Updating the Role of α-Cell Preproglucagon Products on GLP-1 Receptor–Mediated Insulin Secretion

Type 2 diabetes mellitus (T2DM) results in defects in multiple target organs. On the one hand, there is insulin resistance in many tissues including the liver and skeletal muscle. On the other hand, there is both β-cell death and dysfunction resulting in deficient nutrient-induced insulin responses. While the field of islet biology has historically focused its attention on understanding β-cell function and the mechanisms by which these cells become dysfunctional with diabetes, there has been a scientific shift toward greater understanding of other endocrine cells of the islet and their paracrine role in regulating the β-cell.

In recent years, many questions and new data have come forward regarding the role of the α-cell in insulin regulation. Data from our group have questioned the role of intestinally secreted glucagon-like peptide 1 (GLP-1) in regulation of glucose homeostasis, and data from multiple groups suggested a physiological role of pancreatic-produced GLP-1 in regulation of insulin secretion (1–4). In addition, in the last 2 years, a series of critical studies demonstrated a role for glucagon, acting via the GLP-1 receptor (GLP-1R), in paracrine regulation of insulin secretion. The purpose of this article is to discuss these new findings surrounding α-cell preproglucagon products, with a particular focus on challenging the incretin role of GLP-1, in the context of their roles in insulin secretion and consequently glucose metabolism.

In the early 1960s, two studies published the same year found that glucose was lower but insulin was higher after isocaloric loads of glucose when administered orally (5) or directly into the jejunum (6) versus when administered intravenously in humans. Based on these data, the term “incretin effect” was coined to describe this phenomenon of higher insulin responses when nutrients were ingested versus infused. Together with authors of previous studies demonstrating that intestinal mucosa extracts administered to rats could lower blood glucose, these authors hypothesized that a gut-derived humoral substance contributed to the regulation of insulin secretion. Ten years later, it was demonstrated that an intravenous infusion of gastric inhibitory polypeptide (GIP) (aka glucose-dependent insulinotropic peptide) resulted in an increase in insulin immunoreactivity and improvement of glucose tolerance in humans, supporting its role as an incretin (7). However, in anesthetized rats, administration of gut extracts that had GIP removed did not completely ablate nutrient-induced insulin responses, suggesting the presence of additional incretins (8).

More than 10 years after the discovery of GIP, another incretin was finally discovered. GLP-1, a protein coded by the same gene that makes glucagon, was found to stimulate insulin release in rat islets (9) and when infused in humans (10). In fact, the latter study found that GIP was less effective at increasing insulin levels and concluded that GLP-1 was a physiological incretin in humans. More than 30 years later, GLP-1 is effectively targeted to treat the hyperglycemia of T2DM and, due to its additional anorectic properties, is also an effective treatment of obesity.

While the pharmacological benefit of targeting GLP-1 signaling is clear, there have long been questions regarding the physiological GLP-1 system. Anatomically, the greatest concentration of enteroendocrine cells that secrete GLP-1 is located in the distal gut and has limited direct access to nutrients (11). Although much work has been done focused on direct nutrient-induced stimulation of GLP-1 from enteroendocrine cells, many neural or humoral factors (including GIP) secreted from the upper gut have been suggested to provide a feed-forward release of GLP-1 from distal enteroendocrine cells (reviewed in Sandoval and D’Alessio [12]). However, even if feed-forward regulation explains the nutrient-induced release of GLP-1, questions arise as to whether GLP-1 circulates in high enough concentrations or if the half-life is long enough for intestinally secreted GLP-1 to act as a hormone on β-cell GLP-1R. GLP-1 is rapidly degraded by a protease, dipeptidyl peptidase 4 (DPP4), resulting in only ∼10% of intestinally secreted GLP-1 reaching the circulation (13). Whatever is not cleaved by DPP4 is cleared by the liver. Given these clearance challenges and based on the route of blood circulation, it seems unlikely that GLP-1 secreted by the gut could circulate to the heart and back to the pancreas in sufficient amounts to allow it to act as a hormone that regulates insulin secretion.

