1. Introduction

Many imaging methods and tools have been developed to visualize biological structure, such as cells and tissues, in the last decades [1–3]. Among them, Cryo-Soft X-ray tomography (Cryo-SXT) has gained more and more attention due to its high-resolution (tens of nanometer), large penetration depth (several micrometers) and non-destructive three-dimensional (3D) aqueous cell imaging capacity [4–10]. For the purposes of soft X-ray imaging, biological samples need to be imaged under cryogenic conditions. This preserves the natural state of specimens, ensures minimal drift due to biological action and reduces radiation damage during imaging [11]. When imaging photon energy is between the K-absorption edge of oxygen (530 eV, 2.34 nm) and that of carbon (280 eV, 4.38 nm), known as ‘water window’, X-ray can be absorbed by carbon- and nitrogen-containing organic materials more easily than water [12]. For this reason, protein and lipid rich in organelles can be distinguished readily from water.

Although SXT can visualize subcellular structures in 3D with high spatial resolution, it cannot identify biological molecules that perform cellular functions [5]. Alternatively, specific molecules can be labeled by fluorophores and be located by well-established fluorescence microscopy (FM). Common labeling methods include small molecule staining, immunofluorescence labeling and genetically encoded fluorescent tags labeling. FM is an attractive complementary method to SXT. Therefore, SXT-FM correlative imaging realizes the superiorities of two imaging modes concurrently.

In SXT-FM correlative imaging, the specimen is sequentially imaged by FM and SXT. Then the two sets of data are combined to acquire morphological and functional information, which is inaccessible by individual imaging method, of the subcellular organelle. It can be applied to a variety of cell types, including small bacteria and large eukaryotic cell. Different FM methods, including wide-field FM, confocal fluorescence microscopy and super-resolution fluorescence microscopy (SRFM), can be used in SXT-FM correlative imaging. The correlative imaging with wide-field FM has already solved some significant scientific problems, such as identification of organelles, the interaction mechanism of superparamagnetic iron oxide nanoparticles with a breast cell line, and the interrelationship between endoplasmic reticulum (ER) and endosomes, autophagosomes [7,13,14].

Compared with SXT, the spatial resolution of wild-field FM is too low. Confocal fluorescence microscopy was adopted to improve the spatial resolution [15]. With this higher spatial resolution FM imaging, the ER tubules wrapped around and constricted mitochondria before a fission event was revealed [16], and the characterization of the inactive X chromosome topological arrangement in near-native state cells was discovered [17]. Even though, the mismatch in spatial resolution between traditional FM (diffraction limit of light, 200 nm) and SXT (tens of nanometer) still existed and limited mapping of small structures, such as membranes.

Fortunately, the recent emerging of SRFM, has broken the diffraction barrier, and made tremendous impacts in the scientific research [18–23]. There are two main kinds of SRFMs. Ones rely on specific structured illumination, for example, structured illumination microscopy (SIM) [24] and stimulated emission depletion (STED) microscopy [25,26]. The others rely on locating and overlying the centers of sparse blinking fluorescent molecules, for example, photo-activated localization microscopy (PALM) [27], stochastic optical reconstruction microscopy (STORM) [28], ground state depletion microscopy followed by individual molecule return (GSDIM) [29], and super-resolution optical fluctuation imaging (SOFI) [30], collectively termed single-molecule localization microscopy (SMLM) [2]. Complex optical setups are required to generate specific structured illumination for SIM or STED, although super resolution is realized (two-fold enhancement in SIM and 20–60 nm in STED) [31]. By comparison, the setups of SMLM are relatively simple. Even though the high-intensity illumination may generate cellular phototoxicity and the temporal resolution is low, the spatial resolution can reach 2–30 nm [2,31].

