ATP-Sensitive Potassium Channels in Hyperinsulinism and Type 2 Diabetes: Inconvenient Paradox or New Paradigm?

Insulin secretion from pancreatic β-cells is complex, but the dominant role of electrical activity in physiological control of glucose-stimulated insulin secretion (GSIS) is now well understood. Glucose entry leads to a metabolically generated rise in ATP/ADP concentration that initiates β-cell K_ATP channel closure. This results in action potential firing and elevation of cytosolic calcium (Cai) concentration, which in turn triggers the fusion of exocytotic vesicles (1). Consistent with the paradigm, loss-of-function (LOF) mutations in either the pore-forming Kir6.2 (KCNJ11) or the regulatory SUR1 (ABCC8) subunit of K_ATP channels cause increased electrical excitability and hyperinsulinism (HI), both in human congenital HI (CHI) (2) and in mice that have incomplete genetic knockout (KO), or knockdown, of K_ATP channels (3,4). Conversely, K_ATP gain-of-function (GOF) mutations cause neonatal diabetes mellitus (NDM) as a result of underexcitability and undersecretion (5,6). Accordingly, diazoxide, which activates K_ATP channels, and sulfonylureas (SUs), which inhibit K_ATP channels, have proven to be viable therapies for CHI and NDM, respectively (7,8).

Thus, we have developed a very clear understanding that loss of K_ATP activity causes HI, as in human CHI, whereas increased K_ATP activity causes reduced β-cell excitability, undersecretion, and diabetes, as in human NDM (Fig. 1).

However, these apparently straightforward yin-yang consequences of K_ATP LOF and GOF and the simple therapeutic solutions that are provided by K_ATP activators and inhibitors are complicated by an inconvenient and unexplained realization regarding HI. The K_ATP closure→elevated Cai→secretion paradigm implies that reduction of K_ATP levels will shift the insulin secretion curve to the left, i.e., shift the threshold for triggering insulin secretion to increasingly lower glucose concentrations and increase secretion at higher glucose concentrations (Fig. 1). The more K_ATP is lost, the further the shift and the more severe the hypersecretion will be. When K_ATP is lost completely, electrical activity will be continuous at all glucose concentrations (β-cells will be “glucose blind”), and there will be maximal secretion of insulin at all glucose concentrations.

Here’s the rub. In mice, partial loss of K_ATP indeed increases excitability and insulin secretion at lower glucose concentrations (3,9). However, as multiple groups have demonstrated, there is a crossing over to undersecretion, and in vivo insulin secretion is actually dramatically reduced at essentially all glucose concentrations in mice with more complete loss of K_ATP (10–14). Isolated islets from either SUR1^−/− or Kir6.2^−/− K_ATP KO mice are glucose blind, i.e., they are electrically active at all glucose concentrations and lack any marked glucose dependence of secretion, as predicted, but rather than this being because secretion is maximally elevated, almost all studies show it is because secretion stays at a low level, close to the basal level of secretion from normal islets (11–13). Significantly, this undersecretion means that, rather than being hypoglycemic, these animals are actually glucose intolerant as adults (Fig. 1) and become overtly diabetic after high-fat diet (10). The phenomenon is reiterated in mice that express dominant-negative subunits only in β-cells (10,15), obviating the effects of loss of K_ATP in other cell types. This crossover requires ∼70% loss of K_ATP activity, since mice with ∼50–70% overall loss of β-cell K_ATP are hyperinsulinemic (3,4,16) and do not cross over to undersecretion unless challenged with a high-fat diet (10), which may itself cause further lowering of K_ATP activity.

As fully expected for secretion in the absence of K_ATP channel regulation, humans with CHI are glucose blind, i.e., a rise in blood glucose does not stimulate a marked rise in insulin secretion (17). However, this is not because secretion is already maximal at all glucose concentrations. Insulin content may be reduced in CHI islets, and at elevated glucose, insulin secretion may actually be lower than that in normal individuals (17,18). Thus, CHI individuals show essentially the same phenomenon of suppressed maximal secretion as K_ATP KO mice. Treatment options for CHI are limited, and normoglycemia can frequently only be achieved by near-total pancreatectomy (19). Most of these individuals then eventually develop diabetes, logically attributable to secondary consequences of this drastic surgical intervention. In other cases, particularly in cases that do not cause total loss of K_ATP, the K_ATP channel opener diazoxide is effective, and somatostatin analogs can also be effective (20).

