Three-dimensional super-resolution imaging of live whole cells using galvanometer-based structured illumination microscopy

Wenjie Liu; Qiulan Liu; Zhimin Zhang; Yubing Han; Cuifang Kuang; Liang Xu; Hongqin Yang; Xu Liu

doi:10.1364/OE.27.007237

1. Introduction

The development of optical microscopy, especially super-resolution fluorescence microscopy, has extended biological research to the nanoscale. It has the potential of allowing noninvasive, fast and three-dimensional (3D) imaging of the subcellular structures and dynamics of live cells with high molecular specificity [1–5].

However, achieving 3D super-resolution imaging is still challenging, especially for live-cell imaging because of the intrinsic trade-off between spatial and temporal resolution. Different super-resolution techniques have been proposed to overcome the 3D diffraction limit of conventional optical microscopy. One method, called 3D-stimulated emission depletion (STED) microscopy, improves the spatial resolution by depleting the periphery of the illumination point spread function (PSF) based on stimulated emission effect [6]. Although STED has been widely used in biological research to study complex molecular organizations at high spatial resolution, its point-scanning nature and high requirement on the illumination intensities (10³–10⁸Wcm⁻²) may cause photodamage and photobleaching, thus confining the application of volumetric live-cell imaging over a large field of view [6]. By reducing the field of view, improving the scanning speed and modifying the labelling strategy, STED has been applied to time-lapse imaging, but still been limited to small cell region of several square micros [6,7]. Recently, Tønnesen et al. have demonstrated the combination of 3D-STED and “negative” and “exchangeable” fluorescence labelling scheme for tissue-level live imaging of the brain extracellular space [8].

Single molecule localization microscopy, such as photoactivation localization microscopy, stochastic optical reconstruction microscopy and points accumulation for imaging in nanoscale topography, breaks the lateral diffraction limit by sparsely exciting and localizing individual photo-switchable fluorophores over a long image acquisition process [9]. By introducing depth-related parameters, such as the shape and brightness, into the PSF, the axial positons of emitters can also be precisely localized to achieve 3D super-resolution imaging [4,5]. The live-cell imaging application of localization microscopy is hampered by the requirement of the specific fluorescent dyes, cell-toxic imaging buffer, high illumination intensity, and the great number of raw images. Sacrificing the spatial resolution, i.e., reducing the raw image number, and optimizing the staining approach [10] can improve the imaging speed of these techniques to some extent to acquire time-lapse images.

A more promising technique for live-cell imaging is structured illumination microscopy (SIM), which improves the spatial resolution via shifting the high-frequency information of a sample into the acceptable passband of the imaging system [11,12]. The frequency mixing and the subsequent de-mixing can extend the optical transfer function (OTF) and double the spatial resolution. In conventional and commercial SIM systems, a diffraction grating is used to diffract the incident light interfering at the sample plane to produce the structured illumination pattern which is then translated in different phases to demodulate the frequency and rotated in different orientations to obtain isotropic two-dimensional (2D) resolution [13]. If two diffracted beams interfere, a 2D sinusoidal pattern will be produced, resulting in the resolution improvement only in two dimensions. Three-beam interference can form 3D structured illumination pattern and thus improve 3D resolution and optical sectioning capacity [14]. Obviously, mechanical translation and rotation lengthens acquisition time and may cause errors. Faster SIM imaging can be achieved by replacing the grating with a spatial light modulator (SLM) [15,16]. In recent years, the SIM theory and SLM-based SIM system have been greatly developed to enhance lateral resolution from ~120 nm [13] to ~50 nm [17,18] and temporal resolution from the second level to 100–300 Hz over hundreds to thousands of time points [19,20], making it widely adapted to biological studies, including the observation of the actin cytoskeleton remodeling [18], the characterization of the endoplasmic reticulum matrix [21], the identification of the mitochondrial cristae fusion [20], and the confirmation of the extensive hitchhiking interactions among different organelles [19]. However, these developments were mainly focused on two dimensions. Only a few follow-up studies have been conducted with 3D SIM, especially for live-cell imaging.

With routine fast biological applications in mind, we present here an alternative and more flexible system for 3D SIM based on two galvanometer sets conveniently controlling the incident angle (i.e. pattern period) and azimuth angle (i.e. pattern orientation) of the illumination light. The phase of the 3D structured illumination pattern is shifted by two piezoelectric stages. The proposed 3D SIM system has high flexibility in angle and intensity adjustments, pattern generation, and mode switching. As the galvanometer and piezoelectric stage allow the pattern to be rotated and phase-shifted on the order of kHz, the imaging speed of our system is high enough to observe diverse sub-cellular dynamics, such as mitochondria evolution.

