However, an extrinsic tag such as a fluorescent molecule or a quantum dot always changes the properties of a host macromolecule. The significance of the change is often not known a priori. This is particularly relevant when studying properties of proteins [3]. Subtle changes in binding affinities and associated kinetics of protein molecules, by added physical properties of an extrinsic tag or through tag-induced conformational changes in protein molecules, can have a profound influence on some functions of protein molecules. Recognition of stereo-chemically modified double stranded DNA by specialized proteins in a living system is an example [4]. It is thus sensible to develop label-free detection techniques with adequate sensitivities to complement the fluorescence-based detection methods.

Imaging Surface Plasmon Resonance Spectroscopy (SPR) [5–7], Imaging Optical ellipsometry (OE) [8–11], and Reflectometric Interference Spectroscopy (RIFS) [12] are some of the optical techniques that have been explored to meet such a need. In their respective ways, these three label-free optical techniques in essence measure the same optical dielectric response of a thin film, consisting of molecules of biological significance on a solid substrate. As s result they detect same physical or chemical properties of the thin film such as thickness and mass density (including surface coverage) or changes of them during biochemical reactions. So far only Imaging SPR and Imaging Ellipsometry have been explored for high-throughput detection of microarrays of biological molecules [5–11].

In an SPR microscope equipped with a CCD camera, one typically uses the reflectance of a collimated, monochromatic light at a fixed incidence angle as the contrast and obtains an SPR image of an illuminated area on a solid surface covered with microarrays of biomolecules of interest. To maximize the sensitivity, the incidence angle is chosen to be on the falling or rising slope of the surface plasmon resonance peak. Over 2 orders of magnitude, the reflectance changes linearly with the thickness and the mass density (through the optical dielectric constant or refractive index of the film). The proportionality constant can vary from one gold-coated substrate to another and thus needs to be calibrated if quantitative information is required from such an image. Shumaker-Parry and Campbell were able to quantify the performance of such an imaging SPR microscope and achieved a detection limit of 2 × 10⁻⁵ RIU (refractive index unit) or 0.01 nm in detected protein thickness simultaneously over an area of 5mm × 5mm. It remains to be explored how such a microscope may be extended to examine microarrays over a much larger area.

Optical ellipsometry, in one form or another, measures changes in magnitude and phase of complex optical reflectivity (i.e., Fresnel reflection coefficient) in response to changes on solid or liquid surfaces [8]. If one is not concerned with magnetic and chiral properties of the surface layer, relevant reflectivity changes are those for p-polarized (transverse magnetic) and s-polarized (transverse electric) components of an optical beam. At oblique incidence, in response to a surface-bound change such as a biochemical reaction, the complex reflectivity changes disproportionately for p- and s-polarized light at a fixed optical frequency. As a result the magnitude and the phase of the ratio of the Fresnel coefficient for p-polarized light to that for s-polarized light, r_p/r_s ρ tanψ exp(iδ), change. Optical ellipsometry measures such changes.

In a typical imaging ellipsometer, one uses the polarizer-compensator-sample-analyzer (PCSA) scheme in which the phase compensator (C) is fixed and the polarizer (P) and the analyzer (A) are variable. The light beam reflected from an illuminated area on the sample surface (S) passes through the analyzer (A) and is subsequently imaged onto a CCD camera. P and A are rotated until the photocurrents are more or less minimized across the image region on the CCD to yield ψ and δ maps before for example a biochemical reaction takes place on the sample surface. During a subsequent biochemical reaction on the sample surface that results in a thickness change Δd in a surface-bound microarray feature, the corresponding change in the photocurrent under this off-null condition is proportional to (Δψ)² and (Δδ)², thus proportional to (Δd/λ)². The off-null photocurrents are monitored in real-time to detect the biochemical reactions. However the quadratic dependence of the off-null photocurrent on already small quantities Δψ and Δδ sets the detection limit of this type of imaging ellipsometer to roughly Δψ ~ 0.01° and Δδ ~ 0.01° (i.e., ~ 0.0002 radians).

