1. Introduction

Organelles, the nanostructure within live cells, are highly active and morphologically dynamic during a wide spectrum of biological processes [1,2]. Thus, probing and tracking these organelles’ dynamics in their natural state is essential to understanding the physiological status and pathological process of organisms. Electron microscopy can detect the sub-nanometer structure of organelles with the resolution ranging from nanometer to the atomic scale [3,4]. However, the complex preparation and processing of samples make electron microscopy not capable of capturing the dynamic process of live samples. Fluorescence microscopy can selectively render the structures of interest by tagging with fluorescent markers, with spatial resolution on the scale of tens of nanometers achieved via super-resolution optical imaging techniques [5–9]. Yet, fluorescence labeling is hard to have no impact on cellular processes. Moreover, the phototoxicity and photobleaching make it difficult to continuously observe the live cells for a long time, and the number of channels that can be observed simultaneously is usually limited to four to five. Thus, there is a strong need for label-free imaging tools to follow specific organelles in live cells in their natural status.

Quantitative phase microscopy (QPM), as a label-free and non-invasive imaging method, can perform not only high-contrast imaging of transparent samples but also achieve much quantitative information such as the 3D shape, inner structures, or refractive index distribution of transparent and translucent samples [10–12]. There are many QPM approaches reported during the past decades, including digital holographic microscopy (DHM) [13–16], gradient light interference microscopy (GLIM) [17,18], beam-propagation-based QPM [19], Fourier ptychographic microscopy (FPM) [20–24], and QPM based on transport of intensity equation (TIE) [25–27]. In general, interferometric approaches have higher phase measurement accuracy, while single-beam QPMs are experimentally simpler and more robust despite it needs recording of many intensity images and a complicated recovery process to reproduce the sample’s information, restricting their imaging speed to a sub-second level. Zernike phase contrast microscopy has a long history in providing phase-contrast images for bio-samples, featuring a common-path configuration and fast imaging speed. When being incorporated with phase-shifting operation between the dc term and the diffracted term of an object wave, this approach provides quantitative phase images of the sample. For instance, spatial light interference microscopy (SLIM) has achieved a lateral resolution of 350 nm and imaging speed of 16 Hz using an objective with the numerical aperture (NA) of 1.4 and a halogen lamp with the central wavelength of 590 nm [28,29]. Recently, the authors enhanced SLIM by introducing ultra-oblique (UO) partially coherent illumination into SLIM (entitled UO-QPM), realizing the spatio-temporal resolution of 270 nm and 250 Hz [30]. In addition, the mitochondria inside live cells have been distinguished with UO-QPM when assisted by a neural network, which was trained by phase-fluorescence image pairs [31]. Despite being strong already, UO-QPM still has many aspects to be improved. Firstly, the fuzzy and lumpy overall outline of the live cells in the traditional phase-contrast microscope needs to be removed to visualize the organelles clearly. Secondly, the ultra-oblique illumination used in UO-QPM requires a water-immersed condenser objective with high NA, making the system complex and unfriendly for installing samples between the condenser and imaging objective. Thirdly, UO-QPM is expected to be equipped with an incubation system maintaining an environment of 5% CO₂ and 37 °C, which is significantly important for the long-time imaging of slow dynamic processes of organelles inside live cells.

In this study, we enhanced the UO-QPM system by solving the above-mentioned problems, resulting in a live-organelle imaging platform featuring a high spatiotemporal resolution of 245 nm and 250 Hz. Firstly, a phase filter is designed for flat-fielding imaging. Secondly, direct illumination of the sample with LEDs at ultra-oblique angles is employed, without the need for a high NA condenser objective. Thirdly, the UO-QPCM system is equipped with intermittent fluorescence imaging and with an environment of 37 °C and 5% CO₂. With the enhanced UO-QPCM system, we have captured with the investigated imaging platform the generation and fusion process of a new organelle called vesicle-like structure (VLS), and the mitochondrial vacuolization during necrosis for the first time.

