Mechanisms to Elevate Endogenous GLP-1 Beyond Injectable GLP-1 Analogs and Metabolic Surgery

The finding that metabolic surgery rapidly promotes remission of type 2 diabetes (1) provides a unique opportunity to investigate further the role of the gastrointestinal tract in regulating glucose metabolism. A remarkable attribute of this procedure is that improved glucose control in patients precedes and may even be independent of weight loss (2). The mechanistic basis for this is not fully established (3,4), but the profound increase in postprandial concentrations of the incretin hormone glucagon-like peptide 1 (GLP-1) after surgery seems to mediate much of the metabolic benefit (5–8). This hypothesis is supported by studies showing infusion of the GLP-1 receptor (GLP-1R) antagonist exendin-4_(9–39) blunts heightened insulin secretion in patients who have undergone the bypass procedure (9). In a similar way, the postsurgical improvement in pancreatic β-cell function and glucose tolerance observed in patients with type 2 diabetes is decreased by exendin-4_(9–39) (10). Further, in mice, conditional deletion of Glp-1r in β-cells limits the positive effect of metabolic surgery on glucose tolerance (11).

Evidence implicating elevated GLP-1 as the central factor mediating the benefits of metabolic surgery is consistent with the clinical experience of GLP-1–based therapeutics in that the magnitude of GLP-1 increase correlates with the efficacious response. Among these agents, dipeptidyl peptidase-4 (DPP4) inhibitors, which prevent proteolytic breakdown and inactivation of incretins, increase postprandial concentrations of active GLP-1 by approximately twofold, up to 8–10 pmol/L, and lead to modest improvement in glycemic control without reducing body weight (12,13). However, use of injectable GLP-1R agonists enables the level of GLP-1 equivalents to be increased to concentrations of 50–65 pmol/L, or >10-fold above endogenous amounts (14,15). In line with the insulinotropic and anorectic effects of increased GLP-1R activation, the higher exposure achieved by these agents results in more robust glucose lowering accompanied by moderate weight loss (13); body weight can be further reduced by increasing the dose (16). While a range of GLP-1 concentrations have been reported in various procedures and assays, the levels of circulating GLP-1 in patients who have undergone metabolic surgery (30–65 pmol/L) (17–19) are similar to levels achieved by treatment with exogenous GLP-1R agonists.

The unanticipated outcome that bariatric surgical procedures result in GLP-1 concentrations in range of the exogenously delivered GLP-1R agonists reaffirms that targeting the gastrointestinal tract is a viable therapeutic strategy. While GLP-1R agonists have shown strong efficacy and are generally well tolerated, these molecules require subcutaneous injection, and aggressive dose escalation is limited by nausea (20), a phenomenon that varies among agents, possibly because of differences in peak concentrations of parenterally administered drugs (21). Although metabolic surgery can induce disease remission, life-threatening risks are associated with major surgery, and unique complications have presented after the procedure (22). In addition, because metabolic surgery restricts nutrient absorption, long-term dietary supplements are often required (23).

Oral agents that recapitulate the antidiabetic effects of metabolic surgery may avoid some risks associated with this treatment option. Because higher GLP-1 levels likely occur as a result of the surgical procedure bypassing portions of the upper intestine so nutrients more readily flow over L cell–rich segments of the lower gastrointestinal tract, targets expressed in these areas should be prioritized. The studies presented herein investigate various pathways that control L-cell function and describe use of an in vivo model to better assess GLP-1R–mediated pharmacology. In addition, combinations of orally administered small molecules targeting complementary pathways of GLP-1 regulation are shown to increase endogenous GLP-1 in mice to levels surpassing those achieved by injectable GLP-1 analogs or bariatric surgery. These findings should lay the groundwork for discovering next-generation GLP-1–based oral medications.

Animals were studied and maintained in accordance with the Institutional Animal Care and Use Committee of Eli Lilly and Company, and the Guide for the Use and Care of Laboratory Animals by the National Institutes of Health. All animal studies described herein were approved by the Institutional Animal Care and Use Committee of Eli Lilly and Company.