The rapid development of super-resolution microscopy provides new opportunities to correlative imaging. Cryo-SIM has been employed to construct 3D correlative imaging with SXT, which enabled to observe the process of reovirus infection in mammalian cells [32,33]. This strategy can produce 3D images of volumes in depth of 10 µm with spatial resolution up to 200 nm under cryogenic conditions. Indeed, the spatial resolution of SIM is still mismatch with SXT. Compare to SIM, the SMLM provides a relatively simple setups and higher spatial resolution. Such superiority driven the development of correlative SXT and STORM, which has been applied to localize the cholesterol crystals in macrophage-like RAW 264.7 cells [34,35]. This correlative method achieves high resolution of 70 nm, but the penetration depth is limited in 1 µm. Thus, correlative approach able to image thick samples with high resolution is required and exciting.

In this work, GSDIM was utilized for the super-resolution fluorescence imaging due to its super-resolution, facile and cost-effective. A GSDIM was built at National Synchrotron Radiation Laboratory (NSRL), University of Science and Technology of China. With the epi-illumination method, it can observe thicker sample. Thus, the correlative imaging combining SXT with GSDIM was developed to locate biological molecules inside the cell with high spatial resolution. This imaging platform can deliver comprehensive views of both cellular structure and the location of molecules. Furthermore, spatial resolution of 30 nm, which matches with that of SXT, was achieved in this SRFM. The following parts are organized as follows: In Sec. 2 we introduce the correlative imaging method. Sec. 3 presents details of the experiments and Sec. 4 displays the experimental results and demonstrates the identification of mitochondria and lysosomes inside 4T1 cells. Finally, it is concluded in Sec. 5.

2. Correlative Cryo-SXT and GSDIM imaging method and set-up

2.1 Home-built GSDIM

SRFM has broken the diffraction barrier which limits the spatial resolution of visible light microscopy to 200–300 nm in the lateral direction and 500–700 nm in the axial direction [31]. SMLM is a super-resolution method locating and overlying the centroid position of spatially isolated fluorophores. As one of SMLM methods, GSDIM is based on switching the majority of the fluorophores to their triplet state T₁ or another metastable dark state while locating the centroid positions of those that are still left or have spontaneously returned to the ground state S₀ [29,36]. Fluorophores are excited from the ground state S₀ to the fluorescent state S₁ which has a short lifetime (τ_fl), then most of them will emit fluorescence and return to S₀, a few of them (with a typical probability Φ_isc≤0.1%) will switch to the triplet state T₁ which have a long lifetime (τ). Fluorophores may switch from T₁ to another metastable dark state D with a similar or even longer lifetime. After the lifetime of T₁ or D, fluorophores will return to S₀ spontaneously without emitting fluorescence and can be repeatedly excited to S₁, thus, there is only one continue laser needed in GSDIM. Most fluorophores will switch to the long dark state T₁ or D after the same repeated excitation in a short time caused by a high illumination intensity laser. Then the fraction of fluorophores in S₀ is

To reduce the fraction, it is necessary to extend lifetime $\tau $ of the dark state. Embedding specimens in poly (vinyl-alcohol) (PVA) solution or Tris buffer containing glucose oxidase to avoid exposing fluorophores to the triplet-quenching oxygen in the environment, effectively increase the lifetime of the triplet state. Due to the long lifetime of the triplet or another metastable dark state, the bright fluorophores, or emitters are sparse in the exposure time of one image. When the emitters are sparse enough that there is only one emitter in a range of the point spread function (PSF), the position of the emitter is determined by localizing the center of the PSF with a Gaussian fitting. The accuracy of the locating, sub-50-nm in most cases, depends on the accuracy of the fitting which theoretically scales inversely with the square root of the number of detected photons [37]. Then, two fluorophores within a PSF can be separated by locating their centers at different times to break the diffraction barrier. The super-resolution image is reconstructed with the located centroid positions of fluorophores in hundreds or thousands of frames acquired in a period of time.

