Nonlinear structured illumination microscopy by surface plasmon enhanced stimulated emission depletion

Han Zhang; Ming Zhao; Leilei Peng

doi:10.1364/OE.19.024783

1. Introduction

1.1 Superresolution microcopy methods

Superresolution fluorescence microscopes greatly expand our ability to study cellular structures [1–4]. Existing methods can be grouped into two categories [5]: the first category, including stochastic optical reconstruction microscopy (STORM) [6,7], photoactivated localization microscopy (PALM & fPALM) [8,9] and ground-state depletion with individual molecule return microscopy (GSDIM) [10], achieves precise localization by detecting single molecules and switching on and off individual fluorophores. The second category, including stimulated emission depletion (STED) microscopy [11,12], reversible saturable optical fluorescence transitions (RESOLFT) [13], and structured illumination microscopy (SIM) [14], treats a fluorescent structure as a continuous density distribution of fluorophores, and uses modified illumination and/or strong nonlinear effects to overcome the diffraction limit. All existing super-resolution methods have their pros and cons. PALM, STORM and GSDIM have high sensitivity and resolution but limited speed. Laser-scanning STED is capable of high-speed imaging, but being a point-scanning technique, has the trade off between field of view and speed. Laser scanning STED has demonstrated live cell imaging [8,15,16]. However, effects of a strongly focused STED laser on cell viability and function have yet to be fully studied [17]. Saturated structured illumination microscopy (SSIM) [18] improves the resolution to 50 nm at the cost of strong photobleaching.

1.2 Nonlinear SIM with STED effect

At present, the best high-resolution microscopy method for live cell imaging is still linear SIM [19], although its resolution is limited. Nonlinear SIM offers an attractive way to improve resolution without modifying the SIM microscope. In theory, any nonlinear effects can be used in nonlinear SIM to achieve <100 nm resolution. STED, among several nonlinear effects, was suggested previously for nonlinear SIM [18]. Theoretical study of standing-wave total internal reflection microscopy (SW-TIRM) also suggest that adding an standing-wave STED field to a standing-wave excitation field can further increase the resolution of SW-TIRM [20]. At present, only saturation of fluorescence excitation (SSIM) [18] and photoswitching [21] were applied to nonlinear full-field superresolution imaging. However both nonlinear effects are unsuitable for live cell imaging. Nonlinear SIM requires tens to hundreds of raw image frames taken under different illumination pattern. Thus, the photoswitching time limits the imaging speed of photoswitching SIM, as experienced in RESOLFT, PALM and STORM. SSIM imaging can be performed at higher speed, because fluorophore response to excitation saturation is switched within one excitation-emission cycle, a few nanoseconds for most common fluorophores. However, SSIM has limited use in biological study due to its severe photobleaching. In SSIM, fading of the sample prevents weak structures from being seen. At the saturation excitation level, single fluorophores bleach within milliseconds, much faster than the current imaging speed of SSIM. Therefore SSIM is not applicable to single fluorophore imaging in its current form.

In this paper, we examine the feasibility of applying STED to nonlinear SIM. Similar to excitation saturation, STED effect has a switching time of ~200 ps [13], which allows high frame rate acquisition. Different from SSIM, STED effect is generally observed under an excitation intensity that is two magnitudes lower than the saturation level, which allows repeated image acquisition [16]. STED-SIM will also have better resolution than SSIM under similar signal-to-noise conditions. The resolution of a nonlinear SIM technique is strongly influenced by the behavior of the nonlinear effect. The diffraction limit of a microscope acts as a low-pass filter on the Fourier domain of the image, which limits the lateral resolution by cutting off high spatial frequency components of the image. SIM uses the Moiré pattern principle to acquire Fourier components beyond the cut-off [5]. Under linear SIM, the Fourier bandwidth is expanded by 2, thus the resolution is improved by 2 times to ~100 nm. Nonlinear SIM benefits from higher order harmonics of the interference patterned field produced by the nonlinearity, which allow further expansion of the Fourier bandwidth. The SNR in the expanded Fourier band is dependent on the strength of the high order harmonics. SSIM detects illumination harmonics up to the second order, resulting in a 50 nm resolution [18]. However, it will be difficult to expand SSIM to higher harmonics, because at saturation, SNR of higher harmonics effects are self-limiting. The self-limiting behavior is evident in Fig. 1(a) , which is normalized harmonic strengths of saturation, calculated by the same method used in the original SSIM report [18]. Since at the shot noise limit, the photon noise follows the square root of the total photon count, harmonic strengths normalized with the average photon level provide a way to predict the theoretical SNR of harmonic Moiré effects at different illumination strength. In SSIM, 0^th~4^th harmonics decrease if the illumination strength is over a threshold, and higher harmonics level off quickly as the illumination intensity increases, making increasing SNR at higher harmonics impossible. To improve resolution in SSIM, the only solution is to prolong exposure, boost the image photon count to decrease the SNR, which is not practical in most cases because the specimen is already excited at the saturation level and under severe photobleaching. In comparison, harmonic strengths of an interference patterned STED field, plotted in Fig. 1(b), always increase as the STED intensity increases. Although the slope of gain is less at high STED intensity, in theory STED-SIM has unlimited capability for improving resolution. The same principle had been demonstrated in point-scanning STED, where 6 nm resolution was achieved under an extremely intense STED field intensity [22]. The difference in resolution scaling of saturation vs. STED is due to their different noise behavior under increasing pattern intensity. Increasing the intensity of the saturated illumination pattern generates an overall brighter image. Thus, as the saturation nonlinearity increases, the DC component photon shot-noise also increases, resulting in the leveling-off in SNR. On the contrary, increasing the intensity of the STED pattern will generate a darker image. As STED nonlinearity increases, the photon shot-noise decreases, giving more room to improve resolution until the SNR is limited by the camera noise floor.

