View all hidden authors and organizations

Edited by Anders Sejr Hansen, Massachusetts Institute of Technology, Cambridge, Massachusetts, accepted by the editorial board on October 11, 2021 (received August 27, 2021)

We recommend combining a chip-based optical nanomirror with a highly spatially sensitive quantitative phase microscope to obtain the three-dimensional (3D) morphology of hepatic sinusoidal endothelial cells (LSEC). LSEC contains a large number of transcellular nanopores in the plasma membrane-"open pores", usually clustered in groups of 10 to 50 in an area called the sieve plate. Determining the diameter and height of the porous area is an important indicator of cell function, and these dimensions will be affected by drugs and other drugs. Our multi-modal microscope provides a solution for the 3D nanoscale characterization of the aperture diameter and the measurement of the optical thickness of the sieve plate.

The visualization of three-dimensional (3D) morphological changes in the subcellular structure of biological specimens is a major challenge in life sciences. Here, we propose an integrated chip-based optical nanomirror combined with quantitative phase microscopy (QPM) to obtain the 3D morphology of hepatic sinusoidal endothelial cells (LSEC). LSEC has a unique morphology, with small nanopores (50-300 nm in diameter) in the plasma membrane, which are called fenestrations. The windows are grouped into discrete clusters with a thickness of about 100 to 200 nanometers. Therefore, the imaging and quantification of the thickness of the aperture and the sieve plate need to have a resolution and sensitivity of less than 100 nm in the lateral and axial directions, respectively. In chip-based nanomicroscopes, optical waveguides are used to carry and illuminate the sample. The fluorescence signal is captured by an upright microscope, which is converted into a Linnik-type interferometer, which sequentially obtains super-resolution images and phase information of the sample. Assuming that the constant refractive index of the cell membrane is 1.38, the multi-modal microscope estimates that the opening diameter is 119 ± 53 nm, and the average thickness of the sieve plate is 136.6 ± 42.4 nm. In addition, LSEC was treated with cytochalasin B to demonstrate the possibility of accurately detecting cell height. It was found that the average phase values ​​of the windowed regions in normal and treated cells were 161 ± 50 mrad and 109 ± 49 mrad, respectively. The proposed multi-modal technology provides nano-scale visualization of lateral size and thickness maps, which will arouse wider interest in the fields of cell biology and bioimaging.

Far-field optical nanotechnology is often used to visualize subcellular structures in biological specimens by going beyond the diffraction limit. Optical nanomicroscopy includes a variety of technologies, including stimulated emission loss microscopy (1), structured illumination microscopy (SIM) (2), different variants of single molecule positioning microscopy (SMLM), such as light activated positioning microscopy (3) and direct Random optical reconstruction microscopy (dSTORM) (4) and techniques based on intensity fluctuations, such as super-resolution optical wave imaging (5). These technologies can help detect the subcellular structures (<200 nm) of biological samples, such as lipids, proteins, membrane structures, microtubules, and nucleic acids through specific fluorescent labels (6). Each technique has its own advantages and disadvantages; for example, SIM is very popular in live cell imaging due to its fast image acquisition time but limited spatial resolution (7). On the other hand, dSTORM is slower but provides high resolution for characterizing viral proteins (8) and imaging actin filaments in mammalian cells (9, 10), for example. In order to reduce the complexity of the typical SMLM setup using total internal reflection fluorescence (TIRF) configuration, a photonic chip-based optical nanosystem (11⇓ –13) has recently been proposed. In chip-based systems, photonic integrated circuits are used to replace the usual free-space optics for excitation. However, the collection is done through free space optics. The main advantage of this configuration is the separation of excitation and collection paths and the miniaturization of the excitation light path of the system. In chip-based nanomicroscopy, TIRF illumination is generated by the evanescent field of the waveguide, rather than using traditional high magnification and high numerical aperture (NA) TIRF lenses. The evanescent field in the waveguide can be generated over a very large area because it is only defined by the waveguide geometry. Compared with the traditional TIRF-based dSTORM (12), the waveguide geometry can use any imaging objective to image arbitrarily large areas, the latter being limited by the TIRF lens's field of view (FOV).