2. Optical setup

All experiments are performed with our home-built 3D SIM system mounted on an inversed microscope (Nikon, Ellipse Ti), as shown in Fig. 1(a). The incident laser is coupled into a polarization maintaining single-mode fiber (PMSF, Oz Optics, core size 3 μm, NA 0.11), collimated and expanded by a beam collimator (BC, Thorlabs, ZC618FC), reflected by a reflecting mirror (M1), and separated into two beams by a polarized beam splitter (PBS1, Thorlabs, CCM1-PBS251/M). An achromatic half-wave plate (HWP1, Thorlabs, AHWP05M-600) is placed before PBS1 to adjust the intensity ratio between three interferring beams.

Fig. 1 The diagram of 3D SIM setup. (a) Experimental setup. PMSF, polarization maintaining single-mode fiber; BC, beam collimator; M1–M4, reflecting mirror; HWP1–HWP3, half-wave plate; PBS1, PBS2, polarized beam splitter; PZT1, PZT2, piezoelectric stage; GM1, GM2, scanning galvanometer; SL1, SL2, scanning lens; PR1, PR2, polarization rotator; L1-L4, lens; QWP, quarter-wave plate; DM, dichroic mirror; BFP, back focal plane; TL, tube lens. (b) The polarizations of the three incident light beams for three pattern orientations at the BFP of the objective lens. The three beams are always maintained at the same azimuthal linear polarization, which ensures the highest possible modulation contrast of the structured illumination pattern. (c) x-z slice of the 3D structured illumination pattern (top), showing both lateral and axial periodic structure, and corresponding frequency components of the pattern in Fourier domain.

Download Full Size | PDF

The transmitted light from PBS1 is separated again by a polarized beam splitter cube (PBS2, Thorlabs, CCM1-PBS251/M) into two beams that serve as the bilateral beams in 3D SIM interference. Similarly, another half-wave plate (HWP2, Thorlabs, AHWP05M-600) is used to equate the intensity of these two beams to ensure the highest modulation contrast of the lateral structured illumination pattern. The two beams are then directed into two identical high-speed galvanometers (GM1, GM2, Cambridge Technology, CTI 8310k) by two reflecting mirrors (M2, M3), respectively. A half-wave plate (HWP3, Thorlabs, AHWP05M-600) is inserted before GM2 to rotate the light polarization. The two scanning lenses (SL1, SL2, Thorlabs, SL50-CLS2) are matched with the corresponding galvanometers to rectify and focus the two beams at the conjugated plane of the back focal plane (BFP) of the objective lens. After the scanning lenses, two beams are combined by a beam splitter (BS1, Thorlabs, CCM1-BS013/M) and pass through a polarization rotator (PR1, Thorlabs, WPV10L). In general, the polarizations of all interfering beams in 2D/3D SIM need to be maintained at s-polarization for each pattern orientation to ensure the maximal modulation contrast of the illumination pattern in the sample plane so that the reconstructed result has minimal artifacts. The PR used here is a vortex phase plate that rotates the polarizations of the bilateral incident beams at the required s-polarization [22].

The reflected light served as the central normal beam in 3D SIM interference is direction-changed by a mirror (M4) and beam-expanded by a pair of achromatic doublet lenses (L2 and L3). The vortex wave plate-based polarization rotation scheme cannot change the polarization of the light through a fixed point of the plate, i.e., the normal beam always keeps a fixed linear polarization at different pattern orientations leading to a maximal contrast at only one orientation and poor contrast at the other two orientations [19,20,22]. Therefore, here, we use another polarization-rotation device consisting of a liquid crystal cell (Meadowlark, LPR-100-λ) and an achromatic quarter-wave plate (Thorlabs, SAQWP05M-700) to keep the center beam at the same polarization as the other two beams at each pattern orientation [16].

The aforementioned three beams are directed to enter into the input port of the microscope and relayed onto the BFP of the objective lens by two 4f systems. The combination of the three beams are accomplished by a beam splitter (BS2, Thorlabs, CCM1-BS013/M) placed at the confocal plane of the 4f systems. Each beam is then re-collimated by the objective (Nikon, 100X/1.49) and interferes with each other at the sample plane to form the 3D structured illumination pattern. The emitted fluorescence signal is collected by the same objective lens and reflected toward the camera by a dichromatic mirror cube (DM, Chroma, C174298). The DM is fabricated to transmit the incident light and reflect the emission light, therefore minimizing the mutual influence. The reflected fluorescence is focused by a tube lens (TL, Thorlabs, TTL200-A) onto an electron-multiplying charged-coupled device (EMCCD) camera (Andor, iXon Ultra 888) connected to the microscope output port.

