The β Cell in Diabetes: Integrating Biomarkers With Functional Measures

Abstract

The pathogenesis of hyperglycemia observed in most forms of diabetes is intimately tied to the islet β cell. Impairments in propeptide processing and secretory function, along with the loss of these vital cells, is demonstrable not only in those in whom the diagnosis is established but typically also in individuals who are at increased risk of developing the disease. Biomarkers are used to inform on the state of a biological process, pathological condition, or response to an intervention and are increasingly being used for predicting, diagnosing, and prognosticating disease. They are also proving to be of use in the different forms of diabetes in both research and clinical settings. This review focuses on the β cell, addressing the potential utility of genetic markers, circulating molecules, immune cell phenotyping, and imaging approaches as biomarkers of cellular function and loss of this critical cell. Further, we consider how these biomarkers complement the more long-established, dynamic, and often complex measurements of β-cell secretory function that themselves could be considered biomarkers.

Graphical Abstract

Graphical Abstract.

ESSENTIAL POINTS.

• The islet β cell is a critical determinant of the development of hyperglycemia in all forms of diabetes.

• Alterations in the processing of proinsulin and insulin secretion as well as the loss of β cells have all been documented as part of the hyperglycemic syndrome and can be demonstrated prior to attainment of the diagnostic thresholds for diabetes.

• Biomarkers are increasingly being used for predicting, diagnosing, and prognosticating disease in both research and clinical settings.

• In the case of diabetes, these potential biomarkers include genetic markers, circulating molecules, and imaging approaches.

• While these biomarkers complement the more long-established, dynamic, and often complex measurements of β-cell secretory function, functional measures are still frequently required to interrogate the β cell.

Biological markers, commonly termed biomarkers, are being used more frequently to provide an indication of the state of a biological process, pathological condition, or response to an intervention. They are considered to fall into 3 broad categories—molecular, cellular, and imaging—and are used in medicine for predicting, diagnosing, and prognosticating disease (Fig. 1).

Figure 1.

Blood- and imaging-based measures for identifying forms of diabetes and examining β-cell function and mass. Blood-based biomarkers include genetic and epigenetic markers, assays of autoimmunity including islet autoantibodies and T cells, measures of the efficiency of β-cell propeptide processing, and dynamic tests of β-cell secretory function. Imaging-based biomarkers include estimates of pancreas size and fat, β-cell mass, islet amyloid, and insulitis.

Diabetes mellitus, one of the world’s most common noncommunicable diseases (1), represents an important condition in which biomarkers have the potential to provide critical information to identify susceptible individuals prior to the onset of the disease, predict those whose disease course may progress more rapidly than others, and recognize who may be at higher risk of developing complications. The ability to do so is a central tenet of precision medicine and would allow for better management of this heterogeneous disorder (2, 3).

At the core of the need for identification and prediction of diabetes and its outcomes is the islet β cell, which by virtue of the fact that it produces the critically essential hormone insulin, plays a vital role in the development of hyperglycemia in all forms of diabetes. In this review we focus on the β cell, addressing the potential utility of different genetic, circulating, and imaging measures and how they, as biomarkers, provide insight into aspects of cellular function and cell loss. In so doing, we will examine their applicability for one or more of the different forms of diabetes. We will also consider how they complement the more traditional and/or complex measurements of β-cell secretory function that are based largely on dynamic testing. Our emphasis is on human data, supplemented by information from nonhuman studies whenever applicable.

Classification of the Different Forms of Diabetes

Diabetes is a complex, multifactorial disease defined by elevated plasma glucose concentrations. Hyperglycemia is driven by insufficient insulin, either in the presence or absence of reduced insulin sensitivity. The vast majority of cases of diabetes comprise the two major subtypes, type 1 and type 2, the latter accounting for 90% to 95% of cases worldwide (1). In the case of type 2 diabetes, there is considerable heterogeneity, as discussed in more detail later. In addition, there are a number of other subtypes that are less prevalent but of relevance to the discussion of β-cell biomarkers and diabetes.

Type 1 diabetes results from autoimmune attack leading to marked loss and dysfunction of β cells. People with type 1 diabetes have antibodies directed at islet-cell proteins and require insulin therapy early in the course of the disease (4). While most common in youth, the disease may also occur in adults. The course of the disease in older individuals is usually milder, with prolonged periods of endogenous insulin secretion and a lower incidence of diabetic ketoacidosis. This condition is often referred to as latent autoimmune diabetes in adults (LADA). People with LADA have detectable autoantibodies, but whether LADA represents the same clinical entity as type 1 diabetes is debated (5).

Type 2 diabetes typically affects older individuals who are often overweight or obese and suffer from obesity-associated insulin resistance. Disturbingly, it is now also being more frequently recognized in youth (6). In this form of diabetes, which involves multiple pathophysiological mechanisms (7-9), the loss of β-cell function is typically more gradual over time. Thus, individuals may remain rather asymptomatic and undiagnosed for a long time. Insulin is usually not required at diagnosis, although many individuals progress to insulin replacement therapy over the course of the disease (4).

Other subtypes of diabetes include gestational diabetes, which is defined as diabetes occurring in women during the second or third trimesters of pregnancy and which was not present prior to gestation. It occurs in up to 10% of pregnancies and frequently resolves after parturition, while remaining a risk factor for future type 2 diabetes (4, 10). Several monogenic diabetes syndromes have also been identified, in which defects in single genes pivotal for normal β-cell function result in hyperglycemia (3, 11-13). The most important clinical entities are mutations inducing (i) neonatal or congenital diabetes and (ii) maturity-onset of diabetes of the young (MODY). Neonatal diabetes is typically diagnosed before 6 months of age in individuals without a genetic susceptibility to type 1 diabetes and who are found to have one of the known genetic β-cell abnormalities associated with this disorder (3). This disease entity is also frequently associated with defects in other organ systems that can be ascribed to the gene mutation (3). MODY is inherited in an autosomal dominant pattern and is typically diagnosed in adolescence and early adulthood, without signs of insulin resistance and autoantibodies. The disease has been linked to a number of gene defects, which predominantly affect β-cell development and function (11). Given the differences in phenotype related to these gene defects, the required treatment can vary from no treatment for glucokinase gene defects to oral agents and/or insulin for the other forms (3, 11, 12). Pancreatic (exocrine) diseases such as cystic fibrosis, pancreatitis, trauma, hemochromatosis, and neoplasia may also induce hyperglycemia (4). With the advances in treatment of cystic fibrosis resulting in increased lifespan, the incidence and prevalence of cystic fibrosis–related diabetes (CFRD) is increasing (14). Neurocognitive disease and diabetes have been linked, with their pathophysiology perhaps related to the deposition of amyloid fibrils in brain and pancreatic islets (15). Finally, several other forms of diabetes are recognized, such as drug-induced diabetes (associated with glucocorticoids (16), calcineurin inhibitors (17), antiretroviral therapy (18, 19)), diabetes secondary to endocrinopathies (20), and new-onset diabetes mellitus after transplantation of solid organs (NODAT) (21).

Although these classifications are based on extensive clinical experience, heterogeneity exists in disease presentation, with many people failing to be readily categorized into one specific form of diabetes or another. For example, autoantibody-negative, lean adults can present with severe insulin deficiency, and people with severe obesity may present with diabetic ketoacidosis. The latter, often called Flatbush or ketone-prone diabetes, occurs particularly in individuals of non-European origin, with many of these individuals becoming insulin independent after initial presentation (22).

An alternative stratification in 5 separate disease clusters in adults with diabetes was recently proposed using parameters that include age, body mass index (BMI), glycated hemoglobin A1c (HbA1c), presence of islet autoantibodies, and a static measure for insulin sensitivity and β-cell function (23). Recently, it was proposed that a similar clustering approach could be utilized in prediabetes (24). While promising, given that the phenotypes may change over time, the true clinical value of these novel classifications remains to be determined.

Given the number of subtypes of diabetes, classifying patients at the time of diagnosis into a specific form of diabetes may be difficult due to significant overlap in phenotype, with the true diagnosis often becoming more obvious as time and the disease progress. With the ongoing identification and refinement of biomarkers, particularly those related to the β cell, we anticipate greatly enhanced understanding of the disease and its heterogeneity. This improved understanding should, in time, allow for better classification and treatment of individuals. What follows is our evaluation of the current state of knowledge and utility of biomarkers, preceded by brief descriptions of the normal physiology of β-cell function and the alterations in β-cell function and mass that result in the development of hyperglycemia.

Normal β-Cell Physiology: An Overview

To contextualize the use of biomarkers and functional tests in assessing the health of the β cell, we provide a brief overview of this specialized endocrine cell’s physiology as it applies to the production and secretion of insulin. However, it should be recognized that islet amyloid polypeptide (IAPP) is another β cell–specific peptide which is normally produced and secreted in parallel with insulin at a more or less consistent molar ratio. Therefore, unless specified otherwise, the following discussion applies to both insulin and IAPP.

Proprotein Biosynthesis and Granule Maturation

The insulin mRNA pool remains stable in β cells due to RNA binding proteins in the 5′ and 3′ untranslated regions (UTRs) of the transcript (25). Glucose stimulation drives transcription of the insulin (INS) gene, but with a sluggish response that takes approximately 1 hour for pre-mRNA levels to increase, and up to 48 hours for mature transcripts to significantly increase (26). In contrast, based on studies in rodent islets, proinsulin translation, along with that of proIAPP and the processing enzymes prohormone convertase 1/3 (PC1/3) and PC2 (PCSK1 and PCSK2, respectively), rapidly increase in response to glucose, suggesting posttranscriptional regulation of insulin expression (27-29). The β cells achieve rapid glucose-induced proinsulin synthesis by storing insulin mRNA in preassembled polysomes that are transported to the endoplasmic reticulum (ER) membrane and initiate translation in response to glucose (30).

Signal peptide cleavage occurs during translation and insertion of preproinsulin into the ER lumen. Formation of 3 intramolecular disulfide bonds, facilitated through the actions of protein disulfide isomerases, are critical to the proper folding of proinsulin in the ER: CysB7-CysA7, CysB19-CysA20, and CysA6-CysA11. The basis for trafficking and sorting of insulin granule content in β cells is not completely solved (31). Proinsulin is subject to folding as it transits the Golgi, where in mature rodent islets, on exit, it is sorted to the regulated secretory pathway with 99% efficiency (32). While in adult islets IAPP is also efficiently sorted to granules in the regulated secretory pathway (33), in neonatal islets about half of it is released via the constitutive secretory pathway (34). In human islets (pro)IAPP trafficking remains targeted to the regulated secretory pathway when cultured under basal glucose conditions; however, culturing human islets for 8 days in high glucose resulted in proIAPP release from a constitutive pathway (35). This suggests the potential for altered secretory product trafficking and release in immature or dysfunctional β cells, which may be detectable in the circulation under appropriate testing conditions.

After exit from the trans-Golgi network, secretory granules sorted to the regulated secretory pathway mature under conditions in which granule pH decreases and granule [Ca²⁺] and [Zn²⁺] increase through the actions of vesicular H⁺-ATPase (36) and SLC30A8 (37). With increased granule H⁺ and Ca²⁺, prohormone convertase activity also increases, initiating the conversion of proinsulin and proIAPP. This process results in the generation of conversion intermediates prior to the production of the mature peptides. These intermediate forms and the mature peptides are further trimmed at their C-terminus by carboxypeptidase E (CPE). In the case of proinsulin, this results in mature insulin and C-peptide. For proIAPP, following trimming by CPE, proIAPP is amidated at the C-terminus by peptidylglycine α-amidating monooxygenase (PAM) to yield mature IAPP (38). Recently, the role of PC2 in proinsulin processing in human β cells has come into question because PC2 immunoreactivity was not readily detectable in β cells (39), although others have reported β-cell expression by transcriptomics (40) and immunohistochemistry (41). This is compatible with the previous description in mice that PC1/3 is more critical for proinsulin processing than is PC2 (42, 43). While PC2 has been shown in mouse islets to be critical for complete proIAPP processing, in human islets the specific roles of PC2 and PC1/3 in processing proIAPP remain to be confirmed.

Not all proinsulin within a β cell will ultimately be secreted, with (pro)insulin degradation by macroautophagy and selective autophagy key to maintaining β-cell proteostasis and function (44, 45). Prior to ER-Golgi transport, misfolded proinsulin can be degraded via ER-coupled autophagy or ER-associated degradation mechanisms (46). Aged granules in β cells are also degraded, so that ultimately newer ones are preferentially secreted (47). Under conditions of nutrient depletion, newly synthesized granules are selectively degraded in the lysosome via a macroautophagy-independent mechanism, and hyperactivation of this degradation mechanism may play a role in the β-cell dysfunction of type 2 diabetes (48). Altered autophagy by β cells has been observed in pancreas samples obtained from people with type 1 and type 2 diabetes (49, 50). Given altered autophagy in these two forms of diabetes and its role in (pro)insulin degradation, it is plausible that altered autophagy may contribute to the biomarkers secreted from dysfunctional β cells in types 1 and 2 diabetes.

Peptide Secretion

Under nonstimulatory conditions, human β cells maintain a negative resting potential of approximately −70 mV. An ATP-sensitive potassium channel (K_ATP), composed of 4 Kir6.2 (KCNJ11) subunits and 4 sulfonylurea receptor 1 (SUR1) subunits, assists in maintaining a hyperpolarized membrane through transport of K⁺ ions against the membrane electrical gradient but with the [K⁺] gradient from the cytosol to the extracellular space.

Glucose, the primary insulin secretagogue, is transported across the plasma membrane in human β cells by GLUT1 (SLC2A1) and GLUT3 (SLC2A3) (51). The canonical model of glucose-induced insulin secretion posits that ATP generated via glycolysis and oxidative phosphorylation increases the cellular ATP/ADP ratio and results in K_ATP channel closure and membrane depolarization via increased cytosolic [K⁺]. At approximately −60 mV (52), voltage-gated calcium channels begin to open, resulting in rapid Ca²⁺ influx and exocytosis of insulin secretory granules. The rate-limiting step controlling insulin secretion is glucose phosphorylation by glucokinase, the enzyme having half-maximal activity at approximately 8 mM glucose (53), which correlates well with the observed half-maximal rate of glucose-induced insulin secretion in isolated human islets (54). Upon glucose stimulation, coupling of stimulus and secretion results in the rapid release of insulin in 2 phases (55, 56). In humans, the first phase begins at the start of glucose administration, is typically complete within 10 minutes, and represents a rapidly-releasable pool of secretory granules. This phase is followed by the second phase, which lasts for as long as the glucose concentration remains elevated, and includes granules containing newly synthesized insulin.

In isolated human islets, in addition to glucose, fatty acids and amino acids stimulate insulin secretion in part through anaplerosis (54, 57). In the case of amino acid–induced insulin secretion, mitochondrial glutamate dehydrogenase (GDH) is the chief enzyme involved in anaplerosis, doing so by catalyzing the deamination of glutamate to produce α-ketoglutarate (58).

Potentiation of insulin secretion can occur through intracellular metabolism or receptor-mediated signaling of exogenous signals. Peptide signals from either gut-derived incretins or islet paracrine signaling are well-recognized potentiators of insulin secretion (59). Glucagon-like peptide-1 (GLP-1), glucose-dependent insulinotropic polypeptide (GIP), and glucagon are well known secretion-amplifying signals acting via the secondary messenger cyclic adenosine monophosphate (cAMP) to activate protein kinase A (PKA) and exchange protein directly activated by cAMP 2 (Epac2) to promote exocytosis (60, 61). Somatostatin, derived primarily from the islet δ cell, is the most recognized peptide-derived paracrine inhibitor of insulin secretion (62). A myriad of other signals derived from additional cell types in the islet, including other endocrine cells, nerve fibers, endothelial cells, pericytes, and immune cells, fine tune insulin secretion to maintain euglycemia (63-66).

Summary

Based on our current knowledge of insulin and IAPP production and secretion, along with advancements in proteomics and immunoassays, it is becoming increasingly possible to analyze posttranslational modifications and β-cell secretomes in both islets and the circulation. Thus, the field is becoming better equipped to utilize biomarkers and functional tests to probe the physiology of the β cell in health and disease.

Pathophysiology of Diabetes: A Disease of Decreased β-Cell Function and Mass

Despite some uncertainties with respect to diabetes classification, virtually all diabetes entities have β-cell dysfunction as their common pathophysiological component. As such, dysfunction of this cell is not only a prerequisite for the development of diabetes, but its progressive nature typically determines the progressive course of the disease. Further, most forms of diabetes include a reduction in the number of β cells, the importance of this mass loss varying by disease type. What follows is a brief overview of the pathophysiology of these different forms of diabetes, the purpose of which is to frame the subsequent discussion of biomarkers.

Type 1 Diabetes

β-cell loss and dysfunction

The pathological hallmark of type 1 diabetes is insulitis, an inflammatory lesion of the islet associated with β-cell loss, with a key role for autoreactive T cells (67). This evidence is derived mainly from examination of type 1 diabetic pancreata obtained at autopsy (68). These human studies have made the critical observation that the degree of insulitis is heterogeneous, affecting only 10% to 30% of islets, and diffuse within an islet (68-71). These findings contrast with those in the nonobese diabetic (NOD) mouse, where nearly every islet exhibits marked T-cell infiltration (72). Human autopsy studies have clearly shown reduced β-cell mass and pancreas weight but have also demonstrated that insulin-positive β cells may persist for many years after diagnosis (68, 71).

Given the discordant relationship between the degree of islet inflammation, number of residual β cells, and severity of hyperglycemia at presentation, there has to be β-cell dysfunction that is beyond the loss of mass per se (73, 74). This dysfunction has been demonstrated in individuals who have diabetes on oral testing but do not otherwise manifest hyperglycemia (75), as well as in first-degree relatives who are at high risk of subsequently developing type 1 diabetes (76). The β-cell dysfunction in type 1 diabetes manifests as reductions in insulin release in response to intravenous and oral stimulation as well as in impaired processing of the β-cell propeptides (77-84). Further, β-cell function can be recovered after a period in which islets from individuals with recent-onset type 1 diabetes have been cultured under euglycemic conditions (85), suggesting that a component of the functional defect is related to the in vivo milieu (86, 87).

Mechanisms of disease pathogenesis

Genetic risk is largely derived from human leukocyte antigen (HLA) class II haplotypes, in particular the DR and DQ genes that are present in up to 80% to 90% of patients (88). In addition, more than 50 loci have been identified that confer risk for the disease and include candidate genes associated with immune function and/or the survival and function of the β cell (89-91).

