Finding alternatives for insulin therapy and making advances in etiology of type 1 diabetes benefits from a full structural and functional insight into Islets of Langerhans. Electron microscopy (EM) can visualize Islet morphology at the highest possible resolution, however, conventional EM only provides biased snapshots and lacks context. We developed and employed large scale EM and compiled a resource of complete cross sections of rat Islets during immuno-destruction to provide unbiased structural insight of thousands of cells at macromolecular resolution. The resource includes six datasets, totalling 25.000 micrographs, annotated for cellular and ultrastructural changes during autoimmune diabetes. Granulocytes are attracted to the endocrine tissue, followed by extravasation of a pleiotrophy of leukocytes. Subcellullar changes in beta cells include endoplasmic reticulum stress, insulin degranulation and glycogen accumulation. Rare findings include erythrocyte extravasation and nuclear actin-like fibers. While we focus on a rat model of autoimmune diabetes, our approach is general applicable.
Diabetes Mellitus is a life-threatening disease, and its incidence continues to increase, now affecting over 8% of mankind1. The two most prevailing forms are type 2 diabetes, which is caused by insulin-resistance combined with a relative deficiency in insulin, and type 1 diabetes. In the latter, insulin-producing beta cells are destroyed by an autoimmune attack. Upon diagnosis, life-long exogenous insulin therapy is immediately initiated2. Fundamental questions in type 1 diabetes remain: What are the triggers? Can it be prevented? Can we diagnose and immune suppress at-onset or near-onset diabetes? Can patients be cured? To address these questions, microscopy is often pivotal, for example to analyze transplanted Islets3,4, or to assess the success of creating insulin-producing cells5. Soon after Gray published the macroscopic anatomy of the pancreas6, a seminal microscopic observation was made by Langerhans7: “cells are small and have an irregular polygonal formation with a completely homogenous content, gleamy and free from granules. Their nucleus is bright, round and middle-sized. The cells are 0.0096 to 0.012 mm, the nuclei 0.0075 to 0.008 mm in diameter. Mostly, several cells are in close proximity, distributed peculiarly in the Parenchym of the gland”. These clustered cells are now known as Islets of Langerhans (Islets), consisting primarily of beta cells that are destroyed in type 1 diabetes8. Although Langerhans' description was accurate, sharing microscopic images is invaluable. Unfortunately, electron microscopists need to make stringent subjective selections prior to publishing high-resolution snapshots of a small field of view. We developed9 and implemented large-scale EM to visualize the nano-anotomy (nanotomy) of Islets at different stages of diabetes. The data are presented in a way that allows unbiased analysis by a broad (scientific) community.
The nanotomy technique
A prerequisite for nanotomy is complete sampling of tissues, preventing that samples become obscured during preparation (Fig. 1a). Since Islets constitute only ~1% of the pancreas, light microscopy is needed to guide Islets section preparation on special carriers (Fig. 1a–e). Following EM quality control, automated EM-acquisition and stitching is applied (See Methods). Nanotomy thus generates large-scale (Fig. 1f) high-resolution maps allowing inspection of macromolecules anywhere in the dataset (Fig. 1g–k).