The natural long-term outcomes in CHI have not been well studied, but it is increasingly apparent that the symptoms of HI typically lessen with age, including in children with ABCC8 or KCNJ11 LOF mutations (21–23), whether dominantly acting homozygous (Hom) or compound heterozygous (Het) or paternally inherited mutations causing focal CHI. One study of patients with dominant K_ATP mutations found that in adult carriers, all of whom would be expected to have suffered HI, essentially half (14/29) were symptomless, which indicates preceding remission, and 4/29 carriers in that study had diabetes (24). The authors noted that this is similar to levels of type 2 diabetes (T2D) in the general population, but the possibility that diabetes was related to prior CHI should not be discounted, given other reports of HI patients with dominant ABCC8 LOF mutations who remitted and then progressed to diabetes later in life (25,26). In one intriguing case, a female with HI carrying a heterozygous ABCC8 LOF mutation progressed to marked obesity-associated diabetes at 10 years old (27), consistent with this crossover being linked to insulin resistance.

A similar progression is also seen in an extensive Finnish pedigree (25,28), members of which gradually progress from HI toward glucose intolerance and even frank diabetes. Strikingly, knock-in mice carrying the same (SUR1[E1506K]) mutation showed HI in the neonatal period but also rapidly progressed to an undersecreting phenotype (14). K_ATP channel activation by MgADP was reduced in β-cells from young Het and more severely from Hom animals, and channel density was reduced significantly as the animals aged (14). Glucose tolerance (GT) was essentially normal in young Het animals, but young Hom animals were glucose intolerant, and GT worsened in both with age. Young Hom E1506K mice exhibited enhanced insulin secretion and lower fasting blood glucose but became diabetic, with greatly reduced insulin secretion and impaired GT by 6 months of age. These mice thus traverse the “inverse U” relationship (Fig. 1) and match the suggestion that progressive reduction of K_ATP levels to ∼30% of normal will cause hypersecretion, but further reduction will cause crossover.

Interestingly, GOF mutations in glucokinase (GCK), which lead to elevated ATP:ADP ratios and consequent suppression of K_ATP activity, also cause CHI but, in at least one case, an affected individual subsequently developed diabetes (29). Similarly, GCK activator drugs, which were initially developed with a view to treatment of diabetes by stimulation of glucose metabolism and therefore insulin secretion, reduce fasted and fed glucose levels in the short term (days) (30), but over months, these effects are not sustained (31,32). Paralleling this, constitutive pancreatic β-cell–specific expression of a GOF GCK mutation initially causes persistent hypoglycemia, but by 8 months of age, mutant mice develop impaired GT that is further exacerbated by high-fat/high-sucrose diet (33). Similarly, mutations in hepatocyte nuclear factors 1α and 4α cause CHI, potentially due to reduced K_ATP channel activity, but this again crosses over to diabetes with age for unknown reasons (34,35). Thus, humans with CHI, due to LOF mutations in K_ATP or other genetic mechanisms that predispose to reduced K_ATP channel activity, as well as mice with marked loss or inhibition of K_ATP all show a very similar progression from hypersecretion to undersecretion, glucose intolerance, and diabetes. As we discuss below, understanding this crossover progression is essential for treatment and management not only of CHI but also potentially of the elephant in the room, T2D itself.

Many β-cell changes have been identified that accompany the crossover to failure of secretion in K_ATP KO mice. Increased apoptotic markers and increased DNA damage response signals have been reported, as well as evidence of α-cell infiltration, but there is no significant change in islet structure or cell identity. One study, using total internal reflection fluorescence microscopy imaging to assess the secretory mechanism itself, showed an increased number of docked insulin granules and a normal level of t-SNAREs in the plasma membrane of SUR1 KO β-cells, but a markedly reduced number of fusion events (36). In some studies, crossover has been associated with reduction of insulin content (10,14), but this is not a consistent finding and may be insufficient to explain the glucose intolerance, suggesting that downregulation of secretion itself is involved (4,10,11).

