1. Introduction

Fluorescence Correlation Spectroscopy (FCS) is a powerful tool for studying molecular dynamics [1]. Investigation of the dynamics of single molecules in a cell or on surfaces [2] requires that the thickness of the detection layer be small enough to minimize the contribution of the background, and that the fluorescence fluctuation associated with the change in a position of a molecule be maximized. In conventional confocal detection the axial and lateral dimension of the detection volume is in the order of microns, very large compared to the size of a biomolecule. Consequently, the background contribution is substantial and the relative fluctuation associated with the molecule leaving and entering this volume is small. Moreover, in confocal detection, a significant background is also contributed by the Raman scattering and reflection/scattering of the excitation light from sample components and glass surfaces.

Development of Total Internal Reflection Fluorescence (TIRF) correlation spectroscopy (TIRF-FCS) went a long way towards resolution of these problems [3]. In TIRF a sample is excited by the evanescent field generated by the total internal reflection of a laser beam reflecting from the interface between the glass coverslip and a sample (water). The intensity of the evanescent field penetrating into the sample drops exponentially with the distance from the interface. The typical penetration distance is a fraction of a wavelength, usually below 200 nm from the interface. The fact that the evanescent penetration is very shallow, much smaller than the Z-resolution in a confocal microscope, has been exploited to investigate ligand-receptor binding on membranes [4] and surface binding of biomolecules to silica surfaces [5]. TIRF was recently also combined with confocal detection [6] to substantially improve the sensitivity of TIRF-FCS [7, 8].

Recently we have shown that it is possible to further minimize the thickness of the detection volume and maximize the change in the fluorescent signal upon axial movement of a molecule by the application of Surface Plasmon Coupled Emission (SPCE). SPCE arises when a fluorophore couples with with a continous metallic surface to create group of oscillating electrons called surface plasmons [9]. This phenomenon has been exploited in a microscope operating in the Kretchmann (KR) configuration (KR-SPAM-FCS)[10]. Intrinsically, by applying a thin metal film (50 nm gold or silver) on a glass/sample interface and illuminating the sample with the laser beam at the Surface Plasmon Resonance (SPR) angle the observational volume can be drastically reduced. At the SPR angle the incident laser beam excites surface plasmons in the metal film, which produce an enhanced but quickly decaying evanescent field on the aqueous side of the interface. The enhanced evanescent field excites fluorophores very close to the metal surface which than couple to the surface plasmons and emit back to the glass through the metal film in a very directional manner [10]. We showed that the thickness of the detection volume is a product of the depth penetration of the evanescent field and distance-dependent fluorescence coupling to surface plasmons [11–15]. The observation volume is further reduced by metal quenching of excited fluorophores at immediate proximity (below 10 nm) to the surface. In microscopic applications, the fluorescent light which was emitted through the metal film at an SPCE angle can be easily collected by a high numerical aperture objective. A confocal aperture inserted in the conjugate image plane of the objective reduced lateral dimensions of the detection volume to about 0.5 µm and the effective detection thickness to about 35 nm [10].

In this report we demonstrate that significant reduction in thickness of the detection volume is also possible with a simple to realize Reverse Kretchmann (RK) configuration of excitation. We made the interesting observation that light originating on the sample side may pass through the opaque metal film only when it couples to surface plasmons in the film. This may happen only through the near-field interactions of excited fluorophores with free electrons in the metal film [10]. Such coupling is only effective at a subwavelength distance up to about 100 nm from the surface. Any other radiation originating above the coupling distance is efficiently reflected by the metallic surface. This allows direct sample excitation without a need to precisely control the incident angle. In addition the excitation from the sample side is very easy to realize in practice. Due to high reflectivity of the metal film [16] this arrangement has better background rejection than TIRF while preserving the thinner detection layer. SPAM offers no significant advantages over TIRF as far as fluorescence collection efficiency and brightness are concerned [17], but RK-SPAM-FCS has other desirable advantages over TIRF-FCS:

In summary: high refractive index coverslips (n=1.77) coated with a thin metal layer together with a high numerical aperture objective (NA=1.65) are the only requirements needed to realize RK-SPAM-FCS as an efficient method for studying molecular diffusion.

2. Material & Methods

2.1 Chemicals and solutions

Fluorescein-labeled microspheres (100 nm diameter) were from Molecular Probes (Eugene, OR). They were supplied at 3.6×10¹³/mL and used at 10x and 100x dilution for SPAM and TIRF experiments, respectively. All the other solvents were from Sigma (St Louis, MO).