One possibility is that GLP-1 is not a hormone but instead is a neurotransmitter. Activation of GLP-1R on nerves that innervate the portal vein lowers fasting and postprandial glucose levels (14–17). We also know that central nervous system (CNS) administration of GLP-1 improves hepatic insulin sensitivity in healthy (16,18) and obese animals (19), increases hepatic glycogen synthesis (20), promotes peripheral glucose uptake (16,20), and increases insulin secretion in healthy and obese rodents (18–20). These data suggest that CNS GLP-1R mediate multiple aspects of glucose control. However, mice with specific CNS or nodose ganglia–specific GLP-1R ablation from birth have normal glucose tolerance on chow diet and the expected impairment on an HFD and also have comparable improvements in glucose tolerance in response to a long-acting GLP-1 agonist as compared with their genetic control mouse littermates (21). Thus, acute manipulation of these neuronal GLP-1R populations does regulate glucose homeostasis, but these receptors are not, in and of themselves, necessary for long-term glucose control. Further, it still remains unclear as to whether intestinally secreted GLP-1 could activate CNS GLP-1R populations, and, if not, what source of GLP-1 might physiologically act on these receptors.

We know that infusing exogenous GLP-1 peripherally increases insulin secretion and that administration of a GLP-1R antagonist reduces nutrient-induced insulin responses, but what evidence do we have that intestinally derived GLP-1 is physiologically involved in regulating insulin secretion? Most of that work has focused on the impact of administration of a GLP-1R antagonist, which impairs glucose tolerance and insulin secretion in both humans and rodents. However, mice null for the GLP-1R, or mice that have been administered the GLP-1R antagonist, exendin 9-39 (Ex9), have impairment in glucose tolerance whether that glucose is delivered by mouth, which increases circulating GLP-1 levels, or via an intraperitoneal (IP) injection, which does not increase circulating GLP-1 (22).

To test the role of intestinal GLP-1 secretion, two studies have looked at the necessity and sufficiency of Gcg expression in the gut. One recent study performed a loss-of-function experiment with Cre lox-P technology targeted to preproglucagon (Gcg), the gene that encodes GLP-1, comparing total gut with distal gut Gcg knockouts (23). Supporting a role of intestinal GLP-1 in regulating glucose homeostasis, the whole-gut Gcg knockout mice have impaired oral, but not IP, glucose tolerance. Consistent with reduced circulating GLP-1 levels in the whole-gut Gcg knockout mice, gastric emptying rate was reduced; however, insulin responses to glucose were intact. In comparison, mice with only distal gut Gcg removal had impaired oral and a slight impairment of IP glucose tolerance with normal circulating nutrient-induced GLP-1 and insulin secretion. Together these data suggest that the primary impact of intestinally derived GLP-1 on glucose tolerance is through regulation of gastric emptying rate.

For testing of sufficiency, in another study intestinal Gcg was selectively reactivated across the whole gut while the remaining tissues (i.e., pancreas and brain) remained devoid of Gcg (24–26). For interrogation of the specific role of GLP-1, these mice were administered glucose with and without Ex9. Ex9 had no impact on glucose tolerance (either oral or IP) in mice with whole-body deficiency of Gcg, indicating that Ex9 is a true GLP-1R antagonist in vivo (vs. an inverse agonist, as has been suggested previously [27]). Mice with selective intestinal Gcg reactivation had normal postprandial circulating levels of both total and active GLP-1 but surprisingly had no impairment of glucose tolerance in response to Ex9. In contrast, selective reactivation of pancreatic Gcg expression was sufficient to completely restore the response to Ex9. This was in response to both oral administration and IP glucose administration. We interpreted these results to mean that intestinal GLP-1 was dispensable for glucose tolerance and that there is a pancreatic source of GLP-1 that regulates insulin secretion in a paracrine fashion.

Although the results from those studies (23,24) seem at odds, it is interesting that removal of Gcg expression from the distal (but not proximal) gut reduced fasting glucose but resulted in normal glucose-stimulated levels of both active and total GLP-1 and yet these animals still had impairments of oral but not IP glucose tolerance (23). GLP-1R knockout mice, on the other hand, have impairments in both oral and IP glucose tolerance (22), while β-cell–specific GLP-1r knockout mice have normal oral but impaired IP glucose tolerance (28). One possibility is that the different sources of GLP-1 regulate different aspects of β-cell function and/or glucose homeostasis. Regardless, together these data indicate that intestinally secreted GLP-1 is partly necessary for glucose tolerance, seemingly through its impact on gastric emptying rate, but not sufficient for the ability of Ex9 to impair glucose tolerance. These data also underscore the potential role of α-cell–derived GLP-1 in insulin secretion.