2. Results

2.1 Construction of FF-QPCM

The schematic diagram of FF-QPCM is shown in Fig. 1(a). 24 LEDs (470/20 nm, central wavelength/spectrum band) simultaneously illuminate the sample at different oblique angles. The spectra of each object wave is comprised of an unscattered term (zero-frequency components, shown with the blue dots in Fig. 1(b)) and scattered terms (high- and low-frequency components, shown as light-blue and cyan discs in Fig. 1(b)). Generally, if only the unscattered terms (the discrete spots) are phase-shifted [32,33], the generated phase-contrast images of subcellular organelles will be modulated by the bulky phase from the cell body, as is shown in Fig. 1(b)-left. In our study, we utilize a broad phase ring to simultaneously modulate the phase of both the unscattered terms and low-frequency components, so that only the subcellular organelles that correspond to high-frequency components become prominent due to selective phase-contrast, as shown in Fig. 1(b)-right and Method part. With this technique, the slow-varying bulky phase from cell bodies is flattened, so that we call this technique flat-fielding quantitative phase-contrast microscopic (FF-QPCM) technique. In the implementation, when both the unscattered term and the low-frequency scattered term are phase-modulated by a thick ring with the phase value of 0, 0.5π, π, and 1.5π, the generated phase-shifted intensity images are captured by the camera. The phase mapping of the live cells will be recovered by using the phase-shifting algorithm (see Eq. (2) in the Methods part and Supplementary note 2 in Supplement 1). Membrane-bound intracellular organelles have a larger refractive index than that of cytoplasm, and therefore, the proposed FF-QPCM can provide high resolution, high-contrast images for subcellular organelles in live cells (Fig. 1(c)), compared with conventional bright-field images. Interestingly, the phase image is highly correlated with the fluorescence image. FF-QPCM excels at live-cell imaging compared with electron microscopy and fluorescence microscopy (Supplementary note 8 in Supplement 1), considering electron microscopy can only be used for fixed dead cells, and fluorescence microscopy suffers from the photobleaching caused during the continuous observation of live cells. FF-QPCM has key advantages over the previous UO-QPM microscope in SLM-based phase filter design for flat-fielding imaging and direct illumination of the sample with LEDs placed at ultra-oblique angles without the need for a high NA condenser or opposing objective. FF-QPCM has been equipped with an independently designed water-gas circulation system to provide live cells the optimal survival conditions of 37 °C and 5% CO₂ (Supplementary note 7 in Supplement 1).

 

Fig. 1. Imaging principle of the proposed FF-QPCM. (a) The schematic diagram of FF-QPCM, and the phase mapping of live cells is obtained by precision phase modulations (0, 0.5π, π, and 1.5π) via SLM to the unscattered waves (the calibration of SLM can be found in Supplementary note 6 in Supplement 1). (b) FF-QPCM retards the phase of the zero- and low-frequency components simultaneously under annular discrete LEDs on the pupil plane, effectively removing the overall uneven contour of live cells (left: QPCM, right: FF-QPCM). (c) Bright-field, FF-QPCM, fluorescence, and phase-fluorescence merged images of a live COS7 cell whose mitochondria are labeled with specific molecular probes (Thermo Fisher, MitoTracker 7512, USA). Obj, objective lens; P, polarizer; TL, tube lens; SLM, spatial light modulator. The scale bars in (b) and (c) represent 5 µm and 4 µm, respectively.

Download Full Size | PDF

2.2 Intermittent fluorescent imaging and FF-QPCM imaging

FF-QPCM is compatible with fluorescence imaging that can selectively visualize structures of interest with high contrast by fluorescent tagging. The two modality images can be obtained simultaneously and correlated by a digital matching approach (Supplementary note 4 in Supplement 1). FF-QPCM can provide structures of overall subcellular organelles, and fluorescence imaging can provide the image of a specific type of organelle using fluorescence labeling. Furthermore, a novel acquisition strategy that fluorescence imaging is intermittently performed during the continuous FF-QPCM imaging was proposed for long-term tracking of subcellular dynamics. The sample was continuously imaged with FF-QPCM over 110 mins at a frame rate of ten frames per second (FPS), while fluorescence images were taken intermittently once per 10 minutes to identify mitochondria by colocalization with the FF-QPCM images. The intermittent fluorescence imaging strategy avoids the phototoxicity and photobleaching by eight folds (when using 5% on/off ratio) during the long-term imaging, compensating for the lack of specificity in FF-QPCM imaging. The images in Fig. 2(a) show the example of the intermittent imaging strategy, where the mitochondria in live COS7 cells were stained with fluorescent dye (Thermo Fisher, MitoTracker 7512, USA). The two-modality images show the dynamics of mitochondria together with all the organelles in the phase images all the time. In subsequent studies, this strategy was used to identify the specific organelle of interest, although the fluorescence images will not be shown repeatedly.

