Dual Actions of Apolipoprotein A-I on Glucose-Stimulated Insulin Secretion and Insulin-Independent Peripheral Tissue Glucose Uptake Lead to Increased Heart and Skeletal Muscle Glucose Disposal
Apolipoprotein A-I (apoA-I) of HDL is central to the transport of cholesterol in circulation. ApoA-I also provides glucose control with described in vitro effects of apoA-I on β-cell insulin secretion and muscle glucose uptake. In addition, apoA-I injections in insulin-resistant diet-induced obese (DIO) mice lead to increased glucose-stimulated insulin secretion (GSIS) and peripheral tissue glucose uptake. However, the relative contribution of apoA-I as an enhancer of GSIS in vivo and as a direct stimulator of insulin-independent glucose uptake is not known. Here, DIO mice with instant and transient blockade of insulin secretion were used in glucose tolerance tests and in positron emission tomography analyses. Data demonstrate that apoA-I to an equal extent enhances GSIS and acts as peripheral tissue activator of insulin-independent glucose uptake and verify skeletal muscle as an apoA-I target tissue. Intriguingly, our analyses also identify the heart as an important target tissue for the apoA-I–stimulated glucose uptake, with potential implications in diabetic cardiomyopathy. Explorations of apoA-I as a novel antidiabetic drug should extend to treatments of diabetic cardiomyopathy and other cardiovascular diseases in patients with diabetes.
Cardiovascular disease (CVD) is the main cause of morbidity and mortality in diabetes, with a twofold to threefold increased risk for CVD in this patient group (1). In fact, patients with diabetes without prior myocardial infarction have the same number of total coronary heart disease events as subjects without diabetes with prior myocardial infarction (2). Detailed knowledge on how diabetes and CVD are connected is lacking, and studies that investigate how lipid and glucose metabolism are associated in patients with diabetes are thus warranted.
Apolipoprotein A-I (apoA-I) of HDL is involved in the reverse cholesterol transport pathway that reduces the atherosclerotic burden in the vascular wall (3). Other beneficial effects of apoA-I on CVD risk include its anti-inflammatory and antioxidative functions (reviewed in Rosenson et al. ). In recent years, apoA-I has also been ascribed a glucose-controlling function, and studies show that apoA-I/HDL induces glucose uptake in cultured muscle myotubes (5–7) and stimulates insulin secretion from cultured β-cells (8). Using an insulin-resistant diet-induced obesity (DIO) mouse model, we recently showed (9) that short-term apoA-I treatment leads to a significant increase in blood glucose clearance. The increased glucose uptake in peripheral tissues coincided with an increase in glucose-stimulated insulin secretion (GSIS), which suggests that the apoA-I treatment primes the β-cells to be more responsive to the glucose load. Although providing important knowledge about the integrative in vivo physiology, the relative contributions of increased GSIS versus direct and insulin-independent apoA-I effects on peripheral tissues are not known.
Here we used insulin-resistant DIO mice to investigate the action of apoA-I on glucose control in the absence of insulin secretion. With significance for multiaction therapeutic agents to control hyperglycemia in patients with diabetes and for treating diabetic cardiomyopathy, the findings describe apoA-I as directly inducing glucose uptake in skeletal muscle and heart.
Research Design and Methods
Expression and Purification of Recombinant Human ApoA-I
Human apoA-I containing a His tag at the N terminus was expressed in bacteria and purified using immobilized metal affinity chromatography followed by tobacco etch virus protease treatment to remove the His tag, as previously described (10,11). ApoA-I proteins were produced either in-house or at the Lund University Protein Production Platform (LP3).
Animals and Diets
Male C57BL/6 mice (Taconic, Ry, Denmark) were used at 9–12 weeks of age. Animals were on a 12-h light/dark cycle with nonrestricted food and water. The animals were fed a high-fat diet (HFD) (60% fat content; catalog #D12492; Research Diets, New Brunswick, NJ) for 2 weeks. The Malmö/Lund Committee for Animal Experiment Ethics, Lund, Sweden, approved all animal procedures.
Glucose Tolerance Test
Mice fasted overnight (12 h) were injected intraperitoneally with apoA-I protein (14 mg/kg in PBS, pH 7.4; control animals received NaCl). Glucose (50 mg/mouse, or as indicated) was injected intraperitoneally 3 h after treatments, followed by the collection of serum samples at the indicated times. Epinephrine at 0.1 or 1 mg/kg (Mylan, Stockholm, Sweden) and propranolol at 5 mg/kg (catalog #P0884; Sigma-Aldrich, Stockholm, Sweden) were coadministered with glucose as indicated (12–14). Blood glucose levels were measured (OnetouchUltra2; LifeScan, Milpitas, CA), and insulin and C-peptide levels were assayed in serum using ELISA (Mercodia, Uppsala, Sweden, and Crystal Chem, Downers Grove, IL, respectively).