Prohormone convertase 1/3 (PC1/3), the enzyme that posttranslationally processes GLP-1 (and GLP-2 and oxyntomodulin) from proglucagon, is more highly expressed in the gut and CNS compared with the pancreas (29). However, α-cell PC1/3 activity and/or expression is found in embryonic and neonatal mice, in pregnant dams, and in many models of metabolic stress including streptozotocin-induced and genetically induced diabetes, and detection increases with increasing glucose concentrations (4,30–34). Another mouse model with an α-cell knockout of PC1/3 which resulted in a ∼35% reduction of islet GLP-1 in chow-fed animals lead to impaired glucose tolerance in HFD-fed mice administered a low dose of streptozotocin (1). These data indicate that there is a pool of α-cell–derived GLP-1 that regulates glucose tolerance and increases under conditions of metabolic stress.

However, part of the hesitance about accepting that the α-cell produces physiologically active GLP-1 comes from the fact that pancreatic secretion of intact GLP-1 is low and sometimes undetectable. Generally, GLP-1 is difficult to assess in mice. In addition to degradation by DPP4, mouse GLP-1 has also be reported to undergo fast endoproteolytic metabolism (35). Regardless of these limitations, protein extraction of pancreata from wild-type control mice revealed detectable levels of both total and active GLP-1 (via radioimmunoassay for three isoforms of GLP-1 summed for total GLP-1) in one study (36) but only detectable levels of total (via radioimmunoassay for the carboxy terminus) and not active (via a radioimmunoassay for the N-terminus) GLP-1 in another study (37). Besides potential differences in the assays used, what might account for these differences is unknown. Nutrient status of the animals was not stated in either study, and it is not clear whether this could impact tissue GLP-1 levels.

GLP-1 secreted from islets has also been studied in mouse pancreas perfusion studies. Sampling during perfusion studies is done from the portal vein and under anesthesia. Under these conditions, portal vein levels of active GLP-1 in response to glucose were found to be low and often undetectable (27). For GLP-1 to act in a paracrine fashion does not require its entry into the circulation, so this result, in and of itself, does not rule out a physiological role for α-cell GLP-1. In support of this, mice that only express Gcg in the pancreas have total GLP-1 level in the plasma that is less than half that of animals that express Gcg only in the intestine (24). Further, given the role of the autonomic nervous system in regulating glucagon secretion, it is also possible that islet GLP-1 production is blunted by the anesthesia.

Based on the rationale that quantifying pancreatic GLP-1 production is complicated by the fact that the number of α-cells is quite low, a recent study isolated, and then dispersed, islets into single cells before staining for GLP-1. They found not only GLP-1–producing α-cells but also that a proportion of α-cells produce GLP-1, a proportion produces glucagon, and still another proportion produces both peptides (2). They also revealed that 70% vs. 50% of α-cells produced GLP-1 in healthy human versus mouse islets, indicating that α-cell–produced GLP-1 may be important for humans as well. These data reveal a critical piece of information regarding the heterogeneity of α-cells surrounding proglucagon peptides and support the possibility that at least a portion of α-cells produce GLP-1 and are optimally placed to regulate insulin secretion in a paracrine fashion.

Glucagon has traditionally been thought of as a counterregulatory hormone, with secretion being stimulated when plasma glucose levels are low and inhibited when plasma glucose levels are high such that there is a reciprocal relationship between insulin and glucagon secretion (38). Additionally, Roger Unger has long argued the importance of postprandial suppression of glucagon in normal glucose regulation (39). Further, insulin-induced suppression of glucagon (i.e., intraislet hypothesis) has been thought to play a role in hyperglycemia, as conditions of decreased insulin secretion or α-cell insulin resistance (e.g., T2DM) leads to less postprandial suppression of glucagon (40), with hyperglucagonemia being a characteristic of T2DM (41).