Fifteen raw images consisting of 5 phases × 3 orientations need to be acquired at each focal plane to reconstruct a super-resolution 3D SIM image. The pattern period (i.e. incident angle of the light) and orientation (i.e. azimuth angle of the light) are controlled by two galvanometers. The pattern phase is shifted by two one-dimensional piezoelectric stages (PZT1, PZT2, PI, P-753) mounted under M2 and M4. Of note, the 2π/5 phase shifting in conventional 3D SIM systems is achieved either by directly translating the mechanical grating or the pattern loaded onto the SLM because the three interfering beams are generated and controlled by the same device. However, in our galvanometer-based system, phase shifting of the whole pattern needs to be accomplished by simultaneously moving the central beam π/5 (i. e. M4) and one of the bilateral beams 2π/5 (i. e. M2) because the three beams are controlled by different devices (See Section 3 for further details). In addition to the simple pattern generation approach, different imaging modes, such as 2D and total internal reflection fluorescence (TIRF) SIM, can also be implemented in our flexible imaging system by blocking the central normal light path.

Owing to the inherent parallelism of wide-field illumination method, the excitation intensity in SIM can be very low and is kept at ~1–10 W/cm² in our experiments leading to little photobleaching and photodamage. The exposure time of EMCCD is kept at 10–30 ms depending on laser power and samples. The acquisition order of raw 3D images is varying the pattern phase, the pattern orientation and the focal plane position, rather than varying the pattern phase, the focal plane position and the pattern orientation which is relatively slower [14,16]. The axial scanning step is typically set to 125 nm considering the axial resolution of 3D SIM and Nyquist sampling criteria. The axial scanning range typically varies from 2 μm to 3 μm which means axial scanning points of 16 to 24.

3. Principle

Three beams, one at the center of the BFP and the other two near opposite edges of the BFP, are used in 3D SIM to generate the 3D structured illumination pattern modulating the sample frequency information. To demodulate the high-frequency components that the 3D pattern brings in, two phase shift, one lateral shift $φ_{1}$ and the other axial shift $φ_{2}$ are required to be added into one of the two outer beams and the center beam, respectively [14,23]. The electric fields of each beam in the sample plane read:

{\begin{cases} E_{1} = A_{1} e^{i (k_{x} x + k_{y} y + φ_{1}) + i (k_{z} z)} \\ E_{2} = A_{2} e^{i (k z + φ_{2})} \\ E_{3} = A_{3} e^{- i (k_{x} x + k_{y} y) + i (k_{z} z)} \end{cases}

where

A_{1} - A_{3}

are the amplitudes of the three beams,

k_{x}, k_{y}, k_{z}

are the wave vectors, and

k = \sqrt{k_{x}^{2} + k_{y}^{2} + k_{z}^{2}} = n \cdot \frac{2 π}{λ}

with

n

the medium refractive index and

λ

the illumination wavelength.

(x, y, z)

is the coordinate in the object plane. The intensity of the 3D interference pattern generated by the three beams thus can be written as

\begin{matrix} I = (E_{1} + E_{2} + E_{3}) (E_{1}^{*} + E_{2}^{*} + E_{3}^{*}) \\ = \underset{m = 0}{\underset{︸}{A_{1}^{2} + A_{2}^{2} + A_{3}^{2}}} + \underset{m = 2}{\underset{︸}{A_{1} A_{3} e^{i (2 k_{x} x + 2 k_{y} y + φ_{1})}}} + \underset{m = - 2}{\underset{︸}{A_{3} A_{1} e^{- i (2 k_{x} x + 2 k_{y} y + φ_{1})}}} \\ \underset{m = 1}{\underset{︸}{+ e^{i (k_{x} x + k_{y} y)} (A_{1} A_{2} e^{i (k_{z} z - k z + φ_{1} - φ_{2})} + A_{2} A_{3} e^{i (k z - k_{z} z + φ_{2})})}} \\ \underset{m = - 1}{\underset{︸}{+ e^{- i (k_{x} x + k_{y} y)} (A_{2} A_{1} e^{i (k z - k_{z} z + φ_{2} - φ_{1})} + A_{3} A_{2} e^{i (k_{z} z - k z - φ_{2})})}} \end{matrix}

where

m = 0, \pm 1, \pm 2

is the frequency order of the pattern in Fourier domain and

*

denotes the conjugation transformation. Obviously, this illumination pattern contains seven frequency components, namely, seven delta functions in Fourier domain with different phases. The 0-th and ± 2-th frequency components locate on the plane of

k_{z} = 0

, while ± 1-th components locate on both sides of the plane of

k_{z} = 0

because of the existence of the terms

e^{\pm i (k_{z} - k) z}

(Fig. 1(c)). Compared with the pattern used in 2D SIM, 3D SIM pattern has the additional ± 1-th frequency components which bring in the improved resolution and better optical sectioning capacity along the optical axis.

