Advanced Imaging Techniques for the Characterization of Subcellular Organelle Structure in Pancreatic Islet β Cells

Abstract

Type 2 diabetes (T2D) affects more than 32.3 million individuals in the United States, creating an economic burden of nearly $966 billion in 2021. T2D results from a combination of insulin resistance and inadequate insulin secretion from the pancreatic β cell. However, genetic and physiologic data indicate that defects in β cell function are the chief determinant of whether an individual with insulin resistance will progress to a diagnosis of T2D. The subcellular organelles of the insulin secretory pathway, including the endoplasmic reticulum, Golgi apparatus, and secretory granules, play a critical role in maintaining the heavy biosynthetic burden of insulin production, processing, and secretion. In addition, the mitochondria enable the process of insulin release by integrating the metabolism of nutrients into energy output. Advanced imaging techniques are needed to determine how changes in the structure and composition of these organelles contribute to the loss of insulin secretory capacity in the β cell during T2D. Several microscopy techniques, including electron microscopy, fluorescence microscopy, and soft X-ray tomography, have been utilized to investigate the structure-function relationship within the β cell. In this overview article, we will detail the methodology, strengths, and weaknesses of each approach.

Graphical Abstract

Introduction

Type 2 diabetes (T2D) mellitus is a global health pandemic and accounts for 90% of all forms of diabetes (20). In 2021, the International Diabetes Federation (IDF) estimated that 537 million individuals between the ages of 20 and 79 are living with T2D (20). Within the United States, 32.3 million individuals are affected by the disease, and this number is estimated to increase to 36.3 million by 2045 (20). T2D is multi-factorial, resulting from impaired insulin secretion, peripheral insulin resistance, obesity, overnutrition, lack of physical activity, and aging. Genetic and physiologic data suggest that inadequate insulin secretion from the pancreatic β cell is the chief determinant predicting progression from insulin resistance to overt T2D (83, 148, 166). Insulin, a peptide hormone produced by the pancreatic β cells, enables glucose uptake in peripheral tissues to maintain normoglycemia. During the development of T2D, insulin resistance develops in the skeletal muscle, liver, and adipose tissue. To compensate and maintain normoglycemia, β cells initially secrete increased amounts of insulin. However, over time there is a loss of β cell function, mass, and identity, resulting in metabolic instability, and hyperglycemia (83) (Figure 1).

Figure 1.

Progression of T2D. Obesity, overnutrition, and insulin resistance result in loss of β cell function, mass, and identity leading to hyperglycemia. Created with BioRender.com

To effectively treat and prevent T2D, a comprehensive understanding of β cell biology in health and disease is required. In particular, sophisticated imaging techniques are needed to determine how changes in the structure, composition, and function of subcellular organelles contribute to the loss of insulin secretory capacity. Of particular interest are the organelles of the glucose sensing and secretory pathway [i.e., the endoplasmic reticulum (ER), Golgi apparatus, secretory granules, and mitochondria], where insulin is produced, folded, processed, and trafficked for exocytosis. The ER plays a critical role in insulin production, synthesizing preproinsulin, and converting it to proinsulin (57). Proinsulin is then transported to the Golgi apparatus and sorted into secretory granules. Processing of proinsulin into insulin is completed in the secretory granules, which undergo exocytosis in response to nutrient stimulation (57). Mitochondria enable the process of insulin release by integrating the metabolism of nutrients into energy output (84). Dysfunction in any of these organelles can impair β cell function and health, ultimately leading to loss of insulin production and β cell failure (6, 150). Thus far, a number of microscopy techniques, including electron microscopy (EM), fluorescence microscopy, and soft X-ray tomography (SXT), have been utilized to investigate structure-function relationships within this compartment of the β cell. In this Overview Article, we will detail the methodology, strengths, and weaknesses of each imaging approach.

To accurately determine organelle structure-function relationships, there are several criteria that must be met. First, nanometer resolution should be achieved to visualize alterations in key structures in the glucose sensing and secretory pathway, such as the mitochondrial cristae or the ER microtubules. Without nanometer resolution, changes in these structures will be missed. Secondly, live cells must be imaged to monitor changes in these structures during dynamic cellular processes. Not only is live cell imaging key, but real-time visualization of physiology is also critical. β cells normally reside in a cluster of cells, known as islets, within the pancreas, which receives systemic inputs. However, local (i.e., paracrine) inputs also play a critical regulatory role. Therefore, an ideal imaging method would allow for the β cell to be imaged while in the native pancreas environment to understand physiology in an integrated manner. Lastly, the localization of key molecules and receptors is necessary to visualize and characterize interactions between these molecules and cellular organelles and to understand how these interactions lead to alterations in organelle ultrastructural architecture. Although a perfect imaging method does not exist, this Overview Article will outline how each imaging technique meets or fails to meet the above criteria.

Electron microscopy in the characterization of subcellular organelle structures

Prior to the advent of EM, light microscopes were used for imaging. However, light microscopes are extremely limited in resolution due to the wavelength and diffraction limit of light. Resolution is defined as the smallest distance between two point-like objects where those objects can be distinguished from one another without appearing as one single spot (22, 71). Using this definition, the Rayleigh resolution is defined as 0.61λ/NA, where λ is the average wavelength of illumination of transmitted light and NA is the objective numerical aperture (47). Based on the wavelength of light, the resolution that can be achieved using a light microscope is approximately 200 nm (71). However, the wavelength of electrons is approximately 100,000 times smaller than that of light, which allows for sub-nanometer resolution when using EM-based imaging. This enhanced resolution by EM allows for the imaging of sub-cellular organelles, leading to a greater understanding of the relationship between structure and function in cells (184). Though there are many types of EM techniques, this section will focus on the most common types used to visualize the ultrastructure of intracellular organelles within the secretory pathway of the β cell.

Transmission electron microscopy: methods and application in β cell biology

Transmission electron microscopy (TEM) is a commonly used imaging modality to distinguish intracellular organelles with the greatest resolution (184). TEM functions by shooting a beam of electrons through a vacuum where electromagnetic lenses focus the electrons into a very thin beam that is directed through the sample. The electrons either pass through the sample and hit the fluorescent screen at the bottom of the microscope or scatter if a cellular structure blocks its path. This pattern of electron transmission creates an image based on the density of electrons (189).

A significant application of TEM in β cell biology is to define the ultrastructural nature of intracellular organelles at a nanometer scale. In 1961, Lacy was the first to report the ultrastructural characteristics of the β cell plasma membrane, nucleus, mitochondria, Golgi complex, and ER (91). The plasmalemma, otherwise known as the plasma membrane, is a critical component of the β cell as it protects against external insults and creates a fixed environment within the cell. A number of different channels, including adenosine triphosphate (ATP)-sensitive potassium channels, are located on the plasma membrane and enable the transport of molecules in and out of the cell. Insulin vesicles also fuse with the plasma membrane to allow for insulin exocytosis and release (105). TEM was the first imaging modality that allowed for nanometer resolution of the plasma membrane and therefore characterization of its double membrane, the distribution of proteins and lipids within the membrane, and the docking of secretory granules (39). Regarding the β cell, TEM has enabled researchers to visualize the membrane invaginations that connect endocrine cell types as well as the double-basal membrane that envelops islet capillaries (44). Gap junctions found within the β cell membrane are critical for cellular communication as they allow for the coordination of the oscillatory dynamics of membrane depolarization and calcium (Ca²⁺) pulses (12). In the 1970s, Paolo Meda, Alain Perrelet, and Lelio Orci were the first to utilize TEM to characterize gap junctions in the β cell to better understand how these key structures are altered upon stimulation of insulin secretion. They isolated islets from rats and exposed islets to low glucose conditions (0.5 mg/mL) for 90 min followed by high glucose stimulation (3.0 mg/mL) to induce insulin secretion or treated them with glibenclamide, a drug that inhibits ATP-sensitive K⁺ channels, leading to depolarization of the β cell membrane and insulin secretion. Islets were then fixed for TEM imaging. In the islets incubated with high glucose, there was a 1.7-fold increase in the number of gap junctions, while in the islets treated with glibenclamide, there were 2.1 times more gap junctions than in the nonstimulated islets. Therefore, this foundational work was the first to highlight that gap junctions within the β cell may be modulated by different physiologic cues to increase communication between endocrine cells within the islet (112). For more information on the plasmalemma, we refer you to the article within this journal issue on membrane properties.

To understand the relationship between β cell organelle structure and function during different physiological conditions, a number of studies have applied TEM (32, 74, 87, 94, 114, 138, 187). As one example, our group has focused on defining how Ca²⁺ signaling within the ER impacts β cell function and health. Calcium plays a vital role in normal β cell function, and the ER serves as the dominant intracellular store of Ca²⁺. The sarco-endoplasmic reticulum Ca²⁺ ATPase (SERCA) pump transports Ca²⁺ from the cytosol into the ER (103, 165), whereas ryanodine receptors (RYR) and inositol 1,4,5-triphosphate receptors (IP₃R) transport Ca²⁺ from the ER to the cytosol. In addition, ER Ca²⁺ depletion triggers a rescue mechanism known as store-operated Ca²⁺ entry (SOCE), which acts to replenish ER Ca²⁺ stores through the engagement of the ER-localized Ca²⁺ sensor, stromal interaction molecule 1 (STIM1). Using mouse- and cell-based systems, we have shown that both SERCA2 and STIM1 play critical roles in the maintenance of ER integrity during physiological stressors, including obesity and ER stress (87, 165). TEM imaging of SERCA2- or STIM1-deficient β cells from a rat insulinoma cell line (INS-1) revealed a swollen and dilated ER compared to the regularly spaced stacks of ER observed in wild-type cells, illustrating that dysregulated Ca²⁺ signaling in the β cell results in structural changes in intracellular organelles (87, 165).

Similarly, a group at King’s College London used TEM to evaluate how organelle structure is altered upon β cell ER stress. They noticed that a colony of C57BL/6J mice spontaneously developed hyperglycemia. Upon sequencing, they identified a mutation in the Ins2 gene, which caused a glycine-to-serine substitution at position 32 on the B chain of the preproinsulin-2 molecule. Mice harboring this spontaneous mutation were termed KINGS mice, and this mutation prevented normal proinsulin folding, resulting in severe ER stress. TEM imaging on islets isolated from 10-week-old wild-type and KINGS mice revealed that both female and male KINGS mice had dilated ER and swollen mitochondria with distorted cristae in comparison to the wild-type mice, showing that β cell dysfunction is reflected in structural alterations (9). A study by Yi et al. has revealed how ER stress can lead to alterations in ER ultrastructural architecture; however, they induced ER stress through a high-fat diet (HFD). Male C57BL/6J mice were fed a 40% HFD for 16 weeks, and pancreatic β cells were isolated and imaged using TEM to determine the effects of obesity on ER ultrastructural architecture. In β cells from HFD-fed mice, the rough ER had evidence of dilation and vacuole-like changes and the mitochondria were swollen in comparison to chow-fed mice. These findings highlight that ER stress can lead to alterations in organelle architecture and that dysfunction in one organelle can impact neighboring organelles (188).

