Quantitative, in situ visualization of intracellular insulin vesicles in pancreatic beta cells

The Content of Zn²⁺ in Insulin Vesicles.

Insulin vesicles were extracted from INS-1E cells and nonimmortalized beta cells via ultracentrifugation (16, 17), and then an enzyme-linked immunosorbent assay was performed to examine the insulin contents of the different fractions. Fraction 9 had the highest insulin content in duplicate experiments (SI Appendix, Fig. S1). Optical microscopy revealed that the vesicles in fraction 9 were uniform in shape and their size was consistent with that of insulin vesicles. Therefore, fraction 9 was processed to extract isolated insulin vesicles.

Fig. 1 A and B shows scanning EM (SEM) images of insulin vesicles extracted from INS-1E cells and the corresponding energy dispersive spectrometry (EDS) mapping of Zn²⁺, respectively. As a common component analysis method, EDS can analyze the content of an element within a sample but cannot distinguish between element valences. As the valence state of the element Zn is known to be 2+, EDS was used to measure the content of Zn²⁺ and Ca²⁺ in this experiment. The distribution of Zn²⁺ matched that of the insulin vesicles well, confirming the presence of Zn²⁺ in the secretory vesicles. To further investigate the location of Zn²⁺, EDS line scans (Fig. 1C) were performed across two insulin vesicles (marked with a red dashed line in Fig. 1A). Two obvious peaks of Zn²⁺ appeared at the vesicle positions, while there was no signal at other locations, confirming the enrichment of Zn²⁺ in the vesicles. The EDS line scan also identified Ca²⁺ within the vesicles. To further analyze the contents of Zn²⁺ and Ca²⁺, 22 insulin vesicles from INS-1E cells and 25 insulin vesicles from nonimmortalized beta cells were examined using EDS point scanning. The content of Zn²⁺ was basically proportional to the size of the insulin vesicles, but there was no obvious correlation between the content of Ca²⁺ and the size of the vesicles (Fig. 1 D–G). In addition, 24 insulin vesicles from Min6 cells were examined using EDS point scanning, and the results were consistent for the INS-1E cells and nonimmortalized beta cells (SI Appendix, Fig. S2 C and D). In addition, EDS mapping in scanning transmission electron microscopy was carried out on Min6 cells and nonimmortalized beta cells (SI Appendix, Fig. S2 A and B); Zn signal and Ca signal were obvious in both types of cells.

After insulin is synthesized in the endoplasmic reticulum, it is processed into a biologically active form and stored in secretory vesicles (0.3-µm diameter), which are transported extracellularly. Crystallization of insulin in vesicles is a hot topic in the study of islet beta cells; however, it has not yet been observed directly. Here, high-resolution transmission EM (TEM) was used to image cross-sections of pancreatic beta cells. TEM imaging of INS-1E cells revealed lattice fringes of insulin crystals in a secretory vesicle. As shown in Fig. 2A, two distinct diffraction spots were observed after the Fourier transformation, and the plane spacing of insulin crystals was calculated to be 4.22 nm according to the diffraction images. In Fig. 2B, the red line marks the junction of two INS-1E cells and the red arrows indicate the insulin vesicles. In addition, TEM imaging of nonimmortalized beta cells was performed (Fig. 2C). The number of insulin vesicles and distance between the insulin vesicles and the cell membrane were measured in both cell types (SI Appendix, Fig. S3). As shown in Fig. 2D, a large proportion of vesicles were located 0.1 to 0.5 μm from the plasma membrane.

Examining the Location of Zn²⁺ Accumulation in Vesicles Using STXM and Dual-Energy Contrast X-ray Microscopy.

The existence of Zn²⁺ in insulin vesicles revealed by the EM experiments described above is consistent with previous studies (15). Although the EM experiments revealed the relationships between the element contents and sizes of insulin vesicles, due to the low content (∼1%) of Zn²⁺ they were unable to determine the precise specific location of Zn²⁺-bound insulin in secretory vesicles. Therefore, X-ray absorption spectroscopy (XAS) based on STXM image stacks (18) and dual-energy contrast X-ray microscopy (19, 20) were used to explore the location of insulin aggregates within the insulin vesicles.