To produce a well-conditioned separation matrix with disentangled lateral and axial functions, two conditions need to be met [14,23]. The first condition is that the lateral phase shift should be equal within the 0–2π interval; thus a general choice is $φ_{1} = 2 π / 5$ . The second is that the axial periodic structure of the 3D pattern should be maintained fixed in data acquisition, that is to say, the phases of seven frequency components in Eq. (2) must be able to be separated with z-wave vector, meaning that $φ_{1} - φ_{2} = φ_{2}$ , i.e., $φ_{1} = 2 φ_{2}$ [14]. Combining these two conditions, we have the above-mentioned conclusion that, in our case, the phase shift of the whole 3D illumination pattern needs to be achieved by simultaneously moving the central beam $π / 5$ ( $φ_{2}$ ) and one of the two outer beams $2 π / 5$ ( $φ_{1}$ ). Substituting the phase values into Eq. (2), we finally obtain

\begin{matrix} I = (E_{1} + E_{2} + E_{3}) (E_{1}^{*} + E_{2}^{*} + E_{3}^{*}) \\ = \underset{m = 0}{\underset{︸}{A_{1}^{2} + A_{2}^{2} + A_{3}^{2}}} + \underset{m = 2}{\underset{︸}{A_{1} A_{3} e^{i (2 k_{x} x + 2 k_{y} y + \frac{2 π}{5})}}} + \underset{m = - 2}{\underset{︸}{A_{3} A_{1} e^{- i (2 k_{x} x + 2 k_{y} y + \frac{2 π}{5})}}} \\ \underset{m = 1}{\underset{︸}{+ e^{i (k_{x} x + k_{y} y + \frac{π}{5})} (A_{1} A_{2} e^{i (k_{z} z - k z)} + A_{2} A_{3} e^{i (k z - k_{z} z)})}} \\ \underset{m = - 1}{\underset{︸}{+ e^{- i (k_{x} x + k_{y} y + \frac{π}{5})} (A_{2} A_{1} e^{i (k z - k_{z} z)} + A_{3} A_{2} e^{i (k_{z} z - k z)})}} \end{matrix}

After obtaining the five linearly independent raw images at each pattern direction using the phase shifting approach introduced above, the 3D SIM super-resolution reconstruction begins. Before the formal reconstruction, a judging procedure may be added to check the data quality using the “SIMcheck” plugin in ImageJ [24,25] (See Section 4 for further details). The experiment parameters, such as the objective numerical aperture (NA), light wavelength and pixel size are recorded to generate the system OTF. The data are first deconvolved and then transformed into the Fourier domain using Fast Fourier transform (FFT). In Fourier domain, the frequency components are extracted via auto-correlation and shifted back to their correct position, through which the frequency passband of the system is extended. Finally, the inverse Fourier transform is applied to produce the super-resolution image in spatial space. The output result can be served as a feedback to further optimize the acquisition and reconstruction procedure. Figure 2 illustrates this working flow of 3D SIM reconstruction in Fourier and spatial space.

Fig. 2 Overview of the principle and working flow of 3D SIM reconstruction. After the SIM system initialization, the raw image data is obtained at each axial plane with five pattern phases and three angles spaced by 60°. The “SIMcheck” method is then used to evaluate the data and provide the feedback to adjust the experimental parameters. If passed, the reconstruction is performed mainly in the Fourier domain. The final reconstructed result will further provide the feedback for the experiment acquisition and data processing.