Crossover is not simply the result of any obvious changes in other major ion conductances that lead to a compensatory loss of excitation. As a consequence, isolated islets generally are shown to be excitable and firing action potentials at all glucose concentrations, and Cai essentially does what it is predicted to do in the absence of K_ATP: it stays high at all glucose concentrations (11,37). That said, the work of Henquin and colleagues (38–40) has shown that the Cai signal can be significantly affected by prior culture conditions. When previously cultured in high (10 mmol/L) glucose, K_ATP KO islets exhibit a relatively high duty cycle of Cai oscillations, but fresh or low (1–5 mmol/L)-glucose cultured islets show a comparably lower duty cycle and lower Cai overall. This culture condition dependence extends to secretion, which is much higher from KO islets cultured in high glucose, suggesting a glucose-dependent adaptation through an unexplained mechanism. Interestingly, it has been shown that K_ATP levels in CHI islets can themselves be affected by culture conditions (41). Conceivably, there may be subtle changes in other conductances that lead to lowering of the triggering Cai level, and further studies of the details of β-cell excitability in the absence of K_ATP are warranted.

SU drugs inhibit K_ATP channels, thereby raising Cai and triggering insulin secretion, and have long been prescribed as T2D treatments. However, they eventually fail as sole therapy for unknown reasons (42–44). We have suggested that this gradual failure to respond to SUs is the result of the same crossover mechanism, i.e., chronic SU-induced inhibition of K_ATP activity might, by causing the same electrical consequence as K_ATP KO, cause downregulation of insulin secretory capacity (45). Consistent with this idea, implantation of wild-type mice with pellets that chronically release the SU glibenclamide (glib) initially causes blood glucose to drop as expected, but within a few days, blood glucose rises, and after just 1 week of glib treatment, these mice become as glucose intolerant as adult K_ATP KO mice (45). Freshly isolated islets from these pellet-implanted mice also show as marked a reduction of secretory capacity as K_ATP KO islets, but secretory capacity can be restored after overnight incubation, as the drug washes out and K_ATP channel activity is restored (45). In utilizing an independent pharmacologic route to functionally knock out K_ATP, these striking results reveal not only that, at least in mice, crossover due to loss of K_ATP activity develops in vivo within a few days but also that it is rapidly reversible.

Here, the story has reached something of an impasse. Given that crossover is not a developmental phenomenon and can be rapidly induced and rapidly reversed, it is possible that the mechanism involves a physiological adjustment of Ca-dependent processes, such as Ca/calmodulin-dependent kinases (CaMKs), calcineurin, or other Ca-dependent proteins, which can cause immediate physiological effects and can modulate G-protein-coupled signaling through cAMP, the recognized primary modulator of secretion. It could also involve changes in gene expression: transcriptomic analysis of SUR1 KO islet cells, in normal diet or after high-fat diet, demonstrates extensive gene reprogramming, with 4,000–7,000 genes affected (46,47). In addition to genes associated with mitochondrial biogenesis and metabolism, there are gene changes consistent with altered β-cell identity and regulation of β-cell epigenetic and transcriptional programs. How many, if any, of these changes are a developmental consequence of the global gene KO, and how many relate to secretion crossover specifically, is not clear. To test this, it will be important to assess whether any of these transcript changes are reversibly induced during short-term induction of crossover, i.e., in glib-pelleted mice or from inducible KO animals, and it remains a possibility that the rapidity of induction and reversal reflects a pathophysiological response to chronic depolarization and elevated Cai that involves only a few or no changes in gene expression.

These issues need sorting out. As we discuss below, the implications of this crossover phenomenon are potentially enormous and demand mechanistic explanation.

As discussed above, crossover in response to loss of K_ATP provides a ready scenario for remission and development of diabetes that can occur in the rare individuals with K_ATP-dependent CHI. Could it also have a role in the development of T2D, the overwhelmingly important form of diabetes? The textbook progression of T2D involves development of insulin resistance, which is generally associated with obesity and which is initially compensated by increased insulin secretion from pancreatic β-cells. As insulin resistance increases, β-cells eventually fail to “keep up,” gradually crossing over to an undersecreting phenotype, and hyperglycemia develops, in turn worsening β-cell function. We cannot do justice to it here, but an enormous body of work has identified multiple genetic and nongenetic contributors to the development of β-cell dysfunction in T2D (48,49). In response to insulin resistance, β-cell mass expansion provides one mechanism of compensatory increase in secretion, by cellular proliferation and hypertrophy (50,51).