Despite all this, there are also early data showing that glucagon applied to islets stimulates insulin secretion (42). The question of whether this was a physiological or pharmacological effect and whether glucagon was regulating insulin secretion via its own receptor or via the GLP-1R was debated. At the time of these initial studies, the data suggested that glucagon acted on the GLP-1R only at pharmacological concentrations (43). Competitive binding assays in purified rat islets showed that glucagon binding on β-cells could be reduced by 42% in the presence of GLP-1 (42). However, glucagon-induced cAMP production was inhibited by both a glucagon antagonist and Ex9 (42). In perifusion studies, Ex9 suppressed high-dose glucagon stimulation of insulin secretion in cells exposed to 20 mmol/L glucose, which is a very high glucose concentration for islets but was a standard experimental design at the time, while GLP-1 binding was inhibited by high glucagon concentrations. However, at low doses glucagon action seems to be mediated by the glucagon receptor, not the GLP-1R, as Ex9 does not block its affects (42). The addition of Ex9 alone (in the absence of added glucagon or GLP-1) lowers insulin release stimulated by glucose alone (42), suggesting that β-cell GLP-1R are partially occupied by an agonist.

More recent modeling work has confirmed and extended these findings. Fluorescence resonance energy transfer assays that detect cAMP as a read-out for glucagon and GLP-1R activation revealed that glucagon is a nonconventional GLP-1R agonist that is inhibited by Ex9 (44). Although glucagon was less potent than GLP-1, it had full agonist properties on the GLP-1R. The authors postulated that glucagon is a dual agonist of the glucagon receptor and GLP-1R at the high concentrations that may exist in islets and that the net effect is that glucagon and GLP-1 regulation of the GLP-1R and downstream signaling are coordinated so that β-cell function is enhanced in the healthy state or “repaired” under conditions of metabolic stress. A caveat to this is that the concentrations of glucagon achieved in this study may still reflect pharmacological levels, as estimations of islet glucagon levels that are sufficient to stimulate insulin levels are ∼1,000-fold lower (45).

Recently, three independent research groups explored the role of glucagon action on β-cell using three different genetic tools. The first report, from the Holst laboratory, used mouse pancreatic perfusions in control and in a variety of genetic mice including their α-cell–ablated mouse (27). The importance of this technique is that the islet cytoarchitecture is maintained intact. The next study used perifusion of islets isolated from a variety of genetic models (46), and the last study inhibited α-cell activity using designer receptors exclusively activated by designer drugs (DREADDs) [47]). Despite the different strategies, the findings between these studies and the prior studies were generally similar and demonstrated that glucagon stimulates insulin secretion at high, but not low, glucose levels and the effect is blunted by either genetically or pharmacologically blocking GLP-1R signaling. This general finding was further validated in human islets (46). The nutrient conditions were also important, as the presence of amino acids (potent glucagon secretagogues) magnifies the impact of glucagon action on β-cell GLP-1R. These data demonstrate an additional physiological paracrine role for glucagon in regulating insulin secretion under postprandial conditions. A proposed model illustrating the integration of the role of glucagon and GLP-1 on the GLP-1R in regulation of insulin secretion is illustrated in Fig. 1.

If we presume that α-cell production both of glucagon and of GLP-1 is critical for regulating insulin levels, there are several open questions. One question is, what determines the peptides being produced and are they produced from the same cell or from different α-cells? Much recent work has found a high degree of heterogeneity among β-cell populations in a variety of characteristics including developmental origins and their transcriptomes (48). More research is now accumulating regarding the heterogeneity of α-cells as well. Unlike β-cells, the electrical activity of α-cells is not coupled. One reason for this lack of coordination may be the anatomy of the islet. α-Cells are dispersed throughout (or around for rodents) the islet, and this anatomical variability could generally create a greater degree of heterogeneity in function. In fact, different α-cells within the same islet exhibit different calcium activities across the physiological range of glucose levels, and that calcium activity does not correlate with glucagon secretion (49). This could certainly provide conditions whereby differential α-cells secrete differential proglucagon peptides.

Generally, we know surprisingly little about the intracellular regulation of glucagon secretion. As mentioned above, calcium channel activity is heterogenous among α-cells, leading to no consistent answer about the role of activity of this particular channel in glucagon secretion (50). Direct innervation of islets or circulatory factors such as catecholamines have been proposed to be major regulators of glucagon secretion. However, glucagon responses to glucose when studied in isolated islets are similar to what is seen in vivo, suggesting that intraislet factors rather than nervous system or blood flow influence glucagon regulation (50). Paracrine models have proposed that products from nearby β- or δ-cells, including insulin, Zn²⁺, GABA, and/or serotonin, inhibit glucagon secretion (12). However, none of these factors alone can suppress glucagon in dispersed α-cells (50).