One key resource to assist with obtaining TEM images of β cells to evaluate organelle ultrastructural architecture is the Network of Pancreatic Organ Donors with Diabetes (nPOD), which hosts an open-access nanoscale image data repository of nPOD organ donor islets. Specifically, EM was completed on islets isolated from 47 organ donors, and the images are stored in this online database (http://www.nanotomy.org/OA/nPOD/). Thus far, ultrastructural abnormalities and innate immune cell populations have been identified in islets and exocrine pancreas from donors with type 1 diabetes (T1D) and T2D (37). A representative TEM image of a healthy islet cell can be seen in Figure 2A, whereas a TEM image of an islet from a donor with T1D or T2D can be seen in Figures 2B and 2C. In the donors with T1D and T2D, the ER is swollen and dilated compared to the healthy islet (Figures 2A–2C).

Figure 2.

Comparison of images generated by various techniques used to visualize intracellular organelles of the β cell. (A) Representative TEM image of a healthy human islet cell from the nPOD open-access nanoscale image data repository (37). Scale bar = 500 nm. (B) Representative TEM image of an islet cell from a donor with type 1 diabetes from the nPOD open-access nanoscale image data repository (37). Scale bar = 500 nm. (C) Representative TEM image of an islet cell from a donor with type 2 diabetes from the nPOD open-access nanoscale image data repository (37). Scale bar = 500 nm. (D) Representative super-resolution image of the INS-1 cell line with ER labeled using a Sec61β-GFP plasmid and GFP-Booster Alexa Fluor 647 nanobody (shown in orange). Scale bar = 5 μm. (E) Representative SEM image of the surface of a healthy human islet. Scale bar = 20 μm. (F) Representative airyscan confocal image of the INS-1 cell line with mitochondria labeled using a mitotracker dye (shown in orange) and a mitochondria-targeted mCerulean3-based Cameleon Ca²⁺ fluorescence resonance energy transfer (FRET) sensor 4mtD3mC3+16 construct (shown in green) (63). Scale bar = 10 μm. (G) Representative confocal image of an isolated mouse islet with mitochondrial potential labeled with Rhodamine 123 (green) and nuclei labeled with Hoechst (blue). Scale bar = 30 μm. (H) Representative intravital image of an endogenous wild-type mouse islet expressing GCaMP6s under the control of the rat insulin promoter (green). Vasculature is labeled with dextran (red). To stimulate β cell Ca²⁺oscillations (shown in the graph to the right of the IVM image), a 1 g/kg IP glucose bolus was administered. GCaMP6s dynamics were captured using a Leica TCS SP8 confocal/multiphoton imaging system (Leica Microsystems, Inc., Buffalo Grove, IL), following an intraperitoneal injection of glucose (1 g/kg). Quantification of calcium dynamics of individual cell in islets was conducted using Leica LAS X software (v.3.3) and ImageJ. Scale bar = 75 μm.

Although TEM can provide images with sub-nanometer resolution, there are several considerations to review when using this technique. Live tissue imaging is impossible, as the system is imaged in a vacuum (184). Imaging in real-time may provide valuable insight into changes in β cell organelle ultrastructural architecture during normal physiology and cellular stress but is not possible with TEM. There are several strategies that could be used to overcome this caveat. TEM could be completed on cells after fluorescence live cell imaging has been obtained. This would enable users to gain an understanding of what is occurring in the live cell and then, following fixation, the cellular ultrastructural architecture could be viewed. Another potential solution to the lack of live cell imaging capabilities of TEM would be to fix tissues or cells across specified time points. For instance, to determine how the structure of the ER changes during the process of insulin secretion, cells could be stimulated with glucose, fixed at various timepoints (i.e., 1, 2, 5, 10 min, etc.), and the ER ultrastructural architecture could be visualized by TEM over time. If imaging mouse tissues, a cohort of genetically homogenous mice could be sacrificed at specific time points, and fixed tissue could be examined by TEM to determine how β cell structures are altered upon different physiologic stressors. Additionally, there is a technique known as correlative light and electron microscopy (CLEM) (discussed in detail below) that combines fluorescence microscopy and EM to gain a better understanding of physiological processing occurring in the live cell while maintaining the high resolution of EM.

Another caveat of TEM is that the chemical fixation of tissue during TEM processing can lead to fixation-related artifacts such as distorted organelles (78, 169, 184). Because the overall goal of imaging is to relate alterations in structure with changes in function, alterations that are due to the fixation protocol rather than stressors placed on the cell may lead to inaccurate conclusions. While chemical fixation through the use of glutaraldehyde or formaldehyde often can introduce imaging artifacts that may resemble true structures and therefore lead to inaccurate conclusions regarding cellular structures, advances in cryo-fixation techniques can reduce these artifacts as this technique immobilizes molecules within milliseconds, thus limiting alterations made to the sample (97). This technique, which has been deemed the best method to preserve ultrastructure in biological samples, and its benefits will be discussed in greater detail in the section below on cryo-electron microscopy (46, 97, 136, 146, 161).

Additionally, because TEM uses electrons rather than fluorescence to image cells, specific molecules cannot be individually labeled, making localization experiments difficult (179). Even if an immunocytochemical approach is applied to tissues that have been fixed for TEM imaging, proteins often lose antigenicity or are simply inaccessible due to the fixation procedure (179). To overcome this issue, immunogold labeling with colloidal gold is used to localize biomolecules with TEM (64, 172). Specifically, the tissue sample is labeled with a primary antibody associated with an antigen of interest. Next, colloidal gold particles attached to secondary antibodies bind to the primary antibodies. Because gold has a high electron density, this labeling increases the amount of electron scattering during imaging and produces high contrast, dark spots which indicate the molecule of interest (64, 172). However, fluorescence labeling is preferred to gold labeling because there is significant variation in immunogold labeling efficiency between different cellular structures, perhaps due to epitope availability, which can result in biased and often inaccurate results (172).

Lastly, TEM is extremely limited in acquiring 3D subcellular information because it requires thin sections (<0.5 μm) to allow for the penetrance of electron beams to obtain micrographs (61). Therefore, a key consideration when using TEM is that a single cross-sectional view may provide a biased interpretation compared to imaging the entire β cell through multiple sections.

Scanning electron microscopy: methods and application in β cell biology

Scanning electron microscopy (SEM) methodology is similar to TEM in that imaging occurs by shooting a beam of electrons through a sample. However, whereas TEM imaging is based on the pattern of electrons passing through the sample, SEM imaging involves scattering electron beams across a sample and capturing the electrons reflected from the sample’s surface to generate an image of the surface structures (189).

SEM has been applied to the β cell to understand the structure of the cell’s surface and to determine how surface characteristics are altered during various conditions. A representative SEM image of an isolated human islet cell from a nondiabetic donor is depicted in Figure 2E. To understand structure-function relationships, Lernmark and Winblad stimulated islet cells with and without glucose and found a twofold increase in the number of “blebs” seen in the cells stimulated with glucose. They concluded that glucose induces alterations of surface morphology in the β cell (95). Similar results were also observed by Zimny and Blackard in 1975, where they also utilized SEM to evaluate alterations in surface morphology in islet cells stimulated with or without glucose. Zimney and Blackard observed an increase in the number of blebs and other surface irregularities per unit area of islet cell surface. However, they were also interested in distinguishing surface markers between α and β cells, so they utilized the double antibody radioimmunoassay of Morgan and Lazarow with immunoreactive insulin and insulin-I-125 to differentiate between α and β cells. They found that α cells appeared pyramidal in shape and were approximately 8 μm in diameter, while β cells were round or oval with a diameter of approximately 10 μm (195). Although SEM offers a subnanometer resolution of the surface of the β cell and therefore can provide key structural details that would be missed with light microscopy, SEM can only image the surface of the cell. Intracellular organelles cannot be resolved utilizing this technique. Further, all the considerations described for TEM, such as lack of live cell imaging, fixation artifacts, lack of localization, and limitations in acquiring 3D images, also apply to SEM. Advances in tomography have begun to address the issues associated with these 2D imaging techniques.

Electron tomography: methods and application in β cell biology

Electron tomography (ET) has been used in β cell research to visualize and classify 3D volumes of organelle structures such as mitochondria and the Golgi complex (74, 107, 108, 117, 187). As described above, TEM requires the chemical fixation of thin tissue slices (18). However, ET differs from TEM in that the fixed sample is now tilted when imaged, and multiple images are taken at each rotation (48, 107, 110). This tilt series is subsequently aligned, and 3D reconstructions are calculated using one of the available software packages such as IMOD (88) or SerialEM (192). Ten nanometer gold fiducials ensure the correct alignment of the images (48).

ET has offered insights into key relationships between organelle structure and function. In a study by Noske et al., whole cell ET was utilized on two “equivalent” high-pressure frozen glucose-stimulated pancreatic islets that had responded differently to glucose stimulation: one had discharged greater than 60% of its insulin granules and the other had released less than 10%. The authors concluded that the cells that responded to glucose stimulation had increased mitochondrial and Golgi volumes compared to the nonresponsive cells, highlighting the connection between organelle structure and function (117). Marsh et al. also utilized ET to demonstrate the 3D organellar relationships between the Golgi and insulin secretion. To do this, they used the pancreatic β cell line HIT-T15, which is often used in the study of the insulin secretory pathway due to its ability to secrete insulin in response to glucose stimulation. Cells were cultured under normal conditions (37°C/5% CO₂ with no glucose stimulation) for 2 to 3 days prior to freezing and subsequent imaging. Upon 3D reconstruction, imaging revealed that the ER was a continuous compartment that formed close connections with the mitochondria, multiple trans-Golgi cisternae, and compartments of the endolysosomal system (107). Similarly, they applied this technique to islet cells to better understand the architecture of the Golgi, including its elaborate and unique connections. Specifically, intact islets were obtained from female BALB/c mice, cultured overnight, and stimulated with high glucose concentrations (11 mM) for 60 min before the freezing protocol. Utilizing ET, they demonstrated that upon glucose stimulation, individual Golgi cisternae form intercisternal connections to assist with insulin processing and secretion, demonstrating that organelle structure is altered upon functional changes in the cell (108). Even though ET is an improvement from traditional TEM, there are still limitations. Because ET only captures one snapshot of a fixed cell, it is difficult to draw conclusions about the dynamic nature of the islet. While live cells are not compatible with ET, dynamic processes within the β cell can be studied by visualizing islets fixed at different time points to understand how the organelle structures are altered over a specified time period. Newer techniques, including cryo-electron microscopy (cryo-EM), offer some advantages over ET.

