Narrowband photoblinking InP/ZnSe/ZnS quantum dots for super-resolution multifocal structured illumination microscopy enhanced by optical fluctuation

2.1 Synthesis of InP/ZnSe/ZnS QDs

The synthesis of InP/ZnSe/ZnS QDs included three steps with InP, InP, and InP/ZnSe/ZnS. Figure S1a illustrates the synthesis process and depicts the core–shell-wrapped InP/ZnSe/ZnS QDs. The innermost to outermost layers were InP, ZnSe, and ZnS, and the lattice constants were 5.87, 5.67, and 5.41 Å, respectively [28]. The lattice mismatch ratio was approximately 3%, providing a theoretical basis for the synthesis of InP/ZnSe/ZnS QDs [29]. Figure S1b shows a schematic illustration of ultrasonic oscillation, which assists in obtaining more homogeneous monodisperse particles. In addition, the transient temperature effect of ultrasound enables the construction of QDs surface trap states, which are responsible for photoblinking [30]. The shell suitably coats the InP core layer with stable and controllable modulation. The characterization and analysis of InP/ZnSe/ZnS QDs are shown in Figures S2–S5.

2.2 InP cores

InP cores were synthesized using the one-pot method. In typical synthesis, a molar ratio of 1:3 of 0.6 mmol indium acetate (In(Ac)₃) and 1.8 mmol palmitic acids was mixed in a 50 mL three-necked flask. The mixture was degassed and dried under vacuum (He) at 150 °C for 15 min. Then, 1.0 mmol octadecene, 0.3 mmol trimethylsilyl phosphine ((TMS)₃P), and 0.5 mL trioctylphosphine were injected into the indium solution. After reacting at 140 °C for 1.5 min, the mixture was quickly reduced to room temperature. Finally, all solutions in the flask were drawn using a high-temperature needle.

2.3 InP cluster

We added 10 mL octadecene and 0.15 mmol indium acetate to another three-necked flask (100 mL). The mixture was evacuated at 220 °C for 15 min. Then, the temperature was raised to 290 °C, the InP core was quickly injected into the 100 mL three-necked flask for 15 min, and the flask was quickly cooled down to room temperature. Subsequently, slow pumping was used to inject 0.3 mmol (TMS)₃P with 0.5 mL octadecene solution at 245 °C and 6 min/mL. After 30 min, the absorption peak of the InP cluster was approximately 575 nm, and the cluster was purified with ethyl acetate and high-purity ethanol by centrifugation.

2.4 InP/ZnSe/ZnS QDs

We added 10 mmol zinc acetate (Zn(Ac)₂), 1.5 mmol lauric acid, 18 mmol oleic acid, and 20 mL octadecene to a 250 mL three-necked flask, which was degassed at 170 °C for 30 min. The color of the solution was maintained as clear as possible. Subsequently, the temperature was reduced to 110 °C. The cleaned InP cluster was added to a 250 mL three-necked flask for 5 min. Subsequently, 30 µL of HF was injected into the flask and held for 5 min, and 0.4 mmol selenium powder (Se) trioctylphosphine was quickly injected into the flask at 330 °C for 5 min. The temperature was reduced to 315 °C, and a 0.5 mmol Se-trioctylphosphine-HF solution was injected by slow pumping until reaching 5 mL and held for 10 min, in which the HF concentration was 10 µL. Then, 0.2 mL of oleic acid was quickly injected at 340 °C for 10 min. The slow pumping and oleic acid processes were performed three times. InP/ZnSe was obtained. Then, 0.5 mL 1-dodecanethiol was injected into the flask at 315 °C for 10 min. The temperature controller was then raised to 335 °C and held for 15 min. The emission wavelength stabilized at approximately 625 nm. Finally, InP/ZnSe/ZnS QDs were cleaned and vacuum dried. The InP/ZnSe/ZnS QDs (5 mg/mL) were redissolved in n-hexane (C₆H₁₄) and stored at 4 °C.