Fig. 1 Harmonic strengths of nonlinear effects caused by an interference illumination pattern with an increasing intensity. (a) Saturation (b) STED. The resolution increases when higher harmonic orders of the illumination pattern are detected.

Download Full Size | PDF

1.3 Surface plasmon resonance enhanced full-field STED

Applying full-field STED effects on nonlinear SIM will provide a practical solution for high-speed full-field imaging at resolution <100 nm. However, a major barrier exists on applying STED in nonlinear SIM: the laser power necessary to support a strong full-field STED effect. STED improves imaging resolution by suppressing spontaneous emission with a strong laser beam at a typically red shifted emission wavelength (STED beam). The STED field intensity that “quenches” the spontaneous emission by 50% is termed as the “saturation intensity'” (ISAT), which is typically 3~10 MW/cm2 [23]. Although STED in principle can be generated over a large area, due to the high laser intensity requirement, currently resolution enhancement by STED is only possible though point-by-point scanning of a donut-shaped focused STED field and a focused excitation beam [24]. In point-scanning STED, a peak intensity of 20ISAT allows 50-nm resolution, and >80ISAT is needed for 20-nm resolution [25].

To achieve full-field STED over an area tens of times larger than the donut beam of point-scanning STED, an enhancement mechanism has to be employed so that STED can operate at much lower laser intensity and a total laser power attainable with existing laser technology. Surface plasmon resonance (SPR) [26] is a widely used technique for enhancing the evanescence field at a metal-dielectric interface. SPR enhanced Raman [27], fluorescence [28,29] and two-photon fluorescence [30] have been previously reported. Standing-wave SPR fluorescence demonstrated an enhanced resolution similar to linear SIM [31]. In this paper, we experimentally demonstrate that SPR in a glass-silver-SiO₂-water layered structure enhances STED effect by an order of magnitude. We further conduct simulations of STED-SIM imaging formation and reconstruction using conditions that were proven feasible experimentally. The study showed that SPR enhanced STED-SIM can achieve 30-nm resolution over a >100 μm² area with single fluorophore sensitivity.

2. Surface plasmon enhanced STED

To prove that SPR enhanced full field STED is sufficient for nonlinear SIM, we built a STED-TIRF setup to examine STED effects with and without SPR. Figure 2 shows a sketch of the optical setup. A supercontinuum pulsed laser (Fianium SC-450-PP-HE) was used to provide both the excitation and STED beams. The excitation beam was filter by a 615 ± 15 nm bandpass filter (Semrock FF01-630/20-25, tilted) and attenuated to 3 mW. The STED beam was filtered by a 705 ± 25 nm bandpass filter (Semrock FF01-716/40-25, tilted). A delay stage was placed in the excitation beam path to control the arrival time of the STED pulse to a few picoseconds after the excitation pulse. Because SPR can only be generated by the TE mode incident light, a polarized beam splitter and a half wave plate were used to ensure that only the p-polarized STED beam is passed onto the sample. The fluorescent sample was ATTO 633 water solution at 100 μg/ml with anti-fading agent (ProLong Gold, Invitrogen), placed on top of a silver-SiO₂ coated cover slip. SPR is launched in the Kretschmann configuration. Spontaneous emissions of the fluorophores were recorded by a 15X microscope through a 660 ± 10 nm filter, which removed stray excitation and STED light.

Fig. 2 TIRF-STED setup for measuring SPR enhanced STED. PBS: Polarized beam splitter.

Download Full Size | PDF

To predict the optimal SPR launching structure for STED-SIM, we calculated the field enhancement factor [26] of a four-layer glass-silver-SiO₂-water structure, as shown in Fig. 2 . The SPR was excited from the glass side, and the field enhancement was calculated on the SiO₂-water interface, which will be the imaging surface of future STED-SIM. Silver was chosen instead of gold due to its better SPR performance in the visible range [32]. The SiO₂ layer serves two purposes: to prevent silver oxidation and to limit metal quenching effect on fluorescence emission.