 

Fig. 2. Dual-modality imaging of several organelles in live cells by combining FF-QPCM with fluorescence microscopy. (a) Schematic demonstration of intermittent fluorescence imaging during FF-QPCM imaging. Mitochondria in live COS7 cells were labeled with fluorescent dye (Thermo Fisher, MitoTracker 7512, USA), and were captured with FF-QPCM continuously and fluorescence channel intermittently every 10 minutes. (b) Dual-modality images of lipid droplets by labeling stem cells (ADSC) with the lipid-stain (piHCS LipidTOX). (c) Dual-modality images of cell nucleus by labeling live ADSC cells with nuclear stain. (d) Dual-modality images of mitochondria by labeling live ADSC cells with molecular probes (Thermo Fisher, MitoTracker 7512, USA). (e) Dual-modality images of ER network with ER Tracker staining. The scale bars in (a)-(d) are 5 µm, and 10 µm for (e), respectively.

Download Full Size | PDF

Furthermore, we show the phase images, fluorescence images, and the colocalized image of stained lipid droplets (Thermo Fisher, HCS LipidTOX, USA) in Fig. 2(b). We also show the three panels for nucleus, mitochondria, endoplasmic reticulum (ER) network in adipose tissue-derived stem cells (ADSC) in Figs. 2(c)–2(e), where the ADSC cells were stained with a live cell nuclear stain (Biotium, NucSpot Live 650 Nuclear Stain, USA), MitoTracker (Thermo Fisher, MitoTracker 7512, USA), and ER Tracker (Thermo Fisher, ER Tracker E34250, USA), respectively. Obviously, FF-QPCM can visualize with high contrast all the organelles at one time, while the fluorescence images can help with the identification of a specific organelle.

2.3 Label-free observation of mitochondrial fusion and fission in live cells with high resolution and image contrast

As the classical dynamic properties of mitochondria, fusion and fission play a critical role in controlling mitochondrial morphology and quality [34,35]. Hence, studying the dynamic processes of mitochondria is crucial to understanding the mechanism of many mitochondria-associated diseases. Aiming to this, using the proposed FF-QPCM system, we have captured the mitochondrial dynamics in live cells under the environment of 37 °C and 5% CO₂. Figure 3(a) shows the typical end-to-end fusion among three mitochondria in a PLC/PRF/5 cell (Visualization 1). While, Fig. 3(b) shows a special case: a side-to-side fusion (pointed by a white arrow) of two strip-shaped mitochondria and subsequent fission at several sites (Visualization 2). Figure 3(c) shows the process of a strip-shaped mitochondrion tuning itself into a donut-shaped (toroidal) one (Visualization 2). Visualization 3 shows the transformation of mitochondria between strip-shape and donut-shape. Although it was previously reported that the formation of donut-shaped mitochondria was triggered by the opening of the permeability transition pore or K⁺ channels during hypoxia in glucose-free and during reoxygenation in a glucose-containing medium [36], our results were all performed under normal culture conditions without any intervention. It is meant that there are other mechanisms for the formation of donut-shaped mitochondria, and it needs to be explored in the future.

 

Fig. 3. Investigation of mitochondrial dynamics in different live cells under the environment of 37 °C and 5% CO₂. (a)-(c) The time series of FF-QPCM in a live PLC/PRF/5 cell. (d) Mito-disintegration in a live COS7 cell. Here, s and m represent second and minute, respectively. The scale bars in (a)-(d) are 2 µm, 3 µm, 2 µm, and 3 µm, respectively.

Download Full Size | PDF

Classically, the division of mitochondria is a one-becomes-two process. However, we have captured with FF-QPCM a new type of the complex mitochondrial division process in COS7 cells, as shown in Fig. 3(d). The observed mitochondria in COS7 cells are flat and large, which are quite different from the noodle-like mitochondria in ADSC cells. About 30 minutes after our observation, one irregularly shaped mitochondrion splits into many small and short mitochondria, as we call it mitochondrial disintegration (mito-disintegration). The whole dynamic can be found in Visualization 4. These results point to one of the key advantages of the panoramic and gentle visualizations provided by FF-QPCM, which is that label-free and long-term observations of organelles with high spatiotemporal resolution can allow visualization of rare events that are difficult or improbable to observe using standard fluorescence imaging methods.