Quantitative phase microscopy (QPM) is a label-free optical microscopy technique that facilitates sensitive measurement of the refractive index and thickness of two biological samples (14). So far, various QPM methods have been proposed to extract the optical phase and dynamics of biological cells (15⇓-17). These technologies provide high phase sensitivity (space and time), lateral resolution, and high imaging speed (15). The spatial and temporal phase sensitivity of QPM systems are highly dependent on the type of illumination source and interference geometry (17⇓ –19). For example, the universal path QPM technology provides better time-phase sensitivity and can be used to measure cell membrane fluctuations (20). In addition, the spatial phase sensitivity of the system can be improved by using low-coherence light sources (halogen lamps and light-emitting diodes [LED]), but phase shifting techniques are required to utilize the entire FOV of the camera (21). The latest developments in QPM technology with excellent resolution using structured lighting (22, 23) and three-dimensional (3D) information of the sample have been revealed by measuring the phase across multiple illumination angles. This technique facilitates tomography of various biological specimens, such as red blood cells, HT29 cells, and bovine embryos (17, 24). Since the lateral resolution of QPM technology depends on the NA of the objective lens, imaging beyond the diffraction limit (<200 nm) is still challenging and limits the study of subcellular structures. Therefore, it is useful to develop a multi-modal route in which different microscopy methods can be used to provide supplementary information about biological specimens, such as hepatic sinusoidal endothelial cells (LSEC).

Figure 1 depicts LSEC with a large number of windows. The diameter of these transcellular nanopores varies from 50 to 300 nm, which is just below the diffraction limit of optical microscopy (25⇓ –27). The windows are usually grouped in groups of 5 to 100 in an area called the sieve plate (28). The porous form of LSEC acts as an ultrafilter between the blood and the underlying liver cells, facilitating the two-way exchange between the liver and the blood. For example, smaller viruses and drugs can pass through this barrier, while blood cells remain in the sinusoidal vessel lumen (25, 29). The typical thickness of the sieve is about 100 to 150 nm (30), so the window is nanometer-sized in all three dimensions. As shown in Figure 1, the opening in the sieve plate forms an opening throughout the entire LSEC cell body, so TIRF illumination is very suitable for imaging these structures. It is important to determine the diameter and number of fenestrations and the height of the sieve area, as it can be affected by a variety of drugs and conditions (31, 32). The loss of the porous morphology of LSEC, a process called window opening, impairs the filtering properties of the liver, which can lead to atherosclerosis (33). In addition, aging can cause "pseudocapillation", that is, LSEC loses its window at the same time and becomes thicker (34) (Figure 1). This is believed to be the main factor leading to the age-related need to increase the dose of drugs targeting hepatocytes (such as statins) (35). The use of actin disruptors such as cytochalasin B can increase the number of fenestrations in vitro (27). This treatment reduces the height of the LSEC outside the nuclear zone, which helps to form a new window (36).

Top view (A) and cross-sectional view (B) of LSEC. LSEC has a unique morphology, in which nano-scale windows are gathered in a thin sieve plate. Both the aperture diameter and the thickness of the sieve plate are lower than the diffraction limit of traditional optical microscopes. The number and size of windows, as well as the thickness of LSEC, will be affected by aging and liver disease. In vitro, actin disrupting agents (such as cytochalasin B) can be used to increase the number of window openings (27).

Here, we developed an optical nanomirror based on a multi-modal chip and a high-sensitivity QPM system to visualize 3D morphological changes in LSEC. The proposed system separates the light irradiation path from the collection path, thus realizing the direct integration of dSTORM and QPM. The nano-level phase sensitivity of QPM technology is used to extract the optical thickness of the sieve. In addition, the chip-based dSTORM supports super-resolution imaging as low as 50 nm on ultra-large FOVs up to millimeter level (12). Therefore, the integration of dSTORM and QPM allows super-resolution imaging in the lateral dimension (using dSTORM) and nano-sensitivity in the axial direction (using QPM). In this work, we demonstrate the functionality of the system by imaging LSEC with a diffraction-limited TIRF microscope and dSTORM. Using dSTORM, the window and sieve can be observed, and the diffraction-limited QPM can be used to obtain the average optical thickness of the sieve area. In addition, we used cytochalasin B (10 μg/mL) to treat the cells to study the changes in the internal morphology of the sieve plate. The lack of QPM lateral resolution is compensated by dSTORM, which allows us to locate the sieve area containing sub-diffraction size windows. Therefore, in the cell membrane area away from the nucleus, our multi-modal method can be used to reliably reconstruct the 3D shape of LSEC. The integrated system provides a combination of synchronization function and quantitative imaging of cells with large FOV, providing a compact imaging platform with the potential for high-throughput morphology and nano-imaging for specific biological applications.