Cryo-electron microscopy: methods and application in β cell biology

Cryo-EM is based on the same principles as TEM, where a beam of electrons is directed at a sample to generate an image (184); however, this technique utilizes a different methodology for sample preparation. For cryo-EM, the sample is vitrified through rapid plunging into a cryogen, typically liquid ethane, to prevent water from crystallizing (129). Imaging is performed at either liquid nitrogen or liquid helium temperatures (−196 to −269°C degree window) to reduce radiation damage by approximately sixfold compared to imaging at ambient temperatures (45, 113). With this technique, cells can be imaged in a near-native state with a resolution on the order of 30 Å and with a much higher electron dose without worsening the signal-to-noise ratio (38, 102, 106, 167).

This freezing technique can also be applied to ET in a process known as cryo-electron tomography (cryo-ET) to generate 3D images (86). Nevertheless, this technique still places samples at risk of radiation damage, which has limited the application of cryo-ET in β cells, as only a hundredth of one percent of the overall cell can be imaged due to radiation damaging much of the cell (192). A recent report by Zhang et al. has overcome this limitation by recording multiple tomograms from single cells, resulting in a census of subcellular structures at high resolution from distinct regions along the insulin secretory pathway of individual β cells (192). The authors showed the connection between structure and function in the β cell by illustrating that microtubules penetrate the Golgi during insulin release, suggesting that microtubules play a role in keeping the Golgi cisternae close together and facilitating vesicle movement (192). Compared to TEM and SEM, cryo-ET techniques allow for more of the subcellular architecture of the entire mammalian cell to be visualized (76, 156). However, the scattering of the incident electron beams decreases the signal-to-noise ratio (23, 131). Additionally, like TEM, these samples must be fixed, which does not allow for live cell imaging; however, this caveat can be overcome by fixing multiple samples at different time points. Lastly, the sections must be extremely thin (200 nm), which does not allow for whole cells to be imaged (102, 106). Imaging of thicker sections at a nanometer scale is required if the technique is applied to whole islet cells. Advancements in the section titled on “Focused Ion Beam Scanning Electron Microscopy” begin to address these issues.

Focused ion beam scanning electron microscopy: methods and application in β cell biology

Focused ion beam (FIB) scanning electron microscopy (FIB-SEM) is a technique where a FIB ablates a one-nanometer layer from the top of the tissue block-face to be imaged. This process is followed by SEM imaging of the freshly exposed surface. Once an image is collected, the process repeats with the FIB ablating a new layer from the block face, followed by SEM imaging of the new surface. This process is repeated until the entire 3D volume is ablated and imaged, taking multiple days (186).

Although this technique is newer in the diabetes field, its application has already allowed for valuable insights into the relationship between organelle structure and function in the β cell and in other cell types that are critical for proper metabolic function. To establish structure-function relationships between organelle architecture and metabolic function in the context of diabetes, Parlakgül et al. utilized hepatocytes in their normal physiologic environment of the liver as their cells of interest. Specifically, they applied the technique of FIB-SEM to liver samples from lean and obese mice and obtained 5638 and 7896 consecutive images, respectively. After using machine-based learning and convolutional networks, they automatically segmented the various organelle structures, including the ER, mitochondria, cristae, lipid droplets, peroxisomes, and Golgi, and found that there were many obesity-induced alterations in organelle ultrastructural architecture. For example, there was a decrease in abundance of ER sheets and a disturbance in the organization and parallelism of the ER sheets in the obese liver as compared to the lean liver. This finding suggests that obesity-induced alterations in the function of hepatocytes are associated with marked changes in ultrastructural architecture (127). Regarding the β cell, Müller et al. utilized FIB-SEM to reveal the association between 3D microtubule organization and organelle interaction under low- and highglucose conditions in mouse islet cells. The data from this study suggest that microtubules play an important role in transporting insulin granules for exocytosis, as microtubules were enriched near the plasma membrane after glucose stimulation (115).

One key advantage of FIB-SEM over the 3D reconstruction generated with ET is that the nanometer layers are removed with each step using the FIB. The exactness gained in this slicing process allows the resolution along the z-axis to be only a few nanometers, which is unattainable with other techniques (40, 68, 177). Therefore, FIB-SEM allows for high-resolution imaging in all three dimensions, thus enabling the accurate 3D characterization of β cell organelles with nanometer resolution. However, FIB-SEM is extremely limited by equipment availability and expense. In addition, this technique only allows for a single static snapshot of a cell, as live cell imaging is impossible. As mentioned above, this caveat could be overcome by fixing multiple samples at different time points. FIB-SEM imaging is slower than standard EM because images are obtained pixel-by-pixel (186). Furthermore, once a tissue is imaged, it cannot be re-imaged as the FIB destroys the sample when ablating each layer (186). To solve this issue, when collecting tissues for imaging, a cohort of genetically homogenous mice can be sacrificed at the same exact time so that multiple tissues offering the same insights may be utilized either for additional FIB studies or other complementary studies. Protein localization is also not possible with FIB-SEM.

Fluorescence microscopy in the characterization of subcellular organelle structures

Fluorescence microscopy is one of the commonly used imaging methods that allows for the localization of individual molecules in either fixed or live cells (59, 124). Integrating the resolution of EM with the live cell and localization capabilities of fluorescence microscopy has the potential to monitor β cells in real-time to characterize the relationship between ultrastructural architecture and function. Fluorescence microscopy is based on the principle that a fluorescent indicator targets a specific protein, lipid, or ion, which allows for the visualization of that target in the 3D cellular space due to the absorption and emission of photons (59, 124). During fluorescence microscopy, a light source is used to direct photons at a specified wavelength toward a labeled sample. When the photons hit the fluorescent indicators on the specimen, they absorb the light energy, become excited to a higher energy state, and subsequently emit photons with a longer wavelength than the absorbed photons as they fall back down to their ground state. The emitted photons are then visualized (98, 145). Wide-field fluorescent microscopy, in which the entire field of view is illuminated, is a commonly used tool in cell biology. However, the detection of out-of-focus light compromises the resolution of the overall image, so confocal microscopy is often utilized instead (162, 181). This section will detail the most common types of fluorescence microscopy used to characterize β cell organelle structure.

Confocal microscopy: methods and application in β cell biology

Laser scanning confocal microscopy is one of the current imaging standards in the field of cell biology due to the ability to label specific targets of interest and to construct 3D images up to 100 μm in thickness with a sub-micrometer resolution in the live cell (22, 60). Laser scanning confocal microscopy functions according to the same principles as detailed above; however, a few key additions allow for increased thicknesses of samples to be imaged. An image is obtained by scanning a laser beam continually across and through the sample. At each point in the scanning process, the laser excites the specimen at that location, and the specimen emits a photon at a specific wavelength. From the emitted light, only the in-focus z-plane will pass through the pinhole aperture and be amplified by a photomultiplier tube (PMT) before reaching the detector. The pixel intensity of the image generated is determined by the time-averaged response of the PMT to the emitted fluorescence (145). PMTs are one type of detector that can be used in conjunction with confocal microscopy and have the greatest sensitivity in the blue to ultraviolet (UV) range (5). However, semiconductors such as silicon photomultipliers and avalanche diodes are alternatives to PMTs and function by utilizing the photoelectric effect to convert light into electricity. These semiconductors may be used instead of PMTs as they have high quantum efficiency at longer wavelengths and can be designed with ionization energies that directly coincide with the wavelength of interest (5). However, there are also hybrid detectors which combine PMTs and avalanche diodes. These systems are able to prevent photobleaching by minimizing the laser intensity that is often seen with PMTs in isolation (51). Once the pixel intensity of the image generated is determined through the process described above with the focal plane at one axial z-location, the z-location is moved to a new focal plane. The z-axis slices can be combined to form one 3D stack (21, 60). Because cells do not need to be fixed before laser scanning confocal microscopy, live cell data can be collected and dynamic processes within the β cell can be visualized. A representative confocal image of INS-1 cells with the mitochondria labeled can be seen in Figure 2F.

Spinning disk confocal microscopy scans the sample using a disk that contains multiple pinholes rather than the single pinhole used for laser scanning confocal microscopy. As the disk spins, each pinhole serves as a point source of light that scans across the specimen before following the image acquisition process detailed above (145). Additionally, spinning disk confocal microscopes use camera-based detectors with high sensitivity for fast imaging rates, making this technique optimal for live cell dynamic measurements and for capturing processes that occur very quickly (145).

Several published studies have utilized fluorescent labeling and confocal microscopy to better understand the relationship between β cell function and subcellular organelle dynamics (21, 24, 27, 54, 55, 193) (Table 1). For example, Gerencser et al. computed the mitochondrial volume in primary β cells isolated from rats using MitoTracker fluorescent dye to validate the measurement of mitochondrial membrane potential, which is a major component of the proton motive force that determines the rate of ATP synthesis, the ratio of ATP/adenosine diphosphate (ADP), the closure of ATP-sensitive K⁺ channels, and the eventual secretion of insulin. The authors characterized how membrane potential is altered under different glucose concentrations to understand how glucose concentrations affect variations in glucose-stimulated insulin secretion (55). The authors concluded that the degree of hyperpolarization of the mitochondria after glucose stimulation could modulate the maximum available rate of ATP synthesis and the cytosolic ATP/ADP ratio, suggesting a dynamic connection between these metabolic pathways during glucose-stimulated insulin secretion (55). A representative confocal image of islet cells with mitochondrial potential labeled with Rhodamine 123 can be seen in Figure 2G. Another group, Pétremand et al., labeled the Golgi using the Golgi marker GM130 in the MIN6 β cell line to determine if high-density lipoproteins (HDLs) can promote proper protein trafficking during ER stress. Specifically, they infected MIN6 cells with a temperature-sensitive vesicular stomatatis virus glycoprotein-green fluorescent protein (VSVG-GFP) fusion protein-encoding lentivirus (VSVG) to visualize properly folded proteins. Cells were then incubated with or without palmitate, a compound known to prevent ER protein export, in the presence or absence of HDLs. After incubation, the Golgi was labeled with an antibody recognizing the Golgi marker GM130 to determine if the fluorescent properly folded proteins (VSVG) were colocalized with the Golgi (GM130), signifying proper protein trafficking. The majority of cells treated with palmitate had reduced amounts of successful protein trafficking, as shown by a lack of colocalization. However, HDL co-treatment restored protein export, as illustrated by increased colocalization of the Golgi and protein stains. The authors concluded that HDLs contribute to improved β cell ER homeostasis by maintaining proper protein trafficking (133).