Download Full Size | PDF

4. Experimental results

To demonstrate the performance of the proposed system, we conducted a series of biological experiments. First, we use 100 nm fluorescent nanoparticles as point emitter samples immersed in water solution and fixed on a coverslip. The imaging is performed with the excitation light wavelength of 488 nm and the corresponding fluorescent light wavelength of 520 nm, using an objective lens of 100 × 1.49 NA. The recorded deconvolved wide-field image and 3D SIM super-resolution image at the same x-y plane are shown in Figs. 3(a) and 3(b), illustrating the dramatic improvement of the lateral resolution. The particles overlapped with each other in the diffraction-limited raw image are readily discernible after the super-resolution reconstruction. And all beads in Fig. 3(b) are smaller than that of Fig. 3(a). The bottom-right images in Figs. 3(a) and 3(b) are the corresponding x-z slices of the particles in the blue-boxed regions, which were generated by recording a 3D stack consisting of 20 images with a lateral pixel size of 43 nm and an axial spacing of 125 nm between adjacent layers. Thus the data needs to be interpolated along the axial direction to match the 3D pixel size. The 3D field of view of the whole volume is $44 \times 44 \times 2.5$ μm³. By comparing these two images, we can find that 3D SIM method can indeed improve the axial resolution and remove out-of-focus background, above and below the plane of focus, which is dominant in the conventional wide-filed data. The improved optical sectioning capacity can be also demonstrated by the imaging result that the size and brightness of the out-of-focus nanoparticles decrease much more after 3D SIM reconstruction than that of particles in the focal plane. The lateral resolution improvement is further illustrated from the axial view by distinguishing the two adjacent nanoparticles (green arrowheads). To quantify the 3D resolution, the full width at half maximum (FWHM) of isolated nanoparticles are measured (Figs. 3(c) and 3(e)), yielded a 3D SIM FWHM of 104 nm laterally and 320 nm axially, respectively. We also measure the 120 nm separation distance of two closed particles (unresolved in the wide-field image). All results confirm the 3D resolution improvement of the 3D SIM technique.

Fig. 3 Experimental results of 100 nm fluorescent nanoparticles. (a) Wide-filed image of deconvolution and (b) super-resolution image with 3D SIM, illustrating the improvement of lateral resolution. The bottom-right images in (a) and (b) are the corresponding axial slices of the blue-boxed regions, illustrating the improvement of axial resolution and the removal of out-of-focus background. The green arrowheads indicate the lateral differentiation of two adjacent nanoparticles from axial view. (c) Lateral intensity profiles of the nanoparticles in (a) (gray line) and (b) (orange line). The measurement was performed for 20 particles. (d) Lateral intensity profiles of the orange dotted lines in (a) (gray line) and (b) (orange line). (e) Axial intensity profiles of the yellow dotted lines in the bottom-right images in (a) (gray line) and (b) (orange line). Scale bars, 1 μm in (a, b) and 300 nm in the bottom-right images in (a, b).

Download Full Size | PDF

Then, as a biological performance test, we image mitochondria in bovine pulmonary artery endothelial cells stained with MitoTracker Red. An example data shown in Fig. 4(a) contains five linearly independent raw images with equidistant phase at the pattern orientation of 0° (first row), 60° (second row) and 120° (third row), respectively. The corresponding schematic frequency distribution and extended but overlapped OTF at each direction are also shown. With the raw SIM data, prior to reconstruction, we quasi-quantitatively diagnose the image quality using “SIMcheck” method [25] from three sources: channel intensity profile, modulation contrast-to-noise ratio (MCN) and raw Fourier projection (Fig. 4(b)), which are the key for a good and few-artifact reconstruction and are often used to evaluate SIM data [24]. The first channel intensity profile check provides the average intensity differences between fifteen raw images and thus indicates, for example, the possible three illumination angle difference, fluorescence fluctuation and photobleaching speed. The channel intensity profile of our data shows a negligible variation $< ~ 5 %$ over the whole image sequence, illustrating the high illumination quality. The second MCN check provides the local ratio of the illumination pattern contrast to out-of-focus background, which is crucial for SIM reconstruction as it indicates the ability that the amount of high-frequency components can be accurately reassigned. The average MCN of our data is 8.21 which is a good value for a good reconstruction according to the “SIMcheck” criteria. The final raw Fourier projection provides a visual inspection of the data quality in Fourier domain. The ± 1-th and ± 2-th high-frequency spots are clearly shown in the check result, proving the robustness of our system. Consistent with the nanoparticle experiment, the reconstructed super-resolution SIM image of the mitochondria features not only the volumetric resolution improvement, but also the evident contrast enhancement (the effective suppression of the out-of-focus information), as shown in Fig. 4(c). The intensity profiles shown in Fig. 4(d) further quantitatively demonstrate the strong resolution improvement after 3D SIM reconstruction.

Fig. 4 Experimental results of mitochondria in a bovine pulmonary artery endothelial cell stained with MitoTracker Red. (a) Each line contains five linearly independent raw images at each pattern orientation with equidistant phase, the corresponding frequency distribution and the extended OTF. (b) “SIMcheck” results of the channel intensity profile, modulation contrast-to-noise ratio (MCN) and raw Fourier projection of the raw data shown in (a). (c) Wide-filed image of deconvolution (left) and super-resolution image with 3D SIM (right). The bottom-left images are the OTF before (left) and after reconstruction (right), illustrating the extension of the frequency support. (d) Intensity profiles of the orange dotted lines in (c). Scale bars, 10 μm in (a) and 5 μm in (c).