Positron Emission Tomography and X-Ray Computed Tomography Analysis
Animals treated as described above were anesthetized using isoflurane and analyzed by a nanoscan positron emission tomography (PET)-computed tomography (CT) system (MEDISO, Budapest, Hungary) followed by PET imaging for 60 min and CT imaging for 5 min. Tracer amounts of 18F-labeled fluorodeoxyglucose (18F-FDG; contains the positron-emitting radionuclide 18F with a half-life of 110 min) were added to the glucose prior to injection. During the imaging, animal breathing was monitored to ensure a stable heart rate. The PET signal was integrated in 15-min intervals, as indicated using Interview FUSION software (MEDISO). In parallel experiments, nonanesthetized lean or HFD-fed animals were sacrificed 60 min after the FDG-glucose load, with or without propranolol/epinephrine, as indicated. The animals were dissected for liver, soleus muscle, heart, and adipose tissue (epididymal fat), and the 18F-FDG radioactivity (in counts per minute) of the organs/tissues was analyzed in a Wallac Wizard 1480 Automatic Gamma Counter (PerkinElmer, Waltham, MA). The radioactivity (in counts per minute) was corrected for time-dependent decay. The PET-CT analyses were performed at the Lund University Bioimaging Center (Lund, Sweden).
Heart tissue samples from animals 3 h after apoA-I injection and ±15 min after the glucose load (40 mg/mouse i.p.) were homogenized using an Omni TH Tissue Homogenizer (Omni International, Kennesaw, GA) in lysis buffer (20 mmol/L Tris, 250 mmol/L sucrose, 1 mmol/L EDTA, and 1% NP-40 with protease and phosphatase inhibitors). Lysates were centrifuged for 10 min at 10,000g, and protein concentrations were determined by Bradford protein assay. Lysates were heated at 95°C for 2 min in SDS sample buffer and subjected to PAGE on precast Bio-Rad (Hercules, CA) gradient gels and electrotransfer to nitrocellulose membrane. Ten micrograms of protein was loaded per sample according to Bradford protein quantification. Membranes were blocked for 30 min in 50 mmol/L Tris/HCl pH 7.6, 137 mmol/L NaCl, and 0.1% (weight for volume [w/v]) Tris-buffered saline with Tween-20 containing 10% (w/v) milk. The membranes were then probed with indicated antibodies (AMPK [catalog #2603], pAMPK [phospho-AMPK(Thr172)] [catalog #2535], AKT [catalog #4691], pAKT [phospho-AKT(Ser473)] [catalog #4060], and pGSK3 [phospho-GSK(α/β Ser 21/9)] [catalog #9331] antibodies; Cell Signaling Technology; and GSK3 [catalog #44-610] antibody; Bio-Source) in Tris-buffered saline with Tween-20 containing 5% (w/v) milk or 5% (w/v) BSA for 16 h at 4°C. Detection was performed using horseradish peroxidase–conjugated secondary antibodies and the enhanced chemiluminescence reagent. The signal was visualized using a Bio-Rad Image camera, and band intensities were quantified using Bio-Rad Imaging software.
Blood and Heart Lipid Analyses
Plasma samples and heart tissue samples obtained from animals 3 h after apoA-I injections were analyzed for triglyceride (TG) and free fatty acid (FFA) content. Lipids were extracted from the hearts following the method of Folch et al. (15). Briefly, the heart tissue was homogenized in PBS and protein content measured using the Bradford assay. The homogenate was mixed with 3 mL of 2:1 chloroform/methanol and centrifuged at 3,000g and 4°C for 10 min, and the organic phase was collected. The lipids were dried, resuspended in 500 µL of 1% Triton X-100 in chloroform, further dried, and then dissolved in 100 μL of double-distilled water. Assays for TGs and FFAs were performed using kits (Abcam) following the instructions of the manufacturer. The lipid contents in the tissue samples were normalized to protein content.
All data are displayed as the mean ± SEM. Analysis was performed by one-way ANOVA with Tukey post hoc test or, where indicated, by Student t test, using GraphPad Prism software. P ≤ 0.05 was considered to be significant.