A set of STXM images (i.e., stack) was obtained using photon energies ranging from 1,010 to 1,090 eV (Fig. 3A). Various energy intervals were used within different energy ranges together with relatively larger image scanning steps to improve data collection efficiency. The energy interval between 1,018 and 1,060 eV (which is close to the Zn²⁺ edge) was 0.3 eV, and the interval for the rest of the energy range was 1 eV. The scanning step size of the STXM images was 130 nm. The STXM stack dataset was integrated and imported into the open-source software aXis2000, and a principal component analysis–cluster analysis (PCA-CA) of the insulin vesicles was performed. First, hot spots and glitches within the STXM images were recognized and removed, and then all of the images were aligned using the auto alignment function in aXis2000. For each image, the regions of interest within the insulin vesicle and a blank area were selected manually, and the intensities of these two regions were integrated to calculate the absorption ratio corresponding to the specific photon energy. In this way, the XAS spectra of specific areas could be calculated. The quantitative XAS spectra of two regions within the insulin vesicles are shown in Fig. 3 C and D. An obvious absorption peak was identified at 1,020.8 eV, which is very close to the standard Zn²⁺ L₃-edge (1,021 eV), indicating the existence of Zn²⁺ within the surveyed areas (Fig. 3B). Because Zn²⁺ is an important component of insulin, it can be regarded as an indicator of insulin aggregates within secretory vesicles.

The dual-energy contrast X-ray microscopy technique is based on the abrupt change in absorption when the energy of the incident X-rays changes from being just below the absorption edge of a specific element to being above it (Fig. 4 A and B). The observed differences between images collected at the two energies reflect the locations of element-specific nanomaterials. This method can simultaneously image the ultrastructures of cells and intracellular nanomaterials and thus can be used to investigate the interactions between nanomaterials and biological systems. Consequently, dual-energy contrast X-ray microscopy can theoretically be used to survey the spatial distribution of Zn²⁺, which is used as an indicator of the location of insulin aggregates within insulin vesicles. However, the Zn²⁺ content of insulin vesicles (∼1%) is too low to be detected by conventional dual-energy contrast X-ray microscopy based on STXM or other absorption contrast microscopy. Ptychography can determine the complex-valued refractive index by differentiating between the phase of radiation that has passed through a sample and the absorption of the sample. For biological samples, the phase cross-section is 1,000 times higher than the absorption cross-section over a broad range of X-ray energy (21, 22); consequently, the ptychography phase result has a high signal-to-noise ratio and outstanding resolving power. Here, we used dual-energy contrast X-ray microscopy based on absorption images obtained via phase-to-absorption Kramers–Kronig transformation (KKT) (23–26) of the ptychography phase results to survey the spatial distribution of insulin aggregates within secretory vesicles.

As mentioned above, the XAS spectra of the insulin vesicles indicated that the absorption edge of Zn²⁺ in beta cells is around 1,020.8 eV (Fig. 3 C and D). Therefore, we acquired ptychography datasets at 20 different X-ray photon energies across the Zn²⁺ L₃-edge, from 1,018 eV to 1,028 eV. With the collected ptychography dataset, the phase projections with 20.8-nm resolution at all photon energies were reconstructed and transformed into absorption projections via KKT. Fig. 4C shows a representative transformed absorption projection of an insulin vesicle. Dual-energy contrast X-ray microscopy was then conducted using two transformed absorption projections at photon energies below (1,018 eV) and above (1,020.8 eV) the zinc L₃-edge. As shown in Fig. 4D, the Zn²⁺ signal was concentrated in the upper left corner of the insulin vesicle examined, indicating that insulin is aggregated at this location, which is consistent with the STXM stack result. To verify the accuracy of the ptychography-based dual-energy contrast X-ray microscopy method, we performed a dual-energy contrast microscopy analysis of two randomly selected photon energies below the Zn²⁺ L₃-edge. As shown in SI Appendix, Fig. S4, no obvious abrupt change in absorption was observed. Additional results of the feasibility analysis of ptychography-based dual-energy contrast X-ray microscopy are provided in the supplemental material.

Comparison of Insulin Vesicles and Ordinary Vesicles.

The vesicles in beta cells extracted by ultracentrifugation included mature and immature insulin vesicles, as well as other types of vesicles (9). High-contrast ptychography phase images of two insulin-containing and two other vesicles were obtained to examine the morphological differences between them.

As shown in Fig. 5 A–D, these analyses revealed that insulin vesicles have a clear membrane structure and a concentrated high-density area formed by the accumulation of insulin, while the other vesicles have fuzzy boundaries and relatively dispersed high-density areas. Fig. 5 E–L shows the results of EDS mapping of Zn²⁺ and Ca²⁺ in the four vesicles. Strong Zn²⁺ and Ca²⁺ signals were detected in insulin vesicles due to insulin accumulation, whereas no corresponding signals were identified in the other vesicles. Next, we performed a horizontal line scan across the center of the four vesicles and found that the internal electron density fluctuations of the insulin vesicles and other vesicles were comparable (Fig. 5M). Normally, insulin vesicles contain areas of high density due to insulin accumulation. In order to explore the dense regions inside different vesicles, we also examined the overall phase shift values and numbers of pixels occupied by high-density areas in the two types of vesicles and found that, compared with insulin vesicles, high-density areas make up a larger proportion of the entire vesicle in other types of vesicles (Fig. 5N). In addition to the internal density information, accurate discrimination between insulin vesicles and other kinds of vesicles in beta cells would require further information regarding the nature of the elements contained.