The thickness of the silver layer was optimized under two conditions: ideal collimated beam, and a broadband focusing Gaussian beam that matches the experimental condition. Figure 3 plots field enhancement factors under different silver thicknesses and incident angles. The optimal silver thickness is calculated to be 40 nm with a focusing beam. Compared with an ideal collimated beam at single wavelength, the electric field enhancement of a focused broadband beam is lower, due to partial wavelength and SPR angle mismatch in the broadband focusing beam. Nevertheless, an order of magnitude of STED field enhancement was expected.

Fig. 3 Calculated SPR field enhancement at the SiO₂-water interface of a glass-silver-SiO₂-water structure. The optical field incidents from the glass side. Two conditions were calculated: collimated incident beam at a single wavelength of 705 nm, and a focused Gaussian beam with a 50 nm bandwidth, 10 μm focal size and a 2° focusing angle, which matched the experimental condition. (a) SPR field enhancement factor vs. silver thickness. The SiO₂ layer thickness is 15 nm. The incident angle was optimized to maximal enhancement at each thickness. (b) SPR Field enhancement factor vs. incident angle. The SiO₂ layer thickness is 15 nm. The silver layer thickness is 40 nm.

Download Full Size | PDF

Figure 4 shows experimental STED fluorescence quenching from the STED-TIRF setup. Spontaneous fluorescence emissions under increasing STED beam power were measured by a CCD camera. Quenching of spontaneous fluorescence emission at the peak of the STED beam with and without SPR enhancement were compared as shown in Fig. 4 bottom. STED quenching is much stronger on the SPR substrate than on the plain glass coverslip. The STED quenching effect was fitted with a quenching model of ${(1 + P_{S T E D} / P_{S A T})}^{- 1}$ . Because of the long pulse duration of the laser (≥100 ps, on the same order of magnitude as fluorophore lifetime under STED quenching) and the large temporal overlap between the excitation and STED pulse, the quenching behavior in our case is similar to CW STED [33].

Fig. 4 SPR enhanced STED. Top: spontaneous emission patterns from Atto 633 water solution on a glass-silver-SiO₂ substrate under an increasing STED power. The pattern frame size is 30 × 30 μm². The STED power was increased from 0 to 2.5, 10, and 50 mW respectively. The STED beam had an FWHM size of 18 × 11 μm². Considering the objective lens caused a 40% transmission loss, the estimated instantaneous STED intensities on the sample were approximately 0, 3, 13, and 65 MW/cm² at the beam center. Bottom: Quenching of spontaneous emission at the center of the STED beam on SPR substrate and glass. The SPR substrate was a glass coverslip coated with 40 nm silver followed by 15 nm SiO₂. Due to the long excitation pulse and large temporal overlapping between the excitation and the STED pulse, the STED quenching effect follows ${(1 + P_{S T E D} / P_{S A T})}^{- 1}$ , which is the same as the quenching behavior under CW STED [33]. P_SAT values are listed in Table 1.

Download Full Size | PDF

The electric field enhancement factor varies little when the SiO₂ layer thickness is between 0~30 nm according to the field calculation. However, the SiO₂ layer thickness is a critical factor for metal quenching effect on fluorescence emission. The enhancement on STED quenching effect is the result of STED quenching competing with all radiative and no-radiative decay channels. When using plain glass coverslips, the I_SAT is the STED intensity that makes $k^{}_{S T E D} = k_{n r} + k_{r}$ . The I_SAT under SPR enhancement is when $k^{S P R}_{S T E D} = k_{n r} + k_{r} + k_{m q}$ , where $k_{m q}$ is the metal quenching rate. The metal quenching effect raises I_SAT, and therefore the enhancement factor on observed STED quenching effects will be smaller than the electric field enhancement factor. Besides raising I_SAT, metal quenching also causes decrease in the fluorescence emission. To limit the metal quenching effect, an optimal SiO₂ layer thickness needs to be employed to balance between the decay of STED evanescence field, which is an exponential decay to the SiO₂ thickness, and the decrease of metal quenching. The metal quenching effect is caused by Föster energy transfer between the fluorophore and the metal surface. Thus the metal quenching rate follows ${(1 + R^{4} / R_{0}^{4})}^{- 1}$ , where R is the thickness of the SiO₂ layer, and R₀ is the half quenching distance [34]. The half quenching distanced between a fluorophore and an ideal metal surface is estimated to be 5~7 nm [34], depending on properties of the metal and the fluorophore. Because the metal quenching rate decays much faster than the decay of STED evanescence field, using a proper dielectric spacing layer between the metal and the fluorescent sample minimizes the metal quenching effect without significantly sacrificing the STED field strength. We compared SPR enhanced STED on 40-nm silver layer with 10,15 and 20-nm SiO₂ layers respectively. Spontaneous fluorescence emission intensity at the center of the STED beam was fitted with a quenching curve of ${(1 + P_{S T E D} / P_{S A T})}^{- 1}$ . P_SAT represents the STED beam power that causes 50% quenching at the beam center. Table 1 lists fitted P_SAT on glass and SPR structures with four different SiO₂ thicknesses. Estimated I_SAT values are also listed. The strongest STED effect was observed on 40-nm silver and 15-nm SiO₂. Comparing STED on a glass slide and on a 40-nm silver with 15 nm SiO₂ slide, we found an 8 times enhancement in the STED quenching effect, which, as predicted, is slightly lower than the field enhancement factor obtained through field calculation.