Aside from an individual’s genetic predisposition, a trigger seems to be necessary to initiate the immune response characterizing the disease. The role of an environmental trigger in type 1 diabetes is supported by the discordant incidence rates in monozygotic twins (92) and differences in disease rates that are not simply explained by genetic differences (93). One possibility is viral infection (94), particularly enteroviruses (95), although the link to this latter group of viruses has not been universal (96). Some attention has focused recently on the gut microbiota, as its composition has been reported to differ in people with type 1 diabetes (97) and a small study raised the possibility that transplantation of fecal microbiota may preserve C-peptide responses early in the disease course (98). A number of dietary factors have also been proposed to be a trigger for the immune system (94). Importantly, environmental triggers such as viruses and bacteria activate the innate immune system, which can in turn initiate and intensify activation of the adaptive immune system as well as contribute to β-cell death and dysfunction (99, 100).

In type 1 diabetes, autoreactive CD4+ and CD8+ T cells infiltrate the islet and mediate loss and dysfunction of β cells by production of cytokines as well as cell-cell interactions (101). This effect is likely compounded by impaired function of regulatory T cells that are normally responsible for immunological tolerance (102). Finally, islet autoantibodies, which are useful biomarkers, are considered innocent bystanders (103).

Type 2 Diabetes

Impact of obesity and insulin resistance

Obesity, and particularly central/visceral adiposity, is a key component in the pathogenesis of insulin resistance (104-106). The net effect is that insulin is less effective in stimulating glucose uptake by skeletal muscle, reducing hepatic glucose production, and inhibiting adipose tissue lipolysis (107). To overcome this reduced insulin effectiveness, the β cell releases more insulin, leading to a state of hyperinsulinemia to maintain normal glucose tolerance (NGT). However, when the β-cell response is inadequate, impairments in glucose tolerance and ultimately type 2 diabetes develop (7). The inadequacy of this β-cell response will frequently only be apparent when it is interpreted in the context of the prevailing insulin sensitivity (108-110), a concept discussed in more detail later.

Role of β-cell loss and dysfunction

Type 2 diabetes is characterized by a reduction in the number of β cells as well as secretory dysfunction of those cells that remain. This process is progressive, which is easily discernible when examining function longitudinally, but it is less clear when considering β-cell mass, as repeated biopsies are not ethical and imaging techniques are not sufficiently advanced. In fact, the loss of secretory function can be demonstrated very early on in the pathogenic course of the disease, with reduced responsiveness to secretagogues evident well before glucose levels reach diagnostic thresholds. As the disease progresses, increasing doses and number of glucose-lowering medications are typically required, with progression to a need for insulin being an indicator of the β cell having reached a state of near total “failure.”

It is clear that β-cell mass is decreased in most people with type 2 diabetes, with this loss ranging from about 40% to 60% in matched subjects (111-113). A 40% deficit has even been reported in those with impaired fasting glucose (IFG) vs 63% in type 2 diabetes (111). However, the degree of loss varies tremendously among individuals, with a large degree of overlap in the proportion of β cells between healthy individuals and those with type 2 diabetes (112, 113). This reduction in β cells is the result of cell death and possibly dedifferentiation compounded by the fact that adult β cells do not readily replicate (111, 114-117). Importantly, human studies have quantified β cells in different ways and reported “β-cell mass” as area relative to exocrine or islet area on a tissue slice. This approach is important for a number of reasons. First, pancreatic weight is not always available and both pancreatic weight and volume have been shown to be lower in people with type 2 diabetes compared with nondiabetic controls (113, 118). Second, islet density may vary within the different regions of the pancreas (119). Third, the proportion of the islet comprising β cells may also vary between pancreas regions (120). Thus, the difference in the quantity of β cells determined simply by histology on pancreatic sections alone may not always provide a true estimate of the deficit (113, 118).

Defective β-cell secretory function is not only present in people with overt type 2 diabetes, but also in people with IFG or impaired glucose tolerance (IGT) (110, 121-124). These deficits affect both pulsatile and oscillatory secretion as well as the different secretagogue-induced dynamic components (125-130), with the first-phase response to intravenous glucose being essentially absent when fasting glucose exceeds 115 mg/dL (125). These secretory abnormalities exist despite immunostaining showing that insulin is still present in the islet. In addition to the defect in β-cell secretory function, the cell is also incapable of efficiently processing proinsulin to mature insulin, the magnitude of the defect linked to the degree of secretory dysfunction and glycemia (131-137). These considerations are discussed in greater detail subsequently.

The magnitude of the β-cell defect in type 2 diabetes differs between individuals and this could explain, at least in part, the heterogeneity in terms of disease progression and development of complications (23). Age appears to be an important factor in disease heterogeneity, with older age typically being associated with milder hyperglycemia, while individuals diagnosed at a younger age and particularly in adolescence, manifest a more rapid decline in β-cell function (23, 138). For reasons that are still not understood, adolescents tend to be more insulin resistant and have hyperresponsive β cells compared with body adiposity–matched, middle-aged adults (139, 140). Given the differences in β-cell function, some people can be successfully treated with oral agents, while others progress rapidly and require insulin therapy. Other contributors to the heterogeneity of glucose metabolism include sex hormones, and particularly estrogen status, medications such as steroids, and socio-economic status (141-144). Differences in the gradation of gene-based scores (genetic risk score [GRS] and partitioned polygenic score) to predict glucose concentrations and the processes that contribute to diabetes development further highlight the heterogeneity of type 2 diabetes (145-148). Finally, heterogeneity in the function of individual β cells within the islet, differences in the mass of β cells between individuals with and without type 2 diabetes, and the effect of gene variants on gene expression are also all likely to contribute to variation in β-cell function (113, 149-153). Future exploration of novel β-cell biomarkers, with or without functional tests, should better recognize this heterogeneity and could potentially reveal new differences and improve their utility in type 2 diabetes populations.

Mechanisms of disease pathogenesis

While environmental factors, most notably excessive food intake and obesity, play a key role in the rising prevalence of type 2 diabetes, heritability is a key factor. Recent advances in sequencing have identified more than 400 gene variants associated with type 2 diabetes; based on work by Mahajan et al (154) and Udler et al (155), 128 of these 400 variants can be linked to a phenotype comprising either β-cell function, obesity/adiposity, lipodystrophy-like/insulin action, or lipid metabolism/liver (156). Of these, nearly 70% had a β-cell phenotype related to growth, development, and/or function (148, 154-157). Evidence is also accumulating suggesting that epigenetic changes (DNA methylation, histone modification, microRNA [miRNA], long non-coding RNAs [lncRNAs]), possibly starting as early as in utero, are an additional factor (158, 159).

Islet amyloid deposition has long been recognized to occur in the majority of people with type 2 diabetes (151, 160, 161). These deposits are formed by aggregation of the normal β-cell secretory product IAPP (162, 163). While its physiological role remains uncertain, in its native form IAPP is not harmful. However, the process of oligomer formation renders the peptide cytotoxic resulting in β-cell apoptosis and dysfunction; interestingly, the end product amyloid appears to be largely inert (164). Based largely on studies in animal models, the magnitude of amyloid formation appears to be related to the degree of secretory demand placed on the β cell (165-167). With progress in imaging, it is now possible to demonstrate the presence of islet amyloid in vivo in animals (168-171), and in time it is hoped this will be possible in humans and will provide further insight into the pathogenesis of type 2 diabetes.

A role for chronic inflammation in type 2 diabetes has been firmly established, although the primary trigger(s) remain unclear and the utility of individual molecules as biomarkers of the status of the β cell is not well established. Postulated triggers of islet inflammation in type 2 diabetes, characterized by increased number of activated proinflammatory macrophages, cytokine production, and β-cell dysfunction, include dyslipidemia and IAPP aggregates (172-176). Islet-resident macrophages are emerging as important players both in β-cell health and regeneration (177), and in mediating β-cell dysfunction in type 2 diabetes via proinflammatory cytokines such as interleukin (IL)-1 β, IL-6, and tumor necrosis factor (TNF)-α (178). A role of islet inflammation in type 2 diabetes was supported by a small clinical trial in which administration of the IL-1 receptor antagonist anakinra improved glycemic control and β-cell function (179). However, in a larger, long-term clinical trial, IL-1 antagonism with a human anti-IL-1β monoclonal antibody (canakinumab) did not reduce diabetes incidence (180), although some initial improvement in glycemia was observed in those with preexisting diabetes. Whether the adaptive immune system contributes to islet dysfunction in type 2 diabetes is more debated (181).

Monogenic Forms of Diabetes

Maturity-onset diabetes of the young

MODY, first described in 1974 (182), is a cluster of 11 different autosomal dominant forms of diabetes, which in many instances affect transcription factors, resulting in impaired insulin production and release (3). The 4 most common forms affect (i) glucokinase (MODY2), the rate-limiting step in β-cell glucose metabolism, that is characterized by β-cell insensitivity to glucose mediating a relatively small impairment in insulin secretion and thus relatively mild hyperglycemia and typically no need for glucose-lowering therapy (11, 183); (ii) HNF-1α (MODY3) and the less common HNF-4α (MODY1), both of which result in hyperglycemia and progressive β-cell dysfunction requiring pharmacological intervention that can initially be sulfonylureas but frequently advances to insulin (184, 185); and (iii) HNF-1β (MODY5), where the progressive loss of β-cell function with an insulinopenic phenotype is frequently accompanied by variable renal abnormalities and developmental defects of the genital tract (186).

Neonatal diabetes

Neonatal diabetes becomes apparent early in life and when it consists of homozygous mutations in certain MODY genes, it induces permanent neonatal diabetes and pancreas agenesis. About 50% of cases are caused by potassium channel gene (KCNJ11 and ABCC8) mutations, resulting in impaired insulin secretion that can be readily restored with sulfonylurea treatment (3, 13). While impairments in β-cell function characterize neonatal diabetes, some of these mutations are also associated with defects in other organ systems (3, 13).

Cystic Fibrosis–Related Diabetes

Cystic fibrosis arises due to mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene and is the most common, lethal, autosomal recessive disorder (187). CFRD is characterized by mild insulin resistance, impaired β-cell function, and reduced β-cell mass along with disordered glucagon release (188). The β-cell dysfunction occurs through distinct mechanisms, including a marked inflammatory response (189, 190), impaired chloride conductance in the β-cell membrane and increased susceptibility to oxidative stress (191-193). There remains some controversy over whether the chloride channel is in fact present in β cells, since small studies utilizing CFTR modulator therapy failed to improve insulin secretion and glucose tolerance (194, 195). Interestingly, islet amyloid has also been observed on autopsy in people with cystic fibrosis (190, 196).

Summary

β-cell dysfunction is a characteristic of all the different forms of diabetes, although differences in pathophysiology exist. In type 1 diabetes, β-cell loss through autoimmune attack is an important contributor to the secretory function defect. On the other hand, in type 2 diabetes, while β-cell loss is important, the functional defect seems to be a more critical component. While risk genes for these 2 major forms of diabetes have been identified, they have not proven as useful as biomarkers as in the monogenic forms of diabetes and CFRD. In time, we expect that identification of additional aspects of the pathogenesis of hyperglycemia will result in the development of new biomarkers that will supplement genetic information and allow for the use of composite measures that better predict disease development and outcomes.

Genetics and Epigenetics

Over the last decade, there has been major progress in our understanding of the genetic basis of the different forms of diabetes. Genome-wide association studies (GWAS) have identified more than 400 sequence variants (single nucleotide polymorphisms [SNPs]) associated with individual risk of type 1 and type 2 diabetes (89, 90, 197). Combining these variants into genetic risk scores (GRS) offers information that remains stable throughout life and improves the prediction of diabetes risk and understanding of diabetes heterogeneity (148). The complex interplay between genetic and environmental factors associated with diabetes has been further highlighted by identification of various epigenetic modifications, such as methylation/acetylation, that can alter gene expression. What follows is a discussion of the utility (and lack thereof) of genetics along with epigenetics as potential biomarkers in diabetes.

Genetics

Type 1 diabetes

Genetics is an important contributor to type 1 diabetes, with an identical twin concordance rate between 30% and 70% (198-200) and a risk of ∼7% for siblings (201). The main genetic drivers underlying this risk are Class II HLA (or major histocompatibility [MHC] system) DR and DQ genes, which are located on chromosome 6 and encode for cell surface proteins typically expressed on antigen-presenting cells (202). The HLA Class II haplotypes DR3-DQ2 and DR4-DQ8, alone or in combination, are known to be associated with the highest genetic risk for type 1 diabetes and are considered to contribute to 50% of type 1 diabetes heritable risk (202). Of note, other HLA Class II haplotypes, such as DRB3, DRB4, and DRB5, have also been associated with an increased risk of type 1 diabetes (203, 204). In contrast, certain HLA Class II haplotypes, such as DR15-DQ6, seem to confer protection from type 1 diabetes (202). Therefore, prospective studies starting early in life have used Class II HLA DR-DQ typing, alone or in combination with family history of type 1 diabetes, to assess the genetic risk for development of islet autoantibodies and progression to clinical type 1 diabetes (205-207). Interestingly, some studies have reported that the specificity and order of appearance of the first islet autoantibody was related to the HLA DR-DQ genotype (208, 209). However, in a cohort of the Finnish Type 1 Diabetes Prediction and Prevention (DIPP) study, while Class II HLA DR-DQ genotype improved estimation of type 1 diabetes risk, it was limited to the development of islet autoimmunity and was not associated with the progression rate from advanced autoimmunity to clinical diabetes (210).

While the HLA Class II DR-DQ region represents the strongest association with type 1 diabetes, other SNPs outside the Class II HLA region impact the risk of and progression of the disease. Some HLA Class I genes (eg, A*24 and B*39 alleles) encoding for presenting peptides for T cells, have also been independently associated with type 1 diabetes susceptibility and progression of β-cell loss (211-213). Further, in recent years, linkage analysis and GWAS have identified more than 50 non-HLA genetic loci contributing to type 1 diabetes risk, including SNPs near the preproinsulin (INS), protein tyrosine phosphatase non receptor type 22 (PTPN22), cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), interleukin 2 receptor α (IL2RA) and CFTR genes (89-91, 214) (Table 1). Studies have confirmed that some of the genetic variants outside of the HLA DR-DQ region can affect seroconversion for islet autoantibodies and/or progression to clinical diabetes (209, 215-217). In a very recent study in children, those with certain susceptibility alleles demonstrated a more rapid decline in β-cell function compared with those without these alleles (213). Furthermore, while many of these type 1 diabetes susceptibility candidate genes are involved in immune function, the emerging concept is that some of them are expressed in islets and may play a role in modulating β-cell function and survival (218, 219). Examples include GLIS3 (220), which may contribute to cytokine-induced cell death, and RNLS (221), which was identified in a CRISPR screen as a modulator of immune-mediated β-cell death.

Table 1.

Major genes with a β-cell phenotype linked to the different types of diabetes

Gene symbol	Gene name	Function of gene product	Type(s) of diabetes
ABCC8	ATP-binding cassette transporter subfamily C member 8	Insulin secretion (modulation of ATP-sensitive potassium channels)	Type 2 diabetes; Monogenic diabetes
ADCY5	Adenylate cyclase 5	Regulation of calcium-dependent insulin secretion	Type 2 diabetes
AP3S2	Adaptor related protein complex 3 subunit sigma 2	Golgi vesicles formation and trafficking to lysosomes	Type 2 diabetes
ARAP1	ArfGAP with RhoGAP domain, ankyrin repeat and PH domain 1	Regulation of Golgi structure and cytoskeleton; Cell migration	Type 2 diabetes
BCAR1	BCAR1 scaffold protein, cas family member	Cell adhesion and migration	Type 2 diabetes
BCL11A	B-cell CLL/Lymphoma 11A (Zinc finger protein)	Lymphopoiesis; Negative regulation of p53 activity (transcriptional repressor)	Type 2 diabetes
C2CD4A	C2 calcium-dependent domain containing 4A	Transcription factor	Type 2 diabetes
CCND2	Cyclin D2	Cell cycle regulation	Type 2 diabetes
CDC123/CAMKID	Cell division cycle 123 homolog/Calmodulin dependent protein kinase ID	Cell cycle regulation	Type 2 diabetes
CDKAL1	CDK5 regulatory subunit associated protein 1-like 1	Growth and development; Proinsulin to insulin conversion	Type 2 diabetes
CDKN2A	Cyclin dependent kinase inhibitor 2A	Cell cycle regulation	Type 2 diabetes
CEL	Carboxyl ester lipase	Cholesterol and lipid-soluble vitamin ester hydrolysis	Monogenic diabetes
CENPW	Centromere protein W	Cell cycle regulation	Type 1 diabetes; Type 2 diabetes
CTSH	Cathepsin H	Lysosomal proteins degradation	Type 1 diabetes
DGKB	Diacylglycerol kinase beta	Cell signaling	Type 2 diabetes
G6PC2	Glucose-6-phosphasase catalytic 2	Glucose metabolism	Type 2 diabetes
GCK	Glucokinase	Glucose metabolism; β-cell growth and development	Monogenic diabetes
GIPR	Gastric inhibitory polypeptide receptor	Potentiation of insulin secretion	Type 2 diabetes
GLIS3	GLIS family zinc finger 3	β-cell growth and development; Transcription factor	Type 1 diabetes; Type 2 diabetes; Monogenic diabetes
GPSM1	G protein signaling modulator 1	Cell signaling	Type 2 diabetes
HHEX	Hematopoietically expressed homeobox	Growth and development; Transcription factor	Type 2 diabetes
HMG20A	High mobility group 20A	Transcription factor; Histone methylation	Type 2 diabetes
HMGA2	High mobility group AT-Hook 2	Transcription factor; Chromatin regulation/acetylation	Type 2 diabetes
HNF1A	Hepatic nuclear factor 1 α	β-cell growth and development; Transcription factor	Type 2 diabetes; Monogenic diabetes
HNF1B	Hepatic nuclear factor 1 β	β-cell growth and development; Transcription factor	Type 2 diabetes; Monogenic diabetes
HNF4A	Hepatic nuclear factor 4 α	β-cell growth and development; Transcription factor	Monogenic diabetes
HSD17B12	Hydroxysteroid 17-β dehydrogenase 12	Metabolism of steroid hormones	Type 2 diabetes
IDE	Insulin-degrading enzyme	Peptide degradation (including insulin, IAPP and glucagon)	Type 2 diabetes
INS	Insulin	Insulin production	Type 1 diabetes; Type 2 diabetes; Monogenic diabetes
IGF2BP2	Insulin-like growth factor 2 binding protein 2	β-cell growth and development; Transcription factor	Type 2 diabetes
JAZF1	Juxta-posed with another zinc finger gene 1	Cell cycle regulation; Transcriptional repressor	Type 2 diabetes
KCNJ11	Potassium voltage-gated channel subfamily J member 11	Insulin secretion	Type 2 diabetes; Monogenic diabetes
KCNQ1	Potassium voltage-gated channel subfamily Q member 1	Insulin secretion	Type 2 diabetes
KLF11	Kruppel like factor 11	Exocrine cell growth and development; Transcription factor	Monogenic diabetes
MTNR1B	Melatonin receptor 1B	Mediation of melatonin actions (including inhibitory effect on insulin secretion)	Type 2 diabetes
NEUROD1	Neurogenic differentiation 1	Growth and development; Transcription factor	Monogenic diabetes
NOTCH2	Neurogenic locus notch homolog protein 2	Growth and development; Transcription factor	Type 2 diabetes
PAM	Peptidylglycine α-amidating monooxygenase	β-cell processing enzyme	Type 2 diabetes
PAX4	Paired box gene 4	β-cell development and differentiation; Transcription factor	Type 2 diabetes; Monogenic diabetes
PDX1	Pancreatic and duodenal homeobox 1	Pancreatic and β-cell growth and development; Transcription factor	Monogenic diabetes
PIM3	Pim-3 proto-oncogene, serine/Threonine kinase	Cell signaling; Cell proliferation and survival	Type 2 diabetes
PTPN2	Protein tyrosine phosphatase non-receptor type 2	Cell survival	Type 1 diabetes
PTPN9	Protein tyrosine phosphatase non-receptor type 9	Cell signaling; Cell growth and differentiation; Cell cycle regulation	Type 2 diabetes
PRC1	Protein regulator of cytokinesis 1	Cell cycle regulation	Type 2 diabetes
PROX1	Prospero homeobox 1	Transcription factor	Type 2 diabetes
RNLS	Renalase	Modulator of immune-mediated β-cell death	Type 1 diabetes
RREB1	Ras responsive element binding protein 1	Transcription factor	Type 2 diabetes
		Cell differentiation
SLC30A8	Solute carrier family 30 member 8	Proinsulin and proIAPP conversion	Type 2 diabetes
SPRY2	Sprouty RTK signaling antagonist 2	Cell signaling	Type 2 diabetes
TCF7L2	Transcription factor 7 like 2	Blood glucose homeostasis; Transcription factor	Type 2 diabetes; Cystic fibrosis–related diabetes
THADA	Thyroid adenoma associated protein	Cell survival	Type 2 diabetes
WFS1	Wolframin ER transmembrane glycoprotein	Regulation of cellular calcium homeostasis	Type 2 diabetes; Monogenic diabetes
ZBED3	Zinc finger BED-type containing 3	Transcription factor	Type 2 diabetes

Genes were identified and selected from the GWAS and candidate genes listings in (89-91, 154, 155, 157, 218-221, 243).