2.3 Preparation of coverslips

A 48 nm thick layer of gold was deposited on the high refractive index coverglasses from Olympus by EMF Corp. (Ithaca, NY). 5 nm layer of silica was deposited on the top of gold to protect it from oxidation. 2 nm chromium undercoat was used as an adhesive background. For TIRF experiments, glass coverslides were used.

2.4Microscopic measurements

The schematic of the microscope is shown in Fig. 1. The beam of light from Ar laser (Model IMA101, Melles Griot, Carlsbad, CA) excites the sample placed on the stage of the inverted microscope (Olympus IX71) from the top (sample side). The sample is placed on a coverslip that is coated with gold film. Excited fluorophores near the surface interact with free electrons in the metal and produce Surface Plasmons propagating along the surface of the metal film. The high refractive index glass under the metal film constitutes proper conditions for far-field plasmons decoupling. In effect, fluorescent light decouples from the plasmons to emerge at the bottom of the coverslip at a precisely defined SPCE angle (which is smaller than SPR angle). Essentially, only the excitation energy (fluorescence) from molecules excited within 10 to 100 nm of the metal layer can couple to the plasmons and penetrate the metal layer via surface plasmons (see the insert). Obviously the excitation light also gives rise to fluorescence from molecules outside this layer, but the emission is unable to penetrate the metal layer (is reflected by the metal layer) and cannot enter the objective. This provides excellent background rejection and is the basis for volume reduction. The fluorescent light, emitted at SPCE angle, is collected by the objective (Olympus Apo 100x, 1.65 NA). The slide rests on a moveable piezo stage (Nano-H100, Mad City Labs, Madison, WI) controlled by a Nano-Drive. The fluorescent light is projected onto a tube lens, which focuses it at the conjugate image plane. A 50 µm confocal aperture or an optical fiber (whose core acts as a confocal aperture) is inserted at this plane. A pair of Avalanche Photodiodes (APD, Perkin-Elmer SPCM-AQR-15-FC) collects light emerging from the aperture. The TTL signal from the diodes is fed to correlator (Flex02-08D, Correlator Inc, Bridgewater, NJ) and the autocorrelation function is displayed on a PC.

 

Fig. 1. RK-SPAM-FCS microscope. The insert at right presents the concept for surface plasmon coupled emission. The excited fluorophores close to the surface couple (via near-field interactions) to surface plasmons in the metal. Propagated surface plasmons decouple on opposite side of the metal film as a far-field radiation and emit in directional manner.

Download Full Size | PDF

To demonstrate the superior background rejection by RK-SPAM we used excitation at 633 nm from 35 mW HeNe laser (Coherent 31-2140-000). The fluorescent light was collected by the objective (60x, NA=1.45, PlanApo, Olympus). Figure 2 compares TIRF and SPAM images of skeletal muscle myofibrils labeled with 100 nM of Alexa647-phalloidin (+10 µM unlabeled phalloidin acting as competition for actin) in the presence of the background contributed by 0.5 mM of Rhodamine 800 added to the sample. Control experiments showed that myofibrils looked normal under Nomarski illumination (not shown). The TIRF image (A) was completely dominated by the background, but the RK-SPAM fluorescence image (B) clearly shows the prominent features of myofabrils.

 

Fig. 2. Background rejection by SPAM. 0.5 mM Rhodamine 800 added as background obscures the image in ordinary TIRF (A). SPAM in RK configuration eliminates much of the background contribution (B). Myofibrils (0.1 mg/mL) were labeled with 100 nM Alexa647-phalloidin +10 µM unlabeled phalloidin for 5 min at room temperature, then extensively washed with buffer containing 50 mM KCl, 2 mM MgCl₂, 1 mM DTT, 10 mM TRIS pH 7.0. 633 nm excitation, 1.65 NA×100 Olympus objective, sapphire substrate, 1.78 Refractive Index immersion oil. The bars are 5 µm and 10 µm in A & B, respectively.

Download Full Size | PDF

2.5 Calculations

It is assumed that the exciting field is continuous in time and weak enough such that the time between excitations is much longer than the time between excitation and emission of the fluorophore. This means that the average number of photons emitted per unit time (average total emitted power) does not depend on the fluorophore lifetime, but only on the average time between excitations. The refractive indices of the metals used in the calculations are interpolated from [19].