The GSDIM was built at the beamline BL07W, NSRL. The home-built GSDIM (Fig. 1) was based on a wide-field FM (Eclipse Ci, Nikon). Laser (OBIS 488 nm LS 100 mW Laser, Coherent) with wavelength of 488 nm was expanded and focused by two lenses (Thorlabs) onto the back focal plane of an air objective (50×, 0.6 NA, CFI60 TU Plan Epi ELWD, Nikon), after being reflected by a dichroic mirror inside the filter cube (FITC filter cube, excitation filter: 480/30 nm, dichroic mirror: 505 nm, barrier filter 535/45 nm, Nikon). The excitation light on the sample plane was parallel light with a peak illumination intensity value of about 7.8 kW/cm² and a diameter of 50 µm. The specimen-containing grid was immersed into the Tris buffer containing glucose oxidase and suspended 7 mm from the bottom of the dish to reduce background fluorescence. The emitted light from the sample was collected by the same objective, sent through an emission filter inside the filter cube, and focused onto a sCMOS camera (Hamamatsu Orca Flash 4.0 V3). The fluorescence image stack with hundreds or thousands of frames of the same focal plane was acquired continuously by the sCMOS and then reconstructed by ThunderSTORM [38], a plugin of ImageJ, into a two dimensional GSDIM image.

 

Fig. 1. Home-built GSDIM for correlative imaging. (a) Schematic drawing of the home-built GSDIM. (b) The picture of home-built GSDIM.

Download Full Size | PDF

2.2 Soft X-ray microscopy

The soft X-ray microscopy (SXM) end-station is installed at the beamline BL07W, Nation Synchrotron Radiation Laboratory (NSRL, Hefei, China) [7,39], and delivers imaging by absorption contrast model. The soft X-rays are focused by an elliptical capillary condenser. Coupled with a micro-zone plate, the whole system can obtain images with an X-ray energy ranging from 280 eV to 700 eV with a spatial resolution of 30 nm. In imaging biological specimens, the X-ray photon energy of 520 eV was selected to obtain high contrast image. Biological samples on standard grids, generally used in electron microscopy, were rapidly frozen and transferred into SXM by a transfer chamber cooled by liquid nitrogen. Biological samples in the imaging chamber of SXM were stored in a vacuum and at the temperature about 100 K to keep the vitreous ice. Regions of interest (ROI) were located by an on-line light microscopy with a 10 ${\times} $ objective and further located by high magnification mosaics of SXM projections. Soft X-ray projections of the sample in the ROI were collected in viewing angle range of ±65° and field of view of 12.4 µm ${\times} $ 12.4 µm. The projections were subsequently reconstructed by an in-house developed Matlab script to obtain a 3D visualization of the sample [40]. Each voxel in the 3D image is a direct measurement of the linear absorption coefficient (LAC), relating to the composition and density of biomolecules. Then, the 3D image was segmented based on the LAC and rendered using the software Amira.

2.3 Correlative method

The GSDIM and SXT images of the same sample are fused in one image to provide comprehensive information, which is the correlative method. In general, specimens are firstly imaged by GSDIM, then imaged by SXM, because the fluorophores can be bleached by the soft X-ray [11]. After imaged by GSDIM, Tris buffer containing glucose oxidase on the specimens is rinsed by water, then specimens are rapidly frozen by plunging into liquid ethane and then stored in liquid nitrogen, and put into the transfer chamber cooled by liquid nitrogen and transferred into the SXM to be imaged.

For precise alignment of images obtained by GSDIM and SXM, a soft X-ray mosaics projection and a wide-field fluorescence image with a larger field of view, than that of correlative imaging of SXT and GSDIM, are used for the alignment by a Matlab script based on previous studies (Please refer to Supplement 1 for more details) [7]. The SXT reconstruction is aligned with the soft X-ray mosaics projection, while the super-resolution reconstruction is aligned with the wide-field fluorescence image. Then the SXT reconstruction and super-resolution reconstruction are matched based on the result of the alignment of the soft X-ray mosaics projection and the wide-field fluorescence image. Lastly, the subcellular structure in the SXT reconstruction which is matched with the GSDIM image is located along the depth. Alignment of images of the same specimen obtained from different methods is vital to the correlative method. Fiducials, such as organelles and fluorescence spheres, which have obvious contrast in different imaging methods are used to match these images. In general, the specimen is regarded as a rigid body, and the alignment of these images is to calculate the transformation of rigid body by the positions of fiducials [41]. Firstly, the fiducials in images obtained by different methods were located manually, the fiducial centers in soft X-ray projections and wide-field fluorescence images were located by object-pixel-value-weighted algorithm and Gaussian fitting respectively. Then the transformation (including translation, rotation, and scaling) of the two groups of centers was calculated by the least-squares method. The detailed progress is shown in previous studies [7].