In addition, GLP-1, insulin, and somatostatin, peptides that all increase postprandially, all suppress glucagon. Conversely, another postprandial hormone, GIP, has been found to increase glucagon (51). There are also mixed results in terms of the impact of various nutrients on glucagon secretion, especially in humans. Ingestion of glucose alone (52,53) suppresses glucagon, but other macronutrients or their metabolites also regulate glucagon secretion. For example, free fatty acids (54) and ketones (55) have been reported to decrease glucagon, but proteins (amino acids) potently stimulate glucagon secretion (52). Free fatty acids have also been reported to suppress the ability of arginine to stimulate glucagon secretion in humans (54), suggesting that glucagon responses to a mixed meal should be restrained compared with when only arginine is administered in humans. Further emphasizing the complex interactions of nutrients on the α-cell, in isolated islets exposed to low glucose conditions, free fatty acid metabolism provides fuel to sustain glucagon levels (56), and long-term exposure to free fatty acids potentiates glucagon release (57). Lastly, the three recent studies that found that glucagon stimulates insulin secretion through the β-cell GLP-1R all found this effect to be magnified in the presence of amino acids under ex vivo conditions (27,46,47). Using in vivo studies in mice, it has been shown that plasma and portal glucagon increases in response to a liquid mixed meal but not in response to oral glucose alone (58). However, the insulin levels appeared (this was not statistically compared) to be higher with the glucose versus the liquid meal administration despite the fact that rough calculations suggest the caloric load was higher with the liquid mixed meal. Altogether, these data highlight both the potential nutrient specificity and the complexity of the system. The varied impacts of neural, paracrine, endocrine, and/or nutrient signals on glucagon secretion may relate to the α-cell heterogeneity discussed above but also emphasize how understanding regulation of glucagon secretion is critical for understanding its role in regulating insulin secretion and glucose homeostasis.

It is critical to consider how these new data should be integrated into a new “incretin” model of glucoregulation. One way to address this is by examining the phenotype of animals devoid of both glucagon and GLP-1. If GLP-1 and glucagon are both important for increasing postprandial insulin, then glucose tolerance and insulin secretion should be greatly impaired in animals that are devoid of both peptides. An initial study that used diphtheria toxin ablation of Gcg in enteroendocrine L cells or α-cells found that combined L- and α-cell ablation reduced the postgavage glucose peak and had minimal overall impact on oral glucose tolerance but did impair IP glucose tolerance (59). Conversely, specific α-cell ablation improved oral but had minimal impact on IP glucose tolerance (59). Using a slightly different diphtheria toxin–inducible α-cell ablation, Traub et al. (1) demonstrated that these mice have an age-induced improvement of IP and oral glucose tolerance but normally elevated insulin. Although the removal of glucagon was certainly not complete, the data suggest that reduction of circulating glucagon reduces its capability to increase hepatic glucose production and leads to an overall improvement of glucose tolerance. Thus, these models suggest that ablation of α-cells in total, and consequently glucagon levels, leads to improved and not impaired glucose homeostasis.

One complication of these models is that an increase in another glucoregulatory peptide could compensate for the loss of another critical glucoregulatory peptide. The increase in plasma GLP-1 in the glucagon receptor knockout mouse is one such example. However, with GLP-1 and GIP, there is theoretically no need for one or the other to increase in compensation, as physiological levels of GIP are sufficient to allow a full improvement in glucose tolerance in response to a DPP4 inhibitor and vice versa (26). This concept of GIP compensating for the lack of GLP-1 with DPP4 inhibitor action may apply to physiological regulation of insulin secretion as well (60). Insulin responses to an intravenous (24) and oral glucose (D.S., unpublished observations) load are normal in animals devoid of both GLP-1 and glucagon, but islets isolated from these mice had reduced basal and stimulated insulin levels (46). In this case, as GLP-1 and glucagon are absent, the normal physiological increases in intestinal GIP may be sufficient to retain normal insulin responses. These data also highlight the critical interaction of the variety of signals that regulate insulin secretion.

