by thermal metal evaporation. Gold evaporation was carried out using a thermal evaporator (E306A, Edwards, Burgess Hill, UK). Gold wire was placed in a filament and the chamber was pumped down to 10⁻¹ mbar with a rotary pump, then to 10⁻⁷ mbar with an oil diffusion pump, while being cooled with liquid nitrogen. A current of ~30 A was passed through the filament, so the metal melted and evaporated onto the glass coverslips held above the source with rare earth magnets. An Intellemetrics IL100 quartz crystal microbalance was used to estimate the evaporation rate and film thickness. The thickness of the gold coverslips was measured with a spectroscopic ellipsometer (Alpha-SE, J.A. Woollam, Lincoln, NE, USA) at a fixed incident angle (70°) and with a wavelength range between 380 to 900 nm.

The sensitivity of the planar gold surfaces was experimentally determined using sodium chloride (NaCl) salt-water solutions. Different concentrations of salt solutions were manually added to a chamber on top of a gold-coated coverslip using a 1 ml pipette whilst the reflected SPR light intensity response was measured. Deionized water was added as the first sample to find the reflection gradient maximum and the SLED was positioned here for the entirety of the experiment. While keeping all other parameters fixed, the refractive index of the media was adjusted using different concentrations of salt solution. The change in refractive index was calibrated with an Abbe Refractometer (Abbemat 200, Anton Paar, Graz, Austria). The reflected light from the sample plane was recorded using the EPIX camera (12-bit) at ~286 Hz. To allow the solution to stabilize, 30–60 s was left between each step. The resulting intensities were exported using Image-J and further analysis was performed in MATLAB.

For cell-based experiments on glass, the coverslips were treated with poly-L-lysine (PLL, Sigma Aldrich, P4707). The PLL coated glass coverslips were prepared by immersion in PLL solution for 5 min followed by thorough rinsing with deionized water.

For cell-based experiments on planar gold, the coverslips were treated with 11-Amino-1-undecanethiol (AUT, Sigma Aldrich, 674397), to generate a self-assembled monolayer [39]. First, the gold coverslips were solvent cleaned, followed by an oxygen plasma treatment.

The AUT was dissolved in ACS grade Ethanol at 1 mM, which was then poured into an immersion cylinder to completely cover the gold. Immersion time was not less than 18 h. Following immersion, the solution was drained from the cylinder and the gold was thoroughly rinsed, first with ethanol to remove the bulk AUT solution, and then with distilled water to remove the cytotoxic ethanol. The coverslips were dried thoroughly with N₂ before cell plating.

The 3T3 fibroblast cell line was maintained in Dulbecco's Modified Eagle Medium (DMEM, Gibco, 11960) with 10% fetal bovine serum (FBS, Gibco, 10270). Cells were subcultured and seeded on to AUT treated gold coverslips at a density of 0.3  ×  10⁶ cells.

The stem cell-derived cardiomyocytes were differentiated on AUT treated gold coverslips from human embryonic stem cells (hESC-CM) following the protocol in [40].

Cultures of primary hippocampal neurons were dissected from E18 Wistar rat embryos following a standard protocol [41]. The primary rat hippocampal neurons were plated on the functionalized coverslips with ~150 000 dissociated cells in 500 µl media.

3.1.1. Field of view (FOV).

Functional imaging of excitable cells using SPR microscopy requires performing measurements from a number of single cells to inspect the cell–cell interactions and to replicate single cell measurements. Therefore, the system has been designed with a wide-FOV to allow imaging of multiple single cells with subcellular resolution. This has been achieved by deliberately reducing the magnification of the imaging system.

The FOV was increased by mismatching the objective and tube lens to reduce the effective magnification of the 60×  oil-objective. The Nikon objective lens is designed for a tube lens with a focal length of 200 nm, however, we used a 60 mm tube lens. This reduces the effective magnification to 18×  and enlarges the FOV to up to 500 µm, depending on the size of the camera sensor.

The number of pixels in the EPIX camera sensor is 640  ×  480, at 9.9  ×  9.9 µm each, so the FOV has been increased to 350  ×  250 µm from 105  ×  80 µm. However, the back aperture of the objective lens restricts the usable FOV to a circular diameter of 320 µm. Figure 4 shows SPR images of mouse fibroblast cells (3T3) seeded on gold coverslips, showing the wide FOV and the potential to image relatively large populations of cells, for example, there are ~40 cells within figure 4.