2.4 Detection of mitochondrial vacuolization during cell necrosis

It has been reported that cellular damage resulted from detrimental environments or other adverse effects could lead to mitochondrial dysfunction, which further induces ionic imbalance, elevated reactive oxidative stress, oxidative damage, and finally, the formation of mitochondrial vacuolization [37–40]. The phenomenon is featured by swelling of mitochondria, expansion of intermembrane space, and disorganization or disappearance of cristae. They always come along at the onset of disease or aging, playing important roles in the progression of degenerative disorders.

With FF-QPCM, we have captured the continuous dynamic process of mitochondrial vacuolization during necrosis for the first time, proving the ability of FF-QPCM for imaging organelles with consecutive changed refractive index. Figure 4(a)-left shows the image of the whole cells. And the image series in Fig. 4(a)-right show the typical dynamic process about mitochondrial vacuolization, which goes through three stages, swollen, darkening, and driving cell rupture. It can be seen that the contrast of a swollen mitochondrion (pointed by a white arrow) gradually decreases along with the mitochondrial cristae. Then the mitochondria are completely black inside with their interior refractive index becoming lower than that of surrounding cytosol. The detailed process of mitochondrial vacuolization can be found in Visualization 5 and Visualization 6, and the average time of mitochondrial vacuolization is about 30 minutes. We have labeled mitochondria with fluorescent dye (Thermo Fisher, MitoTracker 7512, USA) and detected them with two channels before and after mitochondrial vacuolization to confirm that the vacuolated structures are mitochondria, and the dual-modality imaging is shown in Fig. 4(d) and Supplementary note 9 in Supplement 1.

 

Fig. 4. Label-free FF-QPCM study on mitochondrial vacuolization. (a) Mitochondrial vacuolization process in a COS7 cells. Right images are time series in the white box on the left. (b) The process of mitochondrial vacuole fusion, and the mitochondrial membrane can be clearly observed. Right images are time series in the white box on the left. (c) The role of mitochondrial vacuoles in necrosis. (d) Whole process of mitochondrial vacuolization in live COS7 cells with mitochondria labeled with MitoTracker 7512 and detected by fluorescence channel at the end (red). (e) is the calculated phase of mitochondria in (b) over time. The scale bars in (a)-(d) are 8 µm.

Download Full Size | PDF

It is further found from Figs. 4(b)–4(c) that mitochondrial vacuolization may promote the necrosis in two ways. First, vacuolated mitochondria stick and fuse together to form a larger one, moving to the perinuclear positions. During this process, the phase (e.g., the density in comparison with cytosol) of mitochondrial vacuolization goes from bright to dark (shown in Fig. 4(e)), indicating that the refractive index of the vacuolization decreases with time (the vacuoles might be filled with cytoplasm or water before the cell bursts), as shown in Fig. 4(b) (Visualization 6, Visualization 7, and Visualization 8). The cells will rupture once the membrane of these vacuoles is clearly visible (the interface between the two mitochondria pointed by the black and white arrows when they stick in Fig. 4(b)) and some of the vacuoles also start to rupture (Visualization 8). The second way is shown in Fig. 4(c) (Visualization 9), where the vacuolated mitochondria cause fissure-like structures (pointed by white arrows) on the cell surface, and the cell ruptures under their combined action. In these two ways, the cell necrosis starts with the nucleus rupture, and later the rest of the cell body exploded into a series of bubbles. The mechanical cause of cell explosion still deserves further study, although we speculate that the imbalance of osmotic pressure caused by mitochondrial vacuolization is a key factor.