Subsequent crossover from an adaptive hypersecreting state to an uncompensated undersecreting state has been the focus of many studies, and multiple pathways have been implicated. Genome-wide association studies show that the majority of T2D-linked genes play a role in β-cell function or mass, including Wnt-regulated transcription factors, cell cycle and division regulators, and pancreatic transcription factors (52,53). In addition to a potentially synergistic glucolipotoxic role of chronic hyperglycemia and elevated free fatty acids (54), elevated amino acids and related metabolites (55) may induce deleterious effects on β-cell mass and function. Generation of reactive oxygen species (ROS), activation of endoplasmic reticulum stress, upregulation of chronic hexosamine signaling pathway/O-GlcNAcylation, and induction of proinflammatory cytokines in islets may all reduce β-cell proliferation, insulin secretion, or insulin gene expression and promote β-cell apoptosis (56–58). In addition, β-cell dedifferentiation and transdifferentiation can occur under glucolipotoxic conditions (59) but may be reversible, as we showed for dedifferentiation seen in inexcitable β-cells with no GSIS, if glucotoxicity is reversed by exogenous control (60).

Thus, at the organism level, progression from compensated hyperinsulinemia to uncompensated diabetes can involve a lot of well-studied processes. Significantly, there is no specific role ascribed to changes in K_ATP channel activity in these proposed mechanisms, and a role for altered K_ATP channel activity, if any, in the typical progression of T2D remains essentially unknown, an important gap given the clear importance of these channels for normal secretion. Multiple lines of evidence lead us to suggest that altered K_ATP activity may in fact be a significant component. Interestingly, KCNJ11 was one of the earliest T2D-associated genes, with the coding variant (Kir6.2[E23K]) repeatedly being identified (61,62), but the mechanistic link is still not clear. The mutation is in linkage disequilibrium with a second missense mutation (encoding SUR1[S1369A]) in the adjacent ABCC8 gene. Which is the relevant variant is not clear; at the level of channel function, SUR1[S1369A] may enhance MgATP activation of the channel (63), while Kir6.2[E23K] causes a small decrease in ATP sensitivity (64–66). Either effect might predispose to diabetes via the same mechanism as K_ATP-dependent NDM, i.e., by reducing excitability and secretion (67). However, a Hom knock-in mouse model of Kir6.2[E23K] (68) showed no effect on body weight or blood glucose, and insignificant reduction of GSIS and GT, on a normal diet. On a high-fat diet, there was sexual divergence, with a trend toward greater weight gain in E23K females, associated with slightly worsened GT but an unexplained enhancement of secretion (68).

Further studies may reveal compounding effects of other variants, but these contradictory results provide little support for the idea that K_ATP GOF contributes directly to failure of secretion in T2D by inhibiting excitation. Moreover, the textbook progression of T2D (Fig. 1) suggests this is not a likely scenario: initial increase of insulin secretion in compensation for insulin resistance is more likely to be associated with reduced K_ATP activity and enhanced excitability (69). Could reduced β-cell K_ATP activity really be a causal component of crossover to secretory failure in T2D? As blood glucose is elevated, glucose metabolism in the β-cell will reduce K_ATP activity by inhibition, but there is also evidence that hyperglycemia can drive a reduction of K_ATP channel levels, and maximal K_ATP density is significantly reduced in wild-type β-cells chronically incubated in high glucose in vitro (70). This suggests a positive feedback relationship between loss of K_ATP and hyperglycemia that could become important under glucolipotoxic conditions. Partly through activation of protein kinase C, high glucose may promote endocytosis of K_ATP channels, reducing levels of SUR1 protein in the membrane and increasing excitability (71). Conversely, under glucose-deprived conditions, or in the presence of leptin, phosphorylated AMPK induces recruitment of K_ATP to the plasma membrane (72,73). Thus, both protein kinase C–mediated endocytosis and loss of AMPK-mediated trafficking to the membrane would be predicted to reduce surface K_ATP density under hyperglycemic conditions. Leptin promotion of K_ATP trafficking to the membrane provides a mechanism by which insulin secretion could normally be suppressed (74,75). Conceivably, this mechanism may be inhibited in obese individuals with leptin resistance and contribute to crossover progression and development of diabetes.