Zoom In Zoom Out Reset image size

3.1.2. Spatial resolution.

The resolution of a conventional light microscope is limited by the diffraction limit [42]. The theoretical diffraction-limited resolution of the optical imaging modalities presented (transmission and reflection microscopy) for a 60×, 1.49 NA microscope objective with 680 nm incident light given by the Abbe limit is 0.28 µm. However, by under magnifying to 18×, the minimum resolvable feature, magnified onto a 2  ×  2 region of the sensor in accordance with the Nyquist criteria, is 1.1 µm. The average measured size of a typical cultured mammalian neuronal soma is ~16 µm (data not shown) [43] and a typical cardiomyocyte cell is about 100 µm long and 10–25 µm in diameter [44]. Even after undersampling the resolution is still more than adequate for imaging both cardiomyocytes and networks of neurons including many of the larger dendritic and axonal processes which are typically 1–2 µm in size [45].

The resolution in SPR sensors, however, is determined by the propagation length of the SPs. The plasmons propagate along the metal/dielectric interface for a distance decided by the propagation length before they decay back into photons. This limits the resolution of the system in the direction parallel to the propagation of SP waves. So, when the propagating plasmon reaches a refractive index boundary it is no longer supported and is re-radiated as a photon. This means that even though in theory the plasmon propagation length in water on a 50 nm gold film is typically ~3.1 µm [18], fine details, such as neuronal axons and dendrites can be resolved. Figure 5(c) shows a primary, hippocampal neuron cultured on a gold sensor imaged at 5 d in vitro (DIV). The fine, elaborating features visible in the figure demonstrate that axons and dendrites can be clearly resolved using SPR imaging in this regime. However, cell processes that are perpendicular to the propagation of the SPR are hardly resolved, due to the resolution limit of the SPR system, while they are clearly visible when they are parallel to the propagation of the SP wave [26].

Zoom In Zoom Out Reset image size

To demonstrate the resolution limitations of the SPR microscopy, we explored an alternative evanescent wave microscopy technique that is based on total internal reflection (figure 5(b)). Primary hippocampal neurons were cultured on glass coverslips treated with PLL. The glass surface was illuminated with a uniform disc of light from the 640 nm LED source objective BFP. This disc of light was stopped down using the aperture diaphragm (D1) to just above the total internal reflection critical angle to gain contrast.

Figure 5(b) shows that TIRM can resolve dendrites in both directions demonstrating a superior resolution compared to SPR. Although TIRM provides resolution and contrast based on the frustrated total internal reflection, it has a limited ability to resolve functional time-resolved information compared to SPR.

3.1.3. Refractive index mapping.

In biological samples, spatial differences in the lipids and proteins present result in a variation of the refractive index. These differences provide the contrast in an SPR image because as previously explained, refractive index discontinuities in the sample result in the propagating plasmons being re-radiated as photons.

To assess the spatial variation of the refractive index on our gold surfaces, the refractive index of each pixel in the FOV of figure 5(d) was determined by scanning the illumination angle, figure 6(a) and finding the angle of the minimum reflection for each pixel, figures 6(b) and (c). Then by solving the Fresnel equations for reflection and transmission in a multi-layer device, the refractive index was calculated (see supplementary information).

Zoom In Zoom Out Reset image size

Figure 6(a) shows the full image at various angles during the scan. The reflection as a function of scan angle at one x-position and one y-position was determined and shown in figure 6(b). The dark line in the projections represents the SPR dip and corresponds to the angle of incidence at the gold layer. The complete angle scan is shown at various points on the sample in figure 6(c). Note that at different angles of incidence plasmons are excited successively either under the cells or on cell-free background areas.

Figure 6(d) shows that the effective refractive index is reasonably homogeneous underneath the soma indicating the cell has adhered to the surface uniformly. The effective refractive index in a multilayer system is a function of the gap distance between the SPR sensor and the cell membrane and is weighted with an exponential decay [8, 10]. A larger effective refractive index indicates that the cell membrane has closely adhered to the surface, conversely a smaller measured refractive index indicates that the cell membrane is further away from the surface. Cell membranes are composed mostly of lipids and proteins which have refractive indices in the range of 1.46–1.54, so the refractive index we have measured here of 1.40 for the cell structure is less than previously reported values [46, 47]. The lower RI can be attributed to the lower RI cytosol (n  =  1.35 [46]) reducing the effective RI alongside the gap distance. It is important that the gap distance is less than the evanescent field distance and this will be addressed next.