The schematic diagram of the system in QPM and dSTORM mode is shown in Figure 2. A 660 nm laser (Cobolt Flamenco, λ = 660 nm) was coupled into the waveguide to generate an evanescent field on the surface of the waveguide for the dSTORM experiment. A highly coherent 561 nm laser (Cobolt Jive, λ = 561 nm) is extended by a microscope objective and passed through a rotating diffuser, and then phase imaging is performed through a multimode fiber (MMFB, M35L02-Ø1000 µm; Thorlabs). Rotating diffusers and MMFB are used to generate spatial and temporal diversity to convert highly coherent lasers into partially spatially coherent light sources. It has been previously shown that the reduction of spatial coherence will lead to speckle-free images and increase the spatial phase sensitivity of the interferometric system (37⇓ –39). Therefore, partial spatially coherent sources (PTLS) can be used to extract the morphological changes of the thinnest biological specimens (such as LSEC). A partially coherent beam with a power of about 10 mW is coupled into a Linnik-type QPM system. In the QPM system, the beams reflected from the sample and reference mirror interfere on the plane of the beam splitter. We use a 60x, 1.2 NA water immersion objective (Olympus) for all QPM measurements, which means that the best achievable lateral resolution is 270 nm. The two-dimensional (2D) interference pattern encodes the information of the sample, which is further captured by a complementary metal oxide semiconductor (CMOS) image sensor (Hamamatsu ORCA-Flash4.0 LT, C11440-42U).

Schematic diagram of integrated partial spatially incoherent QPM and chip-based nanosystem for LSEC morphological imaging. MO1-4: Microscope objective lens; RD: rotating diffuser; L1-5: lens; BS: beam splitter. A 660 nm Cobolt laser is used to generate a high-intensity evanescent field on the top of the waveguide chip for single-molecule fluorescence excitation. The fluorescence signal is captured by an upright microscope and then converted to a Linnik-type interferometer to perform QPM.

The two-dimensional intensity distribution of the interferogram can be expressed as I(x,y)=a(x,y) b(x,y)cos[2i(fxx fyy ϕ(x,y)], [1] where a(x ,y) and b(x,y) denote the background term and modulation term, respectively, fxx and fyy are the spatial frequencies of the interferogram along the x and y directions, and ϕ (x,y) is the difference between the phase object and the reference beam.

Standard Fourier transform analysis (40) and Goldstein phase unwrapping algorithm (41) are used to extract the phase information of the sample. The phase information is the combination of the refractive index and thickness of the sample, which can be written as φ(x,y)=2πλ×2h(x,y)*{ns(x,y)−n0(x,y)}, [2] Where λ is the wavelength of the incident light, h is the geometric thickness of the sample, ns and n0 are the refractive indices of the sample and the surrounding medium, respectively, because the imaging is in the reflection mode. By reformatting the equation, the expression of sample thickness can be derived: h(x,y)= λ*ϕ(x,y)4π*{ns(x,y)−n0(x,y)}. [3]

dSTORM imaging is performed by TIRF excitation based on a waveguide chip. Once the sample is stained and the scintillation buffer (9, 10) is added, place the chip on the sample stage and fix it with a vacuum suction cup. A 50×, 0.5-NA objective lens (Olympus) is used to couple the excitation light from free space through end-fire coupling. The waveguide is multi-mode, producing uneven excitation modes. In order to achieve uniform illumination, the coupling objective is scanned along the input surface in an average mode. The imaging was done using a Hamamatsu Orca scientific complementary metal oxide semiconductor camera with an exposure time of 30 milliseconds. For TIRF images, the exposure time was increased to 100 milliseconds, and about 1,000 frames were used on average. All images used approximately 20 mW of input power; however, at the end of each imaging process, the power was gradually increased to approximately 60 mW to obtain additional positioning. Use the Fiji plugin ThunderSTORM (42) to reconstruct the data. For more detailed information about this type of setting, please refer to the literature (11⇓ –13).

