Understanding the structure of hIAPP and its various conformations is important for understanding how it contributes to T2DM and hence for designing effective therapeutics. In T2DM, evidence suggests that hIAPP can adopt different conformations, including pre-IAPP (an un-cleaved form), monomers, dimers, oligomers and fibrils [34]. Like other protein misfolding disorders (PMDs), it is unclear to what extent each protein conformation contributes to toxicity and disease.

In vitro, synthetic hIAPP monomers dissolved in phosphate buffer or diluted hexafluoro-2-isopropanol form random coil conformations, which misfold into β-sheet structures and fibrils [25]. Analysis of hIAPP fibrils using site-directed spin labelling and electron paramagnetic resonance (EPR) spectroscopy, in combination with simulated annealing molecular dynamics, revealed the fibrils to be composed of stacked misfolded forms of hIAPP. Each hIAPP was bent into a hairpin with each arm having a region of β-sheet structure, opposite and staggered against each other. The fibrils had a left-handed twist and a hydrophobic surface at the fibril ends that is suggested to be the site of fibril growth where subsequent hIAPP monomers can bind, leading to fibril elongation [35]. Molecular dynamics simulations confirmed the random coil overall structure of hIAPP and revealed many metastable conformational states [36]. Some of the states had β-hairpin secondary structure and hydrophobic surfaces, features which facilitate nucleation of aggregation pathways. The formation of insoluble hIAPP fibrils from soluble monomers is proposed to occur via nucleated conformational conversion or conformational selection [36]. Nucleated conformational conversion involves the assembly of fibrils from oligomeric intermediates such as protofibrils [37]. Conformational selection means that the binding of monomers does not involve any change in conformation state but binds to an already formed conformation [38]. NMR results suggest that hIAPP undergoes nucleated conformational selection where the N-terminal α-helical region is involved in monomer “collapse” (hydrophobic or hydrophilic collapse), suggesting that monomers first collapse before further rearranging into ordered proto-fibrils composed of β-sheets [39]. Conformational selection involving this selective collapsing of monomers possessing β-sheet structures has also been supported by molecular dynamic studies [37].

Recently, several cryo-EM structures of recombinant hIAPP have been produced [40,41,42,43] that show fibrils are composed of two protofilaments. The structures observed suggest that the process of fibrillization may be synchronous [43] and the structural similarity of hIAPP fibrils to polymorphs of amyloid-β (Aβ) fibrils [42] may explain how both fibrils can cross-seed one another.

The most detailed, and biologically relevant structures of hIAPP fibrils was published recently by Cao et al., 2021 [44], showing eight morphologically different cryo-EM fibril structures seeded from fibrils extracted from islet cells from a T2DM donor. Four of the structures revealed twisted fibrils comprising two linked protofilament chains each composed of stacked β-strands. Two of the twisted fibrils are homodimers, whilst the other two are heterodimers. Two core folds can be observed in the fibrils which were observed to be capable of interacting in three different ways to form protofibrils, thus allowing the different structures to form. Some of the seeded fibrils do not resemble any of the unseeded ones (derived from peptides only), hence, they could resemble the structure of hIAPP in vivo [44].

Since membranes have been suggested as crucial to hIAPP misfolding and aggregation [45], it is essential to characterise the structure of hIAPP when bound to a membrane-like environment. Studies suggest that membranes accelerate the transition from α-helical to β-sheet pleated structures as revealed by NMR, EPR, TEM and Atomic force microscopy (AFM) studies [46,47,48,49]. NMR and EPR experiments using hIAPP bound to micelles highlighted residues 9–22 as having an unstable α-helical region, which is assumed to become stabilised during the aggregation process [34]. Lipid binding assays and TEM showed that when bound to anionic or zwitterionic membranes, hIAPP transitions from an α-helical structure to a β-sheet structure within minutes [50]. AFM has also been used to observe membrane-bound pre-fibrillar hIAPP, showing structures resembling ion-channels which could damage membranes by forming pores and changing their fluidity [51]. Other AFM studies have shown hIAPP accumulating on membranes in the form of unstructured aggregates. Membranes containing cholesterol revealed less bound amyloid compared to cholesterol-free membranes [52,53], highlighting the importance of membrane composition to amyloidosis.

Importantly, decreased insulin degrading enzyme (IDE) levels can lead to hyperinsulinemia, which may contribute to insulin resistance [54]. IAPP is a substrate for the insulin degrading enzyme (IDE) [55], a zinc metalloproteinase expressed in most cell types and also found in the circulation. Modulation of IDE activity could have consequences for amyloid formation by hIAPP and other amyloidogenic proteins [56]. Inhibition of IDE increased amyloidosis and hIAPP cytotoxicity in a cell culture model [57], but not in cultured islets [58]. In mice models, acute administration of a potent IDE inhibitor demonstrated beneficial effects on glucose tolerance [59], indicating potential for therapy of T2DM. However, given its wide distribution and other biochemical functions such as a chaperone and ubiquitin-proteosome pathway modulator, long term or untargeted use of inhibitors may be problematic [57,60].

