Insulin is secreted from pancreatic β-cells, and the high demand of insulin biosynthesis is known to cause β-cell dysfunction in patients with type 2 diabetes mellitus. The insulin biosynthetic pathway has been extensively studied and is still an exciting area for future studies. In this chapter, first, we focus on proinsulin biosynthetic pathway in the endoplasmic reticulum (ER) and recent progress of our knowledge about ER stress. We discuss about how ER stress is involved in the development of diabetes. Second, we focus on the formation of insulin secretory granules. The biogenesis of secretory granules has been explored for several decades; however, it still has been debated and has yet to be understood. We review the current knowledge about the secretory granules and discuss about the problems for future studies.
- islets of Langerhans
- ER stress
- secretory granules
- biosynthetic pathway
Pancreatic β-cells synthesize insulin and secrete it in response to the increase of blood glucose levels. Insulin is synthesized as proinsulin in the endoplasmic reticulum (ER) and transported to the Golgi apparatus. In the trans-Golgi network (TGN), insulin becomes hexamer, and then packaged into secretory granules (Figure 1). In secretory granules, proinsulin is processed to form mature insulin and C-peptide. By the stimulation of high glucose concentration in blood, insulin granule is exocytosed and insulin and C-peptide are secreted into blood.
Pancreatic β-cells are located in islets of Langerhans. In pancreas, there are islet-like cell clusters that are stained differently from other parts of pancreatic tissues (Figure 2). It was named as islets of Langerhans, from the name of the person who found this structure. Paul Langerhans found this structure in his doctoral thesis. The cell clusters appeared to be different from the cells that secrete pancreatic enzymes, but he did not know what the function of this structure was. As he also found the cells that have dendrites in skin, his name is used for these cells as Langerhans cells, the dendritic cells .
Pancreas contains exocrine cells that secrete digestive enzymes including amylase and trypsin, and endocrine cells that secrete hormones including insulin and glucagon. The ratio of exocrine and endocrine cells is approximately 9:1, meaning that less than 10% of cells are endocrine cells located in islets of Langerhans scattered in pancreas. Islets of Langerhans contain α, β, δ, ε, and PP cells. There are a million of islets of Langerhans in pancreas in human and 200–300 in adult mouse. In islets of Langerhans, 15–20% of cells are α-cells, 75–80% are β-cells, 5% are δ-cells, 1% are ε-cells, and 4% are PP cells. α-cells secrete glucagon that raises the blood glucose levels, β-cells secrete insulin that is the only hormone to decrease the blood glucose levels, δ-cells secrete somatostatin that inhibits the secretion of insulin and glucagon, ε-cells promote appetite and secrete ghrelin that inhibits insulin secretion, and PP cells secrete pancreatic peptide whose function is yet to be understood. In rodents, islets of Langerhans have a clear mantle core structure in which β-cells are located in the center of islets of Langerhans and surrounded by α, δ, ε, and PP cells (Figure 1). When rat islets are trypsinized and maintained in culture medium, the cells reassembled into aggregates that have a similar organization of intact islets; β-cells are located in the center and the other cells are in the periphery . Human islets of Langerhans do not have such clear structures. β-cells are mixed with the other cells. In avian pancreas, α-cells are in the center of islets. In zebrafish, their pancreas shares the basic structure with mammalian pancreas . Recent studies show that zebrafish is a good model to study pancreatic development and diabetes mellitus [4, 5, 6, 7].
2. ER stress and insulin
2.1. Proinsulin translation and folding in the ER
Human has INS gene as only insulin gene, whereas rodents have INS1 and INS2 genes for insulin. Human insulin gene encodes preproinsulin that has 110 amino acids containing N-terminal signal peptide following B chain, C-peptide, and A chain. Preproinsulin mRNA translation begins in the cytosol in pancreatic β-cells, and the signal peptide is recognized by signal recognition particle (SRP) to translate proinsulin across the membrane of the ER. In the ER, signal peptide is cleaved by signal peptidase to produce proinsulin that has 86 amino acids consisting of B chain, C-peptide, and A chain (Figure 3). Proinsulin is folded in the ER by chaperones including protein disulfide isomerase (PDI) family and BiP. Molecular chaperons and PDIs bind to the hydrophobic regions of proteins to promote folding and inhibit the aggregation of proteins . Proinsulin has three disulfide bonds in A6–A11, A7–B7, and A20–B19 (Figure 3) . N-glycosylation is often used as a marker for proper folding of newly synthesized proteins in the ER; however, proinsulin does not have N-glycosylation site.