Download Full Size | PDF

We next image microtubules in whole U2OS cells stained with Alexa Fluor 488. The data shown in Fig. 5 is reconstructed from a raw 3D stack consisting of 24 images with an axial step size of 125 nm, meaning an imaging thickness of 3 μm. Figures 5(a) and 5(b) show the diffraction-limited wide-field images and SIM reconstructed images at axial positions of z = 1.25 μm and 2 μm, respectively, illustrating the improvement of both lateral and axial resolution and the true optical sectioning capacity resulting from rejecting out-of-focus background. From the axial view (Figs. 5(c) and 5(d)), we can more visually find the 3D resolution improvement, especially along the axial direction. After 3D SIM reconstruction, the apophysis morphology of the cell center is revealed, which is hardly to be distinguished with conventional microscopy. The cell center is thicker than its margin area because the nucleus exists in the center.

Fig. 5 Experimental results of microtubules in a U2OS cell stained with Alexa Fluor 488. (a, b) Wide-filed image of deconvolution (left) and super-resolution image with 3D SIM (right) at axial positions of z = 1.25 μm and 2 μm, respectively, illustrating the improvement of 3D resolution and optical sectioning capacity. (c, d) Axial slices cut through the yellow and orange dashed lines in (a). Scale bars, 4 μm in (a, b) and 2 μm in (c, d).

Download Full Size | PDF

Finally, we demonstrate the live-cell imaging capacity of the proposed 3D SIM system by studying the time-lapse movement of mitochondria in U2OS cells stained with Atto 647N. Mitochondria are organelles supplying energy in most living eukaryotic cells and play a crucial role in cell dynamics. In our experiments, several mitochondria forms, such as rodlike, annular and granular mitochondria are observed (Fig. 6). Again, reconstructed 3D SIM result illustrates much improved lateral resolution over conventional microscopy (Fig. 6(a)). By comparing the 3D SIM images reconstructed at different axial positions, we find that the mitochondria at the cell center (Fig. 6(a)) disappear at the large depth (Fig. 6(c)), demonstrating the improved axial resolution and good optical sectioning ability of 3D SIM in live-cell imaging. The cell is scanned 15 times through focus with a step of 0.16 μm and exposure time of 15 ms per image, which means about 4 s is required to record one 3D SIM volume for a whole U2OS cell considering the image transfer and polarization rotation time. The time-lapse 3D SIM results at different planes show the mitochondria dynamics in the whole 3D volume, including extension, retraction, migration, fission, and fusion (Figs. 6(b) and 6(d)).

Fig. 6 Live-cell experimental results of mitochondria in a U2OS cell stained with Atto 647N. (a, c) Super-resolution image with 3D SIM at axial positions of z = 0.8 μm and z = 1.92 μm, respectively. The bottom-right of (a) is the corresponding wide-filed image of deconvolution. (b, d) Time lapse 3D SIM image sequences of the blue boxed regions in (a) and (c), showing mitochondria movement and fusion (orange arrowheads) and fission event (yellow arrowheads). Scale bars, 5 μm in (a, c), 10 μm in the bottom-right image in (a), and 1 μm in (b, d).

Download Full Size | PDF

5. Conclusion

In this paper, we demonstrated our galvanometer-based system as a practical tool for 3D SIM imaging. It not only maintains all the advantages of conventional 3D SIM technique, including double 3D resolution improvement, good optical sectioning ability, low illumination intensity and minimal sample preparation requirement, but also makes further developments on 3D SIM, including improved temporal resolution and more flexible system adjustment (intensity, angle, and azimuth). In fact, the system also has the convenient switching ability among other imaging modes, such as (spinning-azimuth) TIRF [4,26], 2D SIM and TIRF-SIM [15], by simply blocking the center beam and changing the voltages loaded onto two galvanometers without any hardware modifications, which further enhances its practicality in different biological applications.

Although our galvanometer-based and SLM-based 3D SIM systems [16] have the ability to be applied in live-cell imaging studies, they are still difficult for ultrafast imaging applications because an electric liquid crystal device is used to rotate the central beam polarization on the order of milliseconds [19], which extends the imaging time. Note the ultrafast imaging here refers to acquiring raw data on the order of tens to hundreds of images per second, which means the exposure time should be extremely low (sub-millisecond) and thus imposes very rigorous requirements on imaging and reconstruction conditions [19,20]. To facilitate the ultrafast imaging, both the system and the reconstruction algorithm need to be modified, such as designing new fixed-beam polarization rotator requiring no additional setting time and introducing Hessian deconvolution into 3D SIM data reconstruction [20]. On the other hand, the interfering beams in 3D SIM can also be maintained at the circular polarizations at all pattern directions at the cost of the decreased modulation contrast [27]. The optimization of sample labeling strategy, the adoption of brighter fluorescent dyes and the use of faster sCMOS cameras are also indispensable for ultrafast imaging. These developments will further greatly expand the application of 3D SIM technique in live-cell life sciences.