Wild-type, Glp-1r, Gipr, and Gpr40 null mice on a C57BL/6 genetic background were maintained at Taconic (Hudson, NY). Lepr^db/db (BKS.Cg−+Leprdb/+Leprdb/OlaHsd) mice were obtained from Envigo (Indianapolis, IN). Male mice (9–14 weeks of age) were singly housed in microisolator cages on wood chip bedding, and standard food (5008 Teklad Global Diet; Envigo) and deionized water were available ad libitum. Lights were on a 12-h light/12-h dark cycle, and temperature and relative humidity were maintained between 21°C and 23°C and 45% and 65%, respectively.

The DPP4 inhibitor sitagliptin (24), somatostatin receptor subtype 5 (SSTR5) antagonist “compound 3–1” (25), TGR5/GPBAR1 agonist “compound 18” (26), and GPR40/FFA1 agonist “compound 20” (27) were synthesized at Eli Lilly and Company. Sitagliptin was dosed at 10 mg/kg to systemically inhibit DPP4 (24). Doses of other compounds were based on solubility in order to maximize exposure. All small molecules were formulated in 20% Captisol (Ligand Pharmaceuticals, San Diego, CA) w/v in water or 1% hydroxyethylcellulose, 0.25% Tween 80, and 0.05% antifoam, with pH adjusted to 2 (TGR5 agonist) or 10 (GPR40 agonist). Exenatide was formulated in 0.01% albuminized PBS, and the GLP-1R antagonist Jant-4 (28) was formulated in 100 mmol/L Tris HCl buffer (pH 8). Both were synthesized internally. Compound in plasma was quantified using liquid chromatography–tandem mass spectrometry at Quintiles (Plainfield, IN), whereas exenatide was measured at Algorithme Pharma (Quebec, Canada).

Mice fasted overnight were orally dosed with the vehicle or compounds, followed 30–60 min later by an oral bolus of 3 g glucose/kg (oral glucose tolerance test [OGTT]) or an intraperitoneal bolus of 1–2 g glucose/kg (intraperitoneal glucose tolerance test [IPGTT]). Blood glucose concentrations were measured up to 120 min after glucose administration using glucometers. Data were used to calculate the area under the curve (AUC). Plasma was analyzed for insulin (Meso Scale Discovery, Rockville, MD).

Mice fasted overnight were dosed with compounds, and blood was collected at various time points into prechilled EDTA tubes containing aprotinin and a DPP4 inhibitor. Although portal vein sampling would provide the most accurate analyses of GLP-1, because of the large number of animals studied it was only practical to collect blood to measure active GLP-1 either by clipping the tail or immediately upon euthanasia by decapitation (Mesoscale Discovery).

Body composition was determined using quantitative nuclear magnetic resonance analysis (EchoMRI, 3-in-1 Composition Analyzer; Echo Medical Systems, Houston, TX).

Data are represented as the mean ± SD and were compared using one-way ANOVA (vs. the vehicle at each time point) or a t test in GraphPad Prism 7 software. The null hypothesis was rejected at P < 0.05.

To develop an in vivo model to evaluate the metabolic consequences of increasing circulating concentrations of GLP-1, the OGTT was explored in Gipr null mice. Without an intact glucose-dependent insulinotropic polypeptide (GIP) axis, the impact of glucose-stimulated incretin release is largely mediated by GLP-1; therefore, loss of function of Gipr should enable a well-controlled experimental system for studying the physiological effects of increasing endogenous GLP-1. For the initial analysis, studies were performed using either the DPP4 inhibitor sitagliptin to increase endogenous levels of active GLP-1 or injection of the GLP-1 analog exenatide to further stimulate GLP-1R signaling. Improvement in glucose tolerance by sitagliptin treatment was similar in both wild-type (65% improvement in glucose excursion AUC; P < 0.05) (Fig. 1A) and Glp-1r null mice (29) (67% improvement in glucose excursion AUC; P < 0.05) (Fig. 1B). However, in the absence of GIP action, the efficacy of sitagliptin was reduced in Gipr null animals (21% improvement in glucose excursion AUC; P < 0.05) (Fig. 1C). To assess the responsiveness of Gipr-deficient mice to GLP-1, additional animals were administered subcutaneous injections of exenatide (0.3 nmol/kg). This treatment group showed that increasing the levels of circulating GLP-1R agonist significantly lowers plasma glucose over the time course of the experiment (80% improvement in glucose excursion AUC; P < 0.05) (Fig. 1C). The finding that sitagliptin is less efficacious in mice lacking Gipr is in line with data from studies demonstrating the importance of GIP (also a DPP4 substrate) in oral glucose tolerance in mice (30). The difference in the degree of glucose lowering that occurred upon treatment with sitagliptin compared with exenatide corresponds well to measurements of endogenous GLP-1 and the concentration of exogenously administered exenatide (peak plasma concentration [C_max] of 14 and 143 pmol/L for sitagliptin and exenatide, respectively) (Fig. 1D). While Gipr deletion has been associated with modest sensitization to GLP-1 (31,32), the concentration-dependent improvement in glucose tolerance observed by increasing exenatide exposure suggests that approaches aimed at elevating circulating GLP-1 may be evaluated using Gipr null mice. In addition to being sensitive to high levels of exenatide, it is noteworthy that the DPP4 inhibitor studies show that the responsiveness of Gipr null mice to oral glucose is not dysregulated, as the increase in endogenous GLP-1 was similar in wild-type and Gipr null mice (C_max of 14 and 15 pmol/L for wild-type and Gipr null mice, respectively) (Fig. 1D). Further, these levels are consistent with postprandial concentrations of active GLP-1 in patients treated with sitagliptin (12). In accordance with studies indicating Gipr-deficient (30) or Gipr^{−/−βCell} (31) mice do not have altered peripheral glucose metabolism, body weight, percentage body composition measurements, and fasting blood glucose were similar between Gipr null mice and their age-matched wild-type counterparts (Table 1).