Three methods can be used to estimate the accuracy of the alignment. One is the leave-one-out method [42], one fiducial marker is excluded, then the predicted position of this fiducial marker is calculated by the transform of the other fiducial markers, the alignment error is obtained by comparing this fiducial marker and its predicted position. This step is repeated for all fiducial markers, and the average alignment error is regarded as an estimated value of the alignment accuracy. Another method is to directly calculate the difference between the actual positions and transformed positions of the fiducial markers [41]. The third one is to calculate the alignment error map [32,43,44]. The fiducial localization error (FLE) in correlative method is defined by the total fiducial localization error (TFLE) as

3. Cell experiments

4T1 cells were cultured for the cell experiments. Filamentous actin (F-actin), with a diameter of about 10 nm [45], was an ideal cellular structure for super-resolution studies. Therefore, F-actin in 4T1 cells was imaged to ascertain the spatial resolution of home-built GSDIM. Mitochondria and lysosomes were labeled for the correlative imaging of SXT and GSDIM.

3.1 Cell experiments for GSDIM

4T1 cells were cultured on the standard grids generally used in electron microscopy (formvar films coated nickel grids, catalog No. AG100N, Beijing Zhongjingkeyi Technology, China) for 10 h. Then cells-containing grids were rinsed three times in phosphate-buffered saline (PBS) and fixed by incubating with 4% paraformaldehyde (PFA) in PBS at room temperature for 15 minutes. An additional three washes in PBS were performed to remove any excess of fixative. Cells-containing grids were permeabilized in 0.1% Triton X-100 in PBS for 15 minutes and washed three times in PBS. Cells were then stained with Alexa Fluor 488 Phalloidin (Alexa Fluor 488 Phalloidin, excitation/emission: 495/518 nm, Catalog No. A12379, Invitrogen, China) to a final concentration of 16.5 nM in PBS at room temperature for 30 minutes and washed three times in PBS to remove excess fluorescent dyes. Finally, 0.2 µL of solution with TetraSpeck spheres (200 nm diameter, showing four well separated excitation/emission peaks, 360/430 nm (blue), 505/515 nm (green), 560/580 nm (orange) and 660/680 nm (dark red), Catalog No. T7280, Invitrogen) was added directly onto the cells to correct the drift of the sample. The treated cells were then imaged by GSDIM on grids.

3.2 Cell experiments for correlative imaging

4T1 cells were cultured and fixed as described above. For staining mitochondria, cells were permeabilized with 0.5% Triton X-100 for 20 minutes, and washed with PBS for three times. Then cells were blocked in PBS containing 3% bovine serum albumin (BSA) for 30 minutes. After removing the blocking solution, cells were incubated with 1:800 Anti-TOMM20 Antibody (Anti-TOMM20 Antibody, Host/Isotype: Rabbit/IgG, Catalog No. GB111481, Wuhan gugeshengwu Technology, China) diluted in PBS overnight at 4°C. After rinsed in PBS for three times, cultures were incubated 50 minutes with 1:400 Alexa Fluor 488 Goat anti-Rabbit lgG secondary antibody (Alexa Fluor 488 Goat anti-Rabbit, Catalog No. GB25303, Wuhan gugeshengwu Technology, China) diluted in PBS. Then cells-containing grids were washed three times in PBS.