2.2. ER stress
The high demand of insulin synthesis under a high plasma glucose condition causes ER stress that could cause β-cell dysfunction. Generally, the secretory proteins and transmembrane proteins are folded and acquire a variety of modification in the ER. Environmental and genetic factors affect protein folding in the ER. If protein folding is inhibited, unfolded proteins accumulate in the ER, leading to ER stress. Cells that sense ER stress cause unfolded protein response (UPR) that includes the inhibition of general protein translation, the induction of expression of ER chaperons, and ER-associated degradation (ERAD). UPR is a cellular response to recover ER homeostasis. In mammalian cells, there are three ER stress sensors, PERK, IRE1α, and ATF6 (Figure 4).
Protein kinase RNA-like ER kinase (PERK) is type-I transmembrane protein that localizes in the ER. Under ER stress, PERK undergoes autophosphorylation to be activated and oligomerized. Oligomerized PERK phosphorylates translation initiation factor, eIF2 α-subunit to inhibit protein translation. The inhibition of protein translation attenuates the accumulation of unfolded ER proteins as well as ER stress . On the other hand, this inhibition of protein translation promotes the translation of ATF4, a transcription factor that induces genes related to apoptosis, amino acid metabolism, and antioxidants. Still, when cells cannot deal with ER stress even by these measures and ER stress is continued, ATF4 induces the transcription of C/EBP homologous proteins (CHOP/GADD153). CHOP and ATF4 form a heterodimer that induces the transcription of each downstream genes and promotes apoptosis [11, 12].
Inositol-requiring enzyme 1 (IRE1) has two isoforms: IRE1α that is expressed ubiquitously  and IRE1β that is expressed in goblet cells that secrete mucin in the digestive tract and lung [14, 15]. IRE1α is the type-I transmembrane protein localized in the ER. IRE1α has kinase and ribonuclease domains in its cytoplasmic region, and its lumenal domain has the binding site for BiP. In normal condition, IRE1α is in inactive form and binds to BiP, as well as other ER-stress sensors PERK and ATF6. In ER stress condition, BiP is released from IRE1α and IRE1α forms oligomer to be in an active form [16, 17, 18, 19, 20, 21]. Autophosphorylation of the kinase domain in cytoplasmic region of IRE1α activates the ribonuclease (RNase) domain in the C-terminal region of cytoplasmic domain of IRE1α [22, 23, 24, 25]. The activated RNase domain cleaves the precursor form of XBP1 (XBP1 unspliced; XBP1u mRNA) in specific two sites to produce mature XBP1 mRNA (XBP1 spliced; XBP1s mRNA) that produces functional transcription factor to induce the transcription of genes related to ER chaperones, ERAD, and lipid metabolism to recover ER homeostasis [26, 27].
Activating transcription factor 6 (ATF6) is a type-II transmembrane protein functioning as a transcription factor. ATF6 localizes in the ER but has the Golgi-localizing signal in its lumenal region that is inhibited in normal condition by binding to BiP. Under ER-stress condition, BiP is released from the lumenal region of ATF6, and its Golgi-localizing signal is exposed to transport ATF6 to the Golgi apparatus [28, 29]. In the Golgi, site 1 protease (S1P) and site 2 protease (S2P) cleave the transmembrane region of ATF6 and produce ATF6 that has only cytoplasmic region containing DNA-binding site [30, 31]. The cleaved ATF6 is translocated into the nucleus to function as a transcription factor that induces the transcription of genes related to ER chaperones [10, 32]. Prolonged ER stress promotes ATF6 to bind to ATF4 to induce the transcription of CHOP, resulting in apoptosis [33, 34, 35].
2.3. ER stress and insulin
Pancreatic β-cells are specialized cells to synthesize and secrete a large amount of insulin. Insulin biosynthesis in pancreatic β-cells accounts for 10–50% of total protein synthesis [36, 37]. Therefore, the burden to the ER (ER stress) in pancreatic β-cells is constitutively high even in physiological condition. It is also known that pancreatic β-cells are sensitive against oxidative stress and hypoxia  as well as ER stress. The expression level of glutathione peroxidase, an antioxidant, is very low in pancreatic β-cells; therefore, pancreatic β-cells are sensitive to oxidative stress . The islets of Langerhans are surrounded by blood vessels and supplied with nutrients and oxygen. Hypoxia affects insulin secretion of pancreatic islets and the survival rate of grafted islets [40, 41].