The name and function of each gene was determined from GeneCards (https://www.genecards.org), accessed on February 27, 2021.

This high genetic heritability (including HLA and non-HLA variants) provides the opportunity to use the GRS to define and stratify the risk for type 1 diabetes. Most of the non-HLA DR-DQ variants seem to have only modest effects on the total genetic risk of developing type 1 diabetes (222). However, their incorporation together with HLA loci in an integrated GRS increases the ability to predict type 1 diabetes and is more powerful than HLA DR-DQ genotyping alone (223-226). The most recent type 1 diabetes GRS, which included 67 SNPs and accounted for interactions between 18 HLA DR-DQ combinations, when applied in samples in the UK Biobank, performed best in identifying individuals with type 1 diabetes (227). Furthermore, combining this most recent genotyped risk with family history, autoantibodies, and clinical characteristics markedly improved type 1 diabetes prediction among susceptible children compared with measurement of autoantibodies alone (228). Finally, assessment of a type 1 diabetes GRS may also help to discriminate type 1 diabetes from type 2 diabetes (229) or from monogenic forms of diabetes (230).

Type 2 diabetes

There is compelling evidence that genetic predisposition underlies the development of type 2 diabetes. Recent estimates of type 2 diabetes heritability range from 25% to 80%, varying based on study duration, parental history, and sibling history (231, 232). Technological and analytical advances have led to the identification of numerous genes linked with type 2 diabetes. Since the first identification of PPARγ using the candidate gene approach (233), GWAS has now identified more than 400 gene variants for type 2 diabetes susceptibility (154). Very recently, both known and novel loci were also identified in the first GWAS of youth-onset type 2 diabetes, suggesting there is a significant overlap in the genetic architecture of the disease in youth and adults (234). Although most products of these gene variants have not been identified, some have been linked to obesity (eg, FTO) and insulin sensitivity (eg, IRS1 and PPARγ), with most linked to β-cell function (eg, TCF7L2, PAM, SLC30A8, MTNR1B, HNF1A, HNF1B) (154, 155, 157, 219) (Table 1). With increasing sample size (235), it is now also becoming possible to use complementary approaches of whole genome or whole exome sequencing to identify specific gene variants (236). Indeed, classification of these GWAS variants according to their association with diabetes-related metabolic traits has identified robust groups each characterized by a specific pathophysiological process: reduced β-cell function with high proinsulin, reduced β-cell function with low proinsulin, obesity, lipodystrophy, and liver/lipid metabolism (154, 155). These findings suggest that cluster analysis could represent an interesting approach to better define the clinical heterogeneity of type 2 diabetes that contributes to different clinical outcomes.

The first use of the GRS analyzed the combined risk of 16 to 18 SNPs and showed only slightly improved prediction of incident diabetes compared with that of clinical risk factors alone (145, 146). With identification of new loci with successive larger GWAS (154, 237), extended polygenic scores have enhanced, albeit modestly, the ability to predict subsequent type 2 diabetes (154). Altogether they are likely to explain around 20% of the overall variation in type 2 diabetes risk, that is half of the estimated heritability (154). In the prospective Metabolic Syndrome in Men (METSIM) Study, a genetic score for type 2 diabetes that included 76 SNPs was associated with changes in β-cell function (quantified by the disposition index), as well as with a 2-fold increase in risk of type 2 diabetes (147). Interestingly, comparison of the 3 published global, extended GRS for type 2 diabetes reported a similar 2.75-fold increased risk for individuals in the top 5% of the polygenic score distribution vs the remainder of the study population (148); however, the clinical utility of these GRS remains unclear. First, their value to improve prediction of incident diabetes is modest compared with clinical risk factors alone. Second, their ability to capture risk in individuals from non-European origin might be suboptimal since most of the GWAS data are derived from European cohorts. Therefore, more studies are required in other populations to generate equivalent data and GRS that could perform best and apply to those populations.

Monogenic diabetes

In contrast to type 2 diabetes, which has overlapping polygenic susceptibility, the molecular genetics underlying monogenic diabetes subtypes, including transient and permanent neonatal diabetes (developing before 6 months of age) and MODY, is well defined and has profound implications on both treatment and future development of associated clinical features (3, 13) (Table 2). Genetic testing has also been suggested to be cost-effective in patients with high suspicion of monogenic diabetes (238).

Table 2.

MODY genes and their pathophysiology, clinical phenotype, and treatment

Gene symbol	Gene name	Type of familial diabetes	Frequency (% of MODY)	Pathophysiology	Clinical phenotype	Treatment
HNF4A	Hepatic nuclear factor 4 α	MODY1	5%-10%	Progressive β-cell dysfunction (mainly insulin secretory defect)	- Common fetal macrosomia	Diet; SU; Insulin
					- Transient neonatal hyperinsulinemia and hypoglycemia followed by diabetes later in adolescence or adulthood
					- Low triglycerides
					- Microvascular complications
GCK	Glucokinase	MODY2	30%-50%	β-cell dysfunction (glucose sensing defect)	- Stable mild hyperglycemia	No medication (except possibly in pregnancy)
					- Low prevalence of microvascular complications
HNF1A	Hepatic nuclear factor1 α	MODY3	30%-65%	Progressive β-cell dysfunction (mainly insulin secretory defect)	- Progressive hyperglycemia with early onset	Diet; SU (additional glinides, GLP1-RA, DPP-4i); Insulin
					- Glycosuria
					- Transient neonatal hyperinsulinemia and hypoglycemia in some
					- Microvascular complications
PDX1	Pancreatic duodenal homeobox 1	MODY4	<1%	β-cell dysfunction	- Range from impaired glucose tolerance to diabetes	Oral glucose-lowering agents; Insulin
					- Pancreas agenesis (homozygosis form)
HNF1B	Hepatic nuclear factor 1 β	MODY5	<5%	β-cell developmental defect and dysfunction	- Diabetes	Oral glucose-lowering agents (minority respond to SU); Insulin
					- Renal malformations
					- Exocrine pancreas deficiency malformation
					- Female reproductive organ abnormalities
NEUROD1	Neurogenic differentiation 1	MODY6	<1%	β-cell dysfunction	Early onset of diabetes	Oral glucose-lowering agents; Insulin
CEL	Carboxyl-ester lipase	MODY8	<1%	Pancreas endocrine and exocrine dysfunction	- Typically autosomal dominant diabetes	Oral glucose-lowering agents (including SU); Insulin
					- Exocrine pancreatic dysfunction
					- Lipomatosis
INS	Insulin	MODY10	<1%	β-cell dysfunction	Neonatal, child or adult-onset diabetes	Diet;
						Oral glucose-lowering agents (including SU);
						Insulin
ABCC8	ATP-binding cassette transporter subfamily C member 8	MODY12	<1%	Insulin secretion defect (ATP-sensitive potassium channel dysfunction)	Frequently causes neonatal diabetes. Clinical phenotype similar to MODY4	SU; Insulin
KCNJ11	Potassium voltage-gated channel subfamily J member 11	MODY13	<1%	Insulin secretion defect (ATP-sensitive potassium channel dysfunction)	Heterogeneous	SU; Insulin
APPL1	Adaptor protein phosphotyrosine interacting with pH domain and leucin zipper 1	MODY14	<1%	Insulin secretion defect	Dysmorphic phenotype and delay in development	Diet; Oral glucose-lowering agents (including SU); Insulin

Abbreviations: DPP4i, dipeptidyl peptidase 4 inhibitor; GLP-1RA, glucagon-like peptide-1 receptor agonist; MODY, maturity-onset diabetes of the young; SU, sulfonylurea.

^aAlthough classified as MODY genes, KLF11 (MODY7), PAX4 (MODY9) and BLK (MODY11) are not included as they were recently refuted or disputed by a group of experts in monogenic diabetes (239).

^bMultiple loss of function variants in RFX6 have recently been described producing a phenotype similar to other MODY genes but with lower penetrance (573). It is not included in the table as it has not yet been assigned a MODY number.

Over 20 different genetic causes of neonatal diabetes have been described so far, and these predominantly affect the β cell (3, 13). Half of the neonatal diabetes diagnoses are linked to mutations in the potassium channel genes KCNJ11 and ABCC8 and have excellent therapeutic responses to sulfonylureas (3, 13). Insulin is usually required in patients carrying other gene mutations associated with neonatal diabetes. Genetic testing also allows the anticipation and identification of associated clinical features such as heart defects and exocrine pancreas deficiency with mutations in GATA4 and GATA6 (3, 13).

MODY has been associated with relevant mutations in at least 14 genes, again predominantly β-cell related (3, 11, 12), although 3 of these genes (KLF11, PAX4, and BLK) have been recently refuted or disputed by consensus of a group of monogenic diabetes experts (239) (Table 2). Mutations in genes encoding glucokinase (GCK) and the β-cell transcription factors hepatic nuclear factor 1 α (HNF1A), hepatic nuclear factor 4 α (HNF4A), and hepatic nuclear factor 1 β (HNF1B) are the most common (3, 11, 12, 240). Importantly, the identification of the MODY subtype results in different therapeutic strategies (3, 11, 12, 241). For example, low-dose sulfonylureas are effective as treatment for MODY caused by mutations in HNF1A and HNF4A, while insulin is required in HNF1B-MODY (3, 11, 12, 241).

Cystic fibrosis–related diabetes

CFRD has emerged as a common complication of cystic fibrosis and is caused by islet inflammation and β-cell dysfunction and loss. While CFTR mutations have been shown to increase the risk of diabetes independently of other risk factors such as pancreatic exocrine dysfunction (242), a recent GWAS has identified other CFRD risk loci (243). The latter reported a genetic overlap with type 2 diabetes and CFRD (eg, TCF7L2) but also interestingly identified 2 CFRD risk loci, PTMA and SLC26A9, that are unrelated to type 2 diabetes.

Epigenetics

Epigenetics represents changes that do not involve alterations of the ribonucleotide sequence but occur beyond conception (Fig. 2). Epigenetic changes, which comprise DNA methylation, histone modifications (acetylation/deacetylation), and noncoding RNA-mediated gene expression modifications, can occur as a result of genetic and/or environmental factors. The intrauterine environment represents the first potential exposure to some factors that have been linked to type 2 diabetes later in life (244). Examples include low birthweight, high birthweight, maternal obesity, gestational diabetes, smoking, and chemicals. Subsequent to birth, environmental factors can also contribute to epigenetic changes, modifying expression of genes involved in type 1 and type 2 diabetes (245, 246) and thereby increasing susceptibility to develop the disease. Advances in our understanding of epigenetics also provide compelling evidence for dysregulation of islet-specific gene expression in type 2 diabetes (247-250). Recent evidence, as detailed below, suggests that epigenetic biomarkers may be useful in the future for predicting diabetes.

Figure 2.

Epigenetic modifications occurring in the β cell. Environmental factors and/or genetics can contribute to epigenetic alterations in the β cell, thereby modifying expression of genes involved in β-cell function and survival. Epigenetic alterations include DNA methylation, histone modifications, and non-coding RNAs.

DNA methylation

DNA methylation has been found to be altered in pancreatic islets from human donors with type 2 diabetes and associated with impaired insulin secretion (251-253). Of note, some of these changes in DNA methylation involved several SNPs identified in GWAS to associate with type 1 diabetes (eg, HLA, INS, PTPN22) and type 2 diabetes (eg, KCNJ11 and ADCY5). In prospective cohorts, epigenome-wide association studies (EWAS) have also reported changes in the methylation status of whole blood DNA that were associated with an increased risk of incident type 2 diabetes (254-257). However, it is not clear whether these epigenetic changes occurred prior or following the development of hyperglycemia. It has been suggested that a diabetic milieu per se can alter human islet gene expression and methylation: expression of 1855 genes changed, with 1469 demonstrating variations in DNA methylation (258). In very recent studies, maternal dysglycemia was associated with changes in DNA methylation of neonates (259, 260), which appeared to be reduced by lifestyle intervention during pregnancy (260). Whether these associations are causal and linked to the risk of incident diabetes still needs to be elucidated. In mice, exposure to hyperglycemia in utero resulted in changes in gene methylation that were associated with decreased fasting insulin concentrations and glucose intolerance in vivo, and impaired glucose-simulated insulin secretion in vitro in islets from these same mice (261). Finally, changes in DNA methylation could also be associated with autoimmune diabetes. For example, a study has detected changes in DNA methylation in CD4+ T cells from adult patients with LADA (262).

Thus, changes in DNA methylation seem to be linked to impaired β-cell function and/or autoimmune responses. Whether these changes can be used as biomarkers for diabetes in clinical practice still requires more investigation. Although studies have shown an overlap in DNA methylation changes between blood and tissues, most identified DNA methylation loci are tissue specific. More studies are therefore needed to confirm that changes in DNA methylation in pancreatic islets are reflected in the peripheral circulation. Finally, since most studies have used cross-sectional approaches, more prospective studies are required to confirm whether any of these epigenetic changes are predictive of diabetes.

Histone modifications

Histone modifications have been detected in blood cells of individuals with type 2 diabetes (263, 264) and type 1 diabetes (265). However, to date there are no reports of genome-wide histone modifications in pancreatic islets from diabetic subjects. Therefore, there is a need for studies of these modifications in patients with diabetes.

MicroRNAs

Recent work on circulating (plasma/serum) miRNAs has highlighted their potential future use as biomarkers in diabetes (158). miRNAs are small noncoding RNAs that act as key regulators of gene expression and are enriched in specific tissues, including pancreatic islets. They can be modified by environmental factors. For example, miR132, which is normally involved in adaptation of β cells to insulin resistance, is upregulated in mice on a high fat diet and in cultured islets or cells exposed to glucose and palmitate (266).

Genetic alterations in miRNAs are rare but would appear to promote the development of type 2 diabetes by reducing β-cell function (266). Under stress conditions and β-cell death, miRNAs can be produced and released by islets into the circulation. Because studies have reported conflicting results regarding circulating miRNAs in type 2 diabetes (267), their use as biomarkers in type 2 diabetes remains unclear. Reasons for discrepancies include differences in study design and population size. With regard to type 1 diabetes, there is also evidence of a different signature of circulating and blood cell miRNAs in patients, some of which have been associated with immune cell responses (268-270). A recent study has identified serum miR-204, which is highly enriched in β cells and known to regulate critical processes of β-cell biology, as a new biomarker for type 1 diabetes-associated β-cell loss (271). However, while miRNAs offer the advantage of being stable and resistant, current limitations for their use as biomarkers for β-cell dysfunction include the absence of tissue-specificity and standardized procedures used in studies. Thus, large prospective studies will likely be required to identify clinically reliable circulating islet-specific miRNA signatures in type 1 and type 2 diabetes.

Long non-coding RNAs

Long non-coding RNAs are typically greater than 200 nucleotides and do not code for proteins. GWAS has identified a large number of lncRNAs of potential importance in diabetes; for example, LOC157273 is associated with increased liver glycogen and may be relevant to diabetes (272-274). For most, however, their functional characteristics are not well understood. That said, a number of groups have demonstrated that challenging islets with approaches such as elevated glucose, cytokines, a high fat diet, or pregnancy may all result in dysregulation of lncRNAs (159). A recent analysis of expression profiles of circulating lncRNAs in serum from patients with diabetes has revealed differences when compared to control patients (275). Whether these changes can be involved in the pathogenesis of the disease and linked to β-cell function remains to be elucidated.

Cell-free DNA and exosomes

Intracellular DNA from the nucleus or mitochondria can be measured in the circulation and has been used in a number of diseases, including cancer where it can be measured as an indicator of cell death. With this in mind, recent research has focused on the measurement of circulating unmethylated preproinsulin (INS) DNA as a new biomarker for β-cell death in type 1 diabetes (276-279). Indeed, unmethylated INS CpG sites are increased in the β cell, and these fragments can be released into the circulation upon β-cell damage. However, the presence of unmethylated INS in other cell types decreases the specificity of this measurement. Although it has been suggested that the addition of complementary biomarkers, such as unmethylated CHTOP, could increase the confidence of detecting β-cell death in youth with type 1 or type 2 diabetes (279), this was not supported by another study (280). In the latter, the use of an ultrasensitive assay for detection of 6 β cell-specific DNA methylation markers (including INS DNA), did not find any evidence of elevated β-cell-free DNA in patients with type 1 diabetes (280). Potential explanations provided by the authors included (i) an insufficient sensitivity of the assay; (ii) destruction of the β cells preceding the time of sampling; and (iii) differences in degree and dynamics of β-cell destruction between individuals.