2.5 Detection of VLSs’ formation following cell membrane dynamics

Observation of intracellular vesicles is of great importance since the vesicles are continuously dynamic and involved in building blocks and transporting shuttles of the organelles. Using FF-QPCM, we have detected new VLSs that have a lower refractive index than the cytosol. The VLSs were seen as black bubbles with their diameter varying from ∼ 200 nm to ∼ 2 µm in the recovered phase images, as shown in Fig. 5(a), and the same observation has also been detected by optical diffraction tomography (ODT) in the previously published literature [16]. To ascertain that the VLSs captured by FF-QPCM are not the phase inversion of the structures with high refractive index due to the defocus, we have carried out the 3D quantitative phase imaging of lipid droplets and VLSs, whose refractive index is higher and lower than surrounding cytosol, as shown in Fig. 5(f)–5(g). The results show that the phase image of the lipid droplet is a bright spot at the focus plane while it becomes black above and below focus (Fig. 5(f)). However, for VLS, it is black at the focus plane while becoming bright above and below focus (Fig. 5(g)). Figure 5(d) shows that the VLSs have an averaged diameter of 0.73 ± 0.30 µm (mean ± standard deviation). With the proposed FF-QPCM system, we have measured the refractive index of the internal material composition of VLSs in mouse ADSC cells as ${1}\mathrm{.347\;\ \pm \;\ 0}{.003}$. Of note, the VLSs were generated by pinocytosis endocytosis at the edge of the cells. Figures 5(b)–5(c) (Visualization 10 and Visualization 11) show the formation process of the VLSs. To be more specific, the lamellipodia of the cell curled up the extracellular culture medium, forming a VLS. Then the VLS was transported to the place around the nucleus. In addition, several small VLSs can fuse together to form a larger one, as shown in Supplementary note 12 in Supplement 1 (Visualization 10). Some of VLSs tend to be around by and interacted with mitochondria, as shown in Supplementary note 12 in Supplement 1 (Visualization 10). Further study shows that many VLSs appeared during the apoptosis of mouse ADSC cells imaged at room temperature of ∼ 20 °C and normal air concentration (Visualization 12), and the number of the VLSs’ varies with time, as is shown in Fig. 5(e). Specifically, the VLSs’ number decreases drastically at both the initial and end stages featuring an averaged reducing speed of 8 and 9 per minutes respectively, while it keeps a dynamic balance during the intermediate stage. The VLSs’ number decreases at both the initial and end stages might be due to the cell shrinkage. Also, we find the VLS (or lipid droplets) closely contacted with the end of the mitochondria during the cell apoptosis process, after which the mitochondria begin mitophagy (Visualization 13). The above experiments indicate that the proposed FF-QPCM has the capability to image transparent organelles with high contrast and high spatiotemporal resolution in both natural and normal cell-survival conditions.

 

Fig. 5. Investigation of the generation of VLSs in live cells. (a) The VLSs are seen as black bubbles in a mouse ADSC cells. (b) VLSs generated by pinocytosis endocytosis at the edge of the cell via curling up the lamellipodia. (c) Schematic diagram of generating a VLS by pinocytosis endocytosis. (d) Quantification of VLSs’ diameter in (a). The black curve is a Guassian fit. (e) The variation of total VLSs’ number over time in Visualization 12. The red curve is a three-order polynomial fit. (f) 3D phase images of a lipid droplet. (g) 3D phase images of a VLS. The scale bars in (a)-(b) are 2 µm while 1 µm in (f)-(g).

Download Full Size | PDF

2.6 Observation of dynamic mitosis process

As is well known, cell division is fundamental to the survival of life, and it is the necessity for reproduction and life and death alternation of organisms. There are complex dynamic processes ongoing during the interphase, including changes in overall cell morphology and intracellular organelles. It is meaningful to capture the intracellular dynamics for better study of the underlying mechanisms of this complex and highly dynamic process. FF-QPCM was applied to capture the dynamic mitosis process of live mouse stem cells (ADSC) during cell division. For the entire observation process, the environment of 37 °C and 5% CO₂ was precisely maintained by virtue of our temperature and air circulation system (see as in Supplementary note 7 in Supplement 1). Figure 6 shows the image series captured during the dynamic mitosis process of live ADSC cells, and Visualization 14 replays the whole dynamic mitosis process. As is shown in Fig. 6, the cell was observed at the beginning of prometaphase featured by disassembled nuclear envelope. Chromosomes that aligned at equator during metaphase and segregated for chromatid during anaphase were imaged clearly. At t = 12 minutes, a cleavage furrow (pointed by white arrows) appeared in the middle of the dividing cell to accelerate the division. At t = 26 minutes, the nuclear envelope enclosed the chromosomes to form two new independent nuclei (surrounded by the black line) at telophase, meaning that mitosis was complete. Interestingly, we observed classical membrane blebs (protrusion of membrane around cells, pointed by black arrows) in anaphase and telophase of mitosis. Membrane blebs play diverse roles in cellular physiology, especially during cell migration, apoptosis, and cell division [41]. Studies have shown that blebs serve as pressure release valves to regulate intracellular pressure during cytokinesis [42]. However, our results showed that the space for daughter cells adhesion increased with the appearance of blebs, indicating its new function during cytokinesis, of which the mechanisms should be thoroughly studied in the future.