The working process of the system to extract the optical thickness dimension of the opening area in LSEC is shown in Figure 3. All imaging in this work was done using Si3N4 strip waveguides with widths ranging from 200 to 500 μm. The chip was manufactured using the procedure previously described (43). Before any sample preparation, use a two-step method to thoroughly clean the chip. First, immerse the chip in 1% Hellmanex deionized (DI) water solution at 70 °C for 10 minutes, then rinse with deionized water, then immerse in isopropanol, and rinse with deionized water again. Finally, use N2 to dry the chip. Create a hollow rectangular chamber with polydimethylsiloxane (PDMS) and place it on the chip to limit the cell attachment area.

Integrate the workflow of QPM and nano-on-chip systems. (A) The Si3N4 strip waveguide was thoroughly cleaned to perform all imaging experiments. (B) Cells are isolated on top of the chip in a restricted rectangular area created with PDMS. C and D show the data acquisition and registration between quantitative phase imaging (QPI) and super-resolution imaging to calculate the size of the window area and the average thickness of the window area group in LSEC.

A modified standard protocol was used to isolate cells from C57BL/6 male mice and cryopreserved Sprague Dawley male rats (44). In short, Liberase (Roche) was used to perfuse the liver, followed by low-speed differential centrifugation, and then cell separation was performed using superferromagnetic beads coupled with the LSEC specific antibody CD146 (MACS, Miltenyi Biotec). After separation, the cells were seeded on a chip pre-coated with human fibronectin and cultured in RPMI-1640 medium in 5% CO2 at 37°C for 2 hours. In the PDMS chamber, the seeding density is approximately 100,000 LSEC per 0.5 cm2. The selected samples were treated with 10 μg/mL cytochalasin B (Sigma-Aldrich) for 30 minutes. The cells were fixed by incubation in 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) for 10 minutes and left in 1% PFA at 4°C until imaging.

Cells were stained with CellMask Deep Red (CMDR) and Vybrant DiD, and the chip was thoroughly rinsed with PBS before staining. Add a 1:1,000 dilution of CMDR in PBS to the inside of the PDMS chamber and let it incubate for 10 minutes. For Vybrant DiD, add a 1:200 dilution in PBS to the inside of the PDMS chamber for 20 minutes. Then rinse the sample thoroughly with PBS again. Before imaging, use 22.5 μL PBS, 22.5 μL H2O-based oxygen scavenger system solution (45), and 5 μL 1 M β-mercaptoethylamine (MEA) to prepare dSTORM buffer. Then rinse the sample thoroughly with PBS before applying the scintillation buffer, and seal the sample area with a cover glass. Finally, the chip is placed under a microscope to obtain interferometry and dSTORM imaging. The extracted phase map and the super-resolution image are further registered to locate the window area and calculate the size and optical thickness of the sieve plate. The windowing is quantified using an intensity-based threshold method similar to the semi-automatic method described in the reference. 46.

The proposed platform integrates an on-chip nanomicroscope and a high-sensitivity QPM system. The on-chip nanomicroscope provides high-throughput imaging by separating the excitation and emission paths, while the PTLS in QPM provides nano-spatial phase sensitivity to identify nano-morphological changes in the sample. We first characterize the system by calculating the spatial phase noise (that is, the spatial phase sensitivity of the system in QPM mode). In order to measure the phase noise in the system, a standard flat mirror with a surface flatness of λ/10 is used as the object to capture the interference image. Figure 4A shows the interferogram recorded on the mirror when the system is running in QPM mode. Ideally, the calculated phase map without any samples on the plane should be zero. However, spatial noise always exists in any QPM system, which is difficult to avoid due to experimental flaws, such as unwanted vibration or temperature fluctuations. Figure 4B depicts the SD (ie, the spatial noise of the system) with phase changes. The average spatial noise of the system is ±20 mrad, which is significantly lower than using direct laser to perform QPM (18, 47). Figure 4C depicts the temporal noise of the phase microscope system. In order to measure the time phase stability shown in Figure 4C, a 60-second time-lapse movie was obtained by placing a standard flat mirror. The average time phase stability is ± 38 mrad.