Exosomes are small extracellular vesicles that carry bioactive molecules, such as proteins and noncoding DNA and RNA which participate in intercellular crosstalk, including paracrine communication between the different cell types in islets (281). Analysis of the content of islet-released exosomes suggest they are mainly derived from β cells (282). It has also been suggested that these exosomes are associated with the development of type 1 diabetes as they could participate in the initiation of the autoimmune process in the islets (281). Some in vitro, ex vivo, and in vivo studies have reported a specific islet-derived exosomal miRNA signature in individuals with type 1 diabetes (283-285), suggesting they could also serve as novel circulating biomarkers of the disease.

Summary

While the use of GRS improves prediction of type 1 diabetes and helps discriminate it from type 2 diabetes or monogenic diabetes, its clinical utility in type 2 diabetes remains unclear. In addition, the field of epigenetic biomarkers, including circulating cell-free RNA/DNA and exosomes as biomarkers of β-cell dysfunction and death seems exciting and promising, but basic and applied research evaluating their utility is still in the early stage and a great deal more work is clearly needed before their use can be translated to clinical practice in diabetes. Particularly, future work should focus on determining (i) unified methods for their identification and characterization; (ii) specific markers allowing validation of their islet origins (vs other tissues) when detected in biological fluids; and (iii) their power in predicting development of the disease in comparison to other biomarkers.

Markers of Autoimmunity and Inflammation

As touched upon previously, in both type 1 and type 2 diabetes, an inflammatory response contributes to the loss of β-cell function. The question arises as to how we can quantify this immune response and use it to classify diabetes subtype, to predict diabetes development in people at risk, and to monitor disease progression due to continuous loss of β-cell function.

T-Cell Responses in People With Type 1 Diabetes

Given that current evidence suggests that CD4+ and CD8+ autoreactive T cells are the main effectors of β-cell destruction in type 1 diabetes (101), measuring the frequency or function of T cells has the potential to assist in understanding type 1 diabetes pathogenesis, and monitoring disease progression (286) and response to immunotherapy (287). A number of T-cell biomarkers have been developed for use in type 1 diabetes that can be classified in 2 main categories: antigen-specific and antigen-agnostic T-cell biomarkers.

Antigen-specific assays

Antigen-specific assays typically measure T-cell responses when peripheral blood mononuclear cells (PBMCs) are incubated with islet antigens such as preproinsulin (287, 288) (Fig. 3). In addition, HLA class I or II multimers can be loaded with autoantigenic peptides to detect antigen-specific T cells (288). Measures of T-cell function that can be assessed include proliferation and cytokine secretion (289).

Figure 3.

Antigen-specific and antigen-agnostic T cell phenotyping. Antigen-specific T cells are profiled following incubation with islet autoantigens or peptides such as preproinsulin or neoepitopes. Activated T cell populations can be characterized through quantification of cytokine secretion, proliferation, and gene expression. Antigen-agnostic T cells are not first activated by islet autoantigens. Bulk- and single-cell -omics analyses have improved T cell transcriptional and epigenetic characterization. Flow cytometry is used to phenotype T cells based on cytokine expression and cell surface markers as well as (phosphorylation of) intracellular proteins and nuclear transcription factors that are key for regulating T cell function (FoxP3 depicted here).

Using such experimental approaches, autoreactive T cells specific for β-cell antigens have been quantified in peripheral blood in persons with type 1 diabetes. However, T cells reactive to islet autoantigens are often detectable in people without diabetes, including in the pancreas (290, 291). In type 1 diabetes, these cells can be phenotypically different; for example, CD4+ autoreactive T cells collected from people with type 1 diabetes secrete higher concentrations of proinflammatory cytokines including interferon-γ (IFN-γ) and interleukins (292), while secretion of anti-inflammatory factors may be reduced (293). Deep phenotyping of autoreactive T cells on a single-cell omics platform has led to the discovery that autoreactive T cells with a proinflammatory profile are present in children that develop type 1 diabetes prior to the formation of islet autoantibodies (294).

Broad application of autoreactive T cells as reliable biomarkers of disease in type 1 diabetes remains challenging for several reasons. First, deep phenotyping of these cells is necessary to determine their polarization and potential disease impact. Second, relevant autoreactive T cells mostly reside in the pancreas or pancreatic lymph nodes with low frequencies in the peripheral circulation (295). Finally, isolation of β-cell–specific T cells has been challenging due to the low-avidity interactions between β-cell antigens and the T-cell receptor (296). Because of these limitations, the use of autoreactive T cells as biomarkers for type 1 diabetes is currently limited clinically.

Antigen-agnostic assays

Alternative approaches for employing T cells as markers of type 1 diabetes include measurement of T-cell subset frequencies and in-depth phenotypic characterization of circulating T cells using omics approaches (Fig. 3). T-cell subset populations can be detected by flow cytometry using an array of surface and intracellular markers. In addition, both multiple- and single-cell technologies have enabled profiling of T cells at the transcriptional and epigenetic levels (288).

Antigen-agnostic assays have been used in recent clinical trials of immunomodulation in type 1 diabetes. As such, several therapies showed a beneficial effect in those with residual β-cell function determined by C-peptide. Examples include (i) the anti-CD3 monoclonal antibody teplizumab, which induced a population of “exhausted” CD8+ T cells (277, 297); (ii) abatacept, which inhibited the interaction between antigen-presenting cells and T cells, thereby contributing to an increase in the fraction of CD4 memory cells (298); and (iii) low-dose antithymocyte globulin (ATG) treatment, which reduced CD4+ T cells (299). In addition, proinsulin peptide therapy increased FoxP3 expression by regulatory T cells (Tregs; which induce immune tolerance) (300), while modulation of gut microbiota reduced CD4+ CXCR3+ and CD8+ CXCR3+ T cells after 1 year of treatment (98). Not all trials, however, showed a clear link between metabolic benefits of immunomodulatory therapies and beneficial alterations in T-cell markers (301).

While T-cell subset frequency and function remain potentially valuable as biomarkers due to their causal relationship with type 1 diabetes, their clinical use is still in its infancy. This situation is likely to change in the coming years with efforts to improve T-cell phenotyping. For example, a recent study reported a higher frequency of CD4⁺CD25⁺CD127^hi (127-hi) cells being associated with longer partial remission and a favorable response to immunotherapy (302). Thus, assessing T-cell function in type 1 diabetes (both antigen-specific and agnostic) holds great promise for clinical usage, but harmonization and standardization of experimental protocols are needed to enable broad application of T cells as biomarkers in type 1 diabetes (303).

T-Cell Responses in People With Type 2 Diabetes

Studies from one group have raised the possibility that T cells autoreactive to β-cell proteins may also be present in people with type 2 diabetes (304). Using peripheral blood–derived T cells from individuals with phenotypic type 2 diabetes who do or do not have islet autoantibodies, immunoblotting-based assays have detected T cells reactive to islet proteins (304). The magnitude of these cellular responses was related to the degree of β-cell function and the progressive loss of β-cell function over time (305, 306). The reactivity of T cells to islet proteins can be reduced by treatment with the insulin sensitizer rosiglitazone, which also reduces cytokine secretion from PBMCs and improves the C-peptide response to glucagon, an effect not observed with sulfonylurea treatment (307).

It has also been hypothesized that changes in adipose tissue biology in people with type 2 diabetes may drive some of these immunological changes. As such, obesity may induce MHC class II expression in adipose tissue, leading to activation of CD4+ T cells (308). These T-cell populations are further characterized by a shift from an anti-inflammatory (T helper [TH]2 cells) to a proinflammatory (TH1, TH17 cells) phenotype, and may contribute to activation of proinflammatory macrophages in adipose tissue.

While these T-cell data raise the possibility that the adaptive immune system may be involved in β-cell dysfunction in type 2 diabetes, their clinical utility in predicting disease, monitoring progression, and tracking responsiveness to interventions remains rather limited.

Modified T-Cell Autoantigens (neoepitopes)

Although T cells learn tolerance to self-proteins in the thymus, in the presence of stress such as hyperglycemia, cytokines, or infection, β-cell proteins may undergo posttranslational modification to become neoepitopes, to which the immune system is naïve (309). Neoepitopes may be directly presented on the β-cell surface via HLA class I or by antigen-presenting cells via HLA II class triggering an immune response (67, 310). Mechanisms by which proteins may be modified to form neoepitopes in type 1 diabetes include enzymatic citrullination or deamidation, or via nonenzymatic posttranslational modification including oxidation and carbonylation (310). Peptide fusion to form hybrid peptides has recently been recognized as a potential source of potentially important neoepitopes in type 1 diabetes (311, 312). Neoepitope formation may also occur through novel mechanisms such as alternative splicing products and defective ribosomal initiation products (DriPs) (313).

While these changes at a cellular level are likely important in the pathogenesis of type 1 diabetes and T cells specific for neoepitopes can be detected in the circulation, neoepitopes derived from β-cell peptides, if secreted, likely circulate at levels too low for detection in plasma. Identification of new neoepitopes and refinement of assays for their measurement in plasma could lead to circulating neoepitopes having utility as β-cell biomarkers.

Autoantibodies

Type 1 diabetes

Islet autoantibodies are more prevalent in people with type 1 diabetes and can provide a quantifiable risk for the disease development (314); thus, they are widely used biomarkers for disease prediction. Importantly, islet autoantibodies do not appear to play an active role in β-cell loss, but instead are markers of the autoimmune process.

Following discovery of the first islet cell autoantibodies (ICA) in children with type 1 diabetes (315), autoantibodies specific for insulin (IAA) (316), glutamic acid decarboxylase (GAD) (317) and protein tyrosine phosphatase (IA2 or ICA512) (318) were discovered. GAD catalyzes the decarboxylation of glutamate to gamma-aminobutyric acid (GABA) and is expressed in islets in synaptic vesicles, while protein tyrosine phosphatase is an enzymatic transmembrane glycoprotein localized in the endocrine cell secretory granules. The most recently identified autoantibody with clinical utility is directed at the zinc transporter 8 (ZnT8), a pancreatic β-cell secretory granule membrane protein involved in insulin exocytosis (319).

GAD is the most frequently used clinically, with the others employed to a more variable extent, for both diagnosis and prediction of type 1 diabetes (320). In individuals with a suspected diagnosis of autoimmune diabetes, one or more islet autoantibodies are usually assessed and sufficient to confirm a diagnosis of type 1 diabetes. For diabetes prediction, autoantibodies are most commonly employed in epidemiological studies and prevention trials, the latter including immunomodulatory approaches to preserve β-cell function and prevent the development of hyperglycemia.

A positive family history of type 1 diabetes and GRS can identify risk for type 1 diabetes in newborns (228, 321). This predictive strength is increased when combined with the presence of islet autoantibodies and other clinical variables, such as body weight and history of sinusitis (228). The presence of multiple islet autoantibodies conveys a larger risk of developing type 1 diabetes than a single autoantibody (321). As such, children with seroconversion to any 2 autoantibodies have a 70% risk of developing type 1 diabetes during childhood or adolescence (322), with a yearly risk of approximately 10% (314). While higher titers are likely associated with increased risk across a population, titers have less predictive power in the individual. Complicating their use in assessing diabetes risk, autoantibodies may appear in different order, depending on haplotype (208, 209, 323), and may disappear over time both in children and adults regardless of whether they eventually progress to diabetes (324). In general, the younger the age of seroconversion, the higher the risk of diabetes.

Islet autoantibody assessment helps establish a diagnosis of autoimmune diabetes and has reasonable power (in combination with family history and HLA type) to identify children at risk (228). This identification will be of particular importance as therapeutic strategies to prevent type 1 diabetes become available. As with T-cell assays discussed above, strict control of the quality and reproducibility of the autoantibody assays is necessary for them to be more commonly used in the clinical setting. Thus, the measurement of islet autoantibodies is currently important in both research and clinical practice to predict the development of type 1 diabetes. Further, assessment of islet autoantibody status, along with other biomarker measures including risk genes, T-cell signatures, and metabolic measures such as proinsulin, have now enabled the identification of disease heterogeneity in type 1 diabetes that holds promise for optimizing disease management and clinical trial design (325). However, while islet autoantibodies may provide some insight into type 1 diabetes pathophysiology, they are currently not a target for therapy, as they are not thought to contribute to β-cell loss.

Latent autoimmune diabetes in adults

While autoimmune diabetes classically presents in childhood or adolescence, it can also occur in adults. LADA, also termed type 1.5 diabetes, differs from adult-onset type 1 diabetes by the duration of insulin independence after diagnosis (326). This longer period of insulin independence is due to the fact that individuals with LADA are diagnosed later in life and typically have better β-cell function than those diagnosed with type 1 diabetes at a young age (327, 328). Thus, LADA has been proposed to be diagnosed when diabetes has an adult age of onset (>30 years), the presence of any islet autoantibody (most commonly GAD), and no need for insulin treatment for at least 6 months after diagnosis (329). When insulin replacement therapy is required within 3 months, the diagnosis of classic adult-onset type 1 diabetes is proposed, although the American Diabetes Association views all forms of diabetes mediated by autoimmune β-cell loss under the rubric of type 1 diabetes (4).

The prevalence of LADA is likely underestimated, as antibodies are infrequently measured in adults with hyperglycemia, and typically only when the phenotype does not match that of classical type 2 diabetes, with overweight/obesity and other aspects of the metabolic syndrome. However, cohort studies have shown that as many as 4% to 12% of people classified as type 2 diabetes can be autoantibody positive (5).

Islet autoantibodies are less frequently present in LADA compared to type 1 diabetes. While IAA, IA2, and ZnT8 autoantibodies usually become negative over time, GAD autoantibodies are not influenced by age and are the most sensitive marker for both adult-onset type 1 diabetes and LADA (330). Higher antibody titers correlate with the presence of higher-risk HLA genotypes (331) and a more severe phenotype at presentation including insulinopenia, higher HbA1c levels, and lower BMI. Measurement of antibodies may also help guide clinical treatment by predicting which patients will require insulin therapy.

The distinction between LADA and type 2 diabetes remains primarily based on clinical presentation and phenotype, with diagnostic testing for islet autoantibodies rarely carried out in obese individuals. With time and disease progression, the correct diagnosis will usually become apparent for most individuals.

Cytokines/Inflammatory Markers

The most common inflammatory markers linked to β-cell dysfunction include the interleukin family (including IL-1β, IL-6, and IL-8) TNF-α, NF-κB, IFN-γ, and chemokines such as CCL2, MCP1, and CXCL1. The plasma concentrations of these proteins are indeed elevated in people with type 2 diabetes (332); however, several issues arise when applying these biomarkers in clinical or research settings. Importantly, it is unlikely that islet macrophages make a meaningful contribution to circulating cytokine levels, as other tissue-resident and circulating macrophages make much bigger contributions. Moreover, the measurement of these markers of inflammation can be difficult, especially using multiplex approaches, with their measurement being predominantly performed in research settings and not routinely in clinical labs. Thus, while cytokines are involved in β-cell inflammation and dysfunction, their circulating levels are not useful as biomarkers, as they do not reflect what may be occurring at the level of the islet.

Pancreatic and Islet Imaging

Several methods for imaging the pancreas have been proposed as a means to quantify β-cell mass (Fig. 4). The basis for this strategy and the purported advantage is to capture the early stages of β-cell loss prior to any change in blood glucose levels. Also, it enables noninvasive, longitudinal assessment of islet viability in vivo, which may guide treatment strategies to mitigate the progressive decline in β-cell function observed in type 2 diabetes.

Figure 4.

Common imaging biomarkers used to assess β cells in humans. The targets and probes are provided for each type of measurement that are designed to provide measures of pancreas size and fat, β-cell mass, islet amyloid, and insulitis.

Imaging islets (and β cells) is especially challenging due to their small size and their relatively low abundance being spread throughout the pancreas. In humans, islets comprise only ~1% to 3% of the pancreas and approximately 50% of the islet is made up of β cells (113, 149, 333, 334). Thus, in order to successfully image islets in vivo, the technique used must be highly sensitive. Also, as islets comprise several cell types, to detect β cells per se, the imaging technique must also be specific. Positron emission tomography (PET) and single photon emission computed tomography (SPECT) are imaging modalities that are clinically applicable and offer high sensitivity, as well as the ability for signal quantification. Both employ tracer molecules labeled with radioactive nuclides. To date, existing radiolabeled molecules targeting the β cell have proven to be suboptimal for imaging in humans, as they often bind other islet cell types and different tissues.

Vesicular Monoamine Transporter Type 2

Vesicular monoamine transporter type 2 (VMAT2) has been suggested to be one of the more specific biomarkers for imaging β cells. However, there is heterogeneity in the proportion of β cells that express VMAT2; specifically, it has been reported that 81%, 96%, and only 53% of cells in the body, tail, and head of the pancreas, respectively, co-localize VMAT2 and insulin (335). Importantly, VMAT2 is not expressed in islet α or δ cells, although its utility as a specific β-cell marker is limited by expression in islet PP cells (335). Indeed, PP cell expression of VMAT2 may explain the disconnect in studies of type 1 diabetic subjects who had no measurable β-cell function, yet total pancreatic binding of the VMAT2 tracer [¹⁸F]fluoropropyl-(+)-dihydrotetrabenazine was ~40% (336, 337). Such residual background signal likely reflects nonspecific binding and yields data that may overestimate β-cell mass. Despite this, VMAT2 expression and its radioligand binding have been found to strongly correlate with β-cell density (338, 339). Thus, specific targeting of VMAT2 in the pancreatic body and tail, which are regions of lower pancreatic polypeptide (PP) abundance (154, 340, 341), may overcome the abovementioned limitations in order to reflect β-cell mass more accurately.

Using this targeting approach, a recent study assessed VMAT2 tracer density in regions of the pancreas in humans with prediabetes or type 2 diabetes and correlated these data to β-cell function and glycemic control (342). Uptake of the VMAT2 tracer in any of the 3 individual regions of the pancreas was positively correlated with the acute C-peptide response to glucose-potentiated arginine stimulation. The same pattern did not hold true for glycemic control, where increased uptake of VMAT2 tracer only in the head, but not the body or tail of the pancreas, correlated with lower HbA1c. Of note, increased uptake of VMAT2 tracer in the pancreas as a whole also correlated with lower HbA1c. An important observation in this study was that loss of VMAT2 tracer binding was modest in the subjects with type 2 diabetes, whose time since initial diagnosis averaged ~10 years (342). This raises the question of when β-cell imaging would be most useful in the clinical course of type 2 diabetes, in order to inform treatment strategies.