How does the knowledge gained from the studies demonstrating that glucagon stimulates insulin secretion through the GLP-1R (27,47,58) inform the prior data interpretations that pancreatic and not intestinal secreted GLP-1 is necessary for glucose tolerance (24)? First, these findings suggest the possibility that the lack of response to Ex9 in the intestinally reactivated Gcg and the full response to Ex9 in the pancreatic reactivated Gcg mice are due to pancreatic glucagon, rather than GLP-1, signaling on the GLP-1R. This interpretation would mean that the entirety of Ex9 action on impairing glucose tolerance is through postprandial glucagon action on the GLP-1R. In addition, competition binding studies demonstrate that glucagon is readily displaced in the presence of GLP-1 and accordingly, even in the perfused pancreas, GLP-1 is significantly more potent than glucagon on the GLP-1R (27). It is important to note that the nutrient conditions are important in shifting of glucagon from its glucose-stimulating to its insulin-stimulating action. Specifically, ad libitum–fed mice administered a liquid mixed meal are more responsive to glucagon actions on the β-cell (58). The pancreatic Gcg-reactivated animals were administered glucose alone and were studied in the fasted state, conditions of low glucagon action on the β-cell. Regardless, the current body of work suggests that both peptides regulate the β-cell through GLP-1R signaling, but the process of posttranslational preproglucagon processing makes it difficult to generate genetic models that distinguish between the impacts of α-cell–derived glucagon versus GLP-1 on local GLP-1R signaling.

Interestingly, when the phenotypes of the various genetic models targeted to Gcg or GLP-1R signaling that have been generated are discussed in concert, some important findings fall out (Table 1). One is that the impacts on glucose tolerance in the developmental versus the adult genetic manipulations have a much stronger glucose phenotype. The other is that the absence of glucagon has a potent impact on improving oral glucose tolerance (sometimes IP glucose tolerance as well). This does not necessarily mean that glucagon or GLP-1 action on insulin secretion and glucoregulation is not important. However, if the primary role of GLP-1 and glucagon is to improve glucose homeostasis through increased insulin levels, then knockout of both peptides or their signaling should impair, not improve, glucose tolerance. Thus, instead, it is possible that given the balance of the two actions (increasing insulin or increasing glucose), glucagon’s action on raising glucose is more powerful. This highlights a limitation of the currently available experimental models and why the body of literature needs to be considered as a whole rather than as individual findings.

This is a rapidly evolving model, and it is anticipated that the expansion of our knowledge on α-cell heterogeneity will go a long way in driving further evolution in our thinking of how α-cell proglucagon products regulate glucose homeostasis. It seems likely that α-cell heterogeneity contributes to some α-cells secreting GLP-1 in response to glucose and some secreting glucagon when both amino acids and glucose are present. The end result is postprandially stimulated insulin secretion, no matter the macronutrient present.

It is interesting that the function of both GLP-1 and glucagon is tightly coupled to glucose availability. GLP-1 action is blunted during hypoglycemia, and glucagon action on the liver is suppressed by hyperglycemia. Both aberrant glucagon secretion and aberrant GLP-1 secretion are seen in T2DM as well (61). For example, glucose-induced suppression of glucagon and glucagon response to hypoglycemia are both reduced with T2DM (62). Generally, the incretin response is also blunted with T2DM and is suggested to be due to reductions in sensitivity to both GLP-1 and GIP (63). Thus, in T2DM, defects in at least three peptides important for action on the β-cells are dysfunctional. Altogether, these data emphasize the importance of glucose availability and potentially the balance of these two glucoregulatory proglucagon-derived peptides within the islet in regulating insulin secretion and glucose control.

Whether α-cell–derived GLP-1 and/or glucagon is critical for regulating insulin secretion, what these data are telling us is that the traditional incretin model is inadequate. The reality is that there have been several open questions about the role for GLP-1 as an endocrine incretin. The traditional characterization of reciprocal roles of insulin and glucagon, in terms of glucoregulation, is also a simplified assessment of glucagon action. Instead, under postprandial conditions, proglucagon peptides from the α-cell represent redundant signals that respond to the variety of fluctuating nutrient conditions, allowing for fine-tuning of insulin responses and overall glucose control in the face of a variety of metabolic conditions, and all of these feedback signals are critical during a meal and their dysfunction is critical in the disease process.

Funding. This work is supported in part by National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, awards DK117635, DK121995, and DK107282 and by an American Diabetes Association grant (1-19-IBS-252).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.