SPRI apparatus and resolution
The SPRI apparatus is described schematically in Figure 1A. The optical design is fundamentally similar to existing SPRI instruments [7, 15, 16], however, the specific configuration is designed to make long term live cell based measurements. By positioning the SPR sensor surface horizontally, incident light is launched from below the sample into a gold-coated SF-10 glass slide which comprises the cell culture surface of an enclosed chamber. This configuration allows cells to be added into the chamber and imaged on the substrate, and facilitates transfer of the cell chamber between the SPR imaging apparatus and an inverted optical microscope. The protocol for SPRI image collection and analysis is described in the Methods section.
Figure 1B provides an indication of the lateral resolution of the SPRI instrument. Polydimethylsiloxane (PDMS) samples, photolithographically patterned with 5 μm circles or 2 μm lines, were placed directly onto SF-10 slides coated with gold and mounting onto the SF-10 prism. The 2 μm patterned lines are arranged obliquely to the direction of the surface plasmon propagation. The SPR images (Fig. 1B) arise due to the difference in the refractive index of air and the refractive index of the PDMS in contact with the surface. Images were acquired using 470 nm and 630 nm incident light as indicated. The 5 μm circle images and line scans are normalized and the normalized reflectivities are plotted displaying a similar spatial sensitivity. The images of the 2 μm lines and line scans are displayed using the raw reflectivity scale to show that the 630 nm SPR image has a larger reflectivity response than the 470 nm SPR image. These results also demonstrate the lateral resolution of the SPRI system to be at least 2 μm based on the ability to distinguish 2 μm features spaced 2 μm apart. This value is in agreement with an estimate made in a previous report using a similar prism based configuration and numerical aperture lens [17].
Figure 2 shows simulated SPR reflectivity curves to demonstrate the linear relationship between reflectivity and refractive index change. The response is shown for both 470 nm and 630 nm incident light. Reflectivity reaches a minimum when the angle of the incident light is optimal for coupling into plasmons. The SPR reflectivity curves are simulated using published optical constants [18–20] and assuming a four-layer Fresnel model [21] consisting of the prism, the gold layer, adsorbed protein, and aqueous solution that corresponds to the experiments described below. Three curves for each wavelength were used to simulate a compact protein layer of refractive index 1.45 and thickness of 0 nm, 10 nm and 20 nm. The mass/area corresponding to each protein layer is 0 μg/cm2, 1.4 μg/cm2 and 2.7 μg/cm2 respectively. SPR imaging is achieved by using a parallel beam of incident light projected onto the surface at a single angle. The SPR imaging angle probed is as indicated by the vertical line in Figure 2. The reflectivity differences at this angle due to the protein layer thickness is displayed in the inset and shows a linear response to mass of material at the interface. Several published reports have also shown a linear relationship between SPR image reflectivity changes and refractive index changes at the surface within a specific range [7, 13, 16].
In this study, we employed the use of both 470 and 630 nm incident light. According to theory, adjusting the wavelength of the incident light affects spatial resolution, refractive index sensitivity, and evanescent wave penetration depth [22]. However, in practice, we could detect little difference in spatial resolution at the two wavelengths. The main difference was that 470 nm SPR images had a larger depth of field and therefore a larger portion of the image was in focus at the maximal spatial resolution. Also, even though Figure 2 indicates that 630 nm light provided greater sensitivity, the sensitivity at 470 nm was maximized by using a gold thickness (30 nm versus 45 nm) more optimal for 470 nm signal response and image analysis was used to average many pixel values into one measurement. This enabled both wavelengths to be sufficiently sensitive and provide comparable results for protein mass/area. For the data shown here, we used 630 nm incident light for monitoring time-dependent deposition of protein coupled with difference imaging to obtain a signal-to-noise of ~3 ng/cm2. For cell measurements we used 470 nm incident light with a signal-to-noise of ~20 ng/cm2. Because the cell mass provides a very large SPR signal, the smaller response of 470 nm light helped insure linearity of SPR signal to mass. The SPR evanescent wave penetration depth (defined as decay of the field to 1/e) for 470 nm light is calculated to be ~60 nm compared to ~150 nm for 630 nm light. Thus, about 63% of the intensity of the field for the 470 nm light is penetrating to a depth of 60 nm above the gold surface. The lower penetration of 470 nm light allowed us to image cell-matrix interactions while minimizing contributions to the SPR signal from the cell interior.