For these studies, a dosing paradigm of exenatide administered 1 h before a glucose challenge was used to mimic the effect of elevating active GLP-1. Here, wild-type mice were used initially to establish a baseline data set. Animals fasted overnight were injected with a range of exenatide doses (0.01–30 nmol/kg). One hour after dosing, blood was collected to measure exenatide (Fig. 2A), and mice then underwent an IPGTT where additional samples were collected over the next hour to determine plasma glucose concentrations (Fig. 2B). Substantial glucose lowering was observed in these experiments when circulating concentrations of exenatide were ≥39 pmol/L (50% improvement in glucose excursion AUC with 0.3 nmol exenatide/kg; P < 0.05). Because the glucoregulatory effects of GLP-1R activation include mechanisms beyond enhancing glucose-stimulated insulin secretion, especially its ability to delay gastric motility (33), follow-up OGTT assays were performed using a range of exenatide doses based on the IPGTT results. These experiments were performed in Gipr null mice using the following treatment schedule: exenatide was injected 1 h before administration of oral glucose. This paradigm should maximize the antihyperglycemic effects of GLP-1R activation. As anticipated, glucose tolerance was dose-dependently improved (36, 55, and 73% improvement in glucose excursion AUC for 0.025, 0.1, and 0.85 nmol exenatide/kg, respectively; P < 0.05) (Fig. 2C). Together, these results indicate that robust glucose lowering can be achieved by elevating the concentration of GLP-1R agonist or GLP-1 equivalents in the range of 10–100 pmol/L.

Multiple regulatory pathways influence incretin secretion, and agents that promote GLP-1 release recently have been discovered (34). To evaluate the ability of different mechanisms to increase circulating GLP-1, distinct pathways were investigated in Gipr null mice. It is well established that somatostatin in the gut exerts an inhibitory effect on GLP-1 release (35,36), and antagonism of SSTR5 has been shown to enhance GLP-1 secretion (37,38). Here, Gipr null mice orally administered the SSTR5 antagonist compound 3–1 (25) 30 min before an oral glucose challenge showed improved glucose excursion compared with animals receiving the vehicle (29% improvement in glucose excursion AUC; P < 0.05) (Fig. 3A). Consistent with relieving suppressive effects of somatostatin, plasma concentrations of GLP-1 after oral glucose were modestly increased in mice administered the SSTR5 antagonist (Fig. 3B). The second approach investigated was to stimulate secretion of GLP-1 (34) with a known GLP-1 secretagogue, the TGR5 agonist compound 18 (previously shown to be inactive in Tgr5 knockout mice [26]). Here, Gipr null mice were orally administered the compound and then underwent an OGTT. TGR5 agonist–treated mice displayed improved glucose tolerance (77% improvement in glucose excursion AUC; P < 0.05) (Fig. 3C) that was associated with elevated concentrations of GLP-1 (Fig. 3D). Although improvement of glucose lowering by the mechanisms of SSTR5 and TGR5 likely involves effector systems in addition to GLP-1, evaluating these pathways in the absence of GIP is advantageous because of the extraordinary effect of GIP after oral glucose delivery in mice.