In type-II diabetes, it was reported that pancreatic β-cell mass is decreased [42, 43]. Huang et al. reported that the rat model of type-II diabetes expressing human islet amyloid polypeptide (hIAPP) showed the decrease of β-cell mass due to β-cell apoptosis, and the proteins related to ER stress including CHOP is highly expressed in β-cells . The relationship between ER stress and diabetes has been studied by a variety of animal models and human genetic diseases. Akita mouse, another mouse model of diabetes named by Akio Koizumi in Akita University, has a single mutation in insulin 2 gene. Although there are no gross defects in the transcription of the wild-type insulin 2 allele and the two alleles of insulin 1, the phenotype of a single mutation of insulin 2 is dominant. Insulin 2 gene in Akita mouse has tyrosine instead of cysteine 96 (C96Y), and the mutated proinsulin does not form the disulfide bond between A chain and B chain (A7–B7). The mutated proinsulin cannot be transported to the Golgi apparatus and its secretion is inhibited . The mutated proinsulin is accumulated in the ER that causes UPR to result in the induction of the expression of GRP78, XBP1, and CHOP. Eventually, pancreatic β-cells die by apoptosis. The necessity of ER stress for β-cell death was demonstrated by the delay of the development of diabetes in mouse produced by crossing Akita mouse with CHOP–knock-out mouse . It was reported that human also has the same mutation .
Wolcott-Rallison syndrome (WRS) is caused by the malfunction of Eif2ak3 gene that encodes PERK . WRS is an autosomal-recessive disorder that has neonatal diabetes, epiphyseal dysplasia, osteoporosis, and growth retardation. Patients with WRS have the point mutation in the kinase domain of PERK or the mutation that causes the deletion mutant of PERK. The mutation causing kinase dead of PERK develops diabetes after several months of birth, whereas the mutation that still maintains kinase activity of PERK delays the development of diabetes after 30 months. As well as WRS, PERK knock-out mice showed the secretory defects in many tissues causing diabetes and growth defects [49, 50, 51]. Furthermore, the knock-in mice having the mutation of phosphorylation site of eIF2α, the downstream molecule of PERK signaling, are unable to inhibit translation leading to over-synthesis of insulin and resulting in the dysfunction of pancreatic β-cells and β-cell death .
ATF6 knock-out mice do not show gross defects in normal diet, whereas high-fat diet causes the dysfunction of pancreatic β-cells [53, 54]. Furthermore, strong ER stress promotes the death of pancreatic β-cells [55, 56].
The knock-out mice of IRE1α specifically deleted in pancreatic β-cells cause diabetic phenotype [57, 58]. The mRNA levels of preproinsulin are not impaired; however, the protein level of proinsulin and mature insulin decreases, and protein and mRNA levels of five PDI protein families, PDI, PDIR, P5, ERp44, and ERp46, also decrease. These results indicate that these five PDI families are involved in proinsulin folding downstream of IRE1α, and upregulation of these PDI families could be the next approach for the treatment of diabetes.
3. Biogenesis of insulin secretory granules
After reaching the Golgi from the ER, secretory proteins are sorted in the trans-Golgi network (TGN) (Figure 1). One of the secretory pathways is the constitutive pathway in which proteins are constitutively secreted. When there is no sorting signal, proteins are thought to follow this pathway in mammalian cells. By contrast, another secretory pathway is the regulated pathway in which secretory proteins are packaged into the immature secretory granules (ISGs) (Figure 5). ISGs mature into mature secretory granules (MSGs), then MSGs are fused with the plasma membrane (PM) upon the stimulation of secretagogues to secrete the contents of MSGs. Proinsulin follows the regulated secretory pathway after the Golgi apparatus.
Secretory proteins destined for regulated pathway are segregated from other proteins and packaged into ISGs. This is termed as sorting by entry. On the other hand, in the process of formation and maturation of ISGs, other proteins are eliminated from ISGs. It is termed as sorting by exit, or sorting by retention [59, 60, 61].
3.1. Proinsulin transport to immature secretory granules
The molecular mechanisms of proinsulin sorting in the TGN are yet to be understood. It is thought that the selective aggregation of proinsulin occurs in the TGN [62, 63]. Insulin secretory granules contain a clear electron-dense core structure suggesting that insulin is crystallized in the granules. In pituitary AtT-20 cells, insulin granules can be formed by transfecting insulin gene, and hemagglutinin that flows in a constitutive pathway is segregated from the dense-core structure. Hemagglutinin is distributed evenly through the Golgi stacks as well as proinsulin; however, they are segregated after the TGN . Therefore, the proinsulin sorting from constitutive pathway could occur in the TGN.