Glucagon-like Peptide-1 Receptor

Like VMAT2, the glucagon-like peptide-1 receptor (GLP-1R) has been the subject of continued efforts in imaging human β cells in vivo. In the pancreas, GLP-1R is predominantly expressed on β cells, but there are also reports of its expression in exocrine tissue and other endocrine cells of the pancreas, including α and δ cells (343-345). This expression pattern therefore poses some challenges in targeting GLP-1R to specifically image β cells in vivo.

Several agents that bind GLP-1R have been developed for β-cell imaging with PET and SPECT (346). The vast majority are stable GLP-1 analogs based on exendin-3 and -4 peptides, which have been designed to prevent the rapid degradation seen with native GLP-1 (347). Due to their agonistic potency at the GLP-1R, they have the potential to induce hypoglycemia when used in high doses (348). To circumvent this undesirable side effect, during imaging a continuous infusion of glucose can be co-administered (349). Alternatively, a small number of GLP-1R antagonists have been developed for use in imaging so as to avoid receptor activation. There has been varied success with these antagonists, with some exhibiting low specific uptake in rodents or an inability to label β cells from humans (350-352). Moreover, some GLP-1R–targeting radioactive agents (eg, exendin analogs) exhibit proximal duodenal uptake as well as high kidney uptake, the latter likely due to final elimination that occurs almost exclusively in the kidney (353, 354). This poses a problem, as it may obscure the adjacent pancreatic tail under imaging conditions of low resolution, resulting in inaccurate estimation of β-cell mass. To address this, recent advances in radionuclide labeling chemistry have allowed for strategies to markedly reduce kidney uptake using improved agents (eg, NOTA-MVK-Cys⁴⁰-Leu¹⁴-Exendin-4) (355).

To date, the caveats of GLP-1R as a biomarker for β-cell mass are similar to those of VMAT2. For example, nonspecific binding of a radiolabeled GLP-1 analog can occur. This possibility was suggested in a study involving ¹¹¹In-labeled-exendin in which pancreatic uptake was reduced in type 1 diabetic subjects when compared with healthy controls, but pancreatic radioactivity concentration of ¹¹¹In-labeled-exendin among individuals was overlapping between the 2 groups (356). Given that the type 1 diabetic subjects had undetectable nonstimulated and stimulated C-peptide levels, it is probable that binding of ¹¹¹In-labeled-exendin in these subjects was occurring in non-β-cell populations. Thus, in order to infer any relationship with β-cell function, further refinements are needed to accurately quantify β-cell mass with GLP-1R probes in diabetic subjects.

Of note, GLP-1R expression is up to 5 times greater in human insulinomas compared to normal β cells (357); this has enabled noninvasive diagnosis of insulinomas in humans (348, 358, 359). Importantly, the utility of GLP-1R-based imaging extends beyond that of quantifying β-cell mass and detecting insulinomas, since it can also inform on the extent of GLP-1R occupancy of therapeutic compounds that act as GLP-1R agonists (360). This may become useful in time to individualize treatment strategies, where one GLP-1 analog may be chosen over another due to enhanced GLP-1R binding.

Insulitis

In type 1 diabetes, autoimmune destruction of β cells is already markedly advanced by the time clinical symptoms manifest (361). Thus, imaging of β cells may enable early detection of insulitis in type 1 diabetes and inform on treatment strategies to curb β-cell loss.

Ultrasonography and computed tomography are imaging techniques used to visualize the pancreas in humans. While both can detect pancreatic anatomical and structural changes that occur during type 1 diabetes, neither can quantify changes at the level of individual β cells in vivo. More promising are anatomical imaging techniques such as magnetic resonance imaging (MRI), which can be used to infer the extent of islet microvascular dysfunction, as seen in insulitis (362). For example, MRI with a specific magnetic nanoparticle as a contrast agent, has been used to image insulitis in recently diagnosed type 1 diabetic subjects, where it was found that pancreatic volume was already reduced compared with nondiabetic subjects (363). In this case, MRI detects magnetic nanoparticles that migrate from leaky vessels into the surrounding tissue and are phagocytosed by inflammatory cells, especially macrophages. Recent optimization of this method has improved resolution and discrimination of pancreatic inflammation (364), although there is no correlation between MRI signals and either autoantibody titers or the number of autoantibodies detected in type 1 diabetic subjects.

By targeting specific cells or cell antigens involved in β-cell destruction in type 1 diabetes, the ability to image insulitis in humans is improved. For example, lymphomononuclear cell infiltration can be targeted using ^99mTc-labeled polyclonal immunoglobulins that recognize the Fc receptor in infiltrating lymphocytes. In humans, significant accumulation of labeled immunoglobulins in the pancreas was found in 7 out of 15 newly diagnosed type 1 diabetic subjects, where radioactivity in the pancreas correlated with metabolic, immunological, and clinical parameters (365). While encouraging, it is important to recognize that in various stages of insulitis, there may be a relatively low number of lymphocytes and/or only a small number of antigens expressed on lymphocytes. The latter may be a factor in other studies, where pancreatic accumulation of ^99mTc-labeled IL-2 was observed in 61% to 65% of people with newly diagnosed type 1 diabetes (366, 367). Importantly, metabolic or immunologic parameters in these subjects did not differ from that observed in subjects who were negative for pancreatic ^99mTc-labeled IL-2 accumulation (367). Also, pancreatic uptake of ^99mTc-labeled IL-2 did not correlate with autoantibody titer (366). Thus, in some cases, there appears to be a weak relationship between lesions observed via pancreas imaging and the clinical characteristics of diabetic subjects at the time of imaging. That said, there is still value in performing imaging for insulitis, as subjects with positive scintigraphy at diagnosis showed better long-term metabolic control when treated, compared with subjects with negative scintigraphy (367).

Amyloid

Islet amyloid is present in ~90% of subjects with type 2 diabetes and contributes to the loss of β-cell mass and function (151, 160, 161). Detection of islet amyloid in humans may provide a means to assess progression of diabetes in order to determine the most effective treatment strategies to halt β-cell loss. To date, few islet amyloid imaging probes have been tested in humans, despite preclinical studies showing some promise in in vitro studies and in humanized models of islet amyloid deposition (168-171).

In general, these probes are based on those developed for PET detection of amyloid β (Aβ) deposits in the brain, some of which also bind the main amyloidogenic peptide constituent of islet amyloid, namely IAPP. One example is the FDA-approved [¹⁸F]florbetapir (368, 369), which has shown significant qualitative and quantitative correlations between in vivo PET imaging and postmortem histopathologic analysis of amyloid β. In vitro studies show florbetapir binds synthetic human IAPP and endogenous islet amyloid deposits (169). In transgenic mice expressing human IAPP, florbetapir enabled PET detection of islet amyloid in vivo; however, the signal was greater only during the first 5 minutes of the PET scan in human IAPP transgenic mice vs nontransgenic mice that do not develop amyloid (169). Moreover, florbetapir uptake was not zero in mice that did not develop amyloid. The latter is a potential drawback for use of florbetapir in assessing islet amyloid in humans, as it could result in erroneous conclusions regarding the loss of β-cell mass.

In another study, the amyloid β probe, [¹⁸F]FDDNP, was found to selectively stain islet amyloid deposits in autopsy pancreas sections from a type 2 diabetic subject, with no staining evident in pancreas from a nondiabetic subject. When administered to nondiabetic humans, [¹⁸F]FDDNP exhibited a favorable pharmacokinetic profile, and allowed for the pancreas to be easily distinguished from neighboring organs using PET imaging (170). Unfortunately, this work did not include people with diabetes, precluding validation of whether [¹⁸F]FDDNP could be used to quantify islet amyloid in vivo.

Thus, further studies are needed before use of islet amyloid probes in the clinic. An obvious advantage of probes like florbetapir and [¹⁸F]FDDNP is that they enable detection of both islet amyloid and amyloid β and may thus offer the opportunity to simultaneously assess amyloid in both the pancreas and central nervous system in humans.

Islet Transplantation

Several imaging modalities have been used to detect and monitor transplanted islets. In early human studies, islets were labeled with iron nanoparticles prior to transplantation and visualization by MRI (370). Since this labeling approach relies on cellular uptake of nanoparticles via random endocytosis, both nonfunctional and functional islet cells are labeled—this limits the utility of imaging in terms of informing on the function and viability of islet cells. Also, iron nanoparticle labeling does not distinguish between β- and non-β cells. Despite this, clinical-grade iron nanoparticles have been developed and tested in MRI visualization of labeled islets transplanted into rodents and humans (371, 372). While the clinical-grade contrast agent offers improved sensitivity for detection, iron nanoparticles in rodents were shown to persist at the site of transplantation beyond the presence of intact islets (372). The latter is clearly a major drawback for long-term monitoring of islets in vivo, as it can result in overestimation of the number of viable islets present following transplantation. When tested in humans with type 1 diabetes with pretransplant negative C-peptide levels, clinical-grade iron nanoparticles were detectable in vivo up to 24 weeks posttransplant (373). Importantly, all subjects exhibited significant C-peptide production at 24 weeks, suggesting the transplanted islets were still viable.

From a clinical perspective, MRI is an attractive imaging modality, since it does not require ionizing radiation and longitudinal measures can be made, owing to the sustained presence of contrast agents within the body. However, PET is more sensitive than MRI, and β-cell–specific probes can be used with PET to differentiate β cells from non-β cells. The utility of such probes (eg, targeting VMAT2 and GLP-1R) is described above. Another probe commonly used for PET imaging is [¹⁸F]fluorodeoxyglucose, which is a glucose analog that enters cells with high rates of glucose utilization but is not fully metabolized and is thereby trapped. In this way, distribution of [¹⁸F]fluorodeoxyglucose is an indicator of the glycolytic rate of cells (374). In a study of 5 type 1 diabetic subjects, PET was combined with computed tomography to image this radiotracer in islets transplanted intraportally (375). The radioactivity concentration in the liver corresponded to only 75% of the expected dose, suggesting loss of transplanted islets. Also, distribution of radioactivity in the liver was heterogeneous, with wide variations in location and concentration, including in regions that may represent islets trapped in sinusoids or clots in the portal branches. While informative with respect to success of the islet transplant procedure per se, the use of [¹⁸F]fluorodeoxyglucose does not aid in monitoring function of transplanted islets over time, largely due to the short half-life of ¹⁸F (110 minutes) and retention of [¹⁸F]fluorodeoxyglucose in islets (196 minutes). Further, due to its reliance on cell metabolic rates, it lacks specificity to islets vs surrounding tissues.

Pancreatic Size and Fat

Human autopsy studies have demonstrated that pancreas size is reduced in both type 1 and type 2 diabetes (113, 376). Some (377-380), but not all (381, 382), data from pancreas imaging by ultrasound, computed tomography, and MRI have been consistent with findings from autopsy studies. A recent systematic review and meta-analysis of imaging studies suggested the discrepant literature might be due to a small sample size in most studies (383). Related to this point is the inter-individual heterogeneity in pancreas volume; thus, longitudinal assessments within individuals may provide a more robust method for determining disease progression, rather than comparing the magnitude of pancreas volume reduction between nondiabetic and diabetic subjects.

With respect to pancreatic size as a biomarker for β-cell function, a study in type 2 diabetes subjects showed that pancreas volume positively correlated with homeostatic model assessment (HOMA)-B, as an estimate of β-cell function (379). While these data support a correlation between pancreatic size and β-cell function, there is evidence that this does not translate to all situations. Specifically, in subjects who underwent weight loss as a means to reverse their diabetes, there was no detectable increase in pancreas volume despite a return to normal insulin secretion after 6 months (384). The latter suggests pancreas size may not be a useful surrogate for β-cell function, even when assessed longitudinally. Another caveat with pancreas size as a β-cell function biomarker is that it largely reflects changes in the exocrine pancreas, rather than islets, which constitute only ~1% to 3% of the pancreas (113, 333).

Pancreatic fat has similarly been proposed as a potential biomarker in diabetic subjects. Various imaging modalities are used to noninvasively quantify pancreatic fat, and it has now been shown in several studies that pancreatic fat content is increased in type 2 diabetes (383, 385). Specifically, when compared with nondiabetic subjects, pancreatic fat was found to be higher in those with prediabetes and highest in those with type 2 diabetes, indicating that it tracks with disease progression (386). When assessed in subjects with impaired glucose tolerance (IGT) or IFG, pancreatic fat was negatively associated with insulin secretion, suggesting that it could contribute to β-cell dysfunction (387). However, in another study in which subjects were categorized on the basis of an oral glucose tolerance test as having NGT, prediabetes, or type 2 diabetes, the presence of pancreatic fat was not related to the dysglycemic state (388). Indeed, there are other conflicting reports on the association between pancreatic fat accumulation and β-cell function, with some (389), but not all (390-392), showing a significant correlation.

These discrepancies among studies of pancreatic fat may be due to differences in methodology used to assess either pancreatic fat, β-cell function, or both, and/or in how the data analyses were conducted. For example, only in some cases were confounders like age, sex, and BMI accounted for when determining the association between pancreatic fat and β-cell function or glycemic status. Also, many studies were cross-sectional, making it difficult to truly relate accumulation of pancreatic fat with deterioration of β-cell function over time. To this point, in a longitudinal study conducted over a period of 5 years, pancreatic fat was not independently associated with future type 2 diabetes (393). More recently, an additional factor has come into play, namely genetic predisposition to diabetes. That is, a study showed that pancreatic fat only impairs β-cell function in subjects with high genetic risk for diabetes (394). Taken together, without additional (longitudinal) studies, the use of pancreatic fat as a biomarker for development or progression of β-cell failure remains premature.

Summary

In sum, several approaches for imaging the pancreas and islets are being studied for estimating β-cell mass or function. While promising in terms of capturing the early stages of β-cell loss and failure, the techniques largely lack sufficient sensitivity and/or specificity to be clinically useful at this time. With technical advances in the field, it is conceivable that select imaging biomarkers could in the future prove beneficial in preventing and managing diabetes.

Circulating β-Cell Propeptides

The biosynthesis of insulin has been well studied, while that of IAPP much less so. Our knowledge of proinsulin biosynthesis and processing has helped guide research into understanding proIAPP biosynthesis, and while they have extensive overlap, there are distinct differences. In the following discussion, we focus first on proinsulin processing and thereafter on IAPP. Some discussion of the normal biosynthesis of these prohormones is presented in the section describing the normal physiology of the β cell.

Assessment of Proinsulin Processing

Healthy humans

Depending on the initial cleavage step, proinsulin processing can proceed by 1 of 2 intermediate steps, via des-31,32 proinsulin (produced by PC1/3 cleavage) or des-64,65 proinsulin (predominantly produced by PC2 cleavage) (Fig. 5) (395, 396). Cleavage by PC1/3 or PC2 occurs on the C-terminal side of pairs of basic residues, initially producing the split cleavage products, split-32,33 proinsulin or split-65,66 proinsulin, respectively. The remaining basic residue pairs are rapidly trimmed by carboxypeptidase E (CPE) to produce des-31,32 or des-64,65 proinsulin (397). Processing via the des-31,32 proinsulin intermediate predominates, as little, if any, des-64,65 proinsulin is detectable in human circulation in health or in diabetes (136). Evidence from studies in human islets (39) and mouse models of prohormone convertase deficiency (42, 43), indicate that PC1/3 is able to completely process proinsulin to insulin and C-peptide. PC2 expression is low in human β cells, and it appears dispensable for complete proinsulin processing in healthy human β cells (39).

Figure 5.

Pathways for β-cell prohormone processing. A, Pathway for proinsulin processing in β cells. The predominant pathway for proinsulin processing to insulin and C-peptide involves cleavage of proinsulin by prohormone convertase (PC) 1/3 on the C-terminal side of basic residues at positions 31 and 32, followed by the removal of these basic residues by carboxypeptidase E (CPE), leading to the production of the proinsulin intermediate, des-31,32 proinsulin. This intermediate is then cleaved by PC2, or in human β cells may be cleaved by PC1/3, leading to production of one molecule of insulin and one of C-peptide. Another possible pathway involves cleavage of proinsulin by PC2 on the C-terminal side of basic residues at amino acids 64 and 65 which, following removal of these residues by CPE, is predicted to lead to des-64,65 proinsulin, although this pathway is not thought to be active in healthy β cells as the des-64,65 proinsulin intermediate is undetectable in human plasma. B, Pathway for proIAPP processing in β cells. Like proinsulin, proIAPP_1-67 is first processed by PC1/3 on the C-terminal side of a pair of basic amino acids, near the C-terminus of the propeptide. Following removal of these basic residues by CPE, the N-terminally extended proIAPP intermediate proIAPP_1-48 is formed. ProIAPP_1-48 processing to mature IAPP is dependent on PC2 in murine β cells. Mature IAPP (and possibly proIAPP_1-48) is amidated in secretory granules at the C-terminus by peptidyl α-amidating monooxygenase (PAM).

Distinct measurement of intact proinsulin and the proinsulin intermediates in human plasma has provided some insight into the nature of processing defects that exist in diabetes, but current assays do not discriminate well between these forms. The majority of commercially available proinsulin assays measure both intact proinsulin and des-31,32 proinsulin (the predominant intermediate form in human circulation), while excluding des-64,65 proinsulin. While these assays enable measurement of both of the major circulating forms of proinsulin immunoreactivity in the same sample, they do not allow specific quantification of the individual proinsulin species. In several of the existing immunoassays for proinsulin immunoreactivity, little data are provided regarding cross-reactivity among the various proinsulin forms. Our current knowledge regarding these assays and their performance as provided by the manufacturers is listed in Table 3. Further advancements toward understanding β-cell prohormone processing in diabetes will require development of approaches, such as mass spectrometry (80), that are able to better discriminate the different forms in human plasma while at the same time being sensitive and specific.

Table 3.