Table 1. Full-field STED on Different Substrates

View Table

With a 50-mW STED beam, 10I_SAT over a 17 × 10 μm² area or 14I_SAT over an 11 × 6.5 μm² area were achieved on an optimal SPR structure of 40-nm silver and 15-nm SiO₂. As discussed in Section 3 below, the SPR-enhanced STED field is sufficient for achieving 30 nm resolution using the nonlinear SIM method.

3. 2D nonlinear SIM with SPR enhanced STED

To achieve two-dimensional resolution enhancement, SIM and SSIM use a rotating interference fringe pattern to illuminate the sample [14,35]. The same approach is applicable in STED-SIM, however, we propose to use a more energy-efficient two-dimensional interference pattern method in STED-SIM (Fig. 5a ). The same approach had been used in saturated patterned excitation microscopy (SPEM) [36].

Fig. 5 (a) Proposed 2D STED-SIM setup. The excitation beam will be merged with one of the four STED beam through a dichroic mirror. (b) Simulating raw frame formation of 2D STED-SIM. The simulation used an object that contains 10-nm wide features inducing a circle, three pairs of thin bars separated by 20, 30, and 40 nm respectively, and a single dot. The STED quenching grid was simulated as ${1 + 0.25 I_{S T E D} [2 + \cos (k_{0} x + ϕ_{x}) + \cos (k_{0} y + ϕ_{y})]}^{- 1}$ , where I_STED was a uniform field at 40 I_SAT. The fringe spacing of the STED grid was 250 nm. The grid was phase-shifted in 13 × 13 steps along the two dimensions. A set of 169 raw frames was simulated. Raw frame was binned down to a Nyquist sampling of 125 nm/pixel. Poisson shot noise and a 2 e^- camera readout noise was added.

Download Full Size | PDF

SPR can only be generated by a TE-mode optical field. Therefore in Fig. 2, a linear polarizer was used to reject the TM-mode of the STED field. Due to the random polarization of the supercontinuum laser, half of the laser energy was rejected. The rejected laser power can be reused to create a SIM pattern in the orthogonal dimension. Due to the random polarization nature of the supercontinuum laser output, the two separated polarizations are incoherent with each other. This allows the two orthogonal interference patterns to add incoherently. Under a strong structured STED field, only fluorophores that locate at zero STED field can emit signal. The 2D structured STED field effectively will impose a quenching grid onto the sample, as depicted in Fig. 5(a). The 2D STED-SIM will employ STED fields at two polarizations, which will eliminate any potential polarization-selectivity on fluorophores.

The image acquisition and reconstruction method of 2D STED-SIM will be similar to SPEM [36]. However, STED is a quenching effect, whereas excitation saturation is a brightening effect. Furthermore, as discussed in section 1.2, the noise behavior of STED-SIM is different with SPEM and SSIM. Because SPEM and SSIM operate with saturated excitation, the image reconstruction algorithm was never tested under low signal condition, which will be the working condition of STED-SIM. To test whether single molecule and live cell imaging is possible with STED-SIM, we simulated the image acquisition process of STED-SIM under low SNR conditions, and performed image reconstruction without using prior knowledge about the STED pattern.

3.1 Image forming theory of STED-SIM

As shown in Fig. 5(b), a patterned STED quenching field will limit spontaneous emission to a point grid. The raw frame caught by the camera will be the remaining spontaneous emission convolved with the point-spread function of the camera optics. To obtain a super-resolution image, the STED grid will be phase shifted from 0 to 2π in N × N 2D steps, and N × N raw frames will be taken. A raw frame under a STED grid phase $(ϕ_{x} = \frac{2 p π}{N}, ϕ_{y} = \frac{2 q π}{N})$ will be

F (x, y, p, q) = \frac{O (x, y)}{1 + 0.25 I_{S T E D} (x, y) [2 + \cos (k_{0} x + \frac{2 p π}{N}) + \cos (k_{0} y + \frac{2 q π}{N})]} \otimes A (x, y)

where p, q = 1~N are indices of phase shifts, O(x, y) is the object, A(x, y) is the point spread function of the imaging optics, I_STED is the STED intensity distribution in the unit of I_SAT, and k₀ is the absolute value of the fringe vector. The denominator of Eq. (1) can be harmonically expanded through two-dimensional Taylor expansion

\begin{array}{l} \frac{1}{1 + 0.25 I_{S T E D} (x, y) [2 + \cos (k_{0} x + \frac{2 p π}{N}) + \cos (k_{0} y + \frac{2 q π}{N})]} \\ = \sum_{m = 0, n = 0}^{N} α_{m n} (x, y) e x p (i m k_{0} x + i \frac{2 m p π}{N}) e x p (i n k_{0} y + i \frac{2 n p π}{N}) \end{array}

where α_mn is

\begin{matrix} α_{m n} (x, y) = \frac{\partial^{m + n} {[1 + 0.5 I_{S T E D} (x, y) + 0.25 Δ I]}^{- 1}}{\partial Δ I^{m + n}} {[I_{S T E D} (x, y)]}^{m + n} \\ = {(- \frac{1}{4})}^{m + n} {[I_{S T E D} (x, y)]}^{m + n} {[1 + 0.5 I_{S T E D} (x, y)]}^{- 1 - m - n} \end{matrix}