Different mechanisms have been proposed explaining how monomers stack together and assemble into oligomers. Identifying exactly how the different conformations form and the key residues involved is important for developing therapeutics to inhibit aggregate formation, especially at the early stages of the process before extensive cellular damage occurs. NMR experiments using hIAPP peptides in solution suggest the initial events may involve adjacent monomers binding via noncovalent interactions involving residues 26–32 before assembling into oligomers [61]. Another study suggests a similar region, residues 29–33, as key to oligomer assembly, and also highlights residues 20–27 as important in fibril formation [62,63].

Several studies indicate that the amyloidogenic properties of human IAPP are linked to residues 20–29, one of the regions previously identified as key to oligomer assembly [64]. Residues 8–18 have also been identified as important in determining the side chain orientations along the β-strand within fibrils that have been shown to differ in fibril polymorphs produced in vitro [28,65]. A comparison of the primary sequence of human IAPP to that in rats (that do not form amyloid or get T2DM) reveals important differences (Figure 1). There are no proline residues in hIAPP, whereas rat IAPP has prolines at residues 25, 28 and 29. Molecular dynamics simulations demonstrated that proline residues show a low propensity of forming of β-sheet structures [64] and explains why hIAPP can more easily form β-sheet structures compared to rats. This provides compelling evidence associating IAPP aggregation with development of diabetes. This has been further corroborated in studies with transgenic rodent models, where rats that do not normally develop diabetes can become naturally diabetic if the human IAPP gene is inserted [66]. We compared the tertiary structures of human (PDB 5MGQ) and rat IAPP (PDB 2KJ7) solved by NMR and highlighted where in these 3D structures residues differed. Residues 18, 23, 25, 26, 28 and 29 differ between the two species.

A Ser20Gly mutation found in Chinese and Japanese populations has been linked to an increased rate of amyloid aggregation, and an increased propensity to T2DM, further supporting the importance of the 20–29 region in amyloidosis [67]. The quantity of amylin produced is unaffected by the mutation, since mRNA levels observed in individuals with and without the mutation were comparable [68]. Cryo-EM structures reveal differences between wild type and S20G fibrils [40,41,42], both showing two protofilaments and a left-handed twist but differences in features such as repeat and crossover lengths. Differences observed in the backbone conformations could explain how the early onset T2D IAPP genetic polymorphism S20G can aggregate more readily than wildtype and may provide a structural explanation for surface-templated fibril assembly [41].

A conserved disulphide bond found outside the cross β-core is located between residues Cys 2 and Cys 7 of hIAPP. Removal of this bond through mutation or Cys-Cys disulphide reduction was shown to increase hIAPP amyloidosis while also decreasing hIAPP-induced membrane leakage, highlighting the importance of this region in fibril formation and related cytotoxicity [69].

Accumulating evidence suggests that hIAPP oligomers play a crucial role in cytotoxicity. Due to their soluble nature and short lifespan, it has been extremely difficult to isolate them to characterise their structure. Altamirano-Bustamante et al. [70] used various techniques (Immunoblotting, TEM, ultrastructural immunolocalization and CD spectroscopy) to confirm the presence of oligomers in sera of healthy, T1DM, T2DM and obese children, utilising specific anti-oligomer antibodies. Samples from all groups showed the presence of fibrils and trimers, hexamers and dodecamers and oligomers of varying size were also observed (detected with anti-oligomer antibodies). Surprisingly, oligomers were also detected in sera from healthy children, but they had the lowest number of small oligomers out of all the groups, which are thought to be more toxic than the larger counterparts. The fact that oligomers have also been seen in healthy individuals is confusing, demonstrating that they are present in asymptomatic individuals and may not always contribute to disease. The observation of hIAPP oligomers in sera now requires further experiments to optimise purification methods to try to isolate them from sera so that they can be further characterised.

HIAPP has been linked to cytotoxicity of pancreatic β-cells, but more recently other cells such as neuronal cells [71]. Whilst several mechanisms linking hIAPP and toxicity have been suggested (as summarised in Figure 2), the forms of hIAPP that mediate this toxicity is still debated, as are the precise mechanisms involved. Some studies propose that it is the fibrils that directly damage β-cells [72], while others suggest oligomeric intermediates are responsible [67]. However, more than one form of hIAPP may contribute to cytotoxicity, possibly to different extents that may be via a variety of mechanisms. Monomeric hIAPP has been observed to increase membrane fluidity, leading to membrane destabilisation [73], and has also been observed to increase production of reactive oxygen species (ROS) in β-cells [74]. Oligomeric hIAPP has been shown to decrease cell viability as well as increase membrane fluidity. Cell damage involving fibril growth at the membrane is linked to membrane leakage [75].

A crucial question in the understanding of hIAPP aggregate formation is what initiates and accelerates the formation of the fibrils that have been linked to T2DM. If we knew what triggered the misfolding cascade in the first place, then T2DM (and possibly other PMDs) may be preventable. Distinguishing between what initiates and what accelerates the process of amyloidosis is key but may involve the same factors. Several factors may contribute to the formation of these misfolded aggregates, and we have summarised these in Figure 3. These include the concentration of hIAPP, cell stress, the role of proteoglycans, the immune response and the association with other amyloidogenic proteins such as cross-seeding (Figure 3).