Substrate patterning and fibronectin deposition
SPRI allows monitoring of the deposition of the extracellular matrix protein, fibronectin, which serves as the ECM for subsequent cell measurements. Microcontact printing [23] was used to prepare 300 μm by 300 μm square patterns of a methyl terminated alkanethiol monolayer on gold. The intervening areas were backfilled with a protein resistant polyethylene glycol (PEG)-terminated alkanethiol, followed by adsorption of fibronectin to the methyl terminated region. Figure 3A shows the resulting SPR image of the pattern of fibronectin areas (lighter squares), each surrounded by darker (thinner) areas of PEG-alkanethiol. Figure 3B shows the kinetics of fibronectin adsorption to the hexadecanethiol coated surface followed by SPRI using 630 nm incident light and difference imaging [15] as described in the Methods. Figure 3B shows that there is little adsorption to the PEG-thiol coated areas. Fibronectin appears to saturate the surface after exposure of about an hour. Because the refractive index changes are directly proportional to mass of protein adsorbed, the rate of protein adsorption and the total protein coverage can be determined (see Methods section). Fitting the data to a homogeneous Langmuir model yielded a rate constant for adsorption of 2.7 × 104 L mol-1 s-1, and a mass/area of approximately 390 ng/cm2 of fibronectin. This value is in good agreement with radiolabeled fibronectin adsorption onto a 1-dodecanethiol coated surface [24]. At saturation, the standard deviation for protein bound was approximately 3 ng/cm2 of protein, providing an estimate of signal-to-noise ratio of greater than 100.
Following fibronectin adsorption, the patterned sample was exposed to serum-containing cell culture medium and measured by SPRI at 630 nm for over 400 minutes at 37°C. The kinetics of adsorption of serum components to the fibronectin coated surface is displayed in Figure 3C and shows approximately 280 ng/cm2 of serum proteins added to the surface at saturation; this result is consistent with a previous report that observed serum albumin adsorption onto fibronectin coated substrates [25]. Negligible adsorption of protein to the PEG-thiol coated areas was observed. The SPRI reflectivity signals were converted to mass/area of serum proteins bound based on the partial specific volume for globular proteins [26, 27]. Although SPRI cannot independently identify the specific proteins absorbed, the partial specific volume for globular proteins is highly conserved [28], and therefore the amount of protein adsorbed can be estimated with good accuracy, without requiring specific knowledge of the composition.
Fibronectin and serum proteins appear to adsorb homogeneously onto the hexadecanethiol patterns as shown in line scans of SPR intensity across these areas (Fig. 3D). This observation is shown here to provide contrast to the non-uniform distribution of deposited protein following the addition of cells to the matrix as shown in following sections.
SPR imaging of cells and patterned fibronectin
Figure 4 shows the results of cells seeded on the patterned substrates described above. Figure 4A shows an SPR image (right) of vSMC collected using 470 nm incident light and corresponding phase contrast (left) and fluorescence (center) images for the same field of view. The fluorescence image was produced by staining fixed cells with Texas Red maleimide [29]. By comparing the images in Figure 4 it is apparent that of the three imaging modes, only the SPR image allows visualization of both the protein patterned regions as well as the cell-substrate contacts. This highly sensitive discrimination is shown quantitatively in Figure 4B. In each image of Figure 4A, a line is drawn (lower right in all images) that spans a cellular region, a fibronectin region, and a PEG-thiol region; intensities under those lines are plotted in Figure 4B. The line scan for the SPR image shows three intensity plateaus corresponding to each of these regions. The fluorescence and phase contrast line scans, in contrast, show only an intensity change for the cellular regions.
The pseudocolor scale in Figure 4 indicates the reflectivity values associated with areas of different mass. It can be seen that the PEG areas have the lowest mass, with a reflectivity on the order of 0.16. The adsorption of protein to the hexadecanethiol areas results in a higher reflectivity ranging between 0.18 and 0.19, and the greatest mass and highest reflectivity values are associated with regions of cell-substrate interactions. Figure 4 shows that a number of cells appear to span the PEG-thiol regions. Since the cell thickness is much larger than the depth of penetration of the plasmon evanescent wave, SPRI contrast is putatively generated by the distance that the cell membrane is from the surface. Thus, for cells that span PEG locations (Fig. 4A, arrow) the region of the cell above the PEG often has less intensity, suggesting that the part of the cell that spans the PEG-thiol is further from the substrate than the rest of the cell.
The optical contrast provided by SPRI allows label-free segmentation of cell areas. Segmentation of cells (Fig. 4C) based on the SPR signal (right) compared to segmentation based on staining with Texas Red maleimide (left) shows that both SPR and fluorescence images provide similar outlines for cell area. Subtle differences are observed, however, in specific areas that provide a strong fluorescence intensity, but little SPR signal (arrow in Fig. 4C). These observations suggest that portions of cells reside farther away from the interface than is detectable by the SPR evanescent wave.
SPRI of cellular deposited material
It is known that cells modify their extracellular environment and can deposit, attach, secrete, and replace material on the substrate to which they adhere [30–33]. Fluorescence detection can be used to chemically identify individual components in each of these processes. SPRI, in contrast, does not chemically identify specific proteins at an interface, but as we will demonstrate, can be used to generate an overall mass/area measurement of cell-deposited material. The amount of putative cell deposited material attached to the fibronectin patterned surfaces during 24 h of cell culture was quantified by SPRI using 470 nm and 630 nm light. Representative contrast-adjusted images (Fig. 5A) show that protein deposition is greater in fibronectin areas that are more heavily populated by cells (>25% of the area is occupied by cells) than in squares in which there are few cells (<10% of the area is occupied by cells). The SPR signal in 5 regions of high and low cell density were averaged and compared. Using a partial specific volume between that for globular and fibrillar proteins, we determined that, on average, an additional 120 ng/cm2 of protein was associated with regions of high cell occupancy versus low cell occupancy. The PEG-thiol areas remained essentially unchanged during this time. A schematic of the layers of protein deposition detected by SPRI is shown in Figure 5B. Approximately 390 ng/cm2 of fibronectin adsorbed onto hexadecanethiol, 280 ng/cm2 serum proteins adsorbed onto fibronectin, and an additional 120 ng/cm2 protein was deposited onto the fibronectin patterns at regions of high cell density. The layer schematic does not account for possible displacement of fibronectin by serum proteins, which has been observed to be minimal on hydrophobic substrates [25].