When

I {}_{S T E D}{(x, y)} > > 1

and

I {}_{S T E D}{(x, y)}

is a slow varying function within the observation field A, Eq. (3) can be rewritten as

α_{m n} (x, y) \approx β^{m + n} {[1 + 0.5 I_{S T E D} (x, y)]}^{- 1}

where β is the scaling ratio between adjacent harmonic orders

β = - \frac{1}{2 A} \iint_{(x, y) \in A} \frac{I_{S T E D} (x, y)}{1 + 0.5 I_{S T E D} (x, y)} d x d y

The image reconstruction method of STED-SIM is identical to SPEM, except more harmonic orders are used. Steps of the reconstruction are shown in Fig. 6 . First, raw frames are Fourier transformed

\tilde{F} (k_{x}, k_{y}, p, q) = \sum_{m = 1, n = 1}^{N} e x p (\frac{2 i m p π}{N} + \frac{2 i n q π}{N}) β^{m + n} \tilde{O} (k_{x} - m k_{0}, k_{y} - n k_{0}) \tilde{A} (k_{x}, k y)

where

\tilde{O} (k_{x} - m k_{0}, k_{y} - n k_{0})

is the frequency-shifted version of a high resolution Fourier image

\tilde{O} (k_{x}, k_{y})

, which is given by

Fig. 6 Image reconstruction of 2D STED-SIM. Raw data was simulated as in Fig. 5(b). To remove ringing, a Hann window filter was applied to the Fourier image before IFFT.

Download Full Size | PDF

\begin{matrix} \tilde{O} (k_{x}, k_{y}) = F T [\frac{O (x, y)}{1 + 0.5 I_{S T E D} (x, y)}] \\ \approx F T [O (x, y)], when I_{S T E D} (x, y) is slow varying \end{matrix}

Because the grid x-y phase index $(p, q)$ is shifted in N × N steps, Eq. (6) represents N² number of linear equations.

In the second step, $\tilde{O} (k_{x}, k_{y})$ is reconstructed by solving these linear equations

\tilde{O} (k_{x} - m k_{0}, k_{y} - n k_{0}) = \frac{M^{- 1} \tilde{F} (k_{x}, k_{y}, p, q)}{β^{m + n} \tilde{A} (k_{x}, k_{y})}

where

M^{- 1}

refers to the inversion of a 4^th order tensor constant

M (p, q, m, n) = e x p (\frac{2 i m p π}{N} + \frac{2 i n q π}{N})

. Equation (8) allows reconstructing

\tilde{O} (k_{x}, k_{y})

by N² times, of which each time the zero frequency location is shifted by a different

(m k_{0}, n k_{0})

. Because the optical transfer function

\tilde{A} (k_{x}, k_{y})

is a low-pass filter with a cut-off frequency at

| \vec{k} | = k_{0}

, when

{(k_{x} - m k_{0})}^{2} + {(k_{y} - n k_{0})}^{2} > k^{2}_{0}

, the reconstructed

\tilde{O} (k_{x} - m k_{0}, k_{y} - n k_{0})

becomes unreliable due to elevated noise. As a result, at any given

(m k_{0}, n k_{0})

, only a local portion of

\tilde{O} (k_{x}, k_{y})

within a circular window of

{(k_{x} - m k_{0})}^{2} + {(k_{y} - n k_{0})}^{2} \leq k_{0}^{2}

is revealed.

The third step of the reconstruction is to gather all local portions of $\tilde{O} (k_{x}, k_{y})$ , and merge them into a full resolution Fourier image. In four-beam STED-SIM or SPEM, the structured illumination is shifted in Cartesian coordinates. Therefore local Fourier images are spread out in Cartesian coordinates. The frequency offset set between adjacent local portions is k₀. Overlapping regions of adjacent Fourier portions are weighted averaged in the final merged Fourier image using $β^{m + n} \tilde{A} (k_{x}, k_{y})$ as the weighting function. In the last reconstruction step, a Hann window was applied to the Fourier image and the reconstructed image is obtained through an inverse FFT, as shown in Fig. 6, step 4.