An overview of tissue, cells, organelles and macromolecules
Six datasets at different stages of diabetes progression are presented (Fig. 2). Several features have been annotated and selected cells have been false colored to indicate different cell types, such as insulin-producing beta cells (green) and glucagon-producing alpha cells (yellow). Figure 2 shows a low-magnification summary of the data: the full extent of our nanotomy data can only be appreciated online (www.nanotomy.nl). Prior to focusing on the effect of type 1 diabetes, an online tour on ultrastructure interpretation is given for the reader less experienced with EM-analysis, starting with dataset I (diabetic resistant); group 2 (cell types); letter A (acinar cell), or I-2A in short. Acinar cells are part of the exocrine pancreas surrounding the Islet, which also contains the characteristic duct cells (I-2E). The endocrine pancreas contains alpha cells (I-2B), beta cells (I-2C), PP1 (see below) and delta cells (I-2D), as well as multiple capillaries harboring erythrocytes (I-2G) and leukocytes (I-2F). Our large-scale approach significantly increases the chances to observe less-abundant cells, such as eosinophil granulocytes (IV-2I) and basophil granulocytes (IV-2J). Nanotomy allows for sub-cellular analysis of organelles, granules and supramacromolecules, which are indicative for cell identity and function. Nuclei, for instance, are absent in erythrocytes (V-2H), occupy most of the volume in lymphocytes (V-2G) and are horseshoe-like in monocytes (V-2I). Mitotic cells are distinguished by the absence of a nucleus (V-10A) and the condensation of chromatin (V-10E). A shrunk structure-less nucleus is characteristic for pyknotic cells (IV-2H). The Golgi apparatus (V-3D) and rough endoplasmic reticulum (ER; V-3A) are abundant in the secretory cells. Zooming in and out aids to identify the substance of secretory granules: zymogen (V-4A), insulin (V-4B), glucagon (V-4F), somatostatin (V-4G) and pancreatic polypeptide (PP). Several macromolecular complexes can be identified, such as nuclear pores (V-7A side view; V-7B top view; Fig. 1k); polysomes (V-7C); ribosomes (V-7D) and cell-cell junctions, including gap junctions, tight junctions (V-7F) and desmosomes (V-7E; Fig. 1l). Rare macromolecular structures include fibers in a nucleus (V-10F; Fig. 1l), possibly the debated filamentous nuclear actin10. Curled membranes within insulin granules are seen (V-4E), which will likely become exosomes upon insulin secretion. These datasets provide a detailed and interactive view of cellular ultrastructure in Islets of Langerhans.
Islets under attack
To analyze the onset and progression of diabetes with nanotomy, we analyzed Islets in diabetes-prone (DP) Bio Breeding (BB) rats, in which blood glucose (BG) levels are indicative of the (pre)diabetic state. The DP animals are lymphopenic: the shortage of regulatory T cells leads to a spontaneous autoimmune attack of the beta cells in most animals with approximately half of the population affected around 80 days11. In control diabetes resistant (DR) BB rats, beta cells look normal with abundant presence of insulin granules and normal nuclei (I-2C). Signs of leukocytes extravasating the endocrine or exocrine capillaries are absent, though leukocytes are found inside the vasculature (I-2F). The Islet of the DP rat with the lowest BG tested (7.2 mM; II) shows morphological characteristic comparable to the control. Numerous apoptotic beta cells are observed in the 9.9 mM data set (III): these cells contain abundant insulin but have characteristic swollen vesicles, dark cytoplasm and condensed chromatin (III-2H). At this stage, no sign of insulitis12 is present, but an apparent increased leukocyte to erythrocyte ratio in the vasculature compared to DR rats hints at an immune activation. At the periphery of the Islet extravasated leukocytes are present. The Islet of the rat with 12 mM BG (IV) shows extensive insulitis and morphological changes. Healthy beta cells with abundant insulin content as well as exhausted beta cells with empty granules are present. More beta cells are being phagocytosed at this stage of diabetes. In a more advanced stage (23 mM; V), massive endocrine extravasation is present: the Islet is filled with a pleiotrophy of leukocytes. Most beta cells show compromised morphology and are being attacked (V-9A,B), phagocytosed and digested (V-9C to G). Lastly, we show a cross-section of an Islet at the end stage of type 1 diabetes (VI). The majority of this section consists of alpha cells (VI-2B), while only some leukocytes are present (VI-2E). Intact beta cells are no longer detectable within this dataset. Our observations expand on earlier reports of insulitis using mainly light microscopic techniques13,14,15,16,17. First apoptotic cells are present in the pre-diabetic state. At a later stage, a pleiotrophy of leukocytes extravasate, resulting in massive insulitis, possibly attracted by chemokines and macrophage activation. In our rat model, leukocyte extravasation seems to originate directly from the endocrine tissue. Intravital imaging of insulitis induced by viral infection in mice18 showed exocrine leukocyte extravasation followed by random walk migration. The distinct mode of insulitis might differ between species or animal model. At the late stage of Islet destruction, leukocytes are no longer needed and most have left the Islet.
Subcellular changes during type 1 diabetes
Besides cellular information, nanotomy allows analyzing macromolecular and organelle features throughout the datasets. The endocrine tissue has a dense network of micro-vasculature, which is surrounded by a basement membrane. In the diabetic animals, several extravasated leukocytes seem to be surrounded by an additional basement membrane, particular when they are in the vicinity of the vasculature (V-6E). In humans, a double basement membrane is present around the Islets19. Our results suggest that this is also the case in rats: leukocytes are temporarily caught within these two basement membranes during extravasation. Interestingly, multi-layered vascular basement membranes have also been observed in the context of other diseases20.