The sorting receptor that recognizes proinsulin and transport proinsulin into ISGs remains unidentified . Carboxypeptidase E (CPE), an enzyme involved in insulin processing, was proposed to play a role as the sorting receptor ; however, the islets from mice that lost CPE by its mutation showed that insulin is efficiently secreted by secretagogues as well as in control islets, whereas the constitutive secretion of insulin remains as low as 1% similar to that in control islets . Therefore, the possibility that CPE plays a role as a sorting receptor in pancreatic β-cells is questionable . While research to find out the sorting receptor has been going on, another possibility was proposed; cargo aggregation/oligomerization is the sorting signal for ISGs [60, 61, 66].
It is proposed that aggregated proteins directly bind to lipid micro-domain in the TGN membranes and these micro-domains could be recognized by cytosolic machineries [59, 67]. Secretory granules (SGs) contain a high amount of cholesterol, and the depletion or addition of cholesterol affects glucose-stimulated insulin secretion (GSIS) . Secretogranin III is one of the components of SGs and known to bind to cholesterol-rich membranes . Although the role of secretogranin III in SG biogenesis in mice is not clear , it could be important to investigate the role of cholesterol.
There is an interesting phenomenon using SEGFP that is the green fluorescent protein (GFP) having a signal peptide in its N-terminus. In cultured insulin-secreting cells (INS-1 cells), SEGFP is sorted to secretory granules and secreted by secretagogues similar to insulin, whereas secreted alkaline phosphatase (SEAP), a model protein of constitutive pathway, is constitutively secreted . SEGFP forms oligomer by a disulfide bond, and its oligomerization could unexpectedly function as a sorting signal to ISGs. The results may support the idea that cargo oligomerization itself, rather than specific sequences on cargo, is required for sorting. The involvement of lipids or other sorting proteins in this case is unclear.
Zinc and calcium ions play important roles in insulin oligomerization. Structural studies showed that insulin forms a dimer, and in the presence of zinc and calcium ions, it forms a hexamer [72, 73]. It is thought that the concentration of zinc and calcium ions rises in the TGN , and these ions are enriched in SGs [74, 75]. The oligomerization regions of insulin and proinsulin are essentially the same with or without C-peptide . The cleavage of C-peptide from proinsulin hexamer decreases the solubility of insulin hexamer leading to crystallization of insulin in mature secretory granules (MSGs). Insulin crystals are thought to be stable and can be stored in MSGs for a long time without being degraded . ZnT8 zinc transporter, the product of SLC30A8 gene, is highly expressed in pancreatic β-cells, and the combined deletion of ZnT8 and ZnT7 inhibits GSIS . However, ZnT8 mutation is protective against type 2 diabetes . The precise function of ZnT8 in insulin biosynthetic pathway and its relationship with the development of diabetes remains unclarified.
Although the molecular mechanisms of sorting are yet to be understood, recent studies revealed the molecules to be involved in the fission process of SGs from the TGN. Arfaptin 1 has a lipid-binding domain termed Bin/Amphiphysin/Rvs (BAR) domain that binds to a curved membrane structure [78, 79] and implicated in a regulating membrane fission . Arfaptin-1 binds to small GTP-binding proteins, Arf1- and Arf-like protein 1 (Arl1), and recruited to the Golgi membrane by a GTP-bound form of Arf1 and Arl1 [81, 82]. Arfaptin 1 is phosphorylated by Protein Kinase D (PKD) that is activated by diacylglycerol (DAG) enriched in the neck of budding vesicles . Non-phosphorylated mutant of Arfaptin 1 (S132A) or PKD inhibitor blocks insulin SG fission from the TGN . The expression of Arfaptin 1 (S132A) or Arfaptin 1 depletion inhibits GSIS. As Arfaptin-1 was reported to be involved in other transport pathways [82, 85, 86], the specificity of Arfaptin 1 in SG biogenesis needs to be carefully addressed. Although Arfaptin 1 is proposed to play a role in membrane fission [84, 87], it could be interesting to investigate the upstream molecules of Arfaptin 1 to look for the sorting machinery for SG biogenesis.