3.2 Sensitivity and resolution of STED-SIM

The theoretical resolution of 2D STED-SIM is set by N, the number of phase-shifting steps. A N × N phase-shift will allow constructing the Fourier image $\tilde{O} (k_{x}, k_{y})$ up to $k_{x}^{2} + k_{y}^{2} \leq \frac{N + 1}{2} k_{0}^{2}$ . The theoretical resolution will be $\frac{2 d_{0}}{N + 1}$ , where d₀ is the fringe spacing [5]. Experimentally, the image quality and the robustness of the image reconstruction of STED-SIM are strongly dependent on photon levels and STED power. As shown in Eq. (5) and (6), when $I {}_{S T E D}{(x, y)}$ is extremely high, the scaling ratio β will approach 1, and the signal to noise ratio at higher spatial frequency will increase. On the other hand, if I_STED is limited and higher harmonics of the STED quenching pattern are below the photon noise floor, increasing N will only introduce more noise into the reconstructed image. Furthermore, as shown in Eq. (8), reconstructing the Fourier image $\tilde{O} (k_{x}, k_{y})$ requires knowledge about the scaling ratio β, which is related to the strength of the STED quenching effect. Since the STED response of fluorophores vary with environment, the STED quenching effect cannot be predetermined, and β has to be obtained through analysis of raw frames, for example by comparing relative magnitudes in overlapping regions of adjacent local Fourier images [14]. The accuracy of β will be influenced by SNR of raw frames and strengths of harmonics orders.

To test the low SNR performance and the photon budget limit of STED-SIM under an experimentally feasible STED field, noise simulation was added into the simulation of raw frame acquisition in Fig. 5(b). The simulation started from an object that contains 10-nm wide features, including a circle, three pairs of thin bars separated by 20, 30, and 40 nm respectively, and a single dot. The photon emission level from the object was set to be 150 photons per 10 × 10 nm². A 0~40 I_SAT (valley to peak) 2D STED grid was applied to the object to quench the emission. The fringe spacing of the STED pattern was set to 250 nm, according to the SPR condition of silver at 705 nm. The grid was phase-shifted in 13 × 13 steps along two dimensions. A set of 169 raw frames was simulated. Simulated raw frames were binned down to a Nyquist rate of 125 nm/pixel. At last, Poisson shot noise and a random 2 e^- camera readout noise were added.

We then proceeded with the image reconstruction without prior knowledge about the STED pattern. The fringe space was obtained by comparing the shifting of the zero frequency point in $\tilde{O} (k_{x}, k_{y})$ and $\tilde{O} (k_{x} \pm k_{0}, k_{y} \pm k_{0})$ . The scaling ratio β between adjacent harmonic orders was set to be the average ratio between overlapping regions of adjacent local Fourier images. In the reconstructed image (Fig. 6, Step 4), bars with 30-nm separation are resolvable, and the 40-nm separation is clearly shown. Our TIRF-STED experimental result showed that with a commercially available laser, an SPR enhanced STED grid of 0~40 I_SAT over an entire 17 × 10 μm² area is achievable. Therefore, STED-SIM will be able to achieve 30-nm resolution over >100 μm² area.

The single dot in the object was clearly reconstructed with 2.5 × 10⁴ emitted photons. At an estimated 10% system collection efficiency, the total number of photons emitted from a single fluorophore that are needed for the reconstruction is 2.5 × 10⁵, which is within the average 10⁵~10⁶ excitation-emission cycle of a high-quality organic fluorophore. The Reducing and Oxidizing System (ROXS) can further reduce bleaching in STED [23]. Thus, STED-SIM will be able to detect single fluorophore molecules with 30-nm resolutions.

4. Discussion and summary

The simulation study shows that to achieve maximum sensitivity, raw frames of STED-SIM should be sampled at the Nyquist rate of the fringe space. Over sampling on the camera introduces more readout noise to raw frames, and decreases the SNR. The low spatial sampling rate is also advantageous for fast live imaging. For example, an sCMOS camera (Cooke pco.edge) can acquire 3600 frames per second over an 80 × 120 pixel region, corresponding to a 10x15 μm sample region with 125nm/pixel sampling rate. Raw frames for reconstructing a 30-nm resolution image can be acquired in less than 0.1 second, much faster than existing single-molecule super-resolution methods, such as PALM and STORM.

SPR enhanced STED-SIM will operate at the TIRF imaging mode, which although is intrinsically limited to near surface, provides a clean background to observe near surface structures. TIRF is a common imaging modality that is widely used in single molecule study and selective visualization of basal membranes. In the TIRF mode, the laser radiation is strictly confined in the evanescence region, and causes much less disturbance to cell functions. Cell viability study showed that under the same illumination photon dose, TIRF causes less damage than Epi-illumination microscopy [17]. Compared with point-scanning STED, which illuminates the entire depth of the sample as Epi-illumination microscopy, STED-SIM will have less photon toxicity.

In summary, we demonstrated that a strong full-field STED effect is experimentally achievable through SPR field enhancement of the evanescent field. Our analysis predicted nonlinear two-dimensional STED-SIM with 30-nm resolution and sub-second acquisition time is feasible. The STED-SIM super resolution microscopy method would provide a solution for observe single molecule processes in vitro or on the basal membrane of live cells, where many biological questions remain due to the resolution limit of conventional microscopes.

Acknowledgments

The authors thank Arthur Gmitro, Ronald M. Lynch and Urs Utzinger for their insightful comments. This research is supported by the Technology Research Initiative Fund of the State of Arizona.