To test the hypothesis that endogenous concentrations of GLP-1 can be increased to levels approaching those of injectable GLP-1 analogs, a combination of TGR5 agonist, SSTR5 antagonist, and sitagliptin was assessed in proof-of-concept studies. This treatment design is attractive because it uses agents that target complementary mechanisms: the TGR5 agonist stimulates GLP-1 secretion, the SSTR5 antagonist blocks inhibition of GLP-1 release, and the DPP4 inhibitor preserves active GLP-1. To evaluate the effect on glucose tolerance, the combination was orally administered to Gipr null mice undergoing an OGTT. As an impressive result, combined dosing of these compounds ablated the glucose excursion (96% reduced glucose excursion AUC; P < 0.05) (Fig. 4A). It is notable that plasma levels of GLP-1 resulting from the treatment showed more than 12-fold higher concentrations of GLP-1 than those observed with the single agents, demonstrating synergistic effects of simultaneously manipulating the three mechanisms, producing supratherapeutic levels of active GLP-1 (>50 pmol/L; C_max of 172 pmol/L) (Fig. 4B). To evaluate glucose tolerance without inhibiting gastric emptying, the triple combination was administered to Gipr null mice undergoing an IPGTT. Here, the combination also dramatically reduced glucose excursion (73% reduced glucose excursion AUC; P < 0.05) (Fig. 4C). These results are significant because glucose was lowered without the stimulus of glucose orally delivered to the gut lumen. In addition, to demonstrate that these levels of GLP-1 can be produced in normal animals without nutrient stimulus, wild-type mice fasted overnight were administered sitagliptin, the SSTR5 antagonist, the TGR5 agonist, or combination of all three, followed 15–180 min after administration for analysis of GLP-1. In line with the rationale for targeting multiple pathways that modulate GLP-1, the concentration of active GLP-1 produced by the combination greatly exceeded therapeutic levels (C_max of 350 pmol/L) (Fig. 4D). In addition, no changes in proinflammatory cytokines were observed (interleukin-6, interleukin-1β, or tumor necrosis factor-α; data not shown).

To assess the findings regarding GLP-1 in the context of recent results showing that combination treatment of a DPP4 inhibitor with a GPR40 agonist increases active GLP-1 (39), studies were performed to evaluate each agent on a background of sitagliptin therapy. Because the previous report did not disclose the specific GPR40 compound tested, our studies used the GPR40 agonist compound 20 (27). Here, head-to-head experiments were performed wherein wild-type mice received sitagliptin alone or in combination with either the SSTR5 antagonist, the TGR5 or GPR40 agonist, or both the SSTR5 compound plus one of the agonists. These studies confirmed that robust levels of active GLP-1 can be achieved by stimulating an alternate signaling pathway: the combination approach using the GPR40 compound showed similar effects to those of the TGR5 molecule (TGR5 vs. GPR40 combination) (Fig. 5A and B). It is important to note that the ability of compound 20 to stimulate GLP-1 secretion was abolished in Gpr40-deficient mice (Fig. 5B). These experiments also revealed the significance of blocking SSTR5. While combinations of just the DPP4 inhibitor and the TGR5 or GPR40 agonist increased GLP-1 concentrations (31 and 28 pmol/L, respectively; similar to levels reported for the DPP4 inhibitor–GPR40 agonist combination [39]), the addition of the SSTR5 compound is critical to substantially elevating GLP-1 (to 305 and 196 pmol/L, respectively).