Table 1.

Summary of Techniques to Fluorescently Label Organelles

Class	Methodology	Advantages	Disadvantages	Example
Antibody	Fluorophore-conjugated antibody binds specifically to an antigen of interest	High affinity and specificity	Large size decreases the accuracy of localization, expensive	Anti-TOM20 antibody: fluorescently labels the mitochondria by targeting the central component of the receptor complex responsible for the recognition and translocation of cytosolically synthesized mitochondrial preproteins (79)
Fluorescent dye	A dye that has an excitation/emission peak to emit fluorescence. May be chemically altered to bind specific moieties	High photostability, does not require maturation time (28), compatible with live cell and in vivo imaging	May lack specificity, potential for photobleaching	MitoTracker: fluorescently labels the mitochondria within live cells using the mitochondrial membrane potential. It is chemically reactive and links to the thiol groups in the mitochondria (28)
Fluorescent protein-based reporter	A fluorescent protein (i.e., GFP or red fluorescent protein (RFP)) is genetically linked to a target protein	Allows in vivo imaging, do not have to stain cells for imaging	May need to create transgenic mouse model or fluorescent cell line prior to imaging (time consuming, expensive), potential for photobleaching	mEmerald-Sec61β: transfection of cells with mEmerald-tagged Sec61β (where mEmerald is a fluorescent protein) to fluorescently label the ER (79)
Nanotechnology (quantum dots)	Nanocrystals emit fluorescence when stimulated with UV light. May be conjugated to antibodies for labeling a molecule of interest	Size-tunable light emission, high signal brightness, extended photostability, smaller size for greater localization (77)	Highly toxic, may degrade within the cell	Luminescent quantum dot: quantum dot conjugated to cytochrome-c oxidase VIII mitochondria-targeting signal peptides (QD-MTS) to fluorescently label the mitochondria (77)

There are limitations to confocal microscopy. Due to the strongly overlapping emission spectra of fluorophores, there is a limit to the number of structures that can be visualized at once, leading to a lack of cellular context per sample. However, there have been many advancements in this area to allow for simultaneous imaging of fluorophores to visualize a larger number of organelles or proteins of interest at once. For instance, the white light laser emits a white spectrum but has the properties of a laser, which is necessary for confocal microscopy. Therefore, it can be used in conjunction with an acousto-optical tunable filter that can be programmed to separate any desired color out of the white spectrum with a spectral band width of only a few nanometers to prevent spectral overlap (30). Somewhat similarly, spectrally tunable optical detectors utilize a tunable bandpass filter that can be modified to allow a specific wavelength to pass through, enabling distinct separation of fluorophores (33). Additionally, the lifetime fluorescence of fluorophores may be taken advantage of to distinguish one fluorophore from the other to determine their relative contribution to each image pixel. However, several images need to be taken to accurately determine the fluorescent lifetime, which then restricts the speed of image acquisition (132, 175). Lastly, a final advancement to aid in the separation of emission to prevent fluorophore overlap is a technique known as spectral imaging coupled to mathematically linear unmixing of the measured spectral profiles. In this technique, each spectrum collected through the image acquisition process is segregated so that the mixed, overlapping fluorescent signals are separated into their individual parts. Therefore, if some emission spectra of the fluorophores utilized in imaging do overlap, they can be clearly distinguished from one another (194).

Another limitation of confocal microscopy is that the resolution of confocal microscopy (~200 nm) cannot reach that of EM (~0.1 nm) due to the diffraction limit of light. Therefore, the details of subcellular structures cannot be resolved using classical confocal microscopy (71). For instance, in the study described above conducted by Gerencser et al., the mitochondrial “structures” are labeled mitochondria, not the exact ultrastructure, such as the mitochondrial cristae, which cannot be visualized with confocal microscopy. To attain greater resolution to visualize organelle substructures and changes in ultrastructural architecture, the diffraction limit of light must be overcome. If the resolution is too low, these changes may be missed completely. Additionally, single molecules cannot yet be localized via confocal microscopy.

Multiphoton microscopy: methods and application in β cell biology

Whereas confocal microscopy utilizes a single photon to excite the fluorophores within the sample, multiphoton microscopy employs two or more photons to improve the visualization of 3D and live cell events, which may overcome some of the pitfalls of classical confocal microscopy. Multiphoton microscopy is based upon the principle that the energy of a photon is inversely proportional to its wavelength. For example, in two-photon excitation, the two photons that are absorbed by the fluorophore must each have a wavelength that is twice the length required by only one photon and they must both reach the fluorophore at the same time. Because the two photons must reach the fluorophore simultaneously, the laser focal point is the only location in which the photons are close enough to generate this event. Therefore, above or below the focal point, no background fluorescence is generated and the emission region is “intrinsically confocal” (41). This technique allows for improved resolution compared to confocal microscopy, especially in the live cell. Additionally, because there is no photon absorption in the out-of-focus sections of the specimen, there is increased specimen penetration in the areas of interest. In turn, the amount of photobleaching that leads to specimen damage is reduced. This phenomenon can be seen in the 2014 study conducted by Low et al. to determine the relationship between the direction of insulin secretion and the location of vasculature. The authors isolated islets from mice and labeled the β cells within the islets with sulforhodamine B. Using live cell two-photon imaging, they visualized the process of insulin granule fusion and exocytosis in response to high glucose concentrations and found that there was an asymmetric, nonrandom distribution of insulin granule fusion sites specifically localized to the membrane that was facing the vasculature during glucose stimulation. Further, by taking advantage of the high resolution in live cells offered by two-photon microscopy, the authors were able to determine that structural proteins such as liprin, piccolo, and Rab2-interacting molecule, which are normally associated with neuronal presynaptic targeting, are also located in the β cell. Specifically, these structural proteins were facing the vasculature, suggesting that β cells are organized in a polarized manner to assist in insulin secretion (101). Takahashi et al. further leveraged the advantages of two-photon microscopy in live cell imaging to characterize the process of insulin granule exocytosis. They labeled isolated mouse islets with the fluid-phase tracer sulforhodamine B and the membrane tracer FM1–43. Islets were stimulated with 20 mM glucose and imaged using two-photon microscopy. The inherent confocality of two-photon microscopy as well as its ability to image live cell events while maintaining high resolution allowed for the determination of key features of insulin granule dynamics. Specifically, they found that the FM1–43 signal appeared either simultaneously or even slightly prior to the sulforhodamine B signal, suggesting that the fusion pore for insulin exocytosis is formed by membrane lipids. Further, they also explored whether the fusion pore was dynamic and expanded during exocytosis or if it remained one size by utilizing tracers of various molecular size. Through visualization with two-photon microscopy, they found that smaller molecules such as sulforhodamine B and 3-kD dextran permeated earlier than larger molecules such as 10-kD dextran, suggesting that the pore does in fact dynamically expand. Therefore, they concluded that because insulin secretion and granule exocytosis are mediated by an expanding lipid fusion pore, there is the potential that the impaired insulin secretory ability in individuals with T2D may stem from issues with the lipid pore structure during insulin exocytosis, as patients often have abnormal lipid metabolism (163).

To collect the emitted light after excitation in the above studies, either an external detector or charged couple device (CCD) camera is used. The CCD camera is ideal for fast data acquisition systems, lending itself nicely to capturing live cell imaging events such as Ca²⁺ oscillations in the β cell (43). In 1994, Bergsten et al. were one of the first groups to characterize the fast and slow oscillations in isolated mouse islets by taking advantage of the CCD camera. They loaded isolated mouse islets with fura-2 and recorded the Ca²⁺ oscillations with dual-wavelength fluorometry with excitation at 340 and 380 nm and emission at 510 nm. When glucose concentrations were raised from 3 to 11 mM, Ca²⁺ concentrations within the islet rose and Ca²⁺ oscillations began to appear. Oscillations were either classified as fast (2–7/min), slow (0.3–0.9/min), or a combination of the two types. More specifically, the authors demonstrated that the slow Ca²⁺ oscillations correlated with the insulin release from the glucose-stimulated islets, thus highlighting the relationship between Ca²⁺ oscillations and insulin release within the islet (14). Further, Zarkovic and Henquin compared Ca²⁺ isolations between dispersed islets and whole islets isolated from mice by computing the difference in period and synchronization index between the different cell types. Cells were loaded with fura-2 and imaged within a perfusion chamber using an inverted microscope by exciting the specimen with 340 and 380 nm. The fluorescence was emitted at 510 nm and recorded using a CCD camera, similar to Bergsten’s group. Based on the emitted fluorescence, it was concluded that whole islets exhibit regular, synchronized Ca²⁺ oscillations under low glucose conditions, which was in contrast to the single cells which were not synchronized (190). Nadal et al. and their colleagues also took advantage of the CCD camera to record Ca²⁺ oscillations within the islet; however, in this case, they were interested in comparing Ca²⁺ oscillations between the different cell types within the islet such as the β cell, the α cell, and the δ cell. Mouse islets were loaded with fura-2 and imaged using laser scanning confocal microscopy, and fluorescence was recorded through the use of a CCD camera. Following Ca²⁺ imaging, the islets were fixed and immunostained with insulin, glucagon, and somatostatin to determine the specific cell types within the islets to correlate the oscillatory behavior with the specific cell type. Overall, it was found that α cells exhibited Ca²⁺ oscillations in the absence of glucose, δ cells showed Ca²⁺ oscillations when exposed to glucose concentrations as low as 3 mM of glucose, and β cells were the only islet cell type that had synchronized oscillatory behavior across all of the β cells within the entire islet when glucose concentrations were increased from 3 to 11 mM, signifying that not all cell types within the islet behave in a similar manner (116).

Different types of microscopy techniques can be combined with multiphoton microscopy to achieve a greater focal depth, reduce out-of-focus fluorescence, and enhance fluorescence signals (25). These techniques include total internal reflectance fluorescence microscopy (TIRF) and fluorescence lifetime imaging (FLIM), which can be used in isolation or in conjunction with multiphoton microscopy, to measure the ultrafast dynamics of β cells. TIRF microscopy takes advantage of the different refractive indices in liquids (i.e., the cell media/specimen) and solids (i.e., the glass coverslip) to create a thin electromagnetic field known as the evanescent wave that acts to excite the fluorophores in the sample. When the photons contact the interface between the solid glass coverslip and the liquid cell culture media, light is either refracted as it enters the second medium or is reflected at the interface. Whether the light is reflected or refracted depends on the incident angle and the difference in refractive indices of the two mediums, and is governed by Snell’s Law where $n_1$ is the higher refractive index, $n_2$ is the lower refractive index, ${\theta }_1$ is the angle of the incident beam within the higher-index medium, and ${\theta }_2$ is the refracted beam angle within the lower-index medium:

$$n_1\sin {\theta }_1=n_2\mathrm{s}\mathrm{i}\mathrm{n}{\theta }_2$$

If the photon hits the interface of the cell medium (liquid) and the glass slide (solid) at the critical angle, the refraction direction is parallel to the interface, and the light is completely reflected into the first medium with no photons passing through into the second medium. At angles greater than the critical angle, the reflected light generates a highly restricted electromagnetic field adjacent to the interface in the lower-index medium, known as the evanescent field, and extends only a few hundred nanometers into the specimen therefore limiting the number of fluorophores which are able to be excited. Only the fluorophores in the region of the evanescent field are excited and recorded (50, 109).