Another possible development of our system is to extend its application in multi-color imaging. As the system reported here rotates the polarizations of the bilateral interfering beams with a vortex phase plate that is wavelength-dependent. Therefore, the plate needs to be replaced when imaging dyes have different wavelengths, which is not very suitable for multi-color studies, such as co-localization analysis. By using the polarization rotator assembled from azimuthal achromatic half-wave plates [19,20], this limitation can be addressed straightforwardly.

Funding

National Basic Research Program of China (973 Program) (2015CB352003); National Natural Science Foundation of China (NSFC) (61827825, 61427818, and 61735017); Natural Science Foundation of Zhejiang province (LR16F050001); Fundamental Research Funds for the Central Universities (2018FZA5005).

Acknowledgments

The authors would like to thank Mats G. L. Gustafsson (Deceased April 2011) for great contribution to the development of SIM microscopy, Lothar Schermelleh (Oxford University) for great contribution to the development of quantitative and successful SIM (especially 3D SIM), and Youhua Chen (North University of China) and Ruizhi Cao (Caltech) for the kind assistance.

Disclosures

The authors declare that there are no conflicts of interest related to this article.

References

1. L. Schermelleh, A. Ferrand, T. Huser, C. Eggeling, M. Sauer, O. Biehlmaier, and G. P. C. Drummen, “Super-resolution microscopy demystified,” Nat. Cell Biol. 21(1), 72–84 (2019). [CrossRef] [PubMed]

2. S. J. Sahl, S. W. Hell, and S. Jakobs, “Fluorescence nanoscopy in cell biology,” Nat. Rev. Mol. Cell Biol. 18(11), 685–701 (2017). [CrossRef] [PubMed]

3. Y. M. Sigal, R. Zhou, and X. Zhuang, “Visualizing and discovering cellular structures with super-resolution microscopy,” Science 361(6405), 880–887 (2018). [CrossRef] [PubMed]

4. W. Liu, K. C. Toussaint Jr., C. Okoro, D. Zhu, Y. Chen, C. Kuang, and X. Liu, “Breaking the axial diffraction limit: a guide to axial super-resolution fluorescence microscopy,” Laser Photonics Rev. 12(8), 1700333 (2018). [CrossRef]

5. A. von Diezmann, Y. Shechtman, and W. E. Moerner, “Three-dimensional localization of single molecules for super-resolution imaging and single-particle tracking,” Chem. Rev. 117(11), 7244–7275 (2017). [CrossRef] [PubMed]

6. H. Blom and J. Widengren, “Stimulated emission depletion microscopy,” Chem. Rev. 117(11), 7377–7427 (2017). [CrossRef] [PubMed]

7. F. Bottanelli, E. B. Kromann, E. S. Allgeyer, R. S. Erdmann, S. Wood Baguley, G. Sirinakis, A. Schepartz, D. Baddeley, D. K. Toomre, J. E. Rothman, and J. Bewersdorf, “Two-colour live-cell nanoscale imaging of intracellular targets,” Nat. Commun. 7(1), 10778 (2016). [CrossRef] [PubMed]

8. J. Tønnesen, V. V. G. K. Inavalli, and U. V. Nägerl, “Super-resolution imaging of the extracellular space in living brain tissue,” Cell 172(5), 1108–1121 (2018). [CrossRef] [PubMed]

9. M. Sauer and M. Heilemann, “Single-molecule localization microscopy in eukaryotes,” Chem. Rev. 117(11), 7478–7509 (2017). [CrossRef] [PubMed]

10. H. Takakura, Y. Zhang, R. S. Erdmann, A. D. Thompson, Y. Lin, B. McNellis, F. Rivera-Molina, S. N. Uno, M. Kamiya, Y. Urano, J. E. Rothman, J. Bewersdorf, A. Schepartz, and D. Toomre, “Long time-lapse nanoscopy with spontaneously blinking membrane probes,” Nat. Biotechnol. 35(8), 773–780 (2017). [CrossRef] [PubMed]

11. Y. Wu and H. Shroff, “Faster, sharper, and deeper: structured illumination microscopy for biological imaging,” Nat. Methods 15(12), 1011–1019 (2018). [CrossRef] [PubMed]

12. R. Heintzmann and T. Huser, “Super-resolution structured illumination microscopy,” Chem. Rev. 117(23), 13890–13908 (2017). [CrossRef] [PubMed]