For high-resolution imaging, we focus the illumination beam to a spot of 2.4 microns (the full width at 1/e² of the maximum of the intensity profile) on the microarray-covered surface and image the reflected beam from the spot onto a single photodiode detector. Using Rayleigh’s criterion, the image resolution is nominally 1.7 microns. To obtain a 2-D image of a microarray using Δ_p - Δ_s as contrasts, we mechanically move the stage in both x- and y-directions and record the values of Δ_p - Δ_s at each spot. The scan time is long in this configuration. For high-speed imaging with a spatial resolution of 15-micron to capture both the end-point and kinetics of biomolecular reactions on 1000-feature or 10000-feature microarrays (with the feature size in the range of 100 microns and feature separation in the range of 300 microns), we use cylindrical optics to focus the illumination beam into a line on the microarray-covered surface and image the reflected beam from the line onto a multi-element photodiode array (instead of a single detector). The scan along the line direction is then achieved by electronically interrogating the elements of the photodiode array at a rate at least 1000 times faster than the mechanical scan. This has enabled us to obtain an end-point image of 800-feature microarrays in less than 14 minutes as shown in Fig. 2. We should note that in application to microarrays, the spatial resolution in the range of 1.7 microns to 15 microns is more than enough since typical printed microarray features are between 80 to 150 microns, and typical separation between neighboring features are between 200 to 500 microns [1]. There is no observable edge effect arising from the finite size of the microarray features. In our present OI-RD microscopes, we have achieved the sensitivity of 0.1 Å. In term of resonance unit (RU) routinely used in SPR biosensors, we have achieved the sensitivity of 1 RU or 10⁻⁴ deg. Such sensitivity is adequate for our current microarray applications in high-throughput small molecule library screening for protein ligands and in high-throughput small molecule drug screening. And it is comparable to surface plasmon resonance microscopy as reported by Shumaker-Parry and Campbell [6].

In Fig. 3a, we show an Im{Δ_p - Δ_s} image of a 3×3 60-nucleotide (nt) oligomer microarray after it has reacted with a mixture of unlabeled 60-nt oligomers complementary to Column 1 and Cy5-labeled 60-nt oligomers complementary to Column 3. Each column is a triplicate of 120-micron spots of oligomers with a well defined sequence: 5'-TCACAAACCC GTCCTACTCT ACTAGCTGCA GTAGCCCCAC TGGTTCCCGT TTCCGATGTT-3' for Column 1; 5'-CCTTGTACCG CTGAGTTCAC ACCGACACAC CTCACCACAC TTACACCGTC CACAAAGAGA-3' for Column 2; and 5'-TTTCCATGCG GACCTACCAC CGTAGTACCT CGCAATGCCA GTGCAACAAG TACACCTGGA-3' for Column 3. The oligomer microarray is printed on a commercial poly-L-lysine functionalized glass slide. The printed oligomers lie flat on the glass surface, due to the electrostatic interaction between the negatively charged DNA backbone and the positively charged amine group at neutral pH. The excess oligomers on top of the saturated oligomer monolayer are removed by a washing step. The rest of the poly-L-lysine functionalized surface is blocked with succinic anhydride in borate-buffered 1-methyl-2-pyrolidinone before the microarray is subjected to the hybridization reaction. The image shown in Fig. 3a is the difference between the image taken after the reaction and the image taken before the reaction. The scale bar is 100 microns in length. The average optical signal change is 1.5×10⁻³ in Column 1 and indicates that 60% (Θ = 0.6) of the surface-immobilized oligomers have reacted with complementary partners. The hybridization is well resolved without extrinsic labeling. The average optical signal change in Column 3 (the positive control column) is also 1.5×10⁻³ [6]. This is expected since the wavelength of the He-Ne laser used for illumination of the OI-Rd microscope is far from the absorption peaks for Cy5 dye.

In Fig. 3b, we display an Im{Δ_p - Δ_s} image of a 4×4 protein microarray after it has been exposed to un-labeled goat antibody against rabbit IgG. Each column of the microarray is a titration series of one type of proteins: bovine serum albumin (BSA) for Column 1; rabbit IgG (RB) for Column 2; mouse IgG (MS) for Column 3; and human IgG (HM) for Column 4. They are printed as 150-micron spots on an epoxy-functionalized glass slide. The excess of the printed proteins is removed by washing steps and the rest of the epoxy-functionalized surface is blocked with BSA. Again, the image shown in Fig. 3b is the difference between the one taken after the reaction and the one taken before the reaction. The scale bar is 100 microns. Without fluorescent labeling, the differential image reveals clearly the specific antibody-antigen capture with a good signal-to-noise ratio. The change in Im{Δ_p - Δ_s} (4×10⁻³) indicates that roughly 20% (Θ = 0.2) of a saturated monolayer of rabbit IgG have reacted with the goat anti-rabbit IgG.