Ptychography Imaging of Glucose-Stimulated Insulin Secretion.

Pancreatic beta cell lines are useful models to study the complex pathways regulating glucose-stimulated insulin secretion. These cells are glucose responsive over physiological ranges and are characterized by a substantial insulin content and robust metabolism-secretion coupling (27, 28). An increase in the extracellular glucose concentration leads to the induction of electrical activity within beta cells. Examining the molecular densities of organelles in whole beta cells during glucose-stimulated insulin secretion is crucial to our understanding of the functions of these cells. Therefore, we collected ptychography images of beta cells under nonstimulated and high-glucose (11 mM) conditions. As a benefit of the high-contrast property of the technique, the nucleus and insulin vesicles were easily distinguished from other cellular structures (Fig. 6 A–D). Even without staining, filamentous structures at the edge of the islet beta cell were clearly observed. The nonstimulated beta cell (Fig. 6A) contained fewer insulin vesicles than the glucose-stimulated cell (Fig. 6D), indicating that glucose stimulation promotes insulin secretion, resulting in an increase in the number of insulin vesicles in the cell.

To quantitatively analyze the resolution of the ptychography phase projections, we first calculated the pixel size of the reconstructed projection according to the formula d = λ/2sin(θ), where λ is the wavelength and θ is the diffraction angle. In our case, the pixel size was 6.05 nm. Fig. 6E shows the power spectral density (PSD) curve of the 2D diffraction patterns, which indicates that the diffraction signal extended to a spatial frequency of 64 μm⁻¹ (i.e., a resolution of 15.5 nm in real space). Considering that the PSD analysis only indicates the highest spatial frequency of the diffraction signal, the achieved resolution may be lower than the theoretical resolution due to signal-to-noise ratio and other factors. Thus, line scans in real space can be used to further estimate the achieved 2D resolution of the reconstructed ptychography image (29, 30). As shown in Fig. 6 F and G, a resolution of 19 nm along the X and Y axes was achieved.

The 3D Distribution of Insulin Vesicles in Islet Beta Cells.

Previous studies have shown that mature insulin vesicles are mainly distributed at the edge of beta cells, and immature insulin vesicles are mainly distributed near the nucleus (31). To explore the 3D dispersion features of insulin vesicles in beta cells, we performed a ptychographic X-ray computed tomography (PXCT) analysis of a marginal area ($1.48\times 6.2\times 0.9\mathrm{ }\mathrm{\mu }\mathrm{m}$), which is shown in SI Appendix, Fig. S5. This experiment was carried out using the soft X-ray spectromicroscopy beamline (BL08U1A) at the Shanghai Synchrotron Radiation Facility (SSRF). The photon energy of the light source was set to 700 eV (the corresponding wavelength is 1.77 nm). To minimize radiation damage, we used equally sloped tomography (EST) to obtain the 3D structure of a single beta cell, which provided a high-quality tomographic result from a limited number of projections. EST is an iterative method based on the pseudo-polar fast Fourier transform, with the projections collected at equally sloped projection angles; consequently, the effect of a missing wedge in limited-angle tomography and interpolation error can be mitigated. By comparing the experimental projection with the calculated projection, the quality of the reconstructed image is verified (32–34).

Twenty-two ptychography phase projections of the selected part of the cell were obtained at EST projection angles ranging from −45° to +69.44° (SI Appendix, Fig. S6). To obtain high-quality ptychography phase projections, when collecting the ptychography datasets, we set the probe diameter to 3 µm and the scanning step to 600 nm to guarantee a high linear overlap ratio of 80%. Approximately 400 diffraction patterns were collected at each projection angle, and the extended ptychographic iterative engine method was used to reconstruct the ptychography phase projection. The 3D structure of the INS-1E cell was then reconstructed by EST using the 22 ptychography phase projections. We also carried out line scans in three directions to estimate the resolution of the 3D EST reconstruction. As shown in SI Appendix, Fig. S7, the resolution in the X and Y directions (perpendicular to the optical axis) was 22.8 nm, whereas that in the Z direction (along the optical axis) was only 64.6 nm due to the low number of projection angles. At this resolution, individual insulin vesicles could be distinguished clearly, and the Amira software package was used to display and segment the reconstructed beta cells (Fig. 7A). In the segmentation results, insulin vesicles are shown in yellow and mitochondria are shown in pink. The segmentation revealed that some vesicles can acquire a different morphology, such as chains and clusters, as shown previously by Zhang et al. (31). In addition, the distances between the vesicles were variable, as was the distance between the vesicles and other organelles such as mitochondria (Fig. 7 B and C). These findings suggest that insulin synthesis, transport, and secretion are affected by adjacent organelles and the complex environment within beta cells.