Unlike the adsorption of fibronectin and serum proteins from solution, which was largely spatially homogeneous as shown in Figure 3D, the deposition of additional protein after cell seeding was not uniform. Spatial analysis of cell-derived protein was performed by determining the amount of protein deposited as a function of spatial proximity to cell edges. SPR images of fibronectin-coated areas occupied with varying densities of cells are shown in Figure 5C. Cell edges were defined with a single threshold value, and used to create a contour outline of each cell, which was then sequentially dilated in 1-pixel steps (1-pixel = 1 μm) out to 120 μm from the cell edge (Additional File 1). The image processing used to generate these images is described in Methods. In Figure 5C, the contour line is shown at 40 μm from the cell edge, and is colored magenta when it overlays a fibronectin area and green when it overlays a PEG-thiol region. The SPR intensity values under the magenta or green lines were averaged for each contour trace around the cell periphery. Figure 5D shows the plot of deposited protein coverage versus distance from the cell edge for fibronectin coated areas. The amount of material on these regions decreases with distance away from a cell. This observation is consistent with the idea that more cell-derived protein would be deposited closer to cells than further away. The high surface coverage of fibronectin used in this study (390 ng/cm2 = 5200 molecules/μm2) was chosen to minimize cell migration [34]. Hence, we surmise that most of the material detected is likely cell-secreted and/or cell-assembled. Notably, no measurable protein deposition onto the PEG-thiol areas was detected (Fig. 5E, which is a plot of the data depicted by green lines from Fig. 5C). These data demonstrate that the difference in reflectivity with distance-from-cell edge is due to material that deposits onto fibronectin but not onto PEG.
Additional file 1: Quantifying protein deposition at cell periphery. Movie showing image analysis procedure used to quantify protein deposition at the cell periphery after 24 h in culture by dilating the cell contours to extract protein coverage versus distance-from-cell edge. Also, shows image analysis for low (<15%) and high (>25%) cell density regions. (MOV 1 MB)
While the flat response of the PEG distance-to-cell-edge coverage plot in Figure 5E shows that no measurable cell derived material is being deposited onto the PEG regions, it also serves as an image analysis control to insure influence from edge effects is minimized. This also shows that other effects such as lateral decay of SPR signal are not observed to bleed into low signal regions.
We also observed that in areas of low cell density (<15% of area corresponding to cells) and high cell density (>25% of area corresponding to cells) that there was deposition of protein that was dependent on distance from cell edges (Additional File 1). This distance dependence was observed to be more pronounced when applied to regions of lower cell density. The sometimes more subtle distance dependence at high cell density is perhaps due to the influence of material secreted by neighboring cells.
To assess the time-dependent deposition of cellular protein, Figure 6 shows analysis of the addition of protein to the substrate by live cells at different time points (Fig. 6A), These data were generated using the approach applied to the data in Figure 5. Cells were identified by simple thresholding as described earlier, and this threshold provided indication of the cell edges. In this experiment, cells were seeded onto 500 μm × 500 μm fibronectin patterns that had been pre-incubated with serum-containing media overnight to allow serum proteins to adsorb to the fibronectin areas. Immediately after addition of cells, no additional material is observed to adsorb onto the fibronectin substrate and the surface coverage is observed to be evenly distributed spatially (Fig. 6B). At 30 min after addition of cells, increased SPR signals, indicating the presence of newly adsorbed protein, appear in areas close to the cell edges. At 24 h after seeding, additional protein mass is detected and is distributed spatially as a function of proximity to the cell edge. Similar to the analysis on fixed cells, most of the cell-derived additions to the substrate appear within 40 μm of the cell edges. This suggests that this additional material is derived from cells, and could be contributed to by cell secretions, cell material left behind following lamelopodia retraction, and/or cell assembly of ECM proteins from the serum.
As verification that the additional additive material is completely cell derived, vSMC were plated in growth medium containing 100 μg/ml cycloheximide (CHX), a protein synthesis inhibitor [35]. The approach allows us to observe the interaction of the cells with the fibronectin patterned surface in the absence of new protein synthesis. Figure 7A shows a representative image of CHX treated cells that are fixed 24 h after plating and Figure 7B shows an image of untreated cells. The data were generated using the approach applied to the data in Figure 5. At 24 h after plating, additional mass is detected proximal to the cell edge for the untreated cells but no additional material is observed to adsorb onto the fibronectin surface for the CHX treated cells (Fig. 7C). From this, we conclude that the additional material is derived from cells.