2.5 Synthesis of photoblinking InP/ZnSe/ZnS QD fluorescent probes

We sonicated 5 mg/mL InP/ZnSe/ZnS QDs at 180 W for 2 h (four times for 30 min), and the QDs were oscillated (HUXI Vortex-M) for 10 min after ultrasonication (KSJ PL-S30T, 180 W). Then, 1 mL of methanol (CH₄O), 1 mL of chloroform (CHCl₃), 1 mL of InP/ZnSe/ZnS QDs (5 mg/mL), 3 mL of 3-mercaptopropionic acid, and 4 mL of tetramethyl ammonium hydroxide (25% in water) were mixed in a 10 mL glass bottle. The QDs were vacuum freeze-dried after stirring for 2 h. Next, 10 µL of InP/ZnSe/ZnS QDs (5 mg/mL, 1.0 nmol), 2.4 mg of streptavidin (SA) (10 mg/mL, 42.5 nmol), 30 µL of ethyl [3-(dimethylamine) propyl] carbodiimide hydrochloride (EDC) (10 mg/mL, 1665 nmol), and 200 µL of 6.7 mM phosphate-buffered saline (PBS) (PH of 7.4) were added to a 2 mL centrifuge tube at molar ratio QDs:SA:EDC of approximately 1:40:1600. The InP/ZnSe/ZnS QDs were centrifuged in 100 KDa ultrafiltration centrifuge tubes by gently stirring for 12 h with centrifugation at 12,000 rpm for 5 min. The InP/ZnSe/ZnS QDs and SA (QD-SA fluorescent probes) were completely dissolved in 20 µL PBS (PH of 8.4) and stored at 4 °C.

2.6 Cellular and subcellular labeling

For microtubule labeling, African green monkey kidney cells (BS-C-1) were first cultured (37 °C, 5% CO₂, 72 h, 20 mm glass bottom culture dishes) and washed three times with PBS (PH of 7.4) for 1 min. Second, the 1 mL 0.2% Triton X-100 permeate was added to culture dishes at 37 °C for 1 min. Then, 200 µL cytoskeleton buffer (1 × PEM, PH of 6.9 containing 0.1 M PIPES, 1 mM EGTA, and 1 mM MgCl₂), 4% paraformaldehyde, and 0.1% glutaraldehyde were mixed in culture dishes at 37 °C for 10 min. Next, the cells were sequentially treated with 1 mL of NaBH₄ (1 mg/mL), 200 µL of permeation solution (20 min), and 200 µL of blocking solution (40 min) and washed three times with PBS (PH of 7.4) for 5 min each time. The permeation solution (100 µL), blocking solution, and 1 µL of primary antibody (alpha-tubulin monoclonal antibody YL1/2) were mixed in culture dishes at room temperature for 1 h. Cells were washed three times with PBS and 0.5% Triton X-100 for 5 min each time. The permeation solution (100 µL), blocking solution, and 1 µL of biotinylated secondary antibody (biotin goat anti-rat antibody) were mixed in culture dishes at room temperature for 1 h. The cells were washed three times with PBS and 0.5% Triton X-100 for 5 min each time. Then, 200 µL of blocking solution and 2 µL of 40 nM QD-SA were incubated in the cells at room temperature for 1 h, and the cells were washed three times with PBS for 5 min each time. Finally, we used 1 mL of PBS (0.1% glutaraldehyde) in culture dishes for 10 min and washed three times with PBS. The cells were used with 1 mL of 50% glycerol fixed and stored at 4 °C for testing.

For mitochondrial labeling, human osteosarcoma U2OS cells were washed three times and directly fixed with 4% paraformaldehyde for 10 min. The biotinylated secondary antibody was replaced (biotin goat anti-mouse antibody). The remaining steps were the same as those used for microtubule labeling.