Glucagon-like peptide 1 (GLP-1) agonists, but not glucose-dependent insulinotropic polypeptide (GIP) agonists, are effective at maintaining insulin secretion in human T2D, and pharmacological studies indicate that GLP-1–induced insulin secretion is similarly retained, while GIP-induced secretion is severely diminished, in K_ATP KO mice (15,76) due to a relative shift from Gs to Gq signaling. The same shift reportedly occurs when isolated islets are chronically depolarized, whether by inhibiting K_ATP with the SU tolbutamide or by incubation in high glucose (15). These parallels are consistent with a switch in incretin efficiency being driven by loss of K_ATP in both K_ATP KO animals and human T2D. Are there any data to link the development of T2D to loss of K_ATP? First established in the 1950s (77), the KK mouse is an in-bred strain exhibiting obesity, insulin resistance, and predisposition to diabetes. Introduction of the yellow obese (Ay) gene to generate the KK-Ay strain resulted in a mouse that developed severe diabetes from an early age (78), and many studies have since been carried out using these mice as polygenic T2D models. Oduori et al. (15) recently reported that β-cells from nondiabetic KK mice respond normally to high glucose with electrical activity and GSIS, but severely hyperglycemic KK-Ay β-cells lack K_ATP activity and are depolarized under all conditions, with loss of GSIS. Oduori et al. further showed that GLP-1 secretion was amplified in the latter, but GIP was without effect. The authors proposed that the switched incretin sensitivity in diabetic KK-Ay islets was a consequence of the loss of β-cell K_ATP, and that this could explain incretin bias in human T2D. For this to be the case, of course, human T2D must then involve loss of K_ATP (79).

If this implication is correct, it is of the utmost fundamental importance, but is it a reasonable idea? Tentatively, we argue that the answer is yes, given that 1) hyperglycemia will reduce K_ATP levels and that 2) loss of K_ATP activity could be marked enough to cause glucose intolerance and overt diabetes when insulin resistance is imposed. We acknowledge that this is a bold contention, but given the extent of these correlated facts and the implications discussed below, it is one that needs to be carefully examined.

If loss of K_ATP activity does indeed trigger β-cell crossover as one of the pathological mechanisms in the development of T2D, the potential for reversibility by restoration of K_ATP conductance implies a potential mechanism to reverse secretory failure in T2D. By activating β-cell K_ATP, the K_ATP channel opener diazoxide suppresses electrical activity and inhibits insulin secretion. Under normal circumstances, such an action should exacerbate a diabetic state. However, there is evidence for counterintuitively beneficial effects of diazoxide treatment in T2D (80,81). Exactly how diazoxide does this is not clear. It may reduce β-cell apoptosis via increased mitogen-activated protein kinase activity (82) and may affect endogenous glucose production via central nervous system K_ATP channels (83). However, extensive studies of Björklund and colleagues (80) suggest that the most likely scenario is indeed one in which activation of β-cell K_ATP channels reverses crossover. Based on the demonstration that prior in vivo diazoxide treatment preserved GSIS in pancreas taken from hyperglycemic rats (84), treatment of insulin-dependent T2D subjects with additional bedtime diazoxide, at doses too low to cause the typical side effects of high dosing, resulted in considerable improvement in stimulated C-peptide and insulin levels (85). Interestingly, in a second study in which the protocol was repeated on insulin-naive T2D subjects (86), glycemic control was, if anything, slightly worsened. In this case, it is tempting to speculate that these subjects had not yet crossed over, and, therefore, activating K_ATP would simply serve to inappropriately suppress insulin secretion. Similarly, a trial of diazoxide in newly diagnosed type 1 diabetes patients could have failed to result in better preservation of β-cell function for the same reasons (87).

Further consistent with the general idea that restoration of K_ATP channel activity is beneficial in T2D is recent clinical evidence that intermittent fasting (IF) can reverse T2D (88). The primary goal of such dietary interventions is to improve insulin resistance by reducing average caloric intake and causing weight loss, but some animal studies suggest that IF (versus simple caloric restriction with equivalent caloric intake) induces additional beneficial outcomes, and some human studies have shown beneficial effects of IF in the absence of any weight loss (89–91). While simply restricting caloric intake may reduce blood glucose levels and inhibition of K_ATP, the possibility that this also regulates K_ATP levels is unexplored.