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

14. M. G. Gustafsson, L. Shao, P. M. Carlton, C. J. Wang, I. N. Golubovskaya, W. Z. Cande, D. A. Agard, and J. W. Sedat, “Three-dimensional resolution doubling in wide-field fluorescence microscopy by structured illumination,” Biophys. J. 94(12), 4957–4970 (2008). [CrossRef] [PubMed]

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

16. L. Shao, P. Kner, E. H. Rego, and M. G. Gustafsson, “Super-resolution 3D microscopy of live whole cells using structured illumination,” Nat. Methods 8(12), 1044–1046 (2011). [CrossRef] [PubMed]

17. 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]

18. D. Li, L. Shao, B.-C. Chen, X. Zhang, M. Zhang, B. Moses, D. E. Milkie, J. R. Beach, J. A. Hammer 3rd, M. Pasham, T. Kirchhausen, M. A. Baird, M. W. Davidson, P. Xu, and E. Betzig, “Extended-resolution structured illumination imaging of endocytic and cytoskeletal dynamics,” Science 349(6251), aab3500 (2015). [CrossRef] [PubMed]

19. Y. Guo, D. Li, S. Zhang, Y. Yang, J.-J. Liu, X. Wang, C. Liu, D. E. Milkie, R. P. Moore, U. S. Tulu, D. P. Kiehart, J. Hu, J. Lippincott-Schwartz, E. Betzig, and D. Li, “Visualizing intracellular organelle and cytoskeletal interactions at nanoscale resolution on millisecond timescales,” Cell 175(5), 1430–1442 (2018). [CrossRef] [PubMed]

20. X. Huang, J. Fan, L. Li, H. Liu, R. Wu, Y. Wu, L. Wei, H. Mao, A. Lal, P. Xi, L. Tang, Y. Zhang, Y. Liu, S. Tan, and L. Chen, “Fast, long-term, super-resolution imaging with Hessian structured illumination microscopy,” Nat. Biotechnol. 36(5), 451–459 (2018). [CrossRef] [PubMed]

21. J. Nixon-Abell, C. J. Obara, A. V. Weigel, D. Li, W. R. Legant, C. S. Xu, H. A. Pasolli, K. Harvey, H. F. Hess, E. Betzig, C. Blackstone, and J. Lippincott-Schwartz, “Increased spatiotemporal resolution reveals highly dynamic dense tubular matrices in the peripheral ER,” Science 354(6311), aaf3928 (2016). [CrossRef] [PubMed]

22. Y. Chen, R. Cao, W. Liu, D. Zhu, Z. Zhang, C. Kuang, and X. Liu, “Widefield and total internal reflection fluorescent structured illumination microscopy with scanning galvo mirrors,” J. Biomed. Opt. 23(4), 1–9 (2018). [CrossRef] [PubMed]

23. A. Classen, J. von Zanthier, and G. S. Agarwal, “Analysis of super-resolution via 3D structured illumination intensity correlation microscopy,” Opt. Express 26(21), 27492–27503 (2018). [CrossRef] [PubMed]

24. J. Demmerle, C. Innocent, A. J. North, G. Ball, M. Müller, E. Miron, A. Matsuda, I. M. Dobbie, Y. Markaki, and L. Schermelleh, “Strategic and practical guidelines for successful structured illumination microscopy,” Nat. Protoc. 12(5), 988–1010 (2017). [CrossRef] [PubMed]

25. G. Ball, J. Demmerle, R. Kaufmann, I. Davis, I. M. Dobbie, and L. Schermelleh, “SIMcheck: a toolbox for successful super-resolution structured illumination microscopy,” Sci. Rep. 5(1), 15915 (2015). [CrossRef] [PubMed]

26. Y. Chen, W. Liu, Z. Zhang, C. Zheng, Y. Huang, R. Cao, D. Zhu, L. Xu, M. Zhang, Y.-H. Zhang, J. Fan, L. Jin, Y. Xu, C. Kuang, and X. Liu, “Multi-color live-cell super-resolution volume imaging with multi-angle interference microscopy,” Nat. Commun. 9(1), 4818 (2018). [CrossRef] [PubMed]

27. H.-C. Huang, B.-J. Chang, L.-J. Chou, and S.-Y. Chiang, “Three-beam interference with circular polarization for structured illumination microscopy,” Opt. Express 21(20), 23963–23977 (2013). [CrossRef] [PubMed]

Three-dimensional super-resolution imaging of live whole cells using galvanometer-based structured illumination microscopy

Abstract

1. Introduction

2. Optical setup

3. Principle

4. Experimental results

5. Conclusion

Funding

Acknowledgments

Disclosures

References

Cited By

Figures (6)

Equations (3)

Optics Express