2.7 Experimental characterization equipment

The morphology, elemental composition, and size distribution of InP/ZnSe/ZnS QDs were studied by transmission electron microscopy (TEM) (JEOL-F200), atomic force microscopy (SPA-300HV). During TEM characterization, an InP/ZnSe/ZnS QD suspension was drop-cast onto an ultrathin carbon-coated holey support film with 300 mesh copper grids. The phase structure of InP/ZnSe/ZnS QDs was characterized by Raman spectroscopy (LabRAM HR Evolution) using an argon-ion laser with an excitation wavelength of 514 nm at liquid-nitrogen-controlled temperature. The optical properties of InP/ZnSe/ZnS QDs were characterized using an ultraviolet–visible spectrophotometer (PerkinElmer Lambda950) and fluorescence photoluminescence (PL) spectrometer (Horiba Fluorolog-3). Fluorescence images were collected using an objective lens (Nikon, 100 × NA = 1.35 Si oil immersion lens), sCMOS camera (Hamamatsu C5440-20UP), and Z piezoelectric stage (P-561.3CD). A 488 nm laser (SAPPHIRE 488-200 CW CDRH) was used to excite core–multi-shell InP/ZnSe/ZnS QDs, and a fluorescence filter (Thorlabs ET500LP) was used to detect emitted red fluorescence.

2.8 SOFI-MSIM data collection

To investigate the applicability of core–multi-shell InP/ZnSe/ZnS QDs as fluorescent probes for super-resolution imaging, an in-house MSIM system based on digital light innovation (DLP D4100) [31] equipped with a 100 × Si oil immersion objective (Nikon NA 1.35) was used. A 1/150× telescope was used to reduce the DMD pixel size to approximately 72 nm on the sample plane. Sparse multifocal excitation patterns were created and scanned by the DMD. In addition, a 488 nm laser was used to excite the QDs, and the fluorescence signal was captured using an sCMOS camera (Hamamatsu C5440-20UP). Optical filters were selected for fluorescence microscopy (Thorlabs ET500lp). Raw imaging data (224/120 multifocal arrays with 50/30/20 frames per array and super-resolution image) for MSIM reconstruction were acquired at 100 Hz. Core–multi-shell InP/ZnSe/ZnS QDs labeled the mitochondria and microtubules for imaging.

2.9 SOFI-MSIM acquisition

SOFI-MSIM comprises two main steps: Fourier ptychography SOFI [32, 33] and MSIM [14] (pixel reassignment and deconvolution). Fourier ptychography SOFI can improve the resolution by an approximate factor of 2 because it does not require the conventional SOFI high-order calculation but only the fluctuation of fluorescent molecules as scatter illumination. In addition, the MSIM resolution can be improved by more than twice through a large Stokes shift. The ratio of excitation wavelength λ_Ex to detection wavelength λ_Em is λ_Em = k _e λ_Ex (1 < k _e < 2), which is k _e = λ_Em/λ_Ex = 620/488 = 1.27. Therefore, we can obtain an improvement in resolution of more than four times.

2.10 Power-law distribution

The exponentially modified power-law distribution is given by [34]

where A is a constant, α is the power exponent, and 1/μ is the saturation rate of the exponential distribution, which indicates the degree to which the probability density deviates from the power-law distribution.