Characteristics of current human proinsulin and C-peptide assays

			Propeptide cross-reactivity					Mature peptide cross-reactivity
Manufacturer	Catalog number	Sensitivity	Intact proinsulina	split-32,33	des-31,32	split-65,66	des-64,65	Mature insulin	C-peptidea
Proinsulin
Abcam	ab242235	0.75 pmol/L	-	NS	NS	NS	NS	>50 ng/mL	>50 ng/mL
Alpco	80-PINHUT-CH01	0.455 pg/mL	-	NS	100%	NS	100%	<0.6%	<0.1%
EMD Millipore	EZHPI-15K	0.5 pmol/L	-	NS	100%	NS	81%	>200 µU/mL	>10 ng/mL
Invitron	IV2-002	0.02 pmol/L	-	5.6%	1.4%	37%	63%	0%	0%
Mercodia	10-1118-01	1.7 pmol/L	-	95%	95%	90%	84%	<0.03%	<0.006%
Meso Scale Diagnostics	K1516MK	0.05 pmol/L	-	NS	NS	NS	NS	<0.5%	0.7%
R&D Systems	DPINS0	1.43 pmol/L	-	NS	NS	NS	NS	>3450 pmol/L	NS
TECOmedical Group	TE1012	0.3 pmol/L	-	5000 pmol/L	<200 pmol/L	1000 pmol/L	200 pmol/L	<10 000 pmol/L	50 000 pmol/L
C-peptide
Alpco	80-CPTHU-CH01	<4.32 pg/mL	<0.01%	NS	0.3%	NS	33.2%	ND	-
Alpco	80-CPTHU-E01.1	2.95 pmol/L	3%	NS	NS	NS	NS	<0.01%	-
Beckman Coulterb	C33451	0.01 ng/mL	3%	NS	NS	NS	NS	0%	-
Invitron	IV2-004	5 pmol/L	2%	NS	NS	NS	NS	0%	-
Mercodia	10-1136-01	≤25 pmol/L (0.076 μg/L)	2%	2%	3%	10%	74%	<0.0006%	-
Mercodia	10-1141-01	2.5 pmol/L (0.0076 µg/L)	5%	2%	3%	10%	74%	<0.0006%	-
Meso Scale Diagnostics	K1516JK	14 pg/mL	27%	ND	ND	ND	ND	0.50%	-
Meso Scale Diagnostics	K151X5D	4.72 pg/mL	32.4%	NS	NS	NS	NS	0.03%	-
Tosoh	25284	0.2 ng/mL	31.5%	NS	NS	NS	NS	ND	-

Information provided is as stated by the manufacturers and not independently verified. Stated information was accessed from manufacturer websites and supporting technical documents during the period February 10, 2021 to March 7, 2021.

Cross-reactivity values reported as percentages represent (100%) (detected proinsulin concentration)/(non-target analyte concentration tested).

Cross-reactivity values reported as concentrations represent the minimum required concentration of non-target analyte to be detected by the ELISA.

Abbreviations: ND, not detected; NS, not stated.

^aIn the case of proinsulin assays, it is assumed to be 100% and for C-peptide assays similarly it is assumed to be 100%

^bRadioimmunoassay

Using the few approaches that distinguish between these molecular forms of proinsulin, insight has been gained into the 2-step pathway for proinsulin processing. With peptide separation using high performance liquid chromatography (HPLC), some important observations have been made. First, in the basal state, about 15% of total insulin-like immunoreactivity is comprised of proinsulin, while following acute stimulation with arginine it is approximately 4% (136). The latter reflects more closely what is in the secretory granule, with the difference from fasting being a result of differences in the clearance rates of the peptides. Second, approximately 50% of proinsulin immunoreactivity in fasting plasma from healthy humans consisted of intact proinsulin and 50% of des-31,32 proinsulin, with no detectable des-64,65 proinsulin (136). In plasma obtained following acute stimulation, the proportion of proinsulin immunoreactivity comprising des-31,32 proinsulin was approximately 70%. Taken together, these data suggest that a significant proportion of proinsulin immunoreactivity is comprised of the des-31,32 proinsulin intermediate produced by PC1/3 cleavage and therefore in human β cells, the second step in proinsulin processing appears to be rate-limiting. These findings and the conclusions built on them are supported by older studies using immunoassays that allowed for a derivation of the relative amounts of intact and des-31,32 proinsulin (398, 399).

In evaluation of β-cell prohormone processing using currently available immunoassays, the data are often expressed as an absolute value. However, to gain an assessment of the efficiency of proinsulin processing, concentrations need to be presented relative to insulin (proinsulin:insulin) or C-peptide (proinsulin:C-peptide) as absolute proinsulin levels alone will reflect not only processing but will also represent secretory demand imposed by the prevailing glucose and insulin sensitivity (134, 400). A caveat in considering the proinsulin:insulin ratio is the difference in clearance rates of the peptides that could confound interpretation of the ratio, with reduced insulin clearance in, for example obesity, potentially reducing the ratio. This difference in clearance is less of a potential problem when using C-peptide instead of insulin, as the former is not cleared by the liver on first pass (401, 402). The ratio of proinsulin and insulin has its most discriminatory value when examining individuals with more severe β-cell dysfunction (134, 135, 403).

There have been a number of clinical studies on the impact of insulin resistance and increased β-cell secretory demand on the proinsulin:insulin ratio as a measure of processing efficiency in healthy humans. In the absence of a β-cell defect, this ratio has been found to be lower in the fasting state and following acute stimulation in humans with NGT who were obese (134, 404) or had nicotinic acid–induced insulin resistance (405), likely reflecting more efficient proinsulin processing as a marker of enhanced β-cell function in the face of prolonged, elevated secretory demand. Induction of a β-cell defect by streptozotocin treatment in baboons resulted in an increase in the plasma proinsulin:insulin ratio (406). Glucocorticoid treatment not only induces insulin resistance, but also produces β-cell secretory dysfunction and an increase in the proinsulin:insulin ratio (133, 407-410), the latter compatible with a proinsulin processing defect. A similar effect on the ratio has been observed with growth hormone, suggesting that it may also have a deleterious effect on proinsulin processing (410). Prolonged increased secretory demand in the face of reduced β-cell mass created by a hemi-pancreatectomy in humans has been associated with an increase in proinsulin:insulin in those at increased risk of diabetes (411), but not in those without diabetes or a risk thereof (412, 413). Clearly, multiple complex factors impact the proinsulin:insulin and proinsulin:C-peptide ratios in humans, with the available data suggesting that moderately increased β-cell secretory demand (eg, induced by obesity or insulin resistance) does not on its own impair proinsulin processing in individuals with healthy β cells. However, as discussed subsequently, in the presence of impaired β-cell function as with corticosteroid treatment or both major forms of diabetes, an impairment in the efficiency of proinsulin processing is evident.

Type 2 diabetes

Studies of proinsulin in type 2 diabetes have been undertaken for many years and have consistently demonstrated that the proinsulin:insulin ratio is increased in this disease, with the magnitude of increase being associated with glycemia and β-cell dysfunction (131-137). Thus, an increasing ratio suggests a more profound impairment in proinsulin processing in type 2 diabetes, reflecting the presence of a marked β-cell defect. The deleterious impact of impaired proinsulin processing is made more clinically significant, given that the secreted prohormone has only 10% to 15% of the biological activity of insulin (414, 415). The ratio has also been reported to be increased in individuals at high risk of developing type 2 diabetes, including those with impaired glucose tolerance or normoglycemic first-degree relatives of individuals with the disease (398, 416-418). Thus, an alteration in the efficiency of proinsulin processing occurs before the development of frank hyperglycemia and could be a marker for the development of type 2 diabetes (417, 419, 420).

The mechanism underlying impaired proinsulin processing and elevated circulating proinsulin:insulin ratios in type 2 diabetes is not fully understood. The increased ratio cannot be simply explained by differences in the clearance of proinsulin vs insulin in people with and without type 2 diabetes, given that when β cells are stimulated to release their granule content, the ratio (closely reflecting β-cell content) remains 3- to 4-fold higher in type 2 diabetes (136). Thus, there has to be an intrinsic abnormality that results in inefficient proinsulin processing along the secretory pathway.

While understanding at a cellular level in humans what may be occurring with proinsulin processing is difficult, clues have been obtained from measurement of the different forms of proinsulin in the circulation. If the proinsulin processing defect in type 2 diabetes resides early in the secretory pathway or is due primarily to loss of PC1/3 activity, one might expect an increase in intact proinsulin relative to the subsequent processing intermediate (des-31,32 proinsulin). Conversely, if the processing defect is more distal in the secretory pathway, or perhaps dependent on increased secretory demand or PC2 cleavage, we might expect a relative increase in this intermediate form vs intact proinsulin. Clues to where the defect may exist have come from limited human studies examining circulating proinsulin and its conversion intermediates. Following HPLC separation of samples obtained immediately following arginine stimulation, the proportion of total proinsulin immunoreactivity consisting of intact proinsulin was 40% in people with type 2 diabetes compared with 30% in healthy controls (136). This observation was made in a small number of subjects and has also been observed in studies using immunoassays for the different forms of proinsulin (137, 399). Collectively, these observations suggest the impairment in proinsulin processing likely involves PC1/3 and is early in the processing pathway. Single cell transcriptomic and immunostaining studies have not shown PC1/3 mRNA or protein concentrations to be markedly decreased in β cells in type 2 diabetes (40, 41), suggesting that the processing defect is probably due to reduced PC1/3 enzyme activity. A reduction in enzyme activity could be due to a change(s) occurring anywhere in the secretory pathway, including the Golgi and secretory granule where pH and calcium are critical determinants of PC1/3 activity (36, 421).

Aside from an intrinsic processing defect, a contribution from β-cell secretory stress to the elevation in the proinsulin:insulin ratio in type 2 diabetes is also likely. This concept is supported by human studies in which β-cell rest, by inhibition of secretion with somatostatin, partially lowered the proinsulin:insulin ratio in type 2 diabetic subjects (422, 423).

Type 1 diabetes

While type 1 diabetes has for many years been thought to be a disease of absolute β-cell loss, studies of human type 1 diabetes donor pancreas from limited autopsy series (73, 424), the Network of Pancreas Organ Donors in Diabetes (nPOD) repository (69, 425), and the Diabetes Virus Detection (DiViD) Study (426), have revealed considerable and variable residual β-cell mass remaining in most type 1 diabetes cases, particularly those with later-age onset of disease. The remaining β cells are insufficient to maintain normoglycemia, clearly indicating they are dysfunctional (427).

A number of clinical studies have demonstrated that, despite near loss of insulin secretion and detectable C-peptide, people with type 1 diabetes can continue to have persistent secretion of proinsulin, even in longstanding disease and in the absence of detectable C-peptide (80-84). Thus, the proinsulin:C-peptide ratio can be markedly elevated when severe secretory dysfunction is present in type 1 diabetes. Furthermore, in autoantibody positive humans, the proinsulin:C-peptide ratio is elevated, more so in those who subsequently progressed to type 1 diabetes (428-430). Impairments in proinsulin processing may therefore also be an early occurrence in the pathophysiology of type 1 diabetes and have value as a predictive biomarker.

Elucidation of the mechanisms underlying continued secretion of proinsulin in type 1 diabetes is an active area of investigation that may provide insight into type 1 diabetes pathogenesis as well as the value of proinsulin as a type 1 diabetes biomarker. Immunostaining studies of pancreas from type 1 diabetes donors from the DiViD Study (431) and from the nPOD biobank (425, 431, 432) have revealed islet cell populations in type 1 diabetes that are rich in proinsulin immunoreactivity but devoid of immunoreactivity for mature insulin, suggesting that populations of residual cells in type 1 diabetes are able to synthesize but not process proinsulin. It has been proposed that these cells may be “sleeping,” “degranulated,” or “dedifferentiated” and therefore have suboptimal function (433-435). Further, this state may enable them to escape autoimmune attack because of decreased production and presentation of key autoantigens. Mass spectrometry analysis of laser-dissected tissue revealed decreased PC1/3 and CPE in type 1 diabetes islet cells, suggesting that β cells in this disease may lack optimal machinery for prohormone processing (432). In support of this, analysis of gene expression in RNA extracted from frozen sections of type 1 diabetic pancreas indicated lower PC1/3 expression, associated with elevated proinsulin:C-peptide protein ratios in the extracted pancreas tissue (425). Although loss of PC1/3 activity might be predicted to lead to increased levels of intact proinsulin, to date there are no data that provide insight into the relative proportion of intact proinsulin vs the des-31,32 proinsulin conversion intermediate in type 1 diabetes.

One plausible driver of the loss of prohormone convertase expression and impaired proinsulin processing could be the impact of proinflammatory cytokines such as IL-1β and TNF-α (436), which have been implicated in type 1 diabetes (437-439). These cytokines induce the loss of processing enzymes in human islets, which is associated with impaired proinsulin processing (432, 436). Another possible contributor is ER stress, markers of which are detectable in type 1 diabetic pancreas (440) and which is hypothesized to contribute to β-cell dysfunction and persistent proinsulin secretion in type 1 diabetes (441). This form of cellular stress has been shown in rodent islets to impair proinsulin processing (442). Further, in the nonobese diabetic (NOD) mouse, a model of type 1 diabetes, ER stress markers have been observed in islets and are associated with elevations in the proinsulin:insulin ratio even before disease onset (443). Lastly, islets treated with proinflammatory cytokines showed reduced sarcoendoplasmic reticulum pump Ca²⁺ ATPase 2b (SERCA2b) transcript (444). Such loss of SERCA2b and reduced ER calcium levels may contribute to changes that impede vesicle transport and proinsulin processing (445).

Pancreas and islet transplantation

The plasma proinsulin:C-peptide ratio has been reported to be elevated in type 1 diabetic recipients of islet transplants (446), suggesting that transplanted islets harbor a processing defect. The defect appears to be exacerbated in those who received less islet mass, likely related to them having increased secretory stress on the graft. In keeping with this, the proinsulin:insulin ratio has been shown to be normal in islet transplant and whole pancreas recipients that maintained insulin independence (447-449). However, the use of insulin assays in determining the efficiency of proinsulin processing in transplant recipients is likely to be confounded by exogenous insulin and care must be taken in interpreting these data (449). It is also possible that multiple other factors may contribute to impairments in islet graft function and impaired proinsulin processing, including the toxic effects of immunosuppressive agents, allo- and autoimmune responses to the islet allograft, as well as glycemic control (450-452). Interestingly, the proinsulin:C-peptide ratio was shown to be markedly elevated in a small cohort of recipients of autologous islet transplants (446). These recipients typically receive fewer islets and do not have any anti-islet graft immune response, nor do they receive immunosuppression. This finding suggests that secretory stress may unmask processing inefficiencies in transplanted islets, and moreover points to the importance of transplanting sufficient islet mass. Collectively, these data suggest that the proinsulin:insulin (or C-peptide) ratio has potential as a biomarker of islet graft function; whether it has value in predicting graft failure remains to be determined.

While the mechanism underlying impaired proinsulin processing in islet grafts remains unknown, it may mirror those in diabetes. One study reported lack of PC2 immunostaining in β cells in human islets transplanted into immune-deficient mice (453). As for other forms of diabetes, our understanding of the nature of any processing defect in islet transplants would be enhanced by the ability to more readily measure the different proinsulin forms in peripheral plasma.

Assessment of ProIAPP Processing

Healthy humans

Our knowledge of proIAPP processing to intact IAPP is more rudimentary than that of insulin production from proinsulin. Based on what is known about proIAPP processing from cell culture and mouse studies, IAPP is initially expressed as the 67-amino acid proIAPP (454, 455), which is cleaved to produce mature IAPP by the same proprotein convertases (PC1/3 and PC2), CPE, and PAM (456-459) (Fig. 5). Although the proIAPP processing pathway largely parallels that of proinsulin in the β cell, studies in mice suggest that PC2 is more critical in the final step in proIAPP cleavage (460).

The molar ratio of IAPP to insulin in the circulation of healthy humans is about 1% to 3% (461-463). Accordingly, the abundance of proIAPP in the circulation is lower than proinsulin, making it more difficult to measure and to discriminate the different conversion intermediates. Thus, there have been few studies reporting the physiology of proIAPP in humans. The recent development of discriminatory assays will now open the field for better understanding (464).

Diabetes

Using an enzyme-linked immunosorbent assay (ELISA) specific for a C-terminally processed, N-terminally extended, intermediate form of proIAPP (proIAPP_1-48), the ratio of proIAPP_1-48 to mature IAPP was shown to be increased in the peripheral circulation of people with type 1 diabetes (464). This pattern would suggest a decrease in the efficiency of the second step in proIAPP processing, which based on studies in mice is predicted to be mediated by PC2 (460, 465). Complete understanding of the nature of the proIAPP processing defect in type 1 diabetes awaits measurements of intact proIAPP_1-67 as well as confirmation that the proIAPP processing pathway in human β cells is the same as in mice. The finding of an elevated ratio of proIAPP_1-48:IAPP in type 1 diabetes mirrors findings of elevated proinsulin:C-peptide ratios (83). As mentioned previously for proinsulin, more information on the molecular forms of proinsulin that are altered in type 1 diabetes is required (81). In keeping with the clinical findings, evidence for impairments in proIAPP processing has been observed in human islets under similar conditions (elevated glucose and cytokines) in which proinsulin processing has been shown to be impaired, including elevated glucose and cytokines (466, 467).

While impaired proIAPP processing has been demonstrated in type 1 diabetes, interestingly no change in the ratio of proIAPP_1-48:IAPP was observed in type 2 diabetes (464). These data do not rule out the possibility that other proIAPP forms, in particular intact proIAP_1-67, may be disproportionately elevated in type 2 diabetes. The lack of a difference in this particular ratio contrasts with another study that reported an elevated ratio of proIAPP:IAPP in type 2 diabetes (468). In this latter study, this ratio was 144% in individuals with NGT and 269% in those with type 2 diabetes. Reasons for this disparity are not clear, but the latter study did not report any characteristics of either the proIAPP or IAPP assays employed, including the specificity of the antibodies for the different proIAPP forms or IAPP. Specific measurement of intact proIAPP_1-67 and the proIAPP_1-48 intermediate in type 2 diabetes promises further insight into the nature of the prohormone processing defects in type 2 diabetes and whether the defect in PC1/3 activity is predominant as for proinsulin. Finding increased proIAPP_1-67 in type 2 diabetes would be compatible with a PC1/3 abnormality.

Histological approaches may provide some insight into proIAPP processing defects in diabetes, including whether proIAPP and proinsulin are persistently expressed in the same β cells. Limited work using human pancreatic sections has shown immunoreactivity toward both N- and C-terminal flanking regions of proIAPP in β cells of nondiabetic pancreas donors (469, 470), as well as immunoreactivity to the C-terminal flanking peptide in type 2 diabetes (471). Such immunohistological studies must be interpreted with caution because the antisera used have different specificity toward (pro)IAPP peptides.

Islet transplantation

The level of proIAPP_1-48 has been found to be disproportionately elevated in recipients of islet transplants (464), as was observed for proinsulin (452). However, whether the magnitude of the abnormality is a marker for the long-term outcome of the transplant is not clear. As we make progress toward the use of stem cell–derived β-like cells in transplantation, it is possible that measuring proIAPP and/or IAPP may be a useful marker for maturation of these cells. This possibility is supported by the recent finding that IAPP expression may be a marker of β-cell maturity in transplanted stem cell–derived β-like cells (472).

Summary

Proinsulin processing has been shown to be abnormal in the major forms of diabetes and following transplantation. The proIAPP processing machinery is likely also affected in diabetes. The value of proinsulin as a biomarker is likely to increase as approaches such as mass spectrometry are developed that allow the measurement of intact proinsulin, its conversion intermediates, and insulin in a single assay. The continued development of ELISAs and peptidomic approaches specific for the different proIAPP forms should facilitate a deeper understanding of the biology of proIAPP processing in humans and their potential as biomarkers of β-cell dysfunction in both type 1 and type 2 diabetes.