Because TIRF allows for excitation of fluorophores specifically in the domain close to the plasma membrane due to the limited region of the evanescent field, single granule movement can be visualized and the process of insulin granule exocytosis can be observed (10, 120, 121, 152). Ohara-Imaizumi et al. took advantage of these principles to determine the role of syntaxin-1 and soluble NSF attachment protein (SNAP)-25 in insulin exocytosis using the Goto-Kakizaki rat model of T2D. Goto-Kakizaki rats develop insulin resistance and an insulin secretory defect due to selective inbreeding for a hyperglycemic trait. Specifically, the authors immunostained isolated islets from hyperglycemic and normoglycemic Goto-Kakizaki rats treated with insulin for 2 weeks with anti-syntaxin-1A, anti-SNAP-25, and anti-insulin antibodies. Following labeling, they visualized insulin granule motion with TIRF microscopy and found that the number of syntaxin-1A and SNAP-25 clusters and the overall number of insulin granules docked at the plasma membrane were reduced in rats with untreated T2D and hyperglycemia as compared to diabetic rats treated with insulin. These findings signify the importance of syntaxin-1 and SNAP-25 in insulin granule fusion and exocytosis (121). Whereas Ohara-Imaizumi et al. were interested in the process of docking, Ma et al. utilized TIRF microscopy to better characterize insulin granule fusion. Specifically, they immunostained MIN6 cells and isolated mouse islets using syncollin-GFP and insulin to label vesicle content. Using TIRF imaging, they reported that the majority of insulin granules fully fused with the cellular membrane and 87% completely released their vesicle contents, showing that endocytic retrieval of insulin granule membrane does not involve direct refilling (104). To further understand the process of secretory vesicle docking prior to insulin exocytosis, Gandasi and Barg also utilized TIRF microscopy. INS-1 cells were transfected with the granule marker neuropeptide-Y (NPY)-mCherry as well as Munc13-EGFP, syntaxin-EGFP, EGFP-Rab3a, and SNAP25-EGFP to visualize the insulin granule docking process at the plasma membrane. They concluded that insulin secretory vesicles dock at the plasma membrane by inducing syntaxin/munc18 clustering in the target membrane (53).

Although TIRF is an excellent technique for evaluating insulin granule dynamics, there are several pitfalls. Because the evanescent field is restricted to only a few hundred nanometers, granules cannot be tracked deeper within the cells and tissues (104). One potential solution to this issue would be to utilize TIRF microscopy in conjunction with confocal microscopy which has the ability to resolve granules at greater depths. In this way, the two approaches can provide complimentary information.

FLIM is another microscopy technique that can be used in conjunction with multiphoton microscopy. FLIM utilizes the concept of lifetime fluorescence which is defined as the average time that a fluorophore remains in its excited state. Fluorescence lifetime can be measured either in the time-domain or the frequency-domain (35). When measuring the fluorescence lifetime in the time-domain, a photon pulse is utilized to excite the fluorophores within the sample. As the emitted fluorescence begins to decay, time-correlated single-photon counting is typically used to determine the fluorescence decay curve. After a significant number of events are recorded, a histogram of all the events is made and the lifetime parameter is extracted. When measuring the fluorescence lifetime in the frequency-domain, a phase-modulation method is used. A light source is pulsed at a high frequency and the emitted fluorescence after excitation is demodulated and undergoes a phase shift, both of which are related to the decay times of the fluorophores. From these values, the lifetime of the fluorophore can be computed.

Because FLIM utilizes the fluorescence lifetime of molecules, the fluorescence lifetime of the autofluorescent nicotinamide adenine dinucleotide phosphate (NAD(P)H) can be used as a concentration-independent, label-free tool to monitor the ratio of free and bound NAD(P)H in the β cell, providing key insights into the metabolic status of islets (49). Gregg et al. sought to understand how β cell metabolism changes with age, as aging is often associated with impaired glucose homeostasis and an increased risk of T2D (26, 62, 65, 90). The authors obtained human islets from 31 different human donors ranging from 19 to 64 years of age, conducted a glucose stimulated insulin secretion assay (GSIS) on a subset of these islets, and found that there was a significant decrease in β cell function as the islets increased in age. To determine the cause of this age-related decline in β cell function, they utilized FLIM to determine the status of mitochondrial metabolism within the β cells. They concluded that donor age was positively correlated with the progressive loss of glucose-dependent nicotinamide adenine dinucleotide plus hydrogen (NADH) utilization, suggesting that a reduction in mitochondrial metabolism contributes to the impaired β cell function that occurs with aging (62). To explore the differences in metabolism between α and β cells, Azzarello et al. leveraged the ability of FLIM to measure the variations in the ratio of unbound to bound NAD(P)H after stimulation with glucose. Specifically, they measured a total of 312 α cells and 654 β cells from four healthy human donors. They used FLIM to measure the ratio of unbound to bound NAD(P)H at a baseline glucose concentration of 2.2 mM and after stimulation with 16.7 mM of glucose. Subsequently, the islets were fixed, immunostained for glucagon and insulin, and imaged by multiphoton microscopy to determine which cells were α or β cells. The authors determined that both α and β cells had increased bound NAD(P)H when the donor islets illustrated lower insulin secretion, while α cells had increased unbound NAD(P)H and β cells had increased bound NAD(P)H when the donor islets had higher insulin secretion, suggesting that there is a coordinated response in the mitochondria of α and β cells during insulin secretion (11). To further characterize the differences in metabolism between α and β cells, Wang et al. examined the differences in oxidative phosphorylation and glycolysis in healthy α and β cells versus α and β cells from a mouse model of T2D. Isolated islets from lean wild-type mice and transgenic mice that express human islet amyloid polypeptide (IAPP), a toxic protein that induces misfolded protein stress seen in T2D, were imaged with FLIM using two-photon excitation to determine the NAD(P)H and lipofuscin signal after stimulation with 16 mM of glucose. Following this measurement, the islets were fixed and immunostained with insulin and glucagon to differentiate the α and β cells. In healthy islets, they found that glucose stimulation enhanced oxidative phosphorylation in β cells while suppressing it in α cells. In islets isolated from mice transgenic for IAPP, glucose stimulation increased glycolysis in β cells while only partially suppressing it in α cells. Together, these findings suggest that the failure to suppress glucagon secretion from α cells in response to hyperglycemia in T2D is secondary to β cell dysfunction rather than a primary defect found in the α cells themselves. This idea was confirmed using human islets isolated from healthy donors and donors with T2D (176).

FLIM has many advantages in the evaluation of the metabolic status of β cells, as it allows live islets to be directly monitored in real time without any imaging artifacts caused by fixation. Further, because the lifetime of the fluorophores is independent of laser power, it is not affected by fluorophore concentration or spectral overlap, which is an issue in confocal microscopy (35). However, the FLIM analysis process is slow and time-consuming, and there are a limited number of molecules that can be localized with FLIM, including NAD(P)H and flavin adenine dinucleotide (FAD) (35). Therefore, the localization of precise organelle structures or hormones such as insulin or glucagon is not possible. To overcome this challenge, islets can be fixed following FLIM imaging and immunostaining can be used to localize the molecules of interest, as was done in many of the examples described above.

Intravital microscopy: methods and application in β cell biology

Multiphoton techniques have increased the resolution obtained by classical laser scanning microscopy and have enabled the measurement of dynamic processes in live β cells; however, islets still must be removed from their physiologic environment within the pancreas for this type of imaging. A major limitation of the approaches detailed above is that they can only be used in ex vivo or in vitro settings, thus removing systemic inputs and cell-cell interactions that are critical for in vivo functioning and the understanding of physiological relevance. Therefore, images acquired from isolated islets may not accurately reflect the true architecture and function of the β cell (8, 157). Intravital microscopy (IVM) utilizes fluorescent reporters to study the structure and function of cells in living organisms (93, 139) and offers a potential mechanism to overcome these limitations. Three different techniques have been developed to allow IVM to assay islet structure and function in vivo: islet transplantation into the anterior chamber of the eye (ACE), externalization of the pancreas, and placement of an abdominal window over tissues of interest such as the kidney or pancreas.

ACE approaches allow islets to be visualized and imaged in vivo over time to monitor islet functionality after transplantation (158). Once engraftment occurs, fluorescent imaging can be used (93) to measure the dynamics of vascularization (3, 158), innervation (158), immune cell interactions (1, 158), alterations in β cell mass (158), and signal transduction (158). Although the ACE system allows for easy access for imaging of the islets, the islets are still removed from their native environment of the pancreas. To overcome this caveat, IVM imaging of the externalized pancreas has been utilized. Here, islets expressing a labeled protein of interest can be imaged directly if studying the endogenous pancreas or can be isolated from a transgenic mouse and transplanted into a recipient mouse (119). The tail region of the pancreas is exposed to enable fluorescent imaging. This method enables live, real-time imaging of β cells in a physiologic environment, and it has been utilized to illustrate the dynamic interplay between immune cells and β cells in diabetes development (34) and to characterize islet blood flow as a function of blood glucose level (119). The genetically encoded Ca²⁺ indicator GCaMP6 under the control of the rat insulin promoter can be leveraged using this technique to visualize Ca²⁺ oscillations upon glucose stimulation in the islet cell (Figure 2H). The externalized pancreas method only allows for one-time imaging of islets in vivo; however, the abdominal window technique has recently been developed to perform longitudinal follow-up (140). This “window” can be placed over the abdomen through surgical intervention to expose a tissue of interest, such as the kidney (a common location for islet transplantation) (173) or the pancreas (139). The imaging is performed by monitoring fluorescent markers for several days.