3.1 SOFI-MSIM system

First, a laser beam with 488 nm wavelength was emitted through a 4f system consisting of lenses f₁ (10 mm) and f₂ (75 mm), which expanded the beam diameter by a factor of 7.5. The Gaussian beam was converted into a flat-field beam by a beam shaper, which irradiated the optical surface of a digital micromirror device (DMD) at 24° with respect to the normal of the DMD panel. The beam was reflected into a 4f system (f₃—75 mm and f₄—75 mm) by the DMD on state. An adjustable iris diaphragm was placed in the spectral plane to block the reflected light of the other diffraction orders from the DMD to avoid stray light interference. The reflected beam formed a sparse excitation array and converged on the back focal plane of the f₄ lens in the 4f filter system through DMD modulation. Its position coincided with the front focal plane of the tube lens (f_T1 of 300 mm). The optical reflective surface of the DMD was then conjugated to the front focal plane of the tube lens. Finally, the excitation array was processed using a Nikon filter box (excitation, dichromatic, and emission filters) to the sample surface and then through the tube lens (f_T2 of 200 mm) to an sCMOS (scientific complementary metal–oxide semiconductor) camera (2304 × 2304 pixels, 6.5 × 6.5 µm). In the experiment, 1024 × 768 pixels were considered for the DMD, and the size of each pixel was 10.8 × 10.8 μm. The optical system was reduced to 1/150 of the original one, and the corresponding size was 72 × 72 nm on the sample surface (Figure 1a). Figure 1b illustrates the SOFI-MSIM process. SOFI-MSIM involves three postprocessing steps: data acquisition, Fourier ptychography SOFI, and MSIM. The red shadowed area of the control signal in Figure 1b shows the different exposure times of the DMD and sCMOS camera.

Figure 1:

 Schematics of system configuration. (a) Optical setup. (b) SOFI-MSIM process.

3.2 Single-particle fluorescence fluctuation of InP/ZnSe/ZnS QDs

We evaluated the fluorescence photoblinking of InP/ZnSe/ZnS QDs at the single-particle level. InP/ZnSe/ZnS QDs were excited with a 488 nm laser for MSIM widefield imaging. To highlight the single-particle imaging quality, Figure 2a shows a fluorescence image of the InP/ZnSe/ZnS QDs with no filtering. We observed bright fluorescence spots and a high signal-to-noise ratio. Figure 2b shows typical time traces of single spots for the InP/ZnSe/ZnS QD fluorescence probe for an exposure time of 10 ms. The single-particle fluorescence fluctuation for exposure times of 5 and 20 ms are shown in Figure S6a–d. The fluorescence on/off activity of the InP/ZnSe/ZnS QDs fluctuated intermittently. In addition, we calculated the fitting parameters of the probability density distributions of the on-state durations for 10 single InP/ZnSe/ZnS QDs by logarithmic transformation, obtaining the results shown in Figure 2c. Power exponent α = 0.88 and exponential correction µ = 0.096 were obtained using Equation (1) (see Experimental Section).

Figure 2:

Fluorescence fluctuation of InP/ZnSe/ZnS QDs. (a) Single-particle widefield overlay of 5000 frames. (b) Photoluminescence fluctuation trace (10 ms). (c) Exponential corrected fitting of on-time interval probability density distribution of QDs. (d) Average duty cycle of QDs across several single-particle fluctuation traces.

Hence, the bright state duration of InP/ZnSe/ZnS QDs was large, and the degree of deviation from the power-law distribution was small, being consistent with previous reports [35]. The intersection of the exponential correction and power-law curves of the probability density was defined as the average on-state duration. The blue dashed line in Figure 2c indicates that the InP/ZnSe/ZnS QDs had an average on-state duration of 42 ms. Figure 2d shows that the average duty cycle of the InP/ZnSe/ZnS QDs was 0.3% according to single-particle statistics. These results quantitatively characterized the low duty cycle of InP/ZnSe/ZnS QDs and their excellent photo-fluctuation properties, suggesting that the QDs are promising materials for super-resolution fluorescence fluctuation imaging.

3.3 SOFI-MSIM experiment for single InP/ZnSe/ZnS QD

We analyzed and verified the effectiveness of the proposed approach. Figure 3 shows the results of the improved resolution of SOFI-MSIM imaging using a single QD. We observed resolution improvements (red line) for widefield (Figure 3a), SOFI (Figure 3b), and SOFI-MSIM (Figure 3c) images. The corresponding FWHM values (52 nm) are shown in Figure 3d. To resolution of the algorithm results was determined and analyzed. We used decorrelation analysis for parameter-free resolution estimation of a single microscopy image, namely, Resolution = 2 × pixel size/(normalized cut-off frequency) [36]. This cut-off frequency is obtained from the normalized spatial frequency. The results for widefield, SOFI and SOFI-MSIM images were 293, 132 and 54 nm, respectively (Figure 3e–g). According to the Abbe limit, the widefield resolution was approximately 231 nm, being improved by an approximate factor of 4.2 when using SOFI-MSIM, indicating optimal compliance consistent with the theoretical analysis (see Experimental Section).