Measurement of β-cell Peptides in the Fasting State and Using Dynamic Testing

The development of the radioimmunoassay and its subsequent use for the measurement of insulin heralded a whole new era of understanding in the role of the β cell in the pathogenesis of diabetes. Until that time, it had largely been considered that type 1 diabetes was characterized by absolute insulin deficiency, whereas in the case of type 2 diabetes it was more uncertain. Based on measurements made in healthy subjects and those with “early maturity-onset diabetes,” it became apparent that what we today know as type 2 diabetes was characterized by a reduction in early insulin release in response to oral glucose (473, 474). This knowledge was gleaned from dynamic testing, which is an approach that is still of great benefit today and used commonly in people at risk of and with different forms of the disease.

While the development of the insulin assay has greatly advanced our understanding, today measuring insulin continues to remain a challenge despite efforts to standardize doing so (475, 476). Interestingly and importantly, such differences in insulin values can even occur when using the same assay in different laboratories (475). The general lack of standardization means that currently available assays may give quite different values for insulin on the same sample. Thus, when considering β-cell function based on insulin immunoassays, one has to be cognizant of the impact of this lack of standardization, which can make it extremely difficult to compare functional measures between studies.

Another important consideration in the interpretation of tests that assess the functional status of the β cell is that its responses need to be considered as a component of an integrated system. Doing so can markedly alter the interpretation of β-cell responses and has underscored the importance of this endocrine cell in the pathophysiology of diabetes as well as the outcome of interventional approaches to prevent and treat hyperglycemia.

Lastly, as introductory comments, brief descriptions and the coefficients of variation of selected measures of β-cell response, insulin sensitivity, and β-cell function discussed below are listed in Tables 4 and 5, respectively.

Table 4.

Descriptions of plasma-based measures commonly used in humans to assess β-cell responses, insulin sensitivity, and β-cell function

Measure	Description	References
Measures Based on an OGTT
β-cell responses
∆Insulin0-30/ ∆glucose0-30	Early insulin response, also known as the insulinogenic index and insulin-to-glucose ratio (IGR). It is measured as the ratio of the increment in insulin relative to glucose during the first 30 minutes following glucose ingestion. It represents a mixture of first- and second-phase insulin secretion as initially defined from intravenous glucose administration.	(477-479)
∆C-peptide0-30/ ∆glucose0-30	Early C-peptide response, which is similar to the early insulin response but is less affected by differences in hepatic insulin clearance.
IncAUCinsulin/IncAUCglucose	Ratio of the incremental area under the curve (IncAUC) for the insulin response during the whole OGTT as a function of the IncAUC for the glucose response. It provides an estimate of the efficiency of the β-cell’s responsiveness over a prolonged period of glucose exposure. The incremental insulin response is preferred to the total insulin response because the latter includes fasting insulin and is thus not as reflective of the β-cell’s response to the glucose stimulus.	(478, 480)
IncAUCC-peptide/IncAUCglucose	Similar to the measure obtained using insulin, but less affected by differences in hepatic insulin clearance.
CIR (corrected insulin response)	Calculated as (100 × 30-min insulin)/(30-min glucose × [30-min glucose − 70 mg/dl])	(481)
Insulin sensitivity
HOMA1-IR	Equation simplified to allow calculation as HOMA1-IR = (FPI × FPG)/22.5, where fasting plasma insulin (FPI) is in µU/mL and fasting plasma glucose (FPG) is in mmol/L. This measure provides a fasting-based measure of insulin sensitivity. As it is an index of “insulin resistance,” low values indicate insulin sensitive and high values insulin resistant.	(482)
IS (insulin sensitivity index)	Calculated as 22.5/(FPI x FPG), where fasting plasma insulin (FPI) is in µU/mL and fasting plasma glucose (FPG) in mmol/L]. It is the inverse of HOMA-IR. Thus, high values indicate insulin sensitive and low values insulin resistant.
1/fasting insulin	A surrogate measure of insulin sensitivity in humans that is highly correlated with the insulin sensitivity index determined using the minimal model of glucose kinetics developed by Bergman and colleagues.	(109)
Matsuda index	This measure is calculated from the glucose and insulin values during an OGTT using the formula 104/(I0 × G0 × Im × Gm)1/2, where G0 and Gm are the fasting and mean glucose and I0 and Im are the fasting and mean insulin.	(483)
β-cell function (integrating insulin response and insulin sensitivity)
HOMA1-%B	Based on fasting samples, a simplified equation to calculate HOMA1-%B = (20 × FPI)/(FPG − 3.5), where fasting plasma insulin (FPI) is in µU/mL and fasting plasma glucose (FPG) is in mmol/L. This measure provides a fasting-based measure of the β-cell’s function relative to the prevailing glucose concentration.	(482)
DIo (oral disposition index)	Product of the insulinogenic index and either 1/fasting insulin or HOMA-IR to provide a measure of β-cell function. Importantly, one cannot simply multiply 2 variables together unless it has been proven that the relationship is hyperbolic with a slope not different to −1.	(110)
ISSI-2 (insulin secretion-sensitivity index-2)	Ratio of the area under the insulin curve to the area under the glucose curve (as a measure of the β-cell response) multiplied by the Matsuda index to provide an estimate of β-cell function.	(484)
Model-derived parameters from an OGTT
β-cell responses
HOMA2-%B	Based on fasting samples, a nonlinear model that utilizes either C-peptide or insulin with paired glucose measurement to provide an estimate of the β-cell’s function. This version incorporates an estimate of proinsulin secretion and thus allows use of either total or specific insulin assays. Further, it accounts for renal glucose losses, thus allowing its use in hyperglycemic subjects. Available at https://www.dtu.ox.ac.uk/homacalculator/.	(485)
Glucose sensitivity	The slope of the curve relating the rate of insulin secretion to standardized glucose concentrations during the test. It is calculated from the glucose and C-peptide data using a model developed by Mari and colleagues.	(486)
Rate sensitivity	An index of early insulin release that represents the dependence of the rate of insulin secretion on the rate of change of glucose concentration. It is calculated from the glucose and C-peptide data using a model developed by Mari and colleagues.	(486)
Potentiation	A time-varying factor expressing a potentiation effect upon insulin secretion. It accounts for the physiological processes that can acutely modify insulin secretion. It is calculated from the glucose and C-peptide data using a model developed by Mari and colleagues.	(486)
φ d	“Dynamic sensitivity,” a measure of the stimulatory effect of the rate of increase in glucose on the secretion of stored insulin, calculated using a model developed by Cobelli and colleagues.	(487)
φ s	“Static sensitivity” is a measure of the effect of glucose on β-cell secretion calculated using a model developed by Cobelli and colleagues.	(487)
Insulin sensitivity
HOMA2-%S	Nonlinear model that utilizes either C-peptide or insulin with paired glucose measurement to provide an estimate of insulin sensitivity. Available at https://www.dtu.ox.ac.uk/homacalculator/.	(485)
OGIS (oral glucose insulin sensitivity)	A simple method for calculating insulin sensitivity using glucose and insulin data that is in good agreement with clamp-based measurements. Available at http://webmet.pd.cnr.it/ogis/.	(488)
Measures based on intravenous glucose testing
β-cell responses
First-phase insulin response	The insulin response to an intravenous bolus of glucose, also known as the acute insulin response to glucose (AIRglucose). This response occurs over the first 10 minutes following glucose administration. This measure can also be determined using C-peptide.	(55, 477, 489, 490)
Second-phase insulin response	This insulin response commences shortly after glucose administration, but is not readily discernible until the first-phase response has ended, ie, at about 10 minutes. This measure can also be determined using C-peptide.	(477, 490)
AIRmax	Acute insulin response at maximal glycemic potentiation (≥25 mmol/L), typically measured using arginine as the secretagogue. This measure represents β-cell secretory capacity. It has been considered a measure of β-cell mass, but data in humans demonstrate it can change acutely when mass would not be changing. Arginine-induced responses can also be measured at lower glucose concentrations, but these do not usually represent the maximal β-cell response.	(127)
Insulin sensitivity
SI (insulin sensitivity index)	An index of insulin sensitivity calculated from parameters determined using the minimal model of glucose kinetics developed by Bergman and colleagues. Determined from a frequently sampled intravenous glucose tolerance test lasting 3 hours.	(108)
M/I	A measure of whole-body insulin sensitivity calculated from clamp studies as the rate of glucose infusion (M) corrected for body size/plasma insulin (I) attained at a steady-state typically achieved after about 2 hours. Addition of tracers (eg, [6,6-2H2]-glucose) allow the measurement of Ra and Rd. Ra denotes rate of tracer appearance when making use of labeled glucose, providing a measure of hepatic insulin sensitivity. Rd denotes rate of tracer disappearance when making use of labeled glucose, providing a measure of skeletal muscle insulin sensitivity.	(491, 492)
β-cell function
DI (disposition index)	A measure that examines the β-cell response relative to insulin sensitivity, the latter a major determinant of secretory demand on the β cell. Importantly, one cannot simply multiply 2 variables together unless it has been proven that the relationship is hyperbolic with a slope not different from −1.	(109)

Abbreviations: ∆C-peptide_0-30/∆C-peptide_0-30, C-peptide index/early C-peptide response; ∆Insulin_0-30/∆glucose_0-30, insulinogenic index/early insulin response; HOMA, homeostatic model assessment; OGTT, oral glucose tolerance test.

Table 5.

Coefficient of variation (%) for selected β-cell function measures obtained using oral and intravenous testing in humans

	NGT	IGT	Diabetes	Diabetes (no meds)	All	Reference
Oral glucose tolerance test (OGTT)
Fasting glucose	4.0	5.2	4.3	4.8	4.5	(478)
2-hour glucose	24.9	11.0	8.0	5.9	--	(478)
Fasting insulin	19.3	14.3	16.5	17.1	16.6	(478)
Fasting C-peptide	11.5	10.8	7.2	8.3	9.7	(478)
HOMA-IR	19.1	14.4	16.1	17.0	16.4	(478)
∆Insulin0-30/∆glucose0-30	41.1	62.9	69.1	79.7	57.1	(478)
∆C-peptide0-30/∆glucose0-30	42.3	37.4	27.3	32.3	34.7	(478)
IncAUCinsulin/IncAUCglucose	29.4	27.0	19.2	22.7	24.9	(478)
IncAUCC-peptide/IncAUCglucose	25.5	14.2	10.0	9.0	17.7	(478)
Model-derived parameters from an oral glucose tolerance test (OGTT)
Glucose sensitivity a	17.2	16.5	20.3	20.3	20.3	(478)
Rate sensitivity a	42.1	40.5	114.8	76.2	44.6	(478)
Potentiation a	43.6	26.7	26.9	26.1	33.0	(478)
φ db	31					(493)
φ sb	18					(493)
Intravenous glucose tolerance test (IVGTT)
AIRglucose (first-phase insulin response)	20.6	--	--	--	--	(494)
	20.1	--	--	--	--	(495)
Kg (glucose disappearance constant)	14.5	--	--	--	--	(494)
	51.4	--	--	--	--	(495)
SI (insulin sensitivity index)c	16.9	--	--	--	--	(494)
	20.2	--	--	--	--	(495)
Hyperglycemic clamp
Fasting insulin	13.2	--	--	--	--	(496)
AIRglucose (first-phase insulin response)	10.4	--	--	--	--	(496)
Maximal insulin response to arginine	16.3	--	--	--	--	(496)

Abbreviations: AIRglucose, acute insulin response to glucose; ∆C-peptide_0-30/∆C-peptide_0-30, C-peptide index; ∆Insulin_0-30/∆glucose_0-30, insulinogenic index; IncAUC, incremental area under the curve; Kg, glucose disappearance constant.

^a calculated using a model developed by Mari and colleagues (486)

^b calculated using a model developed by Cobelli and colleagues (487)

^c calculated using the minimal model of glucose kinetics developed by Bergman and colleagues (108)

Importance of Accounting for Insulin Sensitivity in Assessing β-Cell Function

It was recognized soon after the development of the insulin assay that in absolute terms, healthy obese individuals had greater insulin responses compared with healthy lean individuals. However, by relating these responses to the fasting measurement, it was clear that the magnitude of the response was a function of the basal insulin level and no difference in obese and lean individuals was observed (497). While this approach did not gain much traction for many years as numerous investigators continued to consider insulin responses in isolation, it led to the concept of compensatory hyperinsulinemia and laid the groundwork for subsequent work demonstrating the importance of insulin sensitivity in modulating the β-cell response.

The notion of compensatory hyperinsulinemia was advanced by conceptualization of a negative feedback loop that allows the β cell to recognize the degree of tissue insulin sensitivity and appropriately modulate its insulin response (108). The nature of this feedback loop presupposed that the relationship between insulin sensitivity and a β-cell response measure was hyperbolic in nature, an idea that was subsequently demonstrated in humans using intravenous and oral testing (109, 110). This measure became widely known as the disposition index (108, 498) and can be used in cross-sectional and longitudinal studies (Fig. 6). When the relationship between insulin sensitivity and the β-cell response is shown to be represented by a rectangular hyperbola, the disposition index can be simply calculated as the product of these 2 parameters (108). Importantly, to make the case that the 2 parameters can merely be multiplied together requires that the slope of the regression equation relating the log of each measure be not statistically different to −1. Failure to demonstrate such negates using this simple product and requires use of an alternative equation or accounting for insulin sensitivity in regression models of the β-cell response. Unfortunately, proving the existence of a hyperbola for many pairings of insulin sensitivity and a β-cell response has not always been undertaken; thus, interpretation of data that have not demonstrated this must be made with caution. Recently, an alternative approach has been suggested using the logarithmic expression of the disposition index equation and does not require that the slope be −1 (499). This new approach will require further assessment to determine its veracity and utility in examining this integrated system that is a critical determinant of glycemia. Finally, it is important to recognize that because of differences in clearance, the use of a simple product to calculate the disposition index as a measure of β-cell function is not applicable when using C-peptide measurements.

Figure 6.

Schematics illustrating the relationship between insulin sensitivity and β-cell responses as determinants of β-cell function. A, The relationship is nonlinear in nature and defines a hyperbola with insulin resistant individuals having greater β-cell responses than individuals who are insulin sensitive. The heterogeneity of this relationship among people is demonstrated by individuals falling on either side of the line. The line depicts the mean relationship of insulin sensitivity and the β-cell response for these individuals, commonly known as the disposition index, representing β-cell function. If the slope describing the relationship between these 2 variables is not statistically different from −1, the disposition index can be calculated as the simple product of insulin sensitivity and the β-cell response. The area above the line typically represents better glucose tolerance and below the line poorer glucose tolerance. B, Vector plot of the longitudinal effects of intervention(s) on insulin sensitivity and β-cell responses in individuals or groups of individuals. Any change that moves from the starting point to above the line (even if the β-cell response is lower) represents an improvement in β-cell function, while changes that result in the endpoint being below the line (even if the β-cell response is greater) represent a worsening of β-cell function. Movement along the line, whether it be up or down, represents changes in insulin sensitivity and β-cell responses, but not a change in β-cell function (not illustrated).

Our understanding of the importance of interpreting β-cell responses while accounting for insulin sensitivity is now accepted more or less as a sine qua non when evaluating β-cell function. Application of this approach has advanced our understanding of the critical importance of the β cell in glycemic regulation. While a full discussion of findings using this approach is beyond the scope of this review, select examples are provided later to give a sense of what has been achieved.

Assessment of β-Cell function With Fasting Samples

The mainstay of fasting biomarkers to assess β-cell function is the homeostatic model assessment (HOMA). First introduced in 1985, this steady-state measure was developed to use glucose and insulin pairs to provide estimates of β-cell function (HOMA-B) and insulin sensitivity (HOMA-S) (482). While the latter is used more extensively, particularly with the development of a simplified equation approach (HOMA-IR), HOMA-B has provided insights into the functional status of the β-cell. This measure is expressed as a percentage relative to a “normative” value of 100% determined in a healthy population. However, it does not always provide the same estimate of function as do responses quantified following administration of β-cell secretagogues (500).

While the original description of the model identified its nonlinearity, simple mathematical approximations were provided that allowed the use of an equation, namely HOMA1-%B = (20 × FPI)/(FPG – 3.5), where FPI is fasting plasma insulin in µU/mL and FPG is fasting plasma glucose in mmol/L. With this formula, doubling the insulin concentration at the same glucose concentration will double %B HOMA; thus, simple variability between insulin assays can provide very different estimates of β-cell function on the same sample. It is also possible that a difference, for example a %B of 100% vs 200%, may reflect proportionate differences in insulin sensitivity, so concurrent estimation of HOMA-S is always advised. Modification of the original approach gave rise to HOMA2, which was introduced in 1998 (485) and is available as a calculator at https://www.dtu.ox.ac.uk/homacalculator/. It can be used with either fasting insulin or C-peptide along with fasting glucose. Insulin is used more frequently because it is more readily available and can simultaneously be used to estimate insulin sensitivity; however, one has to be aware of collinearity when relating HOMA-B to HOMA-S (or HOMA-IR) and must perform statistical testing to exclude this possibility.

Due to difficulties with the reproducibility of assays, particularly insulin, the major utility of HOMA-B is in studies where there are typically large numbers of participants followed longitudinally using the same assay so that comparisons of change can be made relative to baseline. It clearly has less utility as an isolated measure, especially if normative data have not been determined for the population under study; thus, caution must be applied when interpreting such data. Some examples where the HOMA approach for quantifying β-cell function has proven useful include epidemiological studies examining progression to diabetes in adults (501, 502) and childhood (503) as well as short- and long-term clinical studies in which the effect of interventions on β-cell function was assessed (504-506). Perhaps the most-cited finding using this approach reported that β-cell function in type 2 diabetes is 50% of normal at the time of diabetes diagnosis and the disease likely commenced many years before (507). As discussed in more detail subsequently, at the time of diagnosis, the magnitude of β-cell dysfunction estimated with this measure does not align with findings using dynamic testing that suggest a greater loss of β-cell function.

Overall, the use of fasting samples continues to be valuable for assessing β-cell function when dynamic testing is not feasible. It is not really an appropriate choice when the number of subjects is low, when there are no longitudinal assessments, and when variability in the insulin assay may influence the actual insulin concentrations. Furthermore, it is not appropriate to use these human-based equations for calculating β-cell function (or insulin sensitivity) in animals.