3. Results

3.1 Autocorrelation in RK SPAM

Suspension of 0.1 µm diameter microspheres was diluted 10 times to 3.6×10¹² spheres/mL. The spheres were placed on a coverslip coated with gold. The intensities were measured in 160 µs intervals for 60 sec. Figure 3 shows a typical trace of intensity fluctuations. The fluctuations are caused by spheres entering and leaving the detection volume.

 

Fig. 3. The intensity fluctuations caused by diffusion of 100 nm spheres through detection volume in RK SPAM experiment on gold substrate. Inset: 100 nm microspheres in RK SPAM, 100x, NA=1.65 objective.

Download Full Size | PDF

Figure 4 shows corresponding autocorrelation function.

 

Fig. 4. Autocorrelation function of data shown in Fig. 3.

Download Full Size | PDF

3.2 Autocorrelation in TIRF

Suspension of 0.1 µm diameter microspheres was diluted 100 times to 3.6×10¹¹ spheres/mL. The spheres were placed on a glass coverslip. The intensities were measured in 160 µs intervals for 60 sec. A typical TIRF correlation function is shown in Fig. 5. The TIRF correlation function has a pronounced “tail”. It is most likely caused by particles sticking to the glass. Those particles are visible, even though the sample was exposed to laser light for ~2 min before measuring, in an attempt to bleach out this fluorescence. In contrast, the autocorrelation function in RK-SPAM on gold has a slowly decaying tail. The particles also probably stick, but they are invisible because of metal quenching – hence no tail.

 

Fig. 5. TIRF autocorrelation function.

Download Full Size | PDF

3.3 Data fitting

The emision profiles of RK and TIRF configurations are compared in Fig. 6. The difference in scale on Fig. 6 for TIRF and RK-SPAM shows that the excitation is stronger in TIRF than in RK-SPAM. However, the emission into the prism comes from a thinner layer in the RK-SPAM and also the background emission, originating further away from that layer, is much better rejected in RK-SPAM than in TIRF. Also, because the fluorescence is quenched near the surface in RK-SPAM there is no detected emission adjacent to it. This means that the detected emission comes from a thin detection volume floating slightly above the metal surface. This also means that the rejection of the emission of fluorophores near the metal surface should be better for small single fluorophores than for fluorescent beads which are more extended. In TIRF there is no such rejection of emission near the surface.

 

Fig. 6. Comparison of RK-SPAM on Au and TIRF. The modeling parameters for RK Au are d₀=170 nm, d₁=30 nm, and for TIRF d₀=135 nm. Red: The RK-SPAM detection signal=intensity at emission into the sapphire prism. Normal incidence (θ=0°). Blue: TIRF excitation. Angle of incidence (θ=49.5°) at maximum intensity of the evanescent wave. Metal is 50 nm layer of gold. The red curve is magnified by a factor of 10.

Download Full Size | PDF

The theoretical approach to fitting the curves is the same as used in [10]. This approach is an extension of the theory by Hassler et al [7]. It uses linear combinations of exponentials axially and a Gaussian laterally. The parameters are different but the principle is the same. The parameters d_n, ω_xy, and N (distances from the surface, the radius of the laser beam in x-y plane with Standard Deviation 2σ, number of molecules in the detection volume, respectively) can be fit to the experiments and will generally be different in RK and KR.

The equation for the theoretical correlation function due to multi-exponential detection function is given by:

$G\left(t\right)=1+\frac{1}{2N}\frac{\sum _{n,m}\frac{d_n{d}_m}{d_n+d_m}{A}_n{A}_m{R}_{n,m}\left(t\right)}{\sum _{n,m}\frac{d_n{d}_m}{d_n+d_m}{A}_n{A}_m}$

where

and ω_xy=2σ

The emission intensity is a product of factors of the emission intensities in the three coordinate directions. In the z-direction, i.e. normal to the plane, the factor is approximated by a sum of exponentials, $\sum _m{A}_m\exp \left(\frac{-z}{d_m}\right)$ , where the A_m’s are the weighting coefficients of the respective exponentials. This is discussed in Reference [9], and illustrated in Fig. 6 in that reference. R_mn is defined as the correlation corresponding to exponentials exp(-z/d_m) and exp(-z/d_n) in that sum. The correlation function G is then a linear combination of the mn R_mn ’s plus a constant. In case of bi-exponential:

A₀=1

(zero at metal surface)

A₁=-1

$G\left(t\right)=1+\frac{1}{2N}\frac{d_0{R}_{00}\left(t\right)-4\frac{d_0{d}_1}{d_0+d_1}{R}_{01}\left(t\right)+d_1{R}_{11}\left(t\right)}{d_0-4\frac{d_0{d}_1}{d_0+d_1}+d_1}$

The fit to the above equation for RK-SPAM-FCS and TIRF is shown in Fig. 7. The diffusion coefficient was taken as 4.12×10^-12 m²/s [20].