As described in the introduction, the evanescent field generated at the interface can excite fluorophores placed within it. Exploiting this phenomenon, we were able to determine whether the neurons were close enough to the gold surface for their membranes to lie within the evanescent field by patching a microelectrode filled with dye into the cell. The dye within the cell was excited using both epi- and SPR-excited fluorescence, figures 7(a) and (b), respectively. Figure 7(b) confirms that the 680 nm SLED SPR system can excite the dye-labeled cell and therefore, the cell is within the evanescent field. This is a necessary prerequisite to allow functional imaging using SPR. 100% of cells tested (N  =  12) were found to lie within the evanescent field. This demonstrates that SPR can be used to excite either structural or functional fluorescent dyes in studies of neuronal (or other) cultures.

Zoom In Zoom Out Reset image size

The sensitivity of the gold surfaces on the SPM was experimentally determined by altering the bulk refractive index of the dielectric on a 50 nm planar gold sample.

The concentration of the solution on the surface on the substrate was changed to either increase or decrease the refractive index using NaCl solutions. The concentration values were chosen so that the refractive index increase or decrease was in step sizes of Δn  =  0.1 × 10^–3. To find the sensitivity, the angle of incidence was chosen such that the reflection gradient (ΔR) is at its maximum. Practically, it is difficult to set this parameter, so the dynamic range was first established by angle-scanning across the SPR dip, and setting the SPR illumination angle to the location of the greatest gradient (figure 3(b)). There is a trade-off between greater sensitivity and SNR, as at the angle of the maximum gradient the light intensity is much lower, reducing the Poisson-limited SNR. Considering this, the angle of incidence was fixed at around 30% of the dynamic range.

The results from all the experiments are summarized in figure 8. Figure 8 shows that the normalized light intensity (ΔR) increases linearly with increasing refractive index. The sensitivity of the sensor was characterized by the RIU as 2  ×  10⁻⁵.

Zoom In Zoom Out Reset image size

The experimentally measured sensitivity here is less than the value of ~10⁻⁶ RIU reported previously from intensity modulating SPR techniques in literature [21, 22]. However, the level of detection in these sensitivity experiments is limited by the exposure time because the power of the noise is proportional to the bandwidth. So, for the relatively small bandwidth used for these measurements (286 Hz), the total noise power may be reduced if a lower sample frequency is used thus increasing the SNR.

In order to demonstrate a potential experimental scenario where a multimodal system would be beneficial to the user, hESC-cardiomyocytes were cultured on gold surfaces. First, bright-field microscopy was used to locate a hESC-CM; see figure 9(a). The periodic contractions of a living, beating cell could be visualized by monitoring the change in SPR intensity on the EPIX camera. By taking the difference of each subsequent frame with the first frame, a stack was produced that shows the area where there is the greatest variation in light intensity ('hot-spot'); see figures 9(c) and (d). The background was corrected for by taking the difference of the 'hot-spot' with a sample of the rest of the cell which was not moving.

Zoom In Zoom Out Reset image size

Figure 9(e) shows that the SPR intensity changes in time with the contractions of the stem cell-derived cardiomyocyte ('signal') with a high signal-to-noise ratio. Two control ROIs are plotted. One on the background and a secondary location within the cell. The resulting SPR responses show that only a couple of areas are moving within the cell, with the rest being uniform. The area where there is the greatest change in light intensity appears to be around the nucleus of the cell (see supplementary video 1). It is possible that the source of the signal is from the stress fibres or contractile apparatus within the hESC-CM pressing on the nucleus causing it to move. The observed contraction localized in the centre of the cell is typical for hESC-CM's attached to rigid surfaces [48]. The control ROI on the 'background' was taken directly next to the beating cardiomyocyte, which shows no change in relative SPR intensity. This demonstrates that the change in SPR intensity is localized to the cell of interest and therefore, most likely due to refractive index changes or localized movement within the cell and not from movement causing mechanical waves throughout the volume. To fully utilize the multi-modal system, fluorescent labelling could be used for comparison of the label-free results with well-established labelling techniques.