The above methods are typically used for live cell, functional imaging of the β cell rather than structural imaging (1, 3, 34, 93, 119, 139, 140, 158, 173). For instance, Chen et al. utilized a fluorescent, cell-surface targeted zinc (Zn²⁺) indicator, or zinc indicator for monitoring induced exocytotic release (ZIMIR), to monitor the exocytotic release and secretion of insulin through the intravital imaging of the exteriorized pancreas of a mouse (29). Because insulin granules contain a high level of Zn²⁺ and Zn²⁺ is released in conjunction with insulin during insulin secretion, they were able to utilize this unique fluorescent probe to monitor the dynamics of insulin secretion and exocytosis (96). Islets were labeled by injecting the ZIMIR dye through the splenic artery of the mouse and the exteriorized pancreas was visualized with confocal imaging 10 to 15 min later. It should be noted that because this dye can be injected through the splenic artery, IVM does not always require genetically encoded endogenous fluorescent reporters. After challenging the mice with a bolus of glucose, the authors successfully visualized the rhythmic secretion of insulin and Zn²⁺ within neighboring islets and concluded that the average oscillation period after glucose stimulation was 32.48 ± 8.06s (29). It is significant to note that the studies by Ghazvini Zadeh et al. take advantage of the presence of Zn²⁺ in insulin granules in their fluorescent Zn²⁺ granule indicator (ZIGIR). This molecule binds to Zn²⁺ with submicromolar affinity and has more than 100-fold fluorescence enhancement with Zn²⁺ complexation. It accumulates in acidic secretory granules and responds at both neutral and acidic pH, which is critical since insulin secretory granules are acidic. Further, because it has low Zn²⁺ affinity, its fluorescence is only observed in secretory granules that have high concentrations of Zn²⁺, rather than other cellular compartments which may contain low amounts of Zn²⁺. Thus far, ZIGIR has only been used with ex vivo confocal microscopy techniques to separate heterogenous β cells based on their insulin content or to sort β cells from other islet cell types (58). While ZIGIR has not yet been utilized in conjunction with IVM, it has the potential to be applied similar to ZIMIR.

Recent studies indicate that IVM can also be used to characterize the ultrastructural architecture of β cell organelles. For example, Reissaus et al. fluorescently labeled and imaged the islet mitochondria in C57BL/6J streptozotocin-treated mice using the abdominal window technique. They concluded that β cells had strong mitochondrial labeling, reflecting their high metabolic activity, which allowed them to be easily distinguished from the other cell types in the pancreas (139).

While these IVM techniques enable live cell functional imaging of the β cells within their physiological environment, there are some drawbacks. For example, because the islets are imaged within a live mouse, the mouse is placed under anesthesia. Although it is anesthetized, there still may be slight movements due to breathing or too weak of an anesthetic, and these motion artifacts may impact results. Specifically, while larger islets seem to be mostly unaffected by these movements, smaller islets are susceptible, potentially leading to inaccurate image analyses (52). Additionally, anesthesia can lead to alteration in glucose tolerance and insulin secretion as illustrated by Sato et al., where the use of sevoflurane and propofol in mice leads to inaccurate conclusions about β cell function (149). Further, when performing IVM in the ACE, the islet has still been removed from its physiologic environment and will be exposed to different hormonal or physiologic signals. Additionally, when imaging β cells within the ACE, the glucose concentration that islets are exposed to can only be estimated from the tail vein or through a more invasive approach (i.e., a blood draw from the carotid artery within the neck region). This discrepancy is a potential issue due to the fact that the glucose concentrations within the eye are different than that recorded in the blood (89). IVM approaches involving externalization of the pancreas also have caveats. Only the tail region of the pancreas can be safely imaged so as to not damage other organs such as the duodenum and the stomach (52). Therefore, even though hundreds of islets can be captured at a time, they are all within the tail region of the pancreas. This is an important caveat, as there is significant regional heterogeneity in islet size and function between different regions of the pancreas, and analyzing only islets within the tail region may introduce some bias (52). Specifically, Tasaka et al. illustrated that the glucagon content in rat pancreatic islets was significantly higher in the dorsal lobe (upper head, body, and tail) of the pancreas as compared to the ventral lobe (inferior head and uncinate process), while insulin content was unchanged. The release of insulin was also similar between both lobes, whereas glucagon release was significantly increased in the dorsal lobe as compared to the ventral lobe, highlighting this regional heterogeneity within islets (164). Furthermore, during pancreatic embryonic development, the head, body, and tail do not all develop from the same source: the dorsal pancreas develops into the upper head, body, and tail of the pancreas whereas the ventral pancreas becomes the inferior head and uncinate process of the pancreas (125). These differences in development may lead also to spatial differences between the islet function and/or architecture.

In addition to the issues that arise from imaging within a live animal, there are still considerations that need to be made due to the microscopy techniques used. The diffraction limit of light constrains the resolution achieved with IVM, and exact structures and single molecules cannot be resolved (56, 71). The depth of the tissues imaged is also a concern when using IVM, as not all β cells are located at the surface of the tissue (168). This small field of view with a limited number of β cells may not represent the dynamic communication across cells in the entire islet. Although the mitochondria can be labeled with probes targeting high metabolic activity, Reissaus et al. reported that structures deeper in the cell with fewer targets for labeling, including the ER, currently cannot be examined using IVM (139). Therefore, although this technique allows for some organelles to be visualized in real-time in vivo, the resolution, field of view, and the number of organelles visualized remain limited. To more clearly visualize deeper, more complex tissues and organs within the animal, multiphoton microscopy, rather than classical confocal microscopy, could be utilized (75, 123, 181). Additionally, super-resolution microscopy methods have been developed to overcome limitations related to the diffraction limit of light, thus allowing for sub-nanometer resolution in the live cell.

Bridging the gap between isolated islets and IVM

While IVM offers the advantage of imaging islets in their native environment, it still has its own set of disadvantages (i.e., bias, lack of depth, expensive and elaborate microscope set-ups). Therefore, there may be cases when isolated islets may be more beneficial to use. Further, it is imperative to consider what type of islet cells (i.e., dispersed islets vs. whole islets) are being utilized for imaging and where these islets are located (i.e., in vitro vs. in vivo vs. in situ) as these differences can greatly impact imaging results. It was originally hypothesized that all β cells within an islet behave as a homogenous unit to release insulin; however, it has now been shown that there is significant functional β cell heterogeneity, with specialized β cells known as “hub cells” that are responsible for coordinating insulin release across all β cells within the islet (13, 82, 135, 144, 174, 191). Furthermore, β cells that remain in contact with other α or β cells have more robust glucose-sensing abilities and a higher capacity for insulin release (185). Therefore, the organization and state of the islet cell is imperative to its function. Dispersed islets, in which islets are dispersed into single cells using trypsin, is a commonly used method for imaging experiments (31). This technique allows for specific cell types within the islet (i.e., only β cells or only α cells) to be studied directly; however, when islets are dispersed into single cells, they lose Ca²⁺ oscillatory behavior and therefore insulin release patterns are different than what is seen in the in vivo environment of the pancreas (118). Isolation of whole, undispersed islets preserves structure and intracellular communication within the islets, but the isolation procedure removes systemic inputs and cell-to-cell interactions with the surrounding pancreas tissue that are critical for in vivo functioning, thus altering the true architecture and function of the β cell (8, 157). The development of the pancreatic tissue-slice technique by Speier and Rupnik allows for visualization of both the endocrine and exocrine pancreas without disrupting the gross morphology and spatial relationships within the pancreas. Furthermore, pancreas slice technology allows the local ganglia that are associated with islets to remain intact, and electrophysiological signals can be analyzed (159). Using this technique, where the pancreas is injected with agarose, collected, and sliced into 130-μm-thick sections using a vibratome, Postić et al. demonstrated that β cells have a biphasic Ca²⁺ response during glucose stimulation, with an initial sharp rise in Ca²⁺ concentrations followed by a prolonged plateau phase, demonstrating calcium-induced calcium release-like behavior (137). However, although the pancreatic slice technique allows the islets to be studied in situ, with conditions closer to those seen in the body as compared to isolated or dispersed islets, the tissue has still been removed from the physiological environment.

Correlative light and electron microscopy: methods and application in β cell biology

As mentioned earlier, a major disadvantage of EM is its inability to accurately localize molecules to definitively identify structures and proteins of interest. CLEM combines the 3D high resolution seen in EM with the advantages of fluorescence microscopy to enable protein localization with sub-nanometer resolution (36). This technique can be performed in two ways: (i) samples are imaged first using fluorescence microscopy and then are subsequently fixed and imaged using any of the above EM techniques, or (ii) sections are fluorescently labeled using probes compatible with both fluorescence and electron microscopes, sliced into ultrathin sections, fixed and prepared for EM, and lastly imaged with both fluorescence microscopy and EM (36). After sample preparation and imaging, the regions of interest must be matched between both imaging systems to ensure the images acquired through the systems can be precisely overlaid. Specific sample holders have been created that are compatible with both microscope types and may include a finder grid which allows for accurate alignment of the sample or navigation markers that are recognized by microscope software for automatic region of interest detection (36, 147, 154, 160, 171). However, for high-precision overlay of the fluorescent and EM images, fiducial markers, which are integrated into the sample, are necessary. Examples of fiducial markers include nanoparticles, quantum dots, and polymer beads (36). While CLEM has not yet been leveraged in the β cell to explore structure-function relationships, this technique has been utilized to better differentiate the exocrine and endocrine pancreas as well as differentiate between cell types (β, α, and δ cells). For example, De Boer et al. immunolabeled paraffin-embedded rat pancreas sections with insulin and a nuclear counterstain, imaged the slices with fluorescence microscopy followed by SEM imaging, and then overlaid the images to localize the β cells. Based on the colocalization of the insulin stain with the insulin granules, they were able to clearly identify the β cells in the endocrine pancreas (36). Similarly, Saitoh et al. immunostained human pancreas sections with insulin, glucagon, and the regenerating islet-derived gene 1 α (REG1α) to determine endocrine hormone localization within the islet. Following fluorescence imaging, they imaged the sections with serial block face SEM and overlaid the images to accurately identify the β cells and α cells in the endocrine pancreas through the colocalization of insulin and glucagon in the EM image (143).

CLEM is a promising application to study the structure-function relationship in β cells as it takes the advantage of two different imaging techniques—the subnanometer resolution capabilities of EM and the protein localization abilities of fluorescence microscopy. Thus, CLEM addresses one of the major drawbacks of EM techniques as a whole: lack of protein localization. However, one of the major challenges when utilizing this technique is the limited number of probes that are compatible with both fluorescence microscopy and EM. However, new probes are constantly being developed, including genetically and nongenetically encoded probes for preembedding labeling through photo-oxidation, peroxidases, and metal tagging or postembedding labeling through the Tokuyasu method (36). Further, high-precision overlay of the fluorescence image and the EM image is critical to accurately identify protein localization within the EM image; however, overlay precision is limited by distortions that may be introduced in the fixation steps for EM imaging (36). As mentioned above, fiducial markers integrated into the samples during preparation can offer great assistance in image overlay and provide an accuracy greater than 0.5 μm (36). Lastly, the resolution of the fluorescence image is limited by the resolution of the specific technique used. To overcome this issue and maintain nanometer resolution in the fluorescence image to be overlaid with the EM image, super-resolution nanoscopy should be utilized.