References and links

1. S. W. Hell, “Microscopy and its focal switch,” Nat. Methods 6(1), 24–32 (2009). [PubMed]

2. S. W. Hell, “Far-field optical nanoscopy,” Science 316(5828), 1153–1158 (2007). [CrossRef] [PubMed]

3. B. Huang, H. Babcock, and X. Zhuang, “Breaking the diffraction barrier: super-resolution imaging of cells,” Cell 143(7), 1047–1058 (2010). [CrossRef] [PubMed]

4. B. Huang, M. Bates, and X. Zhuang, “Super-resolution fluorescence microscopy,” Annu. Rev. Biochem. 78(1), 993–1016 (2009). [CrossRef] [PubMed]

5. R. Heintzmann and M. G. L. Gustafsson, “Subdiffraction resolution in continuous samples,” Nat. Photonics 3(7), 362–364 (2009). [CrossRef]

6. B. Huang, W. Wang, M. Bates, and X. Zhuang, “Three-dimensional super-resolution imaging by stochastic optical reconstruction microscopy,” Science 319(5864), 810–813 (2008). [CrossRef] [PubMed]

7. M. J. Rust, M. Bates, and X. Zhuang, “Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM),” Nat. Methods 3(10), 793–796 (2006). [CrossRef] [PubMed]

8. B. Hein, K. I. Willig, and S. W. Hell, “Stimulated emission depletion (STED) nanoscopy of a fluorescent protein-labeled organelle inside a living cell,” Proc. Natl. Acad. Sci. U.S.A. 105(38), 14271–14276 (2008). [CrossRef] [PubMed]

9. S. T. Hess, T. P. K. Girirajan, and M. D. Mason, “Ultra-high resolution imaging by fluorescence photoactivation localization microscopy,” Biophys. J. 91(11), 4258–4272 (2006). [CrossRef] [PubMed]

10. J. Fölling, M. Bossi, H. Bock, R. Medda, C. A. Wurm, B. Hein, S. Jakobs, C. Eggeling, and S. W. Hell, “Fluorescence nanoscopy by ground-state depletion and single-molecule return,” Nat. Methods 5(11), 943–945 (2008). [CrossRef] [PubMed]

11. S. W. Hell and J. Wichmann, “Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy,” Opt. Lett. 19(11), 780–782 (1994). [CrossRef] [PubMed]

12. T. Klar, S. Jakobs, M. Dyba, A. Egner, and S. W. Hell, “Fluorescence microscopy with diffraction resolution barrier broken by stimulated emission,” in PNAS, (2000), p. 8206.

13. M. Hofmann, C. Eggeling, S. Jakobs, and S. W. Hell, “Breaking the diffraction barrier in fluorescence microscopy at low light intensities by using reversibly photoswitchable proteins,” Proc. Natl. Acad. Sci. U.S.A. 102(49), 17565–17569 (2005). [CrossRef] [PubMed]

14. M. G. L. Gustafsson, “Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy,” J. Microsc. 198(2), 82–87 (2000). [CrossRef] [PubMed]

15. U. V. Naegerl, K. I. Willig, B. Hein, S. W. Hell, and T. Bonhoeffer, “Live-cell imaging of dendritic spines by STED microscopy,” Proc. Natl. Acad. Sci. U.S.A. 105(48), 18982–18987 (2008). [CrossRef] [PubMed]

16. V. Westphal, S. O. Rizzoli, M. A. Lauterbach, D. Kamin, R. Jahn, and S. W. Hell, “Video-rate far-field optical nanoscopy dissects synaptic vesicle movement,” Science 320(5873), 246–249 (2008). [CrossRef] [PubMed]

17. M. Wagner, P. Weber, T. Bruns, W. S. Strauss, R. Wittig, and H. Schneckenburger, “Light dose is a limiting factor to maintain cell viability in fluorescence microscopy and single molecule detection,” Int. J. Mol. Sci. 11(3), 956–966 (2010). [CrossRef] [PubMed]

18. M. G. Gustafsson, “Nonlinear structured-illumination microscopy: wide-field fluorescence imaging with theoretically unlimited resolution,” Proc. Natl. Acad. Sci. U.S.A. 102(37), 13081–13086 (2005). [CrossRef] [PubMed]

19. P. Kner, B. B. Chhun, E. R. Griffis, L. Winoto, and M. G. L. Gustafsson, “Super-resolution video microscopy of live cells by structured illumination,” Nat. Methods 6(5), 339–342 (2009). [CrossRef] [PubMed]

20. P. T. C. So, H.-S. Kwon, and C. Y. Dong, “Resolution enhancement in standing-wave total internal reflection microscopy: a point-spread-function engineering approach,” J. Opt. Soc. Am. A 18(11), 2833–2845 (2001). [CrossRef] [PubMed]

21. L. Hirvonen, O. Mandula, K. Wicker, and R. Heintzmann, “Structured illumination microscopy using photoswitchable fluorescent proteins,” in Proceedings of SPIE, (SPIE, 2008), p. 68610L.