3.2. Maturation of secretory granules
3.2.1. Insulin processing
The excursion of C-peptide decreases the solubility of insulin hexamer and causes insulin crystallization within SGs . Proinsulin is processed into mature insulin by prohormone convertases (PC1/3 and PC2) and carboxypeptidase E (CPE)  (Figure 6). PC1 (also known as PC3) cleaves 32–33 junction between B chain and C-peptide, and then CPE removes 31, 32 arginine residues. The intermediate form of proinsulin that is cleaved in B–C junction but is yet to be cleaved in A–C junction is termed as des-31, 32 split proinsulin. PC2 cleaves 65–66 junction between A chain and C peptide, and CPE removes 64, 65 lysine and arginine residues to form another intermediate termed des-64, 65 split proinsulin . The cleavage of B–C junction tends to occur first before the cleavage of A–C junction . PC1/3 and PC2 are Ca2+- and pH-dependent endopeptidases. The optimal pH of both enzymes is pH 5.5 [90, 91]. The pH at the TGN is reported to be ~6.0 . The pH of ISGs varies from 5.5 to 7.0 and the pH of MSGs is ~5.0 . PC1/3 could be active in the TGN as well as in ISGs , whereas PC2 is thought to be active in ISGs and MSGs [91, 94].
3.2.2. Sorting by retention
As ISGs mature, ISGs produce vesicles to remove proteins that are not required for MSGs, whereas insulin is retained in MSGs. For example, ISGs are known to produce constitutive-like vesicles that contain excess C-peptide than insulin to be secreted [95, 96]. Also, in contrast to constitutive secretory proteins, lysosomal enzymes are thought to be segregated from ISGs in pancreatic β-cells [97, 98]. Generally, lysosomal enzymes are synthesized in the ER as well as secretory proteins and transported to the Golgi apparatus (Figure 1). In the TGN and endosomes, lysosomal enzymes are recognized by mannose 6-phosphate receptors (MPRs) and then packaged into clathrin-coated vesicle (CCVs). Clathrin is a coat protein that forms a cage-like structure to produce CCVs in the post-Golgi compartment . AP-1 is a clathrin adaptor that binds to MPRs and clathrin and mediates to form clathrin/AP-1-coated vesicles [100, 101]. Proinsulin ISGs have clathrin and AP-1 on their surfaces as well as MPRs . Also, MSGs lose the signal of cathepsin B, a lysosomal protease, whereas ISGs still have a strong cathepsin B signal. These results suggest that lysosomal enzymes recognized by MPRs are removed from ISGs .
3.2.3. Homotypic fusion
Syntaxin 6, a SNARE protein that is reported to be important for the homotypic fusion of ISGs in neuroendocrine cells , is also removed from insulin ISGs . In neuroendocrine cells, it is thought that homotypic fusion plays an important role in SG maturation, and the fusion machineries required for homotypic fusion are different than that required for MSGs fusion to the PM [67, 102]. It is proposed that membrane fusion machinery is remodeled in the end of ISG maturation. The role of Syntaxin 6 and homotypic fusion in insulin granule maturation is not clear . However, recent study showed that homotypic fusion could also be important in insulin granule maturation . In islets from HID-1 KO mice, Vamp-4, another SNARE protein that is proposed in ISG-derived vesicle fusion to the PM in neuroendocrine cells , is mislocalized. Proinsulin processing and acidification are delayed, and by 3D electron microscopy, there are less homotypic fusion events . Although the role of HID-1 and Vamp-4 in homotypic fusion in pancreatic β-cells should be addressed in the future, it is possible that homotypic fusion might also be important for insulin granule formation.
Because of the importance of insulin in diabetes mellitus, insulin secretory pathway has been extensively studied. Recent advance in the understanding of biosynthetic pathway reveals the importance of ER stress in β-cell dysfunction and novel machineries of secretory granule biogenesis. However, still many questions remain. What are the mechanisms by which ER-stress sensors regulate proinsulin translation and folding? Is it relevant to prevent β-cell death by preventing UPR? The inhibition of CHOP has been studied to prevent β-cell death for the treatment of diabetes [46, 107, 108, 109]; however, it should be addressed carefully that even if β-cells survive by preventing CHOP, and too much accumulation of unfolded proteins in the ER may prevent normal proinsulin folding and would not support the function of islets of Langerhans. Decreasing the continuous high demand of insulin synthesis is anyway the primary importance for diabetes; then thinking about how to support proinsulin folding, packaging proinsulin into secretory granules, and elimination of unfolded proteins from β-cells would help for developing new treatments of diabetes.