Figure 3:

Experiment for single InP/ZnSe/ZnS QD. (a) Widefield, (b) SOFI, and (c) SOFI-MSIM images. (d) FWHM values of the two nearest points enclosed (red line). The shortest distance that could be resolved was 52 nm. (e) Resolution estimation values of widefield image (65 nm per-pixel). (f) Resolution estimation values of SOFI image (32.5 nm per-pixel). (g) Resolution estimation values of SOFI-MSIM image (13 nm per-pixel).

3.4 SOFI-MSIM imaging of biological sample

We demonstrated SOFI-enhanced MSIM in fixed BS-C-1 cell samples using a 488 nm laser for microtubule imaging. Figure 4a–c show the image processing of microtubules for widefield imaging, SOFI, and SOFI-MSIM, respectively. We analyzed the effect of resolution (∼56 nm) on the normalized spectrum, as shown in Figure 4d. The fine structure of each microtubule was observed more intuitively using SOFI-MSIM. We also performed decorrelation analysis for parameter-free resolution estimation of a single microscopy image, obtaining the results shown in Figure 4e–g, respectively. The resolution estimation values were 291 nm for widefield, 137 nm for SOFI, and 56 nm for SOFI-MSIM, indicating an approximate fourfold resolution increase. Hence, the resolution was substantially increased without requiring buffers and with very high power densities.

Figure 4:

Microtubule super-resolution imaging. (a) Widefield, (b) SOFI, and (c) SOFI-MSIM images. (d) Comparison of blue line segments in microtubule after processing using widefield imaging, SOFI, and SOFI-MSIM to distinguish spectrum (52 nm). (e) Resolution estimation values of widefield image (65 nm per-pixel). (f) Resolution estimation values of SOFI image (32.5 nm per-pixel). (g) Resolution estimation values of SOFI-MSIM image (13 nm per-pixel).

In addition, we demonstrated the enhancement in mitochondrial resolution for fixed U2OS cells (Figure 5a–c). We obtained mitochondrial images processed by widefield imaging, SOFI, and SOFI-MSIM. The resolution of mitochondria could be improved to approximately 55 nm (red line), as shown in Figure 5d. We also performed decorrelation analysis for parameter-free resolution estimation of a single microscopy image. Figure 5e–g respectively show that the mitochondria provided resolution estimation values of 317 nm for the widefield image, 141 nm for the SOFI image, and 59 nm for the SOFI-MSIM image. The results indicate an approximate resolution increase by four times when using SOFI-MSIM. In addition, multifocal arrays can effectively improve the signal-to-noise ratio in a widefield, thus reducing the inter-array spacing. Therefore, the number of acquired frames can also improve the resolution by approximately four times. To evaluate this concept, we used 120 multifocal arrays with different frame accumulations (50, 30, and 20) per array for SOFI-MSIM reconstruction. The imaging improvements were verified for both microtubules and mitochondria, as shown in Figures S7 and S8, respectively. Hence, the SOFI-MSIM combination improved image resolution for microtubule or mitochondria imaging compared with SOFI or MSIM alone (see Figure S9).

Figure 5:

Mitochondria super-resolution imaging. (a) Widefield, (b) SOFI, and (c) SOFI-MSIM images. (d) Comparison of blue line segments in mitochondria after processing using widefield imaging, SOFI, and SOFI-MSIM to distinguish spectrum (55 nm). (e) Resolution estimation values of widefield image (65 nm per-pixel). (f) Resolution estimation values of SOFI image (32.5 nm per-pixel). (g) Resolution estimation values of SOFI-MSIM image (13 nm per-pixel).