 

Fig. 6. Label-free phase imaging of the mitosis of a live stem cell (ADSC) in a time series under the environment of 37 °C and 5% CO₂. The contraction ring is pointed by the green arrows and protruding bubbles the red ones. The scale bar represents 8 µm.

Download Full Size | PDF

3. Methods and materials

3.1 Experimental implementation of FF-QPCM

The detailed schematic diagram of FF-QPCM is shown in Fig. 7(a), for which the quantitative phase imaging module and fluorescence imaging module are integrated into a commercial microscope body (DMi8, Leica7, Germany). The ultra-oblique partially coherent illumination is provided from 24 uniformly distributed LEDs embedded in an annular ring, which was fabricated by a 3D printer. The light emitted from the LEDs illuminates the sample at ultra-oblique angles, which will contribute to the resolution enhancement and the optical sectioning capability of FF-QPCM. The LEDs have a moderate wavelength band of 10 nm (half width) around the central wavelength 470 nm, which provide a high signal to noise ratio (SNR) while keeping the bandwidth narrow enough to ensure the phase modulation accuracy of the spatial light modulator (SLM). The sample is placed in the focal plane of the detection objective lens and is magnified by a telescope system comprised of a detection objective lens (100×/1.44 Oil Immersion, WD-0.17, Leica, Germany) and a tube lens (TL). Then, another telescope system L₂-L₃ relays the intermediate image of the sample to a sCMOS (Zyla 4.2, Andor, UK). And, to perform quantitative phase imaging, at the middle focal plane of the telescope system of L₂-L₃, a SLM (MSP1920-400-800-HSP8, Meadowlark Optics, USA) is then carefully installed so that its liquid crystal plane coincides with the rear pupil plane of the detection objective lens. The spectrum of each object wave can be divided into the unscattered (zero-frequency component) and scattered waves (low-frequency and high-frequency components), which are separated on the SLM plane (the Fourier plane). The intracellular organelles that often have a dimension around several micrometers, of which the spectrum corresponds to high-frequency component. When both the dc term and the low-frequency term of the object waves are retarded on the SLM plane by a thick ring with a diameter of 5.66 mm, a thickness of 500 µm, and a phase series of 0, 0.5π, π, and 1.5π, (Supplementary note 1 in Supplement 1), four phase-shifted phase-contrast images of the sample can be recorded by the sCMOS1, yielding a quantitative phase image of the sample eventually, as shown in Fig. 7(b). Considering the phase modulation of SLMs has, in general, the polarization-dependent selectivity on the incident light field, a polarizer P is placed before the telescope system L₂-L₃ to maximize the modulation efficiency of the SLM. Using the rolling shutter mode of the sCMOS, the temporal resolution of the system is 250 frames per second (FPS) for a field of view (FOV) of 800 × 2048 pixels (corresponding to 43 × 110 µm²) after using the interleaved algorithm [43]. In addition, in current FF-QPCM system, we set the illumination angle of 45° (NA_illu= 0.71) to keep the system compact and to maximize the useable space between the sample and illuminator. In the future, if higher resolution is required, oblique illumination with a larger angle could be used, although this may present experimental challenges as the space between the sample and illuminator will be reduced. The system is highly immune to external disturbances due to the common-path configuration, where the scattered and unscattered waves go through exactly the same optical elements during imaging. For long-term imaging of slow dynamics that last for hours, FF-QPCM is integrated with an autofocusing program to keep the sample focused. Specifically, a set of phase-contrast images are taken at different axial layers. The intensity images with unscattered waves retarded by 0.5π are convolved with a Sobel operator, yielding the gradient images at different layers in real-time, and the focused plane is determined by finding the maximum of the averaged gradient.

 

Fig. 7. Experimental implementation of FF-QPCM. (a) Schematic diagram of FF-QPCM setup. (b) Phase-shifting images of a live COS7 cell and its final reconstructed phase map, where the mitochondria around the cell nucleus are resolved clearly. AI, annular illuminator; DM, dichroic mirror; FS, filter set; L, lens; M, mirror; MO, micro-objective; S, sample; SLM, spatial light modulator; sCMOS, scientific complementary metal oxide semiconductor; TL, tube lens; WLS, white light source; P, polarizer. The scale bar in (b) represents 4 μm.