In Fig. 3c, we show an Im{Δ_p - Δ_s} image of another 3×3 protein microarray after it has been exposed to un-labeled streptavidin. Each column is a triplicate of 150-micron spots of same protein: human IgG for the first column; BSA-biotin complex for the second column; and BSA alone for the third column. The BSA-biotin complex is synthesized for the purpose of “immobilizing” small molecules such as biotin to epoxy-functionalized glass slide with BSA as the anchor. A linker molecule is inserted between BSA and biotin to minimize the effect of BSA on the affinity of biotin. The image shown in Fig. 3c is the difference between the one taken after the reaction and the one taken before the reaction. The specific reaction of streptavidin with BSA-biotin complex is clearly shown in the differential OI-RD image. The bright spots in the images shown in Fig. 3 are present in the microarrays instead of noise. They are dust particles from the ambient and residuals from the processing of the microarray. Streptavidin is roughly a spherical molecule with a 5-nm diameter [20, 21] and a mass volume density of ρ_d = 1.35 g/cm³. When packed in square lattice to a full monolayer, the surface number density of streptavidin is 4.0 × 10¹² molecules/cm², and the surface mass density is 4 ng/mm². Assume that the optical dielectric constant for streptavidin at λ = 532 nm is ε_d(streptavidin) = 2.51 (n_d(streptavidin) = 1.584), we can deduce the surface coverage Θ of the reacted streptavidin using Eq. (1) from the change in the OI-RD signal. After taking into account of the effect of the incidence angle at 59°, the change in Im{Δ_p - Δ_s} (1.7×10⁻²) shows that the coverage of the streptavidin is 1.4 × 10¹² molecules/cm², namely, Θ = 0.35. We should note here that the surface coverage obtained this way is an approximate to the true value as the former depends on the assumption of ε_d = 2.51, the packing geometry, and the validity of Eq. (1). The true relation between the OI-RD signal and the surface coverage can be established in our future investigation by a calibration against a fluorescence method or a nuclear method.

To illustrate that the OI-RD technique is capable of measuring protein-small molecule binding reactions on functionalized glass surfaces in real time, we have performed a series of measurements on bovine serum albumin (BSA) on epoxy-coated glass slide. Because of the small dielectric contrast or refractive index difference between glass and water, the reflection from the interface is very weak, making ellipsometry measurements of biomolecular processes at the interface a difficult task. In our experiments, the functionalized side of the glass slide is in contact with the buffer or the BSA solution as a part of a fluid cell. The other side of the slide is in air. The illumination laser beam is incident on the functionalized surface through the air side and the reflected beam is detected as illustrated in Fig. 1 with a single detector. The experiment begins with 1× phosphate buffered saline (PBS) solution in the fluid cell, and then BSA is added and quickly mixed with the 1×PBS (using a magnetic stir in the cell) to make it a 7.2 µM BSA solution (1.0 mg/ml) in less than 6 seconds. In Fig. 4, we show Im{Δ_p - Δ_s} from the glass-solution interface before and after BSA is added (at t = 0) to the 1×PBS. Except for the first 6 seconds when the mixture is being homogenized, the OI-RD signal shows the uptake of one monolayer of BSA that fully covers the epoxy-functionalized glass. The saturation level at 0.008 is not changed when the BSA solution is replaced with 1×PBS, indicating that the uptake or adsorption of BSA is irreversible. This uptake curve compares well with the observation of BSA adsorption from aqueous buffer on a gold-coated substrate reported by Jung et al. using an SPR microscope [7]. By subsequently exposing the BSA-covered glass slide with Cy5-labeled IgG molecules, we were able to confirm that the saturation level at 0.008 in Im{Δ_p - Δ_s} corresponds to one full monolayer of BSA that covers 98% of the epoxy-coated surface. From Eq. (1), we find that the signal level of 0.008 corresponds to a uniform layer of BSA with thickness of 1.4 nm and an effective dielectric constant of ε_d(BSA) = 2.5 or a refractive index of n_d(BSA) = 1.58. It is difficult to determine independently the refractive index and the thickness of the BSA layer. By keeping the magnetic stir on during the entire experiment, we maintain a constant BSA flux toward the glass surface (namely, 6 seconds after the BSA is added.) Fig. 4 shows that the uptake follows the Langmuir kinetics, namely, (a) the uptake rate is proportional to the probability of an impinging BSA to strike an open epoxy-coated surface; (b) the probability of a striking BSA molecule to bind to the open epoxy-coated surface is a constant.