Measuring Dynamic β-Cell Function Responses

While in the beginning, oral glucose ingestion formed the basis of estimating insulin responses (473, 474), it was not long after that it was demonstrated that intravenous glucose elicited insulin responses that could also be used to address how the β cell was performing (477, 508). These 2 different routes of administration also helped delineate the incretin concept that is characterized by a greater β-cell response when glucose is ingested than when matched glucose levels are achieved with intravenous glucose infusion (509-511). Intravenous testing has also been utilized to gain an estimate of β-cell “mass,” given that the major forms of diabetes are both associated with β-cell loss (512, 513). In addition to providing different insights into β-cell function, these various measures differ in their reproducibility, which is an important consideration when designing studies (478, 494-496).

Oral testing

The standard approach uses glucose as the stimulus, as it not only provides insight into β-cell function but also an opportunity to quantify and classify glucose tolerance (4). An alternative is the meal tolerance test, which provides insight into the β-cell response to nutrient stimuli beyond just glucose (514). While from the β-cell function perspective, insulin is classically used in interpreting these tests, it is possible to substitute C-peptide for that purpose. This latter approach has proven to be particularly useful in studying type 1 diabetes. When using C-peptide assays, it is important to consider their performance and cross-reactivity with proinsulin and insulin; information on different C-peptide assays as provided by the manufacturers is listed in Table 3.

Using these oral approaches, a number of responses can be calculated to provide insight into how the β cell is functioning. The early insulin response, also known as the insulinogenic index, is used most frequently and expresses the increment in insulin above fasting over the first 30 minutes of the test relative to the glucose response (∆insulin_0-30/ ∆glucose_0-30) (477-479). The value of this approach is that the increase in glucose over the first half hour is generally similar in magnitude in people with different degrees of glucose tolerance, thus allowing for an assessment of the β-cell response that is less affected by variance in glucose disposal (Fig. 7). Later in the test, the insulin concentration is far more dependent on the prevailing glucose concentration so that the absolute value can be greater in those who have type 2 diabetes compared to those with NGT (473, 474, 478). It is also possible to evaluate the insulin response throughout the test as a measure of the adequacy of β-cell responsiveness, determining its magnitude as an incremental area under the curve (iAUC) above fasting and relating it to the iAUC for glucose over the same period (140).

Figure 7.

Glucose curve shapes during an oral glucose tolerance test (OGTT). Glucose profiles for A, monophasic; B, biphasic; and C, incessant increase curve types with the corresponding insulin profiles for D, monophasic; E, biphasic; and F, incessant increase curve types.

In line with the earlier discussion, it is also important to interpret these responses in the context of the prevailing insulin sensitivity (Fig. 6). Adjustment of the early insulin response for insulin sensitivity led to the designation of the oral disposition index (DIo) (110), in which the relationship was shown to be a rectangular hyperbola (slope of the relationship of the logged measures being −1), thereby allowing the simple product to be calculated. If the relationship is not a rectangular hyperbola, one must still account for insulin sensitivity but use an approach other than the simple product. It is also important to appreciate that calculating different measures (β-cell response and insulin sensitivity) from the same glucose and insulin data and then examining their relationship increases the risk of collinearity and a false outcome. Thus, testing for collinearity is essential when doing so.

Given the longstanding use of oral tests in examining glucose metabolism in humans, it is not possible in this review to provide insight into the numerous novel observations made using this approach. In some instances, these findings have arisen when the test is performed in conjunction with mathematical modeling, as described below. A selection of the numerous insights obtained using oral tests include that (i) loss of β-cell function is a key component of the development of IGT and diabetes in all major racial/ethnic groups in the United States (122, 123); (ii) there is a progressive loss of β-cell function as glucose tolerance declines and this loss continues as the fasting glucose increases, even across the normal glucose range (124, 156); (iii) relatives of individuals with type 2 diabetes have β-cell dysfunction at a time when they are still normoglycemic (515); (iv) genome-wide SNPs are linked to deficient β-cell function in people with IGT (516-521); (v) favorable responses to interventions that prevent progression of prediabetes to diabetes and of type 2 diabetes itself are dependent on better β-cell function at the time of the intervention (522-524); and (vi) the release of C-peptide in established type 1 diabetes provides evidence that there are residual β cells (427, 525-527) and the degree of residual β-cell function determines the ability to achieve tight glycemic control (525) as well as being associated with reduced episodes of severe hypoglycemia (528).

Dynamic testing has also highlighted the magnitude of β-cell dysfunction in type 2 diabetes. While the HOMA-B estimates based on fasting measures suggest that β-cell function is decreased by 50% at the time of diabetes diagnosis, dynamic testing highlights that the deficit is even greater. Analyses using the disposition index in studies of Finns (124), Japanese Americans (156), and other ethnic groups in the United States studied as part of the GENNID Study (122) (N. Esser and S.E. Kahn, unpublished observation) have found that β-cell function declines as fasting glucose increases, commencing even within the normal glucose range. Compared with those with NGT, at the diagnostic threshold for diabetes β-cell function is reduced by about 80%, while in impaired glucose metabolism it is already decreased by 50%.

Finally, the shape of the glucose concentration curve is being used as an indirect approach aimed at identifying physiologically distinct groups and individuals in whom the disease is more likely to progress (Fig. 7). The dominant curve phenotype in both youth and adults and in prediabetes and diabetes is monophasic, with a peak between 30 and 90 minutes that then declines (529-532). Individuals with this curve shape have lower β-cell function than those whose curve manifests as a biphasic pattern in which glucose increases, declines, and then increases again. Individuals with a monophasic curve are at increased risk of developing impaired fasting glucose and type 2 diabetes (530, 531). A third curve, of the incessant increase type in which glucose rises throughout the test, is underscored by a profound loss of β-cell function and predicts accelerated β-cell dysfunction and glycemic failure (533).

Intravenous testing

Administration of glucose intravenously, either as a short infusion or bolus injection, has for greater than 50 years been used to quantify β-cell responses, an approach that has clearly stood the test of time and is still well utilized today. This approach led to the identification of 2 distinct phases of insulin release: first and second (55). It is important to recognize that these 2 phases are defined and distinguishable by the response to intravenous glucose, and cannot be distinguished during an oral test. Subsequently, the hyperglycemic clamp was developed in which glucose is fixed at a predetermined level (491), a method consequently extended to generate a dose-response curve by clamping glucose at multiple concentrations (126, 127) or using graded glucose infusion rates to allow the glucose concentration to spontaneously equilibrate at different levels (534).

In humans, the first-phase insulin response starts immediately with glucose administration and is usually considered to be complete 10 minutes later, particularly when given as a bolus. The second phase also begins early and embodies insulin concentrations while glucose remains elevated. In diabetes, the first phase is essentially absent while the second phase is reduced, but not gone. In fact, the first-phase response is reduced in IGT and IFG, with it seeming to be lost when the fasting glucose is around 115 mg/dL (6.4 mmol/L) (125, 477, 535), well below the diagnostic threshold for diabetes (536). The second-phase response decreases as the β cell’s secretory ability declines (127). Thus, these phases clearly represent different secretory aspects of the cell, with the first-phase response being a valuable early marker in people whose glucose tolerance is still relatively normal.

Assessment of the glucose dose-response curve by attaining a steady state at multiple levels has provided additional insights into the β cell’s responsiveness. Maximal responsiveness is achieved at a glucose concentration >450 mg/dL (25 mmol/L). This approach has frequently been supplemented with the addition of a nonglucose secretagogue, typically arginine (127). Use of this amino acid not only acutely stimulates insulin release, but also releases glucagon and thus can provide insight into α-cell function (127, 537). This maximal response, frequently termed AIRmax, represents the secretory capacity of the β cell (127). It has also been considered to be a marker of β-cell mass as currently, except at autopsy, mass remains an unmeasurable aspect of diabetes pathology (512, 513). However, it is also clear that this secretory measure can change rapidly (within a week or two) (405), a time interval in which mass would not be expected to increase. This realization has played a part in the development of the concept of “functional β-cell mass” (538). We feel it is important to advise caution in simply linking these two together as they do not always associate. For example, β-cell function can be markedly reduced (more than 80%) when the number of β cells (“mass”) is reduced by 50% or less (111, 113, 127). Alternatively, an intervention can increase β-cell responses fairly rapidly in a time frame when the number of β cells would not have changed (405, 409, 539).

The β cell also exhibits oscillatory behavior, both short, rapid and ultradian in nature (540, 541), with these disturbed in people with abnormal glucose tolerance (128, 129). An uncoupling of the relationship between glucose oscillations and insulin secretion is already apparent in IGT, in keeping with a disturbance in the feedback loop of glucose and insulin secretion early in the course of the disease (130).

Collectively, intravenous-based testing measures continue to provide valuable insights into β-cell secretory function, some of which are in line with what can be determined from the oral test. As within-subject variability of intravenous measures is lower than in oral tests, typically fewer subjects will be required when using intravenous testing (478, 494, 495). However, all are modulated by tissue insulin sensitivity and thus need to be interpreted with this in mind. Doing so has been truly informative, with multiple observations made, including (i) the progression from NGT to IGT and diabetes occurs because of a progressive loss of β-cell function (121); (ii) first-degree relatives of individuals with type 2 diabetes (542, 543) and HLA-identical siblings of people with type 1 diabetes have reduced β-cell function (76); (iii) groups at high risk of developing type 2 diabetes, including older individuals (544, 545) and women with a history of gestational diabetes (546, 547) or polycystic ovary syndrome (548, 549) manifest reduced β-cell function even when glucose tolerance may still be normal; (iv) the loss of β-cell function over time underlies diabetes development in those with prior gestational diabetes (550); (v) the long-term, glycemic improvement with metabolic surgery results from improved β-cell function (551, 552); (vi) youth with IGT or recently diagnosed type 2 diabetes have hyperresponsive β cells compared with adults of similar body size who are 40 years their senior (139); and (vii) the decline in β-cell function in these youth is more rapid than in adults and does not abate in response to metformin or insulin glargine (138, 553).

Mathematical Modeling of β-Cell Function

The era of mathematical modeling to quantify aspects of glucose metabolism gained traction in the 1980s and has expanded in approaches and utilization since. Today, modeling is used frequently to assess β-cell function and can be performed with both intravenous and oral testing, the latter where it has been most commonly utilized.

Modeling is particularly convenient with oral testing (486, 493). By using paired glucose and C-peptide concentrations, a number of measures based on derived insulin secretion rates can be determined. To enable better parameter identifiability, it is desirable to sample more frequently and for longer than is typically done with a standard 2-hour oral glucose or meal tolerance test. Minimizing sample number and/or duration of the test may have practical advantages, but it comes at the risk of a type II error or a need for a larger sample size. Accounting for insulin sensitivity in the interpretation of model-derived β-cell parameters is not always performed but would appear to be advisable and has highlighted differences between groups.

An additional advantage obtained from modeling oral tests is that one can obtain a glucose-dose response curve relating insulin secretion rates to the glucose concentration without having to perform clamps over a broad glucose range. The glucose concentrations on which this calculation is based are those achieved over the course of the oral test and typically fall in the range over which insulin secretion is a linear function of glucose, that is, they do not achieve the concentration (>450 mg/dL) required to estimate maximal secretory capacity. As is commonly the case, these models may include a parameter that improves the final fit of the data but whose physiological meaning is less well characterized.

The utility of mathematical modeling is further supported by findings that have used other approaches to evaluate dynamic β-cell responses. Thus, for example, it has been reported that (i) insulin secretion rates are increased in obese individuals (554); (ii) β-cell function decreases progressively as glucose tolerance declines in adults (554) and youth (555-557); (iii) the response to bariatric surgery is dependent on β-cell function at baseline, with a better response to the intervention in those with better baseline function (558-560); (iv) a subset of individuals at high risk of developing diabetes have insulin hypersecretion in the basal state and following oral or intravenous glucose administration (561); and (v) in nondiabetic individuals with cystic fibrosis, who commonly experience asymptomatic hypoglycemia in the third hour following oral glucose ingestion, insulin secretion rates are inappropriately increased when their glucose falls below fasting (562).

Islet Amyloid Polypeptide in the Assessment of β-cell Function

IAPP was originally identified as the unique peptide component of islet amyloid deposits (162, 163). The peptide is co-released with insulin in response to glucose and nonglucose secretagogues (461-463, 563-568). Its physiological function is not well appreciated, but roles for gastric emptying and appetite regulation have been suggested (569, 570).

Measurement of plasma IAPP concentrations in humans has provided additional insight into β-cell function. As discussed earlier, assays that measure the precursor forms of IAPP are being developed and are expected to inform about the normal processing of proIAPP and how diabetes may affect this process (464). In time, these assays may provide new biomarkers for β-cell function. Using a number of different immunoassays that measure the intact peptide, it is now well recognized that the amount of IAPP produced, stored, and released by the β cell is approximately 1% to 3% that of insulin (461-463). Importantly, as the clearance of IAPP is slower than that of insulin, the molar ratio will vary depending on when sampling is performed relative to the administration of the stimulus (566).

Like with the insulin assay, there is variability in IAPP measurements due to different assay methodologies and specificity of the antibodies. In particular, most older radioimmunoassays likely cross-react with proIAPP forms, ELISAs for IAPP immunoreactivity have varying ability to detect putative O-glycosylated and non-glycosylated forms of the peptide (571, 572), and human IAPP is highly fibrillogenic and special care is needed in its handling as a standard in these assays (455). Despite this, among the observations made using human venous plasma with different IAPP assays are the following (i) obesity and/or insulin resistance are associated with increased IAPP concentrations (565, 566) and are reduced with weight loss following Roux-en-Y gastric bypass (568); (ii) IAPP is markedly decreased in type 1 diabetes (563, 564); (iii) the IAPP concentration is reduced in type 2 diabetes and IGT, with the degree of reduction in IGT intermediate between type 2 diabetes and NGT (461); (iv) IAPP is lower in individuals who are at increased risk of developing type 2 diabetes, including older people (462) and first-degree relatives of those with the disease (463); and (v) following pancreas-kidney transplantation with venous drainage, IAPP concentrations are increased (567).

Given the findings that IAPP and insulin are typically released together, measurement of mature IAPP as an independent biomarker of β-cell secretory function is currently out of favor. Studies of proIAPP and its conversion intermediates continue and are expected to add important information to our knowledge base and may, like proinsulin, prove to be useful in determining the efficiency of proIAPP processing and provide insight into β-cell dysfunction.

Concluding Remarks

Our understanding of the importance of the β cell in the normal regulation of glycemia and the significance of β-cell dysfunction and mass loss in the pathogenesis of hyperglycemia in diabetes has evolved markedly over time and continues to do so. While measurement of insulin provided the first true insights and laid the foundation, the evolution of science and scientific method has provided a more in-depth view of the physiology and pathology of this critical endocrine cell.

With these advances, biomarkers have and continue to become available that provide opportunities in the research setting, with some now having applicability in clinical care. Along these lines, we have enhanced our understanding of the role of genetics and epigenetics not only in “garden variety” types 1 and 2 diabetes but also the rarer forms such as monogenic diabetes and CFRD. Assay methodology advances have identified new autoantibodies, provided insight into propeptide processing, and raised the possibility of estimating β-cell destruction while it is occurring. Imaging approaches have advanced from ultrasonography and computed tomography to different forms of positron emission tomography, thus making the goal of quantifying the amount of β cells more feasible. With the generation of large amounts of these and other forms of data from cross-sectional, longitudinal, and intervention studies in different populations, opportunities are being created for bioinformatic and machine learning approaches to further inform us. Patterns identified in these analyses should help further refine our understanding of diabetes subtypes and the basis for disease progression, leading us closer to precision medicine.

As we look forward from today, much remains to be done, but there is hope that we will soon be able to use β-cell biomarkers to predict, diagnose, and prognosticate diabetes and not need to rely as heavily on functional tests as currently required.

Glossary

Abbreviations

AIRmax

acute insulin response at maximal glycemic potentiation (≥25 mmol/L)

BMI

body mass index

CFRD

cystic fibrosis–related diabetes

CFTR

cystic fibrosis transmembrane conductance regulator

CPE

carboxypeptidase E

ELISA

enzyme-linked immunosorbent assay

ER

endoplasmic reticulum

FPG

fasting plasma glucose (mmol/L)

FPI

fasting plasma insulin (µU/mL)

GAD

glutamic acid decarboxylase

GLP-1

glucagon-like peptide-1

GLP-1R

glucagon-like peptide-1 receptor

GRS

genetic risk score

GWAS

genome-wide association studies

HbA1c

glycated hemoglobin A1c

HLA

human leukocyte antigen

HNF1A

hepatic nuclear factor 1 α

HNF1B

hepatic nuclear factor 1 β

HNF4A

hepatic nuclear factor 4 α

HOMA

homeostatic model assessment

HPLC

high performance liquid chromatography

IA2/ICA512

protein tyrosine phosphatase

IAA

insulin autoantibodies

IAPP

islet amyloid polypeptide

iAUC

incremental area under the curve

ICA

islet cell autoantibodies

IFG

impaired fasting glucose

IFN-γ

interferon-γ

IGT

impaired glucose tolerance

IL

interleukin

IL2RA

interleukin 2 receptor α

INS

insulin

ISSI-2

insulin secretion-sensitivity index-2

K_ATP

ATP-sensitive potassium channel

LADA

latent autoimmune diabetes in adults

lncRNA

long non-coding RNA

MHC

major histocompatibility

miRNA

microRNA

MODY

maturity-onset diabetes of the young

MRI

magnetic resonance imaging

NGT

normal glucose tolerance

NODAT

new-onset diabetes mellitus after transplantation of solid organs

nPOD

Network of Pancreas Organ Donors in Diabetes;

PAM

peptidylglycine α-amidating monooxygenase

PBMC

peripheral blood mononuclear cell

PC1/3

prohormone convertase 1/3

PC2

prohormone convertase 2

PET

positron emission tomography

PP

pancreatic polypeptide

PTPN22

protein tyrosine phosphatase non receptor type 22

SNP

single nucleotide polymorphism

SPECT

single photon emission computed tomography

TNF

tumor necrosis factor

VMAT2

vesicular monoamine transporter type 2

ZnT8

zinc transporter 8

Additional Information

Disclosures: Steven E. Kahn has served as a consultant to Bayer, Boehringer Ingelheim, Casma Therapeutics, Eli Lilly, Intarcia, Janssen, Merck, Novo Nordisk, Pfizer, and Third Rock Ventures. Daniël H. van Raalte has served as a consultant to Boehringer Ingelheim-Eli Lilly Diabetes Alliance, Merck, Sanofi, and Astra Zeneca and has received research grant support from Boehringer Ingelheim-Lilly Diabetes Alliance, AstraZeneca, and Merck. Sakeneh Zraika has received research grant support from Novartis. C. Bruce Verchere is a director, scientific advisor, and shareholder in Integrated Nanotherapeutics and has served as a consultant to Sirona Biochem. All other authors have no disclosures.