Integrate the noise and resolution characteristics of QPM and dSTORM systems. (A) Interferogram captured by the proposed QPM mode on a standard mirror with λ/10 surface flatness. (B) The SD of the spatial phase in A indicates the spatial phase noise of the system (±20 mrad). The color bar represents the phase diagram in radians. (C) Time noise of the phase microscope system (±38 mrad). (D) Using FRC to obtain the 61 nm lateral optical resolution of the chip-based system on the sample. (E and F) Interference image of Si3N4 optical waveguide when using direct laser and PTLS respectively. (E1 and F1) Corresponding to the complete FOV phase map reconstructed within the arc of the optical waveguide (H ~ 8 nm) of the laser and PTLS respectively. (Scale bar, 40 microns.)

The high spatial and temporal coherence of the direct laser will cause speckles and false fringes in the final image, thereby reducing the phase sensitivity of the QPM system. By introducing space and time diversity in the laser beam, through the rotating diffuser and subsequent MMFB (18), this unwanted noise can be avoided. The rotating diffuser and MMFB reduce the spatial coherence of the light source, thereby improving the spatial phase sensitivity of the system.

A Fourier ring correlation (FRC) test was performed on the dSTORM data to estimate the resolution of the system in the nanomirror mode, and the results are plotted in Figure 4D. The resolution is given by the normalized cross-correlation between dSTORM images of the same area in the frequency domain. To this end, the frequency spectrum of the two images is divided into bins to generate a series of concentric rings. The correlation value of each bin is used to form the FRC histogram. Mathematically, FRC(ri)=∑r∈riFT1(r). FT1(r)∑r∈riFT12(r). ∑r∈riFT22(r), [4] where FT1(r) and FT2(r) represent the Fourier transform of two images in the same area. The image resolution is defined by the cutoff frequency at which the cross-correlation falls below a preset threshold. In our example, we first separate the odd and even frames of the acquired data set, reconstruct them separately to generate two dSTORM images of the same area, and provide a value of 61 nm based on the FRC analysis. However, the resolution of our system can be improved by replacing the beam splitter in the system (labeled BS in Figure 2) with a flip mirror, because half of the photons passing through it will be lost. We chose to use a beam splitter because it can perform fluorescence and phase imaging at the same time, while a flip mirror restricts the setting to sequential imaging.

In addition, in order to explain the advantages of some spatially coherent sources, we show a comparison between traditional coherent QPM and partially spatially coherent QPM systems to restore the phase map of the optical waveguide. The experiment was performed on a rib-shaped optical waveguide with a silicon nitride (Si3N4) core material, a refractive index of n~2.04, and a rib height of ~8nm. Figure 4 E and F depict interference images of a waveguide sample with a step height of 8 nm using coherent and partially coherent light sources. Here, the difference in fringe quality between laser and PTLS is obvious. The reconstructed phase diagram of the stepped object (waveguide) is shown in Figure 4 E, 1 and F, 1. Due to the coherent noise, the coherent source does not reconstruct the object structure. In contrast, the phase recovery of an object with a step height of 8 nm can be clearly seen through incoherent illumination (Figure 4F, 1). This result further highlights the advantages of using PTLS to recover the phase image of a thin specimen.

Figure 5 shows the complete data set collected for one imaging area of ​​LSEC. It includes 1) brightfield image, 2) phase image of LSEC, 3) diffraction-limited TIRF image, and 4) dSTORM image with visible window. The brightfield image (Figure 5A) provides a clear, qualitative, diffraction-limited imaging of cells. On the other hand, FIG. 5B shows a quantitative phase diagram (that is, the optical thickness of LSEC). In the phase image, the dark yellow higher phase area represents the cell nucleus surrounded by the plasma membrane. The maximum phase value in the nucleus in the lower left corner is 2.3 rad. Figure 5C shows the diffraction-limited TIRF image of LSEC, which provides excellent optical sectioning, showing the morphological and functional characteristics of the cells. However, the visualization of fine features, such as the fenestration present in the plasma membrane (Figure 5D), can only be achieved when using super-resolution imaging. Figure 5 EG shows the illustration of Figure 5D in TIRF, dSTORM and QPM modes. Comparing Figure 5 E and F, the windowing in the film resolved in the dSTORM image is not visible using diffraction limited TIRF imaging.