Download Full Size | PDF

For wide-field fluorescence imaging, it has been combined with the FF-QPCM by the splitting effect of a dichromatic mirror (DM1). To be specific, a white light source comprised of four independently-controllable LED lamps (central wavelength 405/488/561/640 nm) uniformly illuminates the sample under the action of a switchable filter cube through the telescope system L₁-MO, and the resulted fluorescence image is relayed by MO-L1 and recorded by sCMOS2 (Andor Zyla 4.2, UK) by the telescope system MO-TL2. FF-QPCM and fluorescence imaging channel are synchronized by a self-built control system so that they can achieve true dual-modality imaging. FF-QPCM can also be readily combined with other fluorescence imaging techniques.

3.2 Quantitative phase reconstruction from phase-shifted interferograms in FF-QPCM

Once the illumination wave passes through the sample, the generated light field can be approximated as ${f(x,\; y) = A(x,\; y)} \cdot \textrm{exp}(j\varphi (x,\; y))$, where ${(x,\; y)}$ are the coordinates on the sample plane. As mentioned above, when the zero-frequency and the low-frequency components are retarded in phase for 0, 0.5π, π, and 1.5π by the SLM, the interference of the high-frequency component and the low-frequency component produces a series of phase-shifting images, which convert the phase modulation of the object wave into intensity modulation. The sCMOS (Andor Zyla 4.2, UK) is synchronized with the SLM to record the phase-shifting interferograms:

Here, m = 1,2,3,4 indicates the phase-shifting index. ${{\beta }_{0}}{(x,\; y)}$, ${{\beta }_{{c}}}{(x,\; y)}$, and ${{\beta }_{{s}}}{(x,\; y)}$ are the intermediate functions containing the distribution of the amplitude and phase of the sample field and can be reconstructed by the standard linear least-squares techniques, the detailed information can be found in Supplementary note 2 in Supplement 1. And the phase distribution of the total light field associated with the sample that is being illuminated by the annular illuminator can be calculated as:

Here tan^-1{·} denotes the arctangent function, i.e., inverse tangent function, C${(x,\; y)}$ and α${(x,\; y)}$ are the amplitude and phase distribution of the low-frequency sample field that are accurately calculated in Supplementary note 2 in Supplement 1. Notably, φ${(x,\; y)}$ is defined as $\mathrm{2\pi } \cdot {[n(x,\; y) - }{{n}_{{0}}}{]} \cdot \mathit{h(x,\;\ y)/\lambda }$, being the phase difference between the high-frequency components (so to speak, sub-cellular organelles) and the background plateau (cytosol). ${{n}_{{0}}}$ is the refractive index of the cytosol. Therefore, the refractive index distribution $\; n{(x,\; y)}$ of the sample or its height distribution h${(x,\; y)}$ can be resolved when one of them is known. Further, the dry mass density of cells at each pixel is calculated as $\rho \mathrm{(x,\;\ y)\ =\ \lambda } \cdot \varphi \mathrm{(x,\;\ y)/(2\pi \gamma )}$, where $\mathrm{\gamma }$ is the average refractive increment of protein (∼ 0.2 mL/g). Combined with the phase sensitivity in Supplementary note 3 in Supplement 1, our system can detect the subtle change of dry mass density, i.e., 3.404 fg/µm² spatially and 4.077 fg/µm² temporally. It should be noted that the phase map can also be recovered by using three phase-shifting intensity images. However, the phase recovery from four phase-shifting intensity images is more tolerant to phase-shift miscalibration and less sensitive to second-order detector nonlinearity [44].

3.3 Cell culture and staining

Adipose-derived stem cells (ADSCs) were isolated from the inguinal fat pad of a mouse. In brief, the inguinal fat pads were minced and digested in 0.1% collagenase type I (Gibco 17100017, USA) for 30 minutes. Then cell suspensions were centrifuged and washed by PBS. Finally, the cell pellets were resuspended in α-MEM (HyClone SH30265.01, USA) containing 10% fetal bovine serum (FBS, HyClone SV30208.02, USA) and 1% penicillin-streptomycin (HyClone SV30010, USA), followed by being cultured in an incubator at 37℃ with 5% CO₂.