 

Fig. 7. Fit of theoretical FCS curves to experimental data. Bi-exponential for SPAM a)) and single-exponential for TIRF (b). Note that the y-axis is linear.

Download Full Size | PDF

The fitted parameters are (ω_xy fixed at 0.5 µm):

For RK-SPAM:

d₀=170 nm

d₁=30 nm

d=(d_o ²+d₁ ²)^1/2=176 nm (rms value)

N=17.5

For TIRF:

d=135 nm

N=4.8

There is a significant tail for the longer times for gold which does not fit into the theory of Hassler et. al [7]. We think this is due to adsorption of spheres to the metal surface (see Discussion).

4. Discussion

As presented in Fig. 2 the background rejection of RK-SPAM is superior to TIRF. As our calculation shows, at the same time the axial dimension is comparable to TIRF (Fig. 6). The major advantage of the SPAM approach is its excellent background rejection from the sample layers above the coupling range. The TIRF approach provides only a surface confined excitation and fluorescence is collected from the entire sample volume. A scattering of excitation results in sample excitation above the evanescent field layer, and this secondary emission is contributing to the total fluorescence signal. Even in combination with the confocal pinhole the z-axis resolution is over 1 um. In contrast, for SPCE based detection only emission that couples to surface plasmons can be transmitted through the metal film and collected by the objective. All other fluorescence is just reflected by the metal film. This makes RK-SPAM an attractive alternative to TIRF, because it is much simpler and less expensive to implement.

The axial dimension of RK-SPAM-FCS of 0.176 µm is ~2.8 times smaller than the lateral dimension (equal to the diffraction limit ~0.5 µm). This assures that the majority of fluctuations occur through one dimensional diffusion in the axial direction – an important consideration in applications where ligand-receptor kinetics or adsorption kinetics of small molecules at solid/liquid interfaces are measured [4, 5]. RK-SPAM, like TIRF, allows for a small detection volume, and consequently the relative fluctuation associated with a molecule leaving and entering the observational volume is large. The axial dimension of the detection layer is further reduced by quenching of fluorescence near the surface of a metal. An additional advantage of the thin detection volume of SPCE is that it minimizes the contribution of the background. We note that the fluorophores are not point-like. Spheres may not be able to penetrate so well into the quenched region near the metal surface. If the fluorophores were point-like, the detection volume thickness d would have been much smaller for SPCE than for TIRF.

An additional advantage of SPCE-FCS is that only fluorophores 10–100 nm away from the surface contribute to the fluorescence signal. Since the cell membrane is typically 10 nm thick, the method permits observing of the diffusion of fluorophores inside a cell, while avoiding the monitoring of events occurring at a cell membrane. While in cells where the distance of the major part of the membrane is greater than 50 nm from the surface (e.g. kidney cells, rat neurons or fibroblasts [21, 22]), the quenching offers no benefit, however it is a significant advantage in cells in which the distance between a cell and a substrate is small (~10 nm), such as erythrocyte ghosts on polylysine [18].

One disadvantage of RK-SPAM-FCS over TIRF is that a very high (NA=1.65) objective needs to be used for SPAM but that only a relatively high (NA=1.45) objective is sufficient in TIRF. This is because it is impossible to obtain a SPAM image of samples in aqueous solutions using a conventional TIRF objective.

If ω_xy were larger than 0.5 µm, the somewhat worse fit of the RK-SPAM-FCS autocorrelation function at longer delay times (Fig. 7) could have been caused by surface plasmon travel. However, the analysis suggests that this is not the case. The autocorrelation function was rather insensitive to ω_xy suggesting that the tail is most likely due to imperfect quenching of the beads that are adsorbed to the gold surface.

The average number of molecules in the detection volume, N, is a parameter fitted independent of d and ω_xy, because it is determined by the correlation function for small times:

$N=\frac{1}{2}\underset{\tau \to 0}{\lim }\frac{1}{G\left(\tau \right)-1}$

We note that the method can also be applied in bulk (i.e. not in a microscope), where the suppression of background noise can be enhanced by taking advantage of the directional and highly polarized character of a SPCE signal.