For staining lysosomes, fixed 4T1 cells were incubated in 1% BSA / 10% normal goat serum / 0.3 M glycine in 0.1% PBS-Tween for 1 h to permeabilize and block the cells. After removing the blocking solution, cells were incubated with 1:200 Anti-LAMP2A Antibody (Anti-LAMP2A antibody Lysosome Marker, Catalog No. ab18528, Abcam, UK) diluted in PBS overnight at 4°C. After rinsing by PBS for three times, cultures were incubated with 1:1000 Alexa Fluor 488 goat anti-rabbit lgG (H&L) (Goat Anti-Rabbit IgG H&L Alexa Fluor 488, Catalog No. ab150077, Abcam, UK) diluted in PBS for 1h. Then cells-containing grids were washed three times in PBS.

After being stained, the TetraSpeck spheres were added onto the cells-containing grids as fiducial markers. The prepared cells on grids were imaged by GSDIM. After fluorescence image collection, grids were plunged into liquid ethane and stored in quid nitrogen. After being frozen, cells-containing grids were transferred into the SXM device (BL07W beamline at NSRL). The ROIs were fixed by an on-line light microscopy and were imaged by high resolution Cryo-SXM. Tomography of the ROI was performed with a maximum angle of ${\pm} 65^\circ $. with a step of 1°.

4. Results and discussion

4.1 Super-resolution fluorescence imaging

Figure 2 shows the wide-field and super-resolution images of the stained F-actin. The labeled F-actin can be found by the wide-field fluorescence image, the full width at half maximum (FWHM) of the image was measured as 520 nm (in Fig. 2(a) and Fig. 2(b)). The measured value much larger than the diameter reported by literature (about 10 nm) points that the spatial resolution of wide-field FM is about 510 nm (subtract the diameter of the F-actin). The super-resolution image of the same F-actin is shown in Fig. 2(c). The spatial resolution of the reconstructed GSDIM image can be evaluated by two methods. The first one is to calculate the FWHM of the Gaussian distribution, whose standard deviation is the locating accuracy of the emitter [46]. The second one is based on the image decorrelation analysis without any prior knowledge [32,33,47], implemented in ImageJ using default parameters. Figures 2(d) and 2(e) indicate the spatial resolution is about 30 nm, which is also confirmed by the second method. Furthermore, more high frequency signal is contained in super-resolution image from the spectral power density shown in Fig. 2(f), where both cut-off frequencies are consistent with the calculated resolutions (1/1.96 vs 510 nm and 1/33 vs 30 nm). Therefore, the spatial resolution of the home-built GSDIM is about 30 nm. Notably, the achieved resolution is comparable to the best performance of STORM, which has been employed in the correlative imaging [34].

 

Fig. 2. Wide-field and super-resolution images of the stained F-actin. (a) Wide-field fluorescence image of the stained F-actin. (b) The graph of average fluorescence intensity of the stained F-actin in the red rectangle of (a) shows that the FWHM of the F-actin in wide-field image is 520 nm. (c) Super-resolution fluorescence image of the same F-actin shown in (a). (d) The graph of average intensity of the F-actin in the red rectangle of (c) shows that the FWHM of the F-actin in super-resolution image is 40 nm. (e) The histogram of the locating accuracy distribution of the emitters in (c) shows that the minimal locating accuracy is about 12.5 nm excluding the accuracy values with too few counts, which means that the spatial resolution of super-resolution image is about 30 nm (corresponding to the FWHM of the Gaussian fitting with a standard deviation of 12.5 nm). (f) Spectral power density shows the increased high frequency information, compared super-resolution image with wide-field image. The vertical gray bars show the cut-off frequencies in wide-field image (1/1.96) and super-resolution image (1/33).

Download Full Size | PDF

4.2 Correlative imaging of mitochondria and lysosomes in 3D

The constructed correlative imaging strategy was applied for mitochondria and lysosomes imaging, the data is available in [48,49]. The labeled cells were sequentially imaged by GSDIM and SXT. The alignment of GSDIM and SXM images is shown in Fig. 3. The wide-field image (See Fig. 3(a)) and the mosaics of soft X-ray projections (See Fig. 3(b)) with a larger field of view were aligned based on the fiducial markers, and matched (See Fig. 3(c)) with the alignment error shown in Fig. 3(d). Then the super-resolution GSDIM image (See Fig. 3(e)) was matched with a reconstructed slice of SXT (See Fig. 3(f)), shown in Fig. 3(g), based on the alignment of the wide-field image and SXM projection with the alignment error shown in Fig. 3(h).