COS7 cells (Procell Life Science & Technology Co., Ltd, China) were incubated in high glucose DMEM supplemented with 10% FBS (HyClone) and 1% penicillin-streptomycin (HyClone). PLC/PRF/5 cells (Procell CL-0145) were cultured in medium including MEM (GibcoTM 41090036), 10% FBS (HyClone) and 1% penicillin-streptomycin (HyClone). MCF7 cells (Procell CL-0149) were incubated in MEM (GibcoTM 41090036) supplemented with 10% FBS (HyClone), 0.01 mg/ml insulin (Sigma-Aldrich 6279, USA) and 1% penicillin-streptomycin (HyClone). COS7, PLC/PRF/5 cells, and MCF7 cells were cultured in an incubator with the same conditions of ADSCs. All the above cells were cultured for 80% confluency and then passaged in fluorodish (World Precision Instruments, Inc. FD5040-100, USA) for further observation.

For labeling of mitochondria, COS7 cells were incubated with 200 nM MitoTracker probes (Thermo Fisher Scientific M7512, USA) in the growth medium at 37 ℃ for 30minutes. Then the staining solution was replaced with prewarmed media, and cells were observed immediately. To trace lipid droplets, ADSCs were stained in HCS LipidTO Neutral Lipid Stains (Thermo Fisher Scientific H34475, USA) with 1:1000 dilution in growth medium at 37 ℃ for 30minutes. For staining of the nucleus, COS7 cells were incubated in DMEM medium containing 1×nuclear Stain (Biotium, NucSpot Live 650 Nuclear Stain, USA) at room temperature for 15minutes. To tag endoplasmic reticulum (ER), the ER Tracker (Thermo Fisher, ER Tracker E34250, USA) at a concentration of 200nM in Hank's Balanced Salt Solution with calcium and magnesium was used for staining of COS7 cells at 37℃/5% CO2 for 30 minutes. After being washed with PBS, the cells were observed instantly.

3.4 Preparation of polystyrene spheres on a coverslip

Polystyrene spheres immobilized at the bottom of the fluorodish were used for the adjustment of the system and the verification of the system performance. The original solution containing 1% concentration polystyrene spheres was further diluted with deionized water in a ratio of 1:1000, which was shaken by a vortex mixer at 1000 revolutions per minute for 5 minutes. Add 200 µL of the solution to one of the poly-D-lysine coated imaging chambers and incubate for 5 min at room temperature to immobilize the beads on the surface. Wash the coverslip twice with 250 µL MilliQ water and finally add 250 µL MilliQ water for imaging. Place the coverslip onto the microscope slide holder.

4. Conclusion

The highly dynamic spatiotemporal interactions of intracellular organelles always determine fates of cells. Therefore, long-term intuitionistic observation of organelle dynamics with high spatiotemporal resolution is critical for further understanding of mechanisms behind cell activities. However, such intuitionistic imaging of intracellular organelle dynamics still faces many limitations. In this study, we proposed flat-fielding quantitative phase contrast microscopy (FF-QPCM) to realize the label-free, long-term observation of subcellular organelles with unprecedented spatiotemporal resolution of 245 nm and 250 Hz, which can selectively highlight intracellular organelles without being influenced by the bulky phase from the cell body. What’s more, FF-QPCM was equipped with an intermittent fluorescence channel was to assist in organelle specificity identification. Based on the advantages of FF-QPCM, we got new insights into mitochondrial fusion and fission, chromosomes separation, and membrane blebbing during mitosis. Meanwhile, the transport of various substances along ER network following the hitchhiking and transportation route regulation (Visualization 15), morphological changes in organelles themselves, and complex interactions among organelles in PLC/PRF/5 cells (Visualization 16) and mouse ADSC cells (Visualization 17) are clearly recorded, displaying the complicated communications among organelles which are easily missed with fluorescence microscopes. Moreover, the layer-by-layer display of phase and fluorescence channels of mitochondria can be found in (Visualization 18). And the three-dimensional structures of a COS7 cell of ~ 18.6 μm thickness are visualized in (Visualization 19), where the finer structures like cytoskeleton and microfilament at different axial positions can be observed with high resolution and high contrast. Of note, we have detected new type of mitochondria fission named as mitochondrial disintegration, the formation of subcellular structure called vesicle-like structure (VLS), whose refractive index is lower than cytosol. Further, we have also captured the generation process of mitochondrial vacuolization, highlighting the role of mitochondria in cell necrosis. The above experiments further demonstrate that FF-QPCM provides new platform to investigate the complex dynamics and interactions among intracellular organelles simultaneously. And last, label-free and long-term imaging of subcellular structures with high spatiotemporal resolution will open a new window for the exploration of the unknown world in cell biology.