The endoplasmic reticulum (ER) in beta cells (Fig. 3; V-10J) of the diabetic Islets becomes perturbed: their dilation is a sign of ER stress. The ultrastructure of mitochondria is also affected. Cisternae are clearly distinguishable in normal mitochondria (Fig. 3d; I-3E), whereas mitochondria are rearranged in beta cells of the diabetic Islets (Fig. 3g; V-10I), and, in a final stage, become hard to recognize. Fig. 3 shows a comparison of a healthy versus a high BG beta cell. Note that the insulin granules are hardly present in the latter (V-10G): the large demand for insulin for the high BG animal exhausted these granules.
Particles accumulate in beta cells during diabetes progression
Another striking feature in beta cells of diabetic animals is the accumulation of clusters of darkly stained particles, measuring ~35 nm in size (Fig. 3f; V-10H), and often associated with membranes (Fig. 4a; V-10B). For the non-diabetic animals, these particles are found associated with the vasculature (I-3A) and ducts (I-3B), whereas for the diabetic animals isolated spots are found in leukocytes (V-10D) and clustered spots with the beta cells (Fig. 3f; V-10B;). Only at the final stage of diabetes, when most beta cells are destroyed, the particles are also seen in other endocrine cells. Viruses, in particular enteroviruses, have been named as potential triggers for type 1 diabetes12,21, and in man similar particles were presented as viruses21. BB rats were found to be infected with Ljungan virus22, however, our anti-Ljungan virus immunolabeling experiments on these particles were either negative or non-conclusive (data not shown). 3D reconstruction of one of the particle-membrane clusters (Fig. 4b) shows that these particles vary in size and do not appear polyhedral, arguing against the possibility that these may be viruses. Presence of glycogen in beta cells, apparent as particles of 10–30 nm, has been linked to diabetes previously23,24. The degree of glycogenic infiltration varies from one beta cell to the next and an inverse relationship between the degree of granulation and the extent of glycogen deposition has been reported25. The particles in our samples must be glycogen as concluded from positive silver staining against polysaccharides (Fig. 4). Additional physiological studies are needed to address whether glycogen accumulation, a probable consequence of elevated blood glucose levels, plays a role in diabetes development.
Nanotomy provides a wealth of ultrastructural information online and can be applied to all kinds of tissues and diseases. Our analysis of Islets of Langerhans during autoimmune diabetes onset in a rat model reveals, among others, glycogen particles associated with diabetes progression. Other rare features are mitotic cells (V-10A), nuclear fibers that resemble actin (V-10F) and an extravasated erythrocyte (II-3A); many more unique events wait to be discovered, highlighted and interpreted within nanotomy data, aided by open access analysis. Viruses, implicated in triggering type 1 diabetes in man21, present within Islets should be detectable by our technique. The fact that we could not detect viruses argues against the debated endemic presence of Ljungan picornavirus in BB rats26. While the datasets contain information on thousands of cells and a multifold thereof of macromolecules, we note that presently only a single cross-section of each Islet is shown. The typical and rare events observed in the datasets should become significant by generic implementation and expansion of nanotomy, including human Islets that have a very distinct organization than rodent Islets. Moreover, large-scale EM-imaging, possibly combined with 3D approaches27,28 will allow assessing effects of gene-mutations in genetically-modified organisms, within the cellular context at a molecular scale. Annotation might not only be performed by morphological examination, but also by labeling, including correlated microscopy29. Both scientists as well as the public (patients) can interactively access our data to address what goes wrong with Islets, cells, organelles and macromolecules during type 1 diabetes onset in this rat model, which will be followed up in human Islets of Langerhans and may aid in uncovering the trigger(s) and help developing new therapies.