Further studies were performed using the GPR40 agonist in place of the TGR5 compound and showed the combination approach with this molecule also dramatically elevates active GLP-1 over time (C_max of 170 pmol/L) (Fig. 5C). To determine whether the levels of GLP-1 were near maximal, both agonists (TGR5 and GPR40) were combined with the SSTR5 antagonist and sitagliptin. Because of different pH requirements for the salt forms of these agents, sitagliptin, the SSTR5 antagonist, and/or the GPR40 agonist were first dosed together, followed 30 min later by the TGR5 agonist or the vehicle; plasma was then collected and active GLP-1 analyzed. Despite differences in dosing times from earlier combinations, the results replicated the supratherapeutic effects of triple combinations with either the TGR5 or GPR40 agonists (C_max of 225 and 215 pmol/L, respectively). However, inclusion of both agonists in the treatment regimen doubled GLP-1 levels compared with either one alone, attaining elevations of active GLP-1 above any previous combinations attempted (C_max of 406 pmol/L) (Fig. 5D). It is important to note that the synergistic effects observed from a combination of three or four agents were not due to increases in compound exposures produced by coadministration (Table 2). Slight decreases in the concentrations of the SSTR5 and TGR5 compounds were observed when administered in a triple or quadruple combination. Only the DPP4 inhibitor showed somewhat higher exposure when combined with other treatments. Because the 10 mg/kg dose of sitagliptin systemically inhibits DPP4, slight increases in its concentration should not affect levels of GLP-1.

To assess whether targeting complementary mechanisms can also increase circulating GLP-1 in a setting of metabolic disease, diabetic Lepr^db/db mice were administered sitagliptin, the SSTR5 antagonist, the GPR40 agonist, or the compounds together. Plasma was collected 1 h later and, similar to previous findings, supratherapeutic concentrations of active GLP-1 were observed when the agents were combined (1 h = 460 pmol/L) (Fig. 6A). To evaluate whether the combination treatment can improve glucose tolerance in these mice and assess the contribution of elevated GLP-1, Lepr^db/db mice received the combination regimen with or without coadministration of the GLP-1R antagonist Jant-4 (28), and the animals were subjected to an IPGTT 1 h later (Fig. 6B). It is impressive that combined dosing of the compounds significantly improved glucose tolerance (47% reduced glucose excursion AUC; P < 0.05) (Fig. 6C), whereas GLP-1R blockade abolished glucose lowering (3% reduced glucose excursion AUC; no significant difference) (Fig. 6C). As expected, insulin levels were increased by the combination paradigm and reduced by Jant-4 (Fig. 6D). To fully characterize the combination therapy, further studies are needed to investigate vagal signaling or direct activity in accessible areas of the central nervous system that mediate metabolic actions of GLP-1. Portal vein GLP-1 concentrations and effects on glucagon secretion and gastric emptying should also be assessed.

The studies described in this report provide proof of principle that endogenous concentrations of GLP-1 can be increased to levels that offer substantial therapeutic benefit without requiring invasive procedures. Although data from the prototype drug combinations tested in these experiments strongly support the rationale for pursuing treatment regimens that intervene at complementary points in the regulation of GLP-1, several key challenges remain that must be addressed in order to realize a therapeutic advance. The most obvious hurdle is the availability of agents to include in such a treatment paradigm. While the levels of GLP-1 produced by the combination of a DPP4 inhibitor, SSTR5 antagonist, and TGR5 agonist and/or GPR40 agonist seem to be the highest achieved by pharmacological means, only DPP4 inhibitors are currently approved for human use. In preclinical studies, TGR5 agonists are some of the best reported GLP-1 secretagogues; however, these molecules increase gallbladder size, an effect that has not been overcome (26,40). Moreover, although several GPR40 agonists have been described, none have been submitted for marketing authorization. The field awaits the successful development of a GLP-1 secretagogue, which then would enable safety and efficacy studies of a combination therapy.

A major finding of this report is the effectiveness of targeting a negative regulatory pathway to enhance GLP-1 secretion. While molecules such as the TGR5 and GPR40 agonists promote GLP-1 release through Gαs or combined Gαs/Gαq stimulatory mechanisms (26,41), the studies here point to a substantial benefit of blocking activation of Gαi through SSTR5 antagonism, especially in the combination form. Although development of a potent GLP-1 secretagogue is necessary, agents such as SSTR5 antagonists that block counterregulatory signals may be critical to garnering the full therapeutic potential of the enteric system. This is exemplified in our studies showing that supratherapeutic levels of active GLP-1 are attained only when the SSTR5 antagonist is used.