Super-resolution imaging: methods and application in β cell biology

To overcome the diffraction limit of light, super-resolution microscopy has been developed utilizing two different strategies. Current super-resolution techniques are summarized below in Table 2. The first strategy is illumination-based super-resolution microscopy, which uses nonlinear optical approaches to reduce focal spot size, allowing for the resolution of finer structural details such as the shape of the inner mitochondrial membranes (130, 151). Examples of illumination-based super-resolution microscopy include stimulated emission depletion (STED) fluorescence microscopy (15) and saturated structures illumination microscopy (SSIM) (66). While all methods of illumination-based super-resolution techniques are based on nonlinear optical approaches, each approach is slightly different. For instance, STED microscopy utilizes two different laser beams: an excitation laser beam, similar to that used in confocal microscopy, and a second hollow donut-shaped STED beam. When the two beams are used together, there is a zone in which the beams will overlap and the STED beam will be depleted of almost all of the fluorophores located laterally before fluorescence occurs to allow fluorescence from only a sub-diffraction-sized central focus spot to be collected. This technique reduces the focal spot size and increases resolution (70, 72, 81, 85, 180). In contrast, SSIM utilizes a specified pattern of illumination (i.e., stripes or hexagonal patterns) for excitation, and the emitted fluorescence is collected for each focus plane at variable positions and orientations. Because of the varying patterns of excitation light, the sample interacts with the excitation pattern in unique ways that cause moiré fringes to form, leading to the collection of high frequency information at lower spatial frequencies. Fourier transforms then take advantage of these different frequencies to computationally separate the low and high frequency information. Image reconstruction subsequently occurs with two times higher the resolution in all three dimensions (81, 180).

Table 2.

Comparison of Super-resolution Techniques

Super-resolution strategy	Methodology	Advantages	Disadvantages
Stimulated emission depletion fluorescence microscopy (70, 72, 81, 85, 180)	Two laser beams are used to create a zone of overlap where the STED beam depletes almost all of the fluorophores located laterally to allow for only a sub-diffraction-sized central focus spot to be collected	Reasonably fast imaging speed, high resolution (80 nm—XY, 50 nm—Z)	Because many fluorophores are depleted by the depletion laser, there are a limited number of photons captured by the detector, photobleaching may occur due to the intense quenching beam
Saturated structures illumination microscopy (81, 180)	Uses a specific pattern of illumination for excitation. This allows the sample to interact with the excitation pattern in unique ways leading to the collection of high frequency information at lower spatial frequencies. Fourier transformation is used to separate the low and high frequencies for image reconstruction	Special fluorophores are not needed	High light intensities needed for saturation lead to photobleaching and tissue damage, very time intensive, lower resolution compared to other techniques (90–140 nm—XY)
Single-molecule localization microscopy (130)	Stochastic fluorescence switches on individual photoactivatable molecules. Molecules are bleached and the process is repeated for all molecules to enable temporal separation	Able to visualize hundreds of molecules simultaneously, variety of different types of fluorophores can be used, high lateral resolution (10–20 nm—XY)	Slow, requires specific fluorophores and buffer conditions, limited penetration depth, lower Z resolution
Photoactivatable localization microscopy (17, 73)	Uses the same photoswitchable principles as seen in SMLM, but uses a TIRF microscope for image acquisition. Genetically encoded photoswitchable protein fluorophores are used so that the protein of interest is endogenously fluorescent	Maximum lateral resolution (>50 nm—XY, 30 nm—Z)	Lower resolution power because endogenous fluorescent proteins are typically not as bright as organic fluorophores, endogenous fluorophores are necessary for imaging (inconvenient), phototoxicity seen with multiple imaging cycles, very time intensive
Fluorescence photoactivated localization microscopy (17, 73)	Uses the same photoswitchable principles as seen in SMLM, but uses a confocal microscope for image acquisition. Genetically encoded photoswitchable protein fluorophores are used so that the protein of interest is endogenously fluorescent	Maximum lateral resolution (>50 nm—XY, 30 nm—Z)	Lower resolution power because endogenous fluorescent proteins are typically not as bright as organic fluorophores, endogenous fluorophores are necessary for imaging (inconvenient), phototoxicity seen with multiple imaging cycles, very time intensive
Stochastic optical reconstruction microscopy (142, 170)	Uses the same photoswitchable principles as seen in SMLM with a photoswitchable dye or fluorophore that stochastically turns on and off	Maximum lateral resolution (>50 nm—XY, 30 nm—Z)	Phototoxicity seen with multiple imaging cycles, very time intensive
Expansion microscopy (178)	The specimen of interest is expanded and magnified through the use of a hydrogel to allow for increased space between molecules so that molecules in a diffraction-limited region can be separated and imaged with a classical confocal microscope	Compatible with classic confocal microscopes (does not require specific, expensive, high-resolution microscopes)	Sample preparation is long and complicated. If any of the steps are completed incorrectly, the sample may not stain accurately, it may break or expand unevenly, leading to a distorted, and inaccurate image
Whole-cell 4Pi single-molecule switching nanoscopy (79)	Uses the same photoswitchable principles as seen in SMLM with a photoswitchable dye or fluorophore that stochastically turns on and off. The microscope set up has some alterations in the geometry of the optical design described by Aquino et al. to improve resolution in the z-plane	High resolution in all three dimensions (10–20 nm—XY, 10–20 nm—Z)	Time-intensive imaging, limited labeling density, a very specialized microscope constructed in-house is required, limiting its widespread availability

The second super-resolution microscopy strategy is known as single-molecule localization microscopy (SMLM), which utilizes photoswitchable molecules to resolve molecules that may be too close together to differentiate otherwise. Stochastic fluorescence activation is used to “switch on” individual photoactivatable molecules, and an image is taken. The molecules are then bleached, allowing for the molecules to be temporally separated. This process is repeated for all the single-labeled molecules, which are then merged into one image (130). Examples of photoswitchable-based super-resolution microscopy are photoactivatable localization microscopy (PALM) (15), fluorescence photoactivated localization microscopy (FPALM) (73), and stochastic optical reconstruction microscopy (STORM) (141). PALM and STORM utilize the same photoswitchable principles described above for STED and SSIM; however, these two techniques differ in the types of labeling molecules that are employed and how their emission is modulated. PALM (which uses a TIRF microscope, detailed above in the section on multiphoton microscopy techniques) and FPALM (which uses a confocal microscope) utilize genetically encoded photoswitchable protein fluorophores so that the protein of interest is endogenously fluorescent. A small subset of these molecules is activated by an excitation laser, collected, and then bleached with a higher-powered excitation laser so they do not become excited again. This process is repeated until all localizations are merged into one fully formed reconstructed image (17, 73). On the other hand, STORM uses a photoswitchable dye or fluorophores that are able to stochastically turn their emissions on and off. Rather than being endogenously expressed, as in PALM/FPALM, the dye is attached to the protein of interest through antibody interactions. This technique allows for a much higher quantity of photons to be emitted each cycle, thus enabling a higher-resolution image to be reconstructed (142, 170). When an image is acquired through any of the above methods, not only is the object or protein of interest visualized, but the noise of the system is also included in the final reconstruction. Therefore, to remove the noise, a process known as deconvolution is necessary to remove the system noise from the object of interest by enhancing the contrast from the proteins to increase the overall resolution of the image (92).

Expansion microscopy (ExM) also has the ability to overcome the diffraction limit of light and improve the resolution of images; however, unlike super-resolution, it accomplishes this by magnifying the specimen of interest. This expansion results in increased space between the molecules of interest. Therefore, molecules within a diffraction-limited region can be separated so that each molecule can be individually resolved by conventional diffraction-limited microscopes, such as confocal microscopes (178). All protocols for ExM follow a similar workflow. First, molecular handles are attached to specific labels within the specimen so that a swellable hydrogel can bind to the specimen. Next, the specimen is immersed in a monomer solution that undergoes free-radical polymerization. This process results in a densely cross-linked and highly penetrating polyelectrolyte hydrogel that surrounds the labels, which allows for binding of the molecular handles so that the labels are now coupled to the newly formed hydrogel. Following this step, the specimen is homogenized by chemical denaturation either through heat and detergent treatment or enzymatic digestion. Lastly, the specimen is immersed in water, which enters the hydrogel through diffusion and causes the gel and specimen to swell, enabling molecules within the specimen to expand and separate enough so that there are no longer any regions of the specimen that are diffraction-limited (178).

There are a limited number of published studies that utilize super-resolution microscopy to image the β cell (7, 42, 69, 134) in part because it was previously challenging to culture primary islet cells on the surface of the glass coverslips that are required for sub-cellular imaging (19, 67, 128). However, in 2017, following the observation that primary β cells adhere well to glass coverslips when co-cultured with primary neurons, Phelps et al. identified neural cell-secreted basement membrane as a factor promoting β cell adhesion. Subsequently, they developed a method for preparing primary islet cell monolayers on glass surfaces coated with purified laminin or collagen IV, thus allowing STED microscopy to be used to image β cells. In this study, the authors concluded that the resolution achieved with STED microscopy enabled the measurement of insulin granules, microtubules, and actin filaments, and they found that the measurements derived from this technique were consistent with those derived from TEM (134). Therefore, STED microscopy integrates the resolution of EM with the live cell and localization capabilities of fluorescence microscopy to allow β cell dynamics to be monitored in real time to establish the relationship between ultrastructural architecture and function. Following this study, a handful of other groups have applied super-resolution microscopy to investigate the β cell. For instance, in contrast to the mitochondrial “structures” determined by Gerencser et al. using classical confocal microscopy described above, Liu et al. leveraged the nanoscale resolution capabilities of STED nanoscopy to resolve the spacing of the mitochondrial cristae in primary islet tissue through the use of the mitochondrial cristae probe PK Mito Orange (PKMO) (99).