As an alternative to exercise and dietary modification, bariatric surgery to reduce the functional volume of the stomach is now recognized as effective therapy in T2D (92,93). It can result in impressive improvement in HbA_1c levels and reduction of medication needs, associated with rapid improvement of insulin secretion, particularly meal-stimulated secretion. Loss of first-phase insulin release occurs early in development of T2D, and the effectiveness of gastric bypass for correcting T2D clearly involves recovery of β-cell function. In one study, recovery of first-phase insulin secretion rate (ISR) in i.v. glucose tolerance test (GTT) and β-cell glucose sensitivity during oral GTT was found to be greater in subjects with high β-cell function (i.e., with better glycemic control and with shorter duration of T2D) prior to Roux-en-Y gastric bypass surgery than in those with low function prior to surgery (94). Postoperative postprandial GLP-1 concentrations also increased similarly in both groups, supporting the idea that the primary improvement was in β-cell function. Perhaps reversal of crossover underlies the improvement in β-cell function following gastric bypass? Interestingly, one serious potential complication is the occurrence of hypoglycemia with inappropriately high insulin and C-peptide levels, reflecting hyperstimulation of insulin secretion (95–97). As with CHI, many such individuals can be effectively treated by diazoxide, i.e., by activating K_ATP, but partial pancreatectomy may be needed to achieve significant improvements in symptom control (96,98). We suggest the simple possibility that bariatric surgery in an individual with a pre-existing state of reduced K_ATP levels could, by reducing the metabolic signal for K_ATP inhibition, cause a reversal of crossover, back onto the ascending limb of the inverse-U relationship (Fig. 1), explaining inappropriate hypersecretion of insulin until and unless normal levels of K_ATP activity are restored.

By pharmacologically suppressing K_ATP activity, SUs hyperstimulate secretion and can enhance control of glycemia for an extended period in T2D, but SU therapies inevitably fail over months or years (42–44). We suggest that such treatment, by exaggerating the loss of K_ATP activity, accelerates crossover, but at least initially, this may be reversible. In one intriguing clinical study, nonobese T2D patients with secondary failure of SU therapies regained significant insulin secretory activity after 3 months of intensive insulin therapy (99). In these patients, therefore, control of blood glucose with insulin, reducing stimulatory demand on β-cells, restored secretory capacity. This result parallels the observation that removal of a high-fat diet leads to recovery of secretion and glycemic control in mice with loss of K_ATP channels (10).

One final implication, regarding SU therapy, is that although supranormal reduction of β-cell K_ATP and, hence, overstimulation of hyperexcitability will, by causing crossover, be self-limiting in the use of SU to trigger an increase in insulin secretion in T2D, the same need not be true for K_ATP-dependent NDM. In the latter case, the underlying problem is too much β-cell K_ATP activity due to the KCNJ11/ABCC8 GOF mutations. Conceivably, the correct SU dose will only serve to reduce channel activity to an essentially normal level, allowing the glucose-dependent amplifying pathway to produce quasi-normal, glucose-dependent insulin secretion. This can explain how successful SU treatment provides better glucose control than bolus insulin therapy in NDM patients (8). If the dose is carefully titrated to avoid excess K_ATP inhibition (and, hence, avoid hyperelevation of Cai), this can also explain why SU therapy in patients with K_ATP-dependent NDM, in contrast to those with T2D, is proving to be a sustainable, long-term therapy (100).

Despite the paradigmatic role of K_ATP channels in normal control of insulin secretion, reflected in the CHI and neonatal diabetes that result from LOF and GOF K_ATP mutations, respectively, significant loss of K_ATP channel activity paradoxically is associated with reduced insulin secretability that can be significant enough to cause glucose intolerance, and even frank diabetes, in the presence of insulin resistance. The implications of this crossover are significant. First, it provides a potential explanation for the progression of HI patients from HI to impaired GT and even frank diabetes. Second, the possibility that insulin resistance and hyperglycemia leads to downregulation of K_ATP channels provides a potential explanation for the critical transition from glucose intolerance and HI to undersecretion in the development of T2D. Third, rapid reversibility of crossover, back to normal secretion, as a consequence of reducing excitability implies a mechanism for reversing this pathology in T2D. Such a mechanism might underlie the remission of T2D that can follow diazoxide treatment or when demand for insulin secretion is reduced following gastric bypass.

Funding. C.G.N. is supported by the Center for Scientific Review, National Institutes of Health, grant HL140024, and M.S.R. is supported by National Institutes of Health grant DK123163.

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