22. E. Rittweger, K. Y. Han, S. E. Irvine, C. Eggeling, and S. W. Hell, “STED microscopy reveals crystal colour centres with nanometric resolution,” Nat. Photonics 3(3), 144–147 (2009). [CrossRef]

23. R. Kasper, B. Harke, C. Forthmann, P. Tinnefeld, S. W. Hell, and M. Sauer, “Single-molecule STED microscopy with photostable organic fluorophores,” Small 6(13), 1379–1384 (2010). [CrossRef] [PubMed]

24. K. I. Willig, S. O. Rizzoli, V. Westphal, R. Jahn, and S. W. Hell, “STED microscopy reveals that synaptotagmin remains clustered after synaptic vesicle exocytosis,” Nature 440(7086), 935–939 (2006). [CrossRef] [PubMed]

25. B. Harke, J. Keller, C. K. Ullal, V. Westphal, A. Schönle, and S. W. Hell, “Resolution scaling in STED microscopy,” Opt. Express 16(6), 4154–4162 (2008). [CrossRef] [PubMed]

26. J. Homola, “Electromagnetic Theory of Surface Plasmons,” Springer Ser Chem Sens Biosens 4, 3–44 (2006). [CrossRef]

27. K. Kneipp, Y. Wang, H. Kneipp, L. Perelman, I. Itzkan, R. Dasari, and M. Feld, “Single Molecule Detection Using Surface-Enhanced Raman Scattering (SERS),” Phys. Rev. Lett. 78(9), 1667–1670 (1997). [CrossRef]

28. S. Ekgasit, F. Yu, and W. Knoll, “Fluorescence intensity in surface-plasmon ﬁeld-enhanced ﬂuorescence spectroscopy,” Sens. Actuators B Chem. 104(2), 294–301 (2005). [CrossRef]

29. R.-Y. He, G.-L. Chang, H.-L. Wu, C.-H. Lin, K.-C. Chiu, Y.-D. Su, and S. J. Chen, “Enhanced live cell membrane imaging using surface plasmon-enhanced total internal reflection fluorescence microscopy,” Opt. Express 14(20), 9307–9316 (2006). [CrossRef] [PubMed]

30. R.-Y. He, Y.-D. Su, K.-C. Cho, C.-Y. Lin, N.-S. Chang, C.-H. Chang, and S. J. Chen, “Surface plasmon-enhanced two-photon fluorescence microscopy for live cell membrane imaging,” Opt. Express 17(8), 5987–5997 (2009). [CrossRef] [PubMed]

31. E. Chung, Y.-H. Kim, W. T. Tang, C. J. R. Sheppard, and P. T. C. So, “Wide-field extended-resolution fluorescence microscopy with standing surface-plasmon-resonance waves,” Opt. Lett. 34(15), 2366–2368 (2009). [CrossRef] [PubMed]

32. P. R. West, S. Ishii, G. V. Naik, N. K. Emani, V. M. Shalaev, and A. Boltasseva, “Searching for better plasmonic materials,” Laser Photon. Rev. 4(6), 795–808 (2010). [CrossRef]

33. K. I. Willig, B. Harke, R. Medda, and S. W. Hell, “STED microscopy with continuous wave beams,” Nat. Methods 4(11), 915–918 (2007). [CrossRef] [PubMed]

34. T. Liebermann and W. Knoll, “Surface-plasmon ﬁeld-enhanced ﬂuorescence spectroscopy,” Colloids Surf., A 171(1-3), 115–130 (2000). [CrossRef]

35. R. Heintzmann, T. M. Jovin, and C. Cremer, “Saturated patterned excitation microscopy--a concept for optical resolution improvement,” J. Opt. Soc. Am. A 19(8), 1599–1609 (2002). [CrossRef] [PubMed]

36. R. Heintzmann, “Saturated patterned excitation microscopy with two-dimensional excitation patterns,” Micron 34(6-7), 283–291 (2003). [CrossRef] [PubMed]

50% quenching point	Glass	40 nm Silver + 10 nm SiO₂	40 nm Silver + 15 nm SiO₂	40 nm Silver + 20 nm SiO₂	40 nm Silver + 30 nm SiO₂
P_SAT (mW)	14	2.5	1.7	2.0	3.5
I_SAT (MW/cm²) ^a	18	3.3	2.2	2.6	4.6

Nonlinear structured illumination microscopy by surface plasmon enhanced stimulated emission depletion

Abstract

1. Introduction

1.1 Superresolution microcopy methods

1.2 Nonlinear SIM with STED effect

1.3 Surface plasmon resonance enhanced full-field STED

2. Surface plasmon enhanced STED

3. 2D nonlinear SIM with SPR enhanced STED

3.1 Image forming theory of STED-SIM

3.2 Sensitivity and resolution of STED-SIM

4. Discussion and summary

Acknowledgments

References and links

Cited By

Figures (6)

Tables (1)

Equations (8)

Optics Express