Parts of three cells imaged with brightfield (A), QPM (B), TIRF (C) and dSTORM (D). (Scale bar, 10 μm.) The phase map provides information about the morphology of the cell. The maximum phase value in the cell nucleus at the bottom left is 2.3 rad. The dSTORM image clearly shows the plasma membrane window in the upper cell. The TIRF, dSTORM and QPM images of the illustration in D are also displayed in EG respectively. The color bar shows the phase in radians.

In addition, in order to include verification and more statistical analysis, multiple experiments were performed on a total of five different batches of mouse LSEC. To demonstrate the possibility of accurately detecting cell height changes, the samples were treated with cytochalasin B (10 μg/mL). This actin cytoskeleton disruptor has been extensively studied on LSEC and has a clear effect. Cytochalasin B has been shown to increase the number of window openings in LSEC and reduce the height of cells in the distal nuclear region (36).

Use the integrated microscope platform to image the control and processed LSEC, as shown in Figure 6. Figure 6 shows the dSTORM and phase images of the compared and processed LSEC. The difference in porosity between control and treated cells can be clearly seen in the dSTORM image. The cell height in the nucleus area of ​​the two groups remained similar, while the height of the cell periphery (where the sieve plate was located) decreased. It was also found that the average phase value of the frit in the treated cells was lower than the average phase value of the frit in the control LSEC. The average phase values ​​of the windowed area in the control and treated cells were 161 ± 50 mrad and 109 ± 49 mrad, respectively. Therefore, QPM can provide sufficiently accurate and useful results in the axial direction even when the dimensions in the lateral direction are limited by diffraction. The dSTORM images in Figures 6A and B show the membrane area of ​​the control and treated LSEC.

(A and B) Multimodal imaging of control and therapy (cytochalasin B) LSEC using the integrated on-chip nanomirror and high spatially sensitive QPM system. Cytochalasin B increases the number of windows in LSEC and reduces the height outside the nuclear region, making the cells flatter and opening more windows compared to control cells. (Scale bar, 5 μm.) This phase shows that the maximum phase in the nucleus is 2.5 rad. The color code shows the phase in radians.

Figure 7 shows the measured phase values ​​of several different windowed areas of normal and processed LSEC. Although the window is below the diffraction limit and therefore the spatial resolution limit of QPM (Figure 7 C and D), the average phase of the sieve can be calculated. The average optical thickness of the sieve plate can be calculated from the phase diagram. The box plot of the average phase value of the sieve plate is shown in Figure 7E. A total of 85 and 72 perforated areas from 23 control and 21 treated cells were used to show the average phase value, respectively. Taking into account the constant refractive index (n = 1.38) of the entire cell membrane (48, 49), the average thickness of the windowed area in normal and treated cells is 136.6 ± 42.4 nm and 92.36 ± 41.6 nm, respectively. The average thickness calculated using the QPM phase diagram is an approximation because we assume that the film part of the entire unit (48, 49) has a constant refractive index (n = 1.38). The multimodal microscope provides an estimated window diameter of 119 ± 53 nm (using Fiji's Sauvola local threshold algorithm) and the average thickness of the sieve that controls the LSEC is 136.6 ± 42.4.

(A and C) dSTORM and control the phase image of LSEC. (Scale bar, 2 μm.) (B and D) Phase image of dSTORM and processed LSEC. The opening in the plasma membrane is visible throughout the cell. (Scale bar, 5 μm.) This phase shows that the maximum phase in the nucleus is 2.3 rad. The color bar shows the phase in radians. (E) The box plot shows the average phase value of the control and treated LSEC sieve. The average phase values ​​of the porous area of ​​the control and treatment cells were 161 ± 50 mrad and 109 ± 49 mrad, respectively. (F) Assuming that the refractive index of the LSEC film is constant (n = 1.38), the average thickness of the porous area of ​​the normally processed cells is 136.6 ± 42.4 nm and 92.36 ± 41.6 nm, respectively. The lower phase value of the processed cells can be explained from the dSTORM image (ie, there are more windows in LSEC and therefore less scattering from the sample, which reduces the average phase of the processed LSE).

The change in thickness can be explained because the role of cytochalasin B is to increase the number of windows in LSEC and the height of the cells at the far end of the nucleus is reduced, making the cells flatter and more windows compared to normal cells. The cytochalasin effect results in more porous membranes, which are clearly visible in the dSTORM image. On the other hand, it was found that the average phase value of the porous area of ​​the treated cells was lower than that of the control LSEC. The lower phase value of the processed cells can be explained from the dSTORM image (ie, there are more windows in LSEC and therefore less scattering from the sample, which reduces the average phase of the processed LSEC). These results are consistent with previous studies using atomic force microscopy (36), in which it was also observed that the height of the cell periphery was reduced by about 50%.