 

Fig. 3. Alignment of images obtained by SXM and GSDIM. (a) Wide-field fluorescence image of mitochondria in the 4T1 cell. (b) Mosaics of SXM projection, corresponding to the area surrounded by the red square in (a). The matched image of (a) and (b) is shown in (c). (d) is the error map of the matched area in (c). The super-resolution image of the area surrounded by the blue square in (a) is shown in (e). (f) A reconstructed slice of SXT of the sample surrounded by the green square in (b). The matched image of (e) and (f) is shown in (g) and their alignment accuracy is better than 43 nm based on the error map shown in (h).

Download Full Size | PDF

The joint images of mitochondria were shown in Fig. 4. The morphology of mitochondria can be clearly revealed by SXT. The mitochondria appeared as high contrast hollow spheres in the reconstructed slices of SXT (See Fig. 4(a1), Fig. 4(d1)). The hollow spheres of mitochondria originated from the membrane rich in lipids and proteins, which have high soft X-ray absorption. The super-resolution GSDIM images labeling mitochondria clearly exhibited some hollow spheres with a diameter of about 1 µm to 2 µm in Fig. 3(e), Fig. 4(a2) and Fig. 4(d2), similar to the reported mitochondria [50–52]. As the antibody TOMM20 staining primarily occurs in the membrane of mitochondria, the mitochondria exhibited hollow spheres. The reconstructed slices of mitochondria were matched with their super-resolution GSDIM images shown in Fig. 4(a3) and Fig. 4(d3). The alignment accuracies of Fig. 4(a3) estimated by three methods were 71.6 nm, 70.8 nm and 43 nm, and the counterparts of Fig. 4(d3) were respectively 94.0 nm, 93.5 nm and 40 nm. The structures of mitochondria were shown in the reconstructed slices of SXT, while the fluorescence signals labeled the structures and identified the category. In the joint image of mitochondria, the fluorescence signals mainly labeled the membranes of mitochondria, but there were still some fluorescence signals inside the mitochondrion (See Fig. 4(d3)), probably because the mitochondrion was a little out of focus, and only signals of the membrane close to the focal plane were collected. The mitochondria marked by the fluorescence were segmented and rendered using Amira software (See Fig. 4(a4), Fig. 4(d4)), and distinguished from other organelles (See Fig. 4(b), Fig. 4(e)). Apart from morphology, the 3D localization of mitochondria can be realized by correlative imaging method (See Fig. 4(c), Fig. 4(f)). Specifically, the mitochondria in the thicker sample with a depth of 4.5 µm can be observed by both GSDIM and SXT (See Fig. 4(f)).

 

Fig. 4. Identification of mitochondria inside 4T1 cells by correlative imaging method. The reconstructed slices of SXT of mitochondria (a1) and (d1) and the super-resolution images of the mitochondria (a2) and (d2) are merged shown in (a3) and (d3). The line graph shows that the FWHM of the super-resolution image in the yellow rectangle is 50 nm, which means the spatial resolution of the super-resolution image is less than 50 nm. Finally, the mitochondria were recognized and segmented, shown in (a4) and (d4). The mitochondria and other organelles were segmented and rendered in (b) and (e) (mitochondria: green, other organelles: purple, cell membrane: gray and nucleus: blue). The 3D models of the cells are displayed in (c) and (f).