The rats were derived from the Worcester DP-BB and DR-BB strain, and were maintained and bred at our institutional Central Animal Facility30,31. The animals received humane care in compliance with the principles of laboratory animal care (NIH publication no. 85-23; revised 1985) and the Dutch law on experimental animal care. The university ethical board for animal studies approved all animal experiments reported in this study. The animals used in this study were fed ad libitum with food pellets conducted according to the guidelines for Ethics Committee and Animal Experimentation, University of Groningen. The DPBB rat nr 440 (BG 9.9 mM) was fed a hydrolysed casein based diet supplemented with 4% gluten. The other rats were fed a standard cereal based rodent laboratory chow (RmH-B, AB diets, the Netherlands) as described previously30,31.
Upon necroscopy, the pancreata were removed, and the tail of the pancreata, most rich in Islets (albeit poor in PP cells), was isolated. Small pieces (~3–5 mm) were cut, followed by immersion fixation in 4% paraformaldehyde (PFA) and 0.1% glutaraldehyde (GA) -that also allows immunostaining- in 0.1 M sodiumcacodylate buffer, pH 7.4 (cacodylate). Alternatively 2% GA and 0.5% PFA was used (datasets 7.2 and 23 mM). Following overnight fixation at 4°C, tissue was embedded in 4% low melting point agarose in 0.1 M cacodylate (~25°C) and directly transferred to ice. The blocks were mounted using acrylate glue. Sixty μm slices were cut from the collected blocks using a HM 650 Micron tissue slicer, fixed for 24 hours in the same fixatives mentioned above, and washed in 0.1 M cacodylate. The Islets of Langerhans and the peripheral acinar parenchyma were selected under a stereomicroscope. Next, Islets were post fixed for two hours in 1% osmiumtetroxide and 1.5% potassium ferrocyanide in 0.1 M cacodylate for 2 hrs at 4°C, rinsed three times for 30 min., dehydrated in a graded series of ethanol and, finally, flat embedded between silane-coated glass slides in epon. Semi-thin (~1 μm) and ultrathin (~60 nm) were cut on an ultramicrotome with a diamond knife (Diatome Inc., Switzerland). The semi-thin sections were stained with 1% toluidine blue (Fig. 1B) or with a mixture of the latter with 0.05% basic fuchsine (Fig. 1C). The ultrathin sections were carefully positioned on single slot M600 Nickel grids (Veco, The Netherlands), to allow acquisition uninterrupted by grid bars. The sections were supported by formvar. Ultrathin sections were stained for 2 min. in 5% uranylacetate in methanol followed by Reynolds lead citrate for another 2 min32. Samples were prescreened at low magnification in a transmission electron microscope (CM100; FEI Company, The Netherlands). The best quality samples were selected and subjected to nanotomy. Polysaccharide staining was performed following a modified Thierry protocol33.
Data acquisition and stitching
All nanotomy data, a total of 25.434 images, were collected on a FEI Tecnai Spirit BioTWIN microscope equipped with LaB6 filament, employing inhouse developed software tools. Samples were extensively pre-irradiated at low magnification prior to data collection to induce sample stabilization by sublimation. The virtual slides for rat412 (DR), rat378, (BG 6.6 mM), rat375 (BG 7.2 mM), rat440 (BG 9.9 mM), rat172 (BG 12 mM), rat367 (BG 23 mM) and rat131 (BG > 30 mM) were recorded with a magnification at the detector plane of 9460× (pixelsize 3.2 nm). Hardware binned 2 k × 2 k images were collected with a FEI Eagle CCD camera: the number of images and the required time is listed in Table 1. Every 10th image, the FEI autofocus routine was employed to maintain the sample at −1 μm defocus.
Annotation has been performed manually. We note that the colour-coding is an indication of the cell type, and might not be 100% accurate. Additional methods such as labelling should be applied in order to define all cell types unambiguously, as discussed in the main text. Atlases and text books34,35,36,37,38 have been a major source to define cell types based on morphology.
The guarantors of this paper are R.B.G.R. and B.N.G.G. We thank Ulrike Schnell (UMCG) for translation of the part of Langerhans' thesis7; Rogier van der Stijl and Hans van Wattum (UMCG) for help with annotation, Bo Niklasson (Uppsala University, Sweden) for antibodies, and Mieke Mommaas (LUMC) for histological advice. We acknowledge financial support from the Dutch Diabetes Foundation (J.T.J.V., grant 2009.80.009) and the Netherlands Organization for Scientific Research (M.C.A. and R.B.G.R., grant 016.072.321; B.N.G.G. grant 40-00506-98-9021).