In this work, we developed an optical nano- and highly space-sensitive QPM system based on integrated multi-modal chips. To prove the potential of the proposed system, we used dSTORM to locate the plasma membrane window in LSEC, and then used QPM to measure the thickness of the window area. When the system is run in dSTORM mode, it provides nanometer spatial resolution (61 nm) to visualize the small openings present in LSEC. In the proposed system, the same imaging arm is used to capture phase and super-resolution images without any mechanical displacement in the sample; therefore, the same position of the cells can be easily identified. On the other hand, finding the same cells in two different microscopes can be challenging and time-consuming/impractical because the size of the waveguide chip can be very large (ie 25×25 mm). In addition, the angular displacement in the final data set obtained by using two different microscopes will definitely produce sub-pixel mismatches that are unavoidable by the image registration mechanism, thus affecting the measurement accuracy of the study. Both QPM and dSTORM mode imaging arms use common optomechanical components, and the cost of using two different systems is almost doubled, which can be seen as another advantage of the system described here.

The proposed system implements multi-modal imaging in a simple manner, while still being easy to further customize. In order to improve the resolution of dSTORM, a simple flip mirror instead of a beam splitter will help to obtain a better signal and thus achieve positioning. By replacing part of the spatially coherent illumination with a perfect incoherent light source (such as white light or LED), the phase sensitivity can be further improved. White light sources provide the greatest possible spatial phase sensitivity, but due to poor temporal coherence, multiple frames (ie, phase shift interference [PSI]) are required to extract phase information. In addition, PSI can also be used to improve the lateral resolution of the system. In addition, with minor modifications, different modes can be easily added to current systems, such as waveguide-based optical capture (50) and spectroscopy techniques (51). For ease of use, along the route of automatic coupling, it is also possible to significantly reduce the system footprint (52). Chip-based microscopes have also been shown to be useful for live cell imaging of fragile cells (53). In the future, our goal is to adapt the proposed multimodal microscope platform to the imaging dynamics of fenestration in live LSEC (ie, when challenged by chemicals or drugs that change the fenestration and sieve). Being able to obtain both the aperture diameter and the sieve thickness at the same time will make it possible to track changes in a very detailed manner. This will be a particularly useful tool for discovering agents that reverse age-related pseudocapillation, because this method measures two important parameters at the same time, LSEC thickness and window opening, which increase and decrease during aging, respectively.

The processed dSTORM image and original interferogram supporting the results of this article are available from DataverseNO: https://dataverse.no/citation?persistentId=doi:10.18710/AWRGH1.

BSA recognized UiT, the Norwegian Arctic University Tematiske Satsinger funding program, and Diku-Direktoratet for internasjonalisering og kvalitetsutvikling i høyere utdanning (project INCP-2014/10024). The project has been funded by the EU Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement 766181 project "DeLIVER". KA recognizes the European Research Council (804233).

↵1A.B. and DAC have made the same contribution to this work.

Author contributions: DAC, PM, PS, DSM, KA and BSA design research; AB, DAC, KS, AA, J.-CT and PM conducted research; AB, AA and HM contributed new reagents/analysis tools; AB, DAC and KS analyzed the data; AB, DAC and BSA wrote this paper.

Competitive interest statement: BSA has applied for patent GB1606268.9 for chip-based optical nanotechnology. BSA is the co-founder of Chip NanoImaging AS, which commercializes an on-chip super-resolution microscope system.

This article is directly contributed by PNAS. ASH is a guest editor invited by the editorial board.

This open access article is distributed under Creative Commons Attribution 4.0 (CC BY).

Thank you for your interest in advertising on PNAS.

Note: We only ask you to provide your email address so that the people you recommend the page to know that you want them to see it, and that it is not spam. We do not capture any email addresses.

Feedback privacy/legal

Copyright © 2021 National Academy of Sciences. Online ISSN 1091-6490. PNAS is a partner of CHORUS, COPE, CrossRef, ORCID and Research4Life.