In addition to characterizing how organelle ultrastructural architecture is related to cellular function, others have utilized the single molecule localization capabilities of this technique to observe the trafficking patterns of key signaling molecules and receptors within the β cell. For instance, Olaniru et al. utilized a SNAP-tag in conjunction with live cell imaging via instant structured illumination microscopy (iSIM) to determine the trafficking patterns of the G-protein coupled receptor 56 (GPR56), which is known to increase insulin secretion upon its activation by type III collagen. Upon activation of GPR56 in MIN6 cells, the authors found that GPR56 localized initially to the plasma membrane. However, over the time course of about 20 min, GPR56 constitutively internalizes through recycling endosomes and can also be localized to the cytoplasm, illustrating the trafficking pattern of this molecule (122). Similarly, Wilhelmi et al. studied the role of the guanosine-5′-triphosphate (GTP)-ase ARFRP1 in insulin secretion using structured illumination microscopy (SIM) in INS-1 cells. First, they used a luminescence-based mammalian interactome assay to identify that ARFRP1 interacts with the Golgi-associated PDZ and coiled-coil motif-containing protein (GOPC). To confirm this interaction visually, they labeled both ARFRP1 and GOPC in INS-1 cells, treated cells with basal glucose (2.8 mM) or high glucose levels (20 mM) to stimulate insulin secretion, and imaged cells utilizing SIM. Under both conditions, ARFRP1 and GOPC remained colocalized, suggesting that ARFP1 and the Golgi scaffolding protein GOPC may interact to play a critical role in insulin secretion (183). However, a major limitation of the above studies is the small depth of the z-axis of the β cell that can be imaged.

Although techniques such as STED, PALM, and STORM can achieve 20- to 40-nm resolution in the x- and y-planes, the resolution in the z-plane is limited to about 50 to 80 nm (79). However, with the use of a dual-objective “4Pi” detection geometry, more specifically whole-cell 4Pi single-molecule switching nanoscopy (W-4PiSMSN), the resolution can be further improved in the depth direction to be comparable to that of the x- and y-planes (16). Huang et al. developed a W-4PiSMSN system that expanded upon the previous iPALM and 4Pi-SMSN developments (4, 155) to include key additions that allow for 10- to 20-nm 3D resolution across the thickness of the entire mammalian cell. Huang et al. expanded upon the optical design described by Aquino et al. and included deformable mirrors in both arms of the interferometric cavity to correct for imperfections in the instrument beam path, optimize the point spread function quality for samples with different thicknesses, compensate for ample-induced aberration modes, and introduce astigmatism in both arms without adding further complexity to the system (79).

Although the W-4PiSMSN system has not yet been used to image organelles in the β cell, this technique has successfully been used to resolve the structure of the ER in COS-7 cells, including the 3D membrane contour that was previously only resolvable with ET as well as microtubules, bacteriophages, mitochondria, nuclear pore complexes, coat protein complex I (COPI)-coated vesicles, primary cilium, and spermatocytes (79). This technique allows for the same resolution seen with EM in all three dimensions without the elaborate and destructive fixation process, and it provides the ability to localize fluorescently labeled single molecules. Our group has begun to image the ultrastructural architecture of the ER in INS-1 cells transfected with a Sec61β-GFP plasmid and using a GFP-Booster Alexa Fluor 647 nanobody (Figure 2D). Here, the precise meshwork of the ER can be visualized due to the nanometer resolution and localization capabilities of the technique. This approach has the potential to illustrate structure-function relationships in β cells due to its nanometer resolution, localization capabilities, and live cell imaging potential. However, there are some drawbacks to consider. W-4PiSMSN systems require time-intensive imaging and offer limited labeling density. Like FIB-SEM, this technique requires a very specialized microscope constructed in-house, limiting its widespread availability. Furthermore, the number of structures that can be labeled is limited due to spectral overlap, leading to a lack of cellular context, as all the structures in the cell cannot be visualized at once, as is possible using EM. A potential solution to this issue would be to utilize multiple cellular replicates when imaging. One cellular sample could first be imaged using the W-4PiSMSN technique and, because the buffer used during imaging renders the sample unable to be used again, an additional cellular replicate could then be fixed for EM imaging to provide additional cellular context.

Soft X-ray Tomography in the Characterization of β Cell Subcellular Organelle Structures

SXT is an imaging technique that has recently been applied to the β cell to better understand the mesoscale topology rearrangements at a resolution of the whole cell. Rather than using fluorescent probes or electrons as a source for image contrast, SXT uses the inherent contrast of organic material within a sample (182). Samples are cryo-fixed using the fixation methods used for cryo-EM to preserve the near-native structure of cells (182). Cells are then imaged within the water window (284–583 eV photon energy), where carbon-rich material has a higher linear absorption coefficient than the surrounding background (100), allowing intracellular structures to be imaged with a spatial resolution of 35 to 60 nm (111, 153). After fixation, image collection takes 10 min to complete, allowing many cells to be imaged under various experimental conditions (100, 126).

Two key published studies have applied this technique to β cells. In 2020, White et al. utilized SXT to characterize the subcellular rearrangements in a rat insulinoma cell line before and after glucose stimulation and found that glucose increased the density of insulin packaging, increased mitochondrial volume, and led to the closer proximity of the insulin vesicles to the mitochondria (182). Loconte et al. stimulated a rat insulinoma cell line with glucose but instead focused on the cellular rearrangements during insulin secretion. They showed that the interactions between the insulin vesicles and the mitochondria occur near the plasma membrane during the first and second phase of insulin secretion (100). Both studies found that SXT could be utilized to rapidly map organelles, ultrastructural reorganization, and inter-organelle interactions in the β cell (100, 182). However, even though these ultrastructural changes can be visualized using this method, there is a lack of detail at the molecular level since the resolution is only 35 to 60 nm, and molecules of interest cannot be fluorescently labeled for localization (100, 111, 153). Similar to potential solutions listed throughout this Overview Article, multiple cellular replicates should be prepared to enable one cellular sample to be fixed and imaged using SXT, while a second cellular replicate can be fluorescently labeled to define localization within the same experimental conditions. Moreover, because there is not yet an automatic organelle identification and annotation pipeline, the segmentation process follows a manual method, which is very time consuming (100).

Conclusions

Advancements in technology have allowed for the development and utilization of several different imaging techniques to visualize the ultrastructural architecture of the β cell and to relate changes in structure to cell function. Each methodology has its own advantages, disadvantages, and applications (Table 3).

Table 3.

Summary of Techniques to Image the Organelles in the β Cell

Imaging modality	Advantages	Disadvantages	Solutions
Electron microscopy	Sub-nanometer resolution, three-dimensional reconstruction (with tomography)	Fixation protocol may lead to artifacts, lack of localization, expertise needed for sample preparation, increased time for image reconstruction leading to small sample size, expensive, no live cell data	Fix samples for EM following fluorescence live cell imaging (CLEM), utilize immunogold labeling for some localization capabilities, fix tissues or samples at specified time points, utilize a cohort of genetically homogenous mice, and sacrifice mice at specific time points to examine how structures change over time
Fluorescence microscopy	Localization of molecules of interest, nanometer resolution (with super-resolution techniques), three-dimensional data, live cell imaging possible	Spectral overlap limits the number of targets able to be visualized, photobleaching, diffraction limit of light (with confocal microscopy), and labeling may affect cellular processes	For simultaneous imaging of fluorophores, different methods can be used: white light lasers, acousto-optical tunable filters, bandpass filters fluorescence lifetime, spectral imaging coupled to mathematically linear unmixing
			Super-resolution and expansion microscopy can be used to overcome the diffraction limit of light
Soft X-ray tomography	Relies on the inherent contrast of organic material within a sample for imaging, large number of cells can be imaged due to fast imaging time (10 min)	Resolution of only 35 to 60 nm, lack of localization, time consuming segmentation process, no live cell data	Cellular replicates can be prepared at multiple timepoints to examine how structures change over time, cellular fixation for SXT can occur after fluorescence live cell imaging
Intravital microscopy	Live cell imaging, do not need to remove β cells from physiological environment of the pancreas	Diffraction limit of light, limited depth of tissue, limited number of structures that can be imaged, labeling may affect cellular processes	Multiphoton microscopy can be used in conjunction with IVM rather than confocal microscopy to visualize deeper structures within the tissues, super-resolution may be used to achieve higher resolution
			For simultaneous imaging of fluorophores, different methods can be used: white light lasers, acousto-optical tunable filters, bandpass filters fluorescence lifetime, spectral imaging coupled to mathematically linear unmixing

Although the development of techniques such as ET allow for 3D reconstructions of the β cell on a nanometer scale (48) and cryo-EM allows cells to be imaged in a near-native state (38, 102, 106, 167), single molecules are not able to be localized due to fixation processes and the inability to label molecules of interest. Fluorescence microscopy techniques that are possible using confocal microscopy allow for labeling of targets of interest and construction of 3D live cell images and advancements in multiphoton microscopy and TIRF enable key insights into the dynamics of insulin granules; however, these fluorescent techniques remove β cells from their physiologic environment of the pancreas. IVM allows for β cell organelles to be visualized within the native pancreas environment to understand physiology in an integrated manner, but the resolution is still limited due to the diffraction limit of light, as is also seen in confocal microscopy (145). CLEM is a promising application to study the structure-function relationship in β cells as it combines the advantages of EM and fluorescence imaging; however, it is limited by the resolution of the specific fluorescent technique chosen. Super-resolution microscopy techniques have overcome the diffraction limit of light that restricts many of the other fluorescence principles by temporally separating single molecules (130); however, while these techniques can achieve 20- to 40-nm resolution in the lateral dimensions, the resolution in the z-direction is about 50 to 80 nm (79). Thus, although these super-resolution techniques have been applied in the β cell, precise 3D data is unavailable due to the lack of resolution in the z-direction (134). W-4PiSMSN is beginning to be used to image organelles in the β cell and is able to achieve 10- to 20-nm 3D resolution across the thickness of mammalian cells (79). Implementation of this new cellular imaging technique in the context of the β cell has the potential to relate nanoscale ultrastructural alterations with changes in function, thereby highlighting a potential new therapeutic target for T2D: organelle health.

Didactic Synopsis.

Major teaching points

• T2D arises from inadequate insulin secretion from the pancreatic β cell against a background of preexisting insulin resistance. Electron microscopy, fluorescence microscopy, and soft X-ray tomography are techniques used to characterize the structure and function of the intracellular organelles that regulate insulin secretion.

• Electron microscopy allows for sub-nanometer resolution and three-dimensional reconstruction but does not allow for live cell imaging or the localization of individual molecules.

• Fluorescence microscopy allows for molecule localization, live cell imaging, three-dimensional reconstruction, and imaging at nanometer resolution, when combined with super-resolution techniques. However, a restricted number of targets can be imaged simultaneously due to the spectral overlap between fluorophores.

• Soft X-ray tomography preserves physiologic processes by relying on the inherent contrast of organic material within a sample but cannot generate live cell data and has limited resolution.

• Intravital microscopy allows for live cell imaging within the physiological tissue environment but has limited resolution.