Download Full Size | PDF

The correlative images of lysosomes were shown in Fig. 5 and Fig. 6. Different with mitochondria, the lysosomes were high contrast in whole region (See Fig. 5(f), Fig. 6(a) and Fig. 6(b)), as the lysosomes were rich in enzymes inside. The super-resolution images labeling lysosomes also exhibited hollow spheres (See Fig. 5(e), Fig. 6(c) and Fig. 6(d)), because the antibody LAMP2A mainly labeled the membranes of lysosomes. The joint images of the reconstructed slices of lysosomes and their super-resolution GSDIM images was shown in Fig. 5(g), Fig. 6(e) and Fig. 6(f). The alignment accuracies of the correlative image estimated by the first two methods were 106.8 nm and 105.1 nm, while the error map (Fig. 5(h)) shown that the third accuracies were 50 nm for Fig. 6(e) and 65 nm for Fig. 6(f). The fluorescence signals labeled the membranes of lysosomes in the joint images. The segmented and rendered 3D models of lysosomes based on the joint images were displayed in Fig. 6(g) and Fig. 6(h), where morphologies of lysosomes were clearly revealed. The co-localization of mitochondria and lysosomes in GSDIM and SXT confirmed the successful construction of correlative imaging.

 

Fig. 5. Alignment of images of lysosomes. (a) Wide-field fluorescence image of lysosomes in the 4T1 cell. (b) Mosaic of SXM projection, corresponding to the area surrounded by the red square in (a). The matched image of (a) and (b) is shown in (c). (d) is the error map of the matched area in (c). The super-resolution image of the area surrounded by the blue square in (a) is shown in (e). (f) A reconstructed slice of SXT of the sample surrounded by the green square in (b). The matched image of (e) and (f) is shown in (g) and their alignment accuracy is 43–72 nm based on the error map shown in (h).

Download Full Size | PDF

 

Fig. 6. Identification of lysosomes inside the 4T1 cell by correlative imaging method. The reconstructed slices of SXT of lysosomes (a) and (b), which revealed their structures, and the super-resolution images of the lysosomes (c) and (d) were merged into joint images shown in (e) and (f). Therefore, the lysosomes were recognized in (e) and (f). The 3D models of the lysosomes were displayed in (g) and (h).

Download Full Size | PDF

Compared with the cryogenic confocal microscopy used by Advanced light source (ALS) and the cryogenic SIM used by the Diamond light source (DLS), the spatial resolution of GSDIM has largely improved and has been close to the spatial resolution of STORM used by ALBA synchrotron light source (Listed in Table 1). Compared with the STORM, the penetration depth of GSDIM significantly increases, hence possesses the capability to observe subcellular structures inside cells. The correlative imaging method is especially suitable to image adherent cells with a thickness of a few micrometers, locating and recognizing the subcellular structures with high spatial resolution inside the cells.

Table 1. Performance comparison of several FMs used in the correlative imaging.

View Table

The spatial resolution of home-built GSDIM, especially the resolution in the depth of the cells, can be further improved by suppressing the out-of-focus light. To obtain the fluorescence information in the direction of the depth, a 3D SRFM is to be built. GSDIM provides a possible relatively simple solution for further on-line SRFM, which can avoid the deformation and damage of samples during the transport from GSDIM to SXM.

5. Conclusion

We presented a high resolution and large imaging depth correlative imaging method and constructed the imaging device, based on SXT and GSDIM. The spatial resolution of the home-built GSDIM is evaluated to be about 30 nm by two universal methods. Even in the depth of 2 µm, the spatial resolution of 50 nm can be achieved using the epi-illumination method. For the sample with thickness of 4.5 µm, organelles still can be clearly labeled with the super-resolution image. Organelles could be recognized by correlative imaging through matching super-resolution fluorescence images and reconstructed slices of 3D cell attained by the SXT. As an example, the proposed method identifies and locates mitochondria and lysosomes inside 4T1 cells with high accuracy.

In the presented correlative imaging method, the structural information and the functional information are obtained simultaneously, thanks to the combination of the SXT and GSDIM. Due to the versatility of the fluorescence label, the correlative imaging method can be expected in many research areas, such as pharmacology, toxicology and the development of disease on a subcellular scale. Other imaging methods can also be used for the correlative method, such as electron microscopy and label-free imaging method [53]. Electron microscopy can obtain the cellular structure with higher spatial resolution about several nanometers or better, while label-free imaging method can recognize organelles without fluorescence labeling. Therefore, we believe this advanced technique would be a powerful imaging toolkit to provide useful and comprehensive information in bioscience.