Although our results suggest SSTR5 antagonism is key to elevating GLP-1 to robust concentrations, further investigation of this physiology is warranted. While blocking SSTR5 removes the brake on GLP-1 secretion (35–38), more intricate mechanisms may also be involved; somatostatin could play a role in downregulating the maximum potential of incretin-related targets by participating in a broad negative feedback loop. In support of this hypothesis, DPP4 inhibitors have been shown to elevate postprandial active GLP-1 and GIP concentrations while reducing total GLP-1 and GIP levels, suggesting that increasing active GLP-1 downregulates GLP-1 secretion (42,43). In addition, in rat intestinal cultures, GLP-1 stimulates somatostatin secretion, and SSTR5 activation restrains release of both GLP-1 and somatostatin, findings that support the concept that GLP-1 regulates its own secretion via somatostatin (36,37). Patients with extreme weight loss following metabolic surgery have been treated with the somatostatin analog octreotide, which results in dramatically reduced surgery-induced elevation of GLP-1, as well as a return of appetite and increased food intake (44,45), suggesting a dominant role of somatostatin in suppressing GLP-1. Together, the regulation of somatostatin in this complex system, and possibly the control of other Gαi-coupled targets in L cells, should be further studied.

From a drug discovery standpoint, another significant finding presented in this report is the importance of GIP signaling in mice following administration of oral glucose. This phenomenon was highlighted in OGTT studies showing that glucose lowering by treatment with the DPP4 inhibitor is similar in wild-type and Glp-1r null mice but significantly blunted in Gipr null mice. This observation may have strategic implications on efforts to develop new GLP-1 secretagogues. Because of the precedent that some GLP-1 secretagogues also increase GIP (12,42), testing schemes that use normal mice undergoing the OGTT run the risk of advancing compounds showing robust efficacy that is predominantly driven by GIP. Thus, to avoid confounding contributions of GIP, we propose Gipr null mice as a model for studying the effect of increasing endogenous concentrations of GLP-1. Such a model may have translational importance because patients with poorly controlled type 2 diabetes lack a robust response to GIP (46). It is important to note that Gipr null mice were shown here to secrete GLP-1 in response to an oral glucose bolus, similar to wild-type animals, and the mice responded proportionally to various concentrations of GLP-1 equivalents by injections of different doses of exenatide. These experiments validated the model, and its utility was demonstrated in studies testing a combination treatment regimen where active GLP-1 concentrations were elevated over 12-fold and the glucose excursion was completed ablated. It should be noted that while other factors regulating glucose homeostasis are likely in play as a result of the combination therapy, the improvement in glucose tolerance in mice lacking GIP action is, to our knowledge, unprecedented.

In addition to developing molecules to use in a combination treatment approach, a key remaining question is around the long-term ability of the gastrointestinal tract to produce high levels of GLP-1 continuously. While our studies are limited to acute responses, evidence from metabolic surgery suggests durability of the GLP-1 response. Not only are postprandial concentrations of GLP-1 significantly increased 2 days after surgery, levels of GLP-1 progressively increase for up to a year (44,47,48). Moreover, long-term studies in patients who have undergone the surgery demonstrate exaggerated GLP-1 responses to a mixed meal >10 years after the procedure (49,50). While pharmacological intervention strategies must await the clinical development of suitable agents, it seems that the gastrointestinal tract is capable of maintaining secretion of therapeutic levels of GLP-1.

In summary, the overall results in this report justify the development of oral medications targeting multiple pathways of GLP-1 modulation, including but not limited to GLP-1 secretagogues and SSTR5 antagonists. Our findings show pharmacological combinations of orally administered small molecules targeting different control points in GLP-1 biology can achieve GLP-1 equivalents near or above those demonstrated by injectable agonists or metabolic surgery.

Acknowledgments. The authors acknowledge the expert technical assistance of Samreen Syed, Miguel Toledo, Jorge Alsina-Fernandez, Nathan Yumibe, Yanyun Chen, Chafiq Hamdouchi, Jayana Lineswala, Brian Droz, and Amy Cox (Eli Lilly and Company). In addition, the authors acknowledge Ruth Gimeno and Alexander Efanov (Eli Lilly and Company) for providing opinions related to the writing of the manuscript.

Duality of Interest. D.A.B., A.B.B., M.D.M., and K.W.S. are employees of Eli Lilly and Company and may own company stock or possess stock options. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.A.B. designed and performed the experiments, analyzed the data, and wrote and revised the manuscript. A.B.B. and K.W.S. designed the experiments and wrote and revised the manuscript. E.J.G. performed the experiments, analyzed the data, and wrote the manuscript. M.D.M. designed the experiments and reviewed and edited the manuscript. K.W.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. A portion of the results reported here were presented at the 77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13 June 2017.