The distribution of insulin vesicles observed in the PXCT experiments described above was consistent with the results of 2D ptychography imaging, which verified the dense nuclear structure of the insulin vesicles. This phenomenon was also observed in the TEM image of the beta cells. According to the 3D segmentation result, a complete insulin vesicle will span 5 to 10 slices, and the thickness of each slice is about 50 nm; hence, the size of insulin vesicles is ∼200 to 500 nm, which is consistent with the conventional size of these vesicles. We also used the 3D segmentation results to determine the number and size of insulin vesicles and found that most vesicles had a diameter of 500 to 600 nm (Fig. 7E). In parallel, we determined the sizes of the vesicles in the 2D ptychography and TEM images (SI Appendix, Fig. S8) and found that the results were consistent.

To confirm that the distribution of Zn²⁺ can represent the distribution of insulin vesicles in INS-1E cells, we performed fluorescence mapping (35, 36) of pancreatic beta cells using the 101D-1 soft X-ray spectromicroscopy beamline at the Canadian Light Source (CLS). As shown in Fig. 8 B and C, the distributions of Zn²⁺ and insulin vesicles were comparable in the fluorescence mapping image and projected 3D EST result. This finding adds support to the concept that the distribution of Zn²⁺ can be used to determine the precise positioning of insulin vesicles, which provides perspective for future analyses of the maturation of insulin vesicles.

Cell Culture and Insulin Vesicle Isolation.

For general culture, INS-1E cells were seeded into T75 flasks at a density of 8.0 × 10⁴ cells/cm² and were cultured in RPMI medium supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, and 0.05 mM mercaptoethanol. The cells were grown in an incubator at 37 °C and 5% CO₂. For analyses, cells at passage 44–55 were seeded at a density of 10⁴ cells/mL into six-well plates containing Si₃N₄ grids treated with fibronectin. Since X-ray absorption of a 30-nm-thick Si₃N₄ membrane is relatively low, the effect of substrate on the imaging results could be reduced by attaching cells to this kind of membrane. After 24 h, cells that had adhered to the Si₃N₄ membrane were fixed with 2.5% glutaraldehyde for 15 min. Subsequently, the cells were gradually dehydrated in 30%, 50%, 70%, 80%, 90%, and then 100% ethanol for 15 min per concentration (38). Chemical fixation of cells with 2.5% glutaraldehyde has a protective effect on the internal structure and morphology of cells, as well as on the position and morphology of organelles. Moreover, the gradient dehydration method avoids contraction of cells caused by rapid water loss and ensures that cells are completely dry prior to analysis. Thus, these measures were used to prevent the samples from losing water or changing morphology due to X-ray radiation during the experiments (39, 40). Finally, well-isolated INS-1E cells were selected using an optical microscope and used for the experiments.

To obtain nonimmortalized beta cells, islets were extracted from male C57BL/6 mice (8 to 12 wk old) via collagenase digestion, as described previously (41). Subsequently, isolated islets (at least 200) were washed once with precooled phosphate-buffered saline (PBS), resuspended in 0.4 mL ice-cold Accutase cell detachment solution (Thermo Fisher Scientific; Product No. 00–4555-56), and incubated at 37 °C for 5 min with gentle shaking. Digestion was stopped by the addition of 0.8 mL ice-cold tissue culture medium (TCM), comprising RPMI 1640 with 11 mM glucose, 100 IU/mL penicillin/streptomycin, and 10% FBS. Finally, dispersed islet cells were resuspended in TCM after centrifugation for 3 min at 1,000 g and 4 °C and used for further experiments.

For the insulin vesicles isolation experiment, INS-1E cells were collected from eight T75 flasks, and nonimmortalized beta cells were obtained from islets of nude mice as mentioned above. Briefly, the cells were washed with PBS and then dispersed with 0.25% trypsin for 3 min at room temperature. Subsequently, the trypsin was diluted by adding culture medium, and the sample was centrifuged at 1,000 g for 5 min. The cell pellet was resuspended in 4 mL of isolation buffer and centrifuged for 10 min at 1,000 g, saving the supernatant. Next, the centrifugation and resuspension process was repeated, but the centrifugation speed was increased gradually from 1,000 g for 10 min to 100,000 g for 75 min. The resulting precipitate contained the insulin vesicles. The operation steps and composition of the solution used in the experiment are shown in SI Appendix, Fig. S1.

