BASICS OF LSPR TECHNOLOGY
Unlike surface plasmon resonance (SPR), a sister technology that uses flat gold thin films surrounded by sophisticated instrumentation,
LSPR technology resides in the sensor itself and requires only elementary components.7,8 An LSPR instrument schematic is shown in Figure 1A. White light from a tungsten halogen bulb is directed onto the nanostructured
LSPR film. The light interacts with the localized surface plasmons so some wavelengths are absorbed by the film. The reflected
light therefore has certain wavelengths missing and is analyzed by a spectrometer. Figure 1B compares the absorptions of a
gold nanostuctured LSPR film (red) and a regular gold thin film (dark yellow). The insets are camera images of both films.
The regular gold thin film absorption is featureless and can be described by Rayleigh scattering (~1/λ, in which λ is the
wavelength of light).4 In contrast, the nanostructured LSPR film exhibits strong absorption near 550 nm that is the linear superposition of Rayleigh
scattering and localized surface plasmon absorption. The LSPR has a natural width of ~80–100 nm, but its maximum position,
or λmax, can be tracked in real time with a resolution of a few picometers (Figure 1C).
Figure 1. Description of the LSPR technology. A) An instrument is composed of a source of white light, the LSPR film, and
a spectrometer to read the reflected light. For convenience, in this configuration of an instrument, the light arrives at
90° through optical fibers. The reflected light is sent to the linear array CCD spectrometer (Ocean Optics, USB 2000, with
SONY 2048 pixel chip that cover 430–730 nm range) using optical fibers for analysis. B) Absorption spectra of a regular flat
gold surface (dark yellow) and the LSPR nanostructured film (red), along with a true color of both films. Absorption is related
to the reflected signal through Abs = –log10 (ref/normalization). The peak position λmax is computed in real time by a proprietary algorithm. With current hardware, the standard deviation on the peak position is
~6 pm. C) Sensorgram showing the binding and elution of human IgG on a Protein A surface in real time at 3 Hz. D) To test
the robustness and reproducibility of the LSPR technology, the same experiment is performed 8 times on four surfaces simultaneously.
By computing the shift of the IgG injection after 60 sec, the coefficient of variation (CV) of the 32 readings is ~2.1%.
The position of the LSPR absorption, λmax, is sensitive to the nature of the interface between the nanostructured gold surface and its environment. For example, in
Figure 1C, the LSPR surface is used for the label-free and real time detection of a human IgG antibody binding to immobilized
Protein A. Before injection of the antibody, the LSPR signal λmax is constant, thus Δλmax = 0. When the antibody is injected, the resonance starts to red-shift, and reaches 3,200 pm after 3 min. After a brief rinse
with phosphate buffered saline (PBS), the solution pH is lowered to pH 2 so the interaction between the antibody and the Protein
A surface is disrupted. This produces a sudden decrease of the LSPR signal indicating that the antibody has been removed from
the surface. When the pH is re-established to its PBS value, the LSPR signal returns to its original value, i.e., Δλmax = 0, an indication that the sensor has been regenerated. Traces such as this one, also called sensorgrams, represent raw
LSPR data. Unlike SPR, LSPR does not require correction for bulk effect signals that occur during injections and rinses. This
is a result of the plasmons' localized nature and the fact that they extend and sense only 20–30 nm from the surface. Regular
surface plasmons of thin films sense changes up to 200–1,000 nm away from the surface.7–8
To illustrate the performance of current LSPR technology, a 4-channel biochip is used to monitor channels simultaneously in
real time when the same human antibody is injected. Figure 1D superimposes the sensorgrams when 8 cycles and binding-regeneration
are performed. Analysis of reproducibility is performed by computing the shift in λmax after an arbitrary 60 sec time lapse after the injection. For these 32 independent repeats, the shift reading yields a coefficient
of variation (CV) of ~2.1%. A CV of similar magnitude is measured across channels and across different sensors of the same
batch. This underscores the apparent consistencies possible for both the manufacturing and biofunctionalization of the nanostructured