For the glucose stimulation experiment, cells were cultured under normal growth conditions but were switched to low glucose (2.8 mM) medium 24 h before the assay. In addition, 1 h before the experiment, the cells were cultured in glucose-free Krebs solution (0 mM glucose). Subsequently, 11.1 mM glucose was added, and 40 min later the cells were fixed. Preparation of the Krebs solution is shown in SI Appendix, Fig. S1.

EM Experiments.

The EDS point scan, line scan, and mapping experiments were characterized by scanning EM (SEM; JSM-7800 Plus) and TEM (JEM-2100 Plus). In the SEM experiment, the original solution containing the insulin vesicles was dropped onto clean Si₃N₄ membranes that had been treated with fibronectin to increase their hydrophilicity. After natural drying, the samples can be used for SEM experiments. Under a 10-kV acceleration voltage, high-resolution SEM images were obtained by adjusting the focal length, light path, and astigmatism. X-ray energy spectrum probes were then inserted, the working distance was adjusted to an appropriate height, and EDS mapping signal collection was carried out on the appropriate identified area. The Zn²⁺ element to be measured was selected before the collection, and the signal collection time was about 5 min.

In the TEM experiment, the solution made in the same process as the SEM experiment was dropped onto copper grids, and then the high-resolution TEM images of the insulin vesicle were obtained at 200 keV. Different from the EDS mapping experiment in the SEM, the EDS mapping experiment with TEM needed to be switched to STEM mode before it could be carried out, and the signal acquisition duration was also 5 min.

In addition, TEM images of INS-1E cell slices and nonimmortalized beta cell slices were obtained by using methods of staining and sectioning. INS-1E cells and nonimmortalized beta cells were scraped off the surface of the dish with a cell shovel after chemical fixation and gradient dehydration, and the cell cluster was embedded with epoxy resin. The embedded block was cut into 70-nm-thick slices using an ultrathin sectioning machine after 24 h of curing. The ultrathin sections were placed on copper meshes, and after double staining with uranyl acetate and lead citrate (42), the samples were then observed by 80-keV TEM.

STXM Experiments.

The STXM experiment was performed mainly on the 101D-1 soft X-ray spectromicroscopy beamline at the CLS. A Fresnel zone plate with a 240-μm diameter and a 35-nm outer zone width was used in the STXM experiment. The spatially coherent soft X-rays generated by the oscillators were focused into small focal spots by the Fresnel zone plate. The first-order diffracted light was selected and filtered to remove stray light after the focal spot passed through the order-sorting aperture. The photon energy of the light source was set to be 700 eV (corresponding wavelength is 1.77 nm). The INS-1E cells were raster scanned point by point using the 2D motion of the sample stage. X-rays passing through the sample were detected by photomultiplier tubes to form a complete image.

STXM stacks data were also measured in the same environment, with a scanned image size of $9\times 9$ µm². Insulin vesicles PCA-CA analysis was performed on conventional STXM energy-stack datasets with the open-source software aXis2000 (McMaster University), using the number of significant components of 8, identified clusters number of 10, and angle distance measure mode (cutoff 0.01). This method identifies natural groups of spectra and then calculates the average spectra and displays the thickness maps associated with these spectra. The STXM stacks energy range is from 1,010 eV to 1,060 eV, interval 1 eV, with data analysis through aXis2000.

X-ray Scanning Coherent Diffraction Experiments.

Ptychography experiments were carried out on beamline 101D-1 at the CLS and on beamline BL08U1A at the SSRF. During all the ptychography experiments, the photon energy of the incident X-ray was set to 700 eV or at the Zn L-edge for the spectro-ptychography measurement. In the ptychography experiment, the samples were raster scanned point by point with a 2 μm–diameter probe at a step size of 500 nm, which corresponds to a 75% overlap between adjacent scanning areas. Diffraction patterns at each location of the sample were collected with the detector in the far field. The central 760 × 760 pixels of the patterns with diffraction signals were selected, and the mixed-states ePIE algorithm was used to conduct the ptychographic reconstruction (43–45). High-quality sample images were reconstructed with 10-nm pixel resolution after the overall diffraction patterns of the whole sample were collected. The data for 3D reconstruction were collected on beamline 08U1A at the SSRF and a total of 22 angles of projections from −45° to +69.44°.