Relative Contributions of ApoA-I Treatment on Insulin-Dependent and Insulin-Independent Glucose Disposal in Insulin-Resistant DIO Mice
Short-term apoA-I treatment of insulin-resistant DIO mice leads to a remarkable increase in blood glucose clearance (9). To elucidate to what extent this is due to increased GSIS and/or a direct effect of apoA-I on peripheral tissues, such as skeletal muscle (i.e., insulin-independent effects), an experimental approach to block insulin secretion in the short term and transiently in insulin-resistant DIO mice (Fig. 1A) by the administration of low doses of propranolol and epinephrine was used (12–14). To establish the model, we first analyzed the effects of the chemical inhibition of GSIS (Fig. 1B). Although both concentrations of epinephrine (0.1 or 1 mg/kg) tested (at a fixed concentration of propranolol) showed rapid decreases in serum insulin levels to <100 pmol/L, the lower dose of epinephrine (0.1 mg/kg) was deemed sufficient for the experiments. Next, animals were treated with apoA-I, or NaCl solution as control, for 3 h followed by a glucose tolerance test (GTT) with or without epinephrine/propranolol to create insulin release blockade. The glucose clearance capacity of apoA-I–treated mice was markedly improved compared with the NaCl control group, with a 55% reduction of the area under the curve (AUC) for glucose (Fig. 1C and D) and a 2.5-fold increase of the AUC for insulin (Fig. 1E and F), confirming earlier findings (9). The insulin blockade in the NaCl control mice resulted in a small but statistically significant increase in the AUC value for glucose and a trend (statistically not significant) of reduction in circulating insulin levels. This indicates that insulin-stimulated glucose uptake in peripheral target tissues of the DIO mice is absent in the insulin-blocked group. In apoA-I–treated mice, insulin blockade resulted in normalization of the insulin level (i.e., no increase vs. NaCl control group) despite the injected apoA-I (Fig. 1F, AUC values), including a loss of the early-phase insulin secretion (Fig. 1G). The blockade of GSIS in the apoA-I–treated group also led to a decreased capacity to clear plasma glucose; however, it was still significantly enhanced compared with capacities in the two NaCl control groups (Fig. 1D), demonstrating a direct, insulin-independent effect of apoA-I on glucose clearance in peripheral tissues.
The AUC values were further used to quantify the relative contributions. Based on the difference in AUC for glucose between the NaCl group and the apoA-I protein–treated group being set to 100% (i.e., the total “apoA-I effect”), the direct insulin-independent effect of apoA-I protein was calculated to account for 44% of the improved glucose clearance capacity, whereas the increased insulin levels in circulation in the apoA-I protein–treated group constitute 56% of the total effect. Finally, the insulin/C-peptide ratio was analyzed to investigate a potential role of the liver in the altered levels of circulating insulin (Fig. 1H). However, no apparent difference was observed, suggesting that the clearance of insulin by the liver is not affected by apoA-I.
In Vivo Glucose Distribution in Insulin-Resistant DIO mice: PET-CT Scanning
PET combined with X-ray CT scan analysis is a noninvasive in vivo imaging approach that has been used to study glucose disposal in lean healthy adults (16), and in insulin-resistant individuals (17) and in individuals with type 2 diabetes (18), as well as to study glucose disposal in rodents (19). We here used PET-CT scanning together with the glucose tracer molecule 18F-FDG to study tissue-specific accumulation of glucose in HFD-fed animals. The glucose tracer molecule is taken up by cells via facilitated glucose transport and can be phosphorylated by hexokinase in the cytosol but not further metabolized (18). PET-CT scan images (Fig. 2A) showed that apoA-I–treated animals displayed an increase of 18F-FDG in the heart, which was particularly pronounced in the insulin-blocked animals. To enable direct quantification, parallel groups of animals were injected with glucose/18F-FDG and sacrificed after 60 min of incubation, followed by dissection and γ-quantification of radioactive signal. Short-term treatment with apoA-I protein did not affect glucose uptake in the liver (Fig. 2B) or in epididymal fat (Fig. 2C), and insulin blockade had modest effects on glucose uptake in these tissues. Soleus muscle, which is known to have a decreased insulin-stimulated glucose uptake capacity in insulin-resistant human individuals (17), displayed a reduction in glucose accumulation (44% less accumulated 18F-FDG) in the insulin-blocked animals versus controls (Fig. 2D). Treatment with apoA-I resulted in a significant increase (1.6-fold increase) in glucose uptake in soleus muscle in mice with preserved GSIS (Fig. 2D, apoA-I vs. NaCl). In insulin-blocked animals, apoA-I protein treatment led to a 2.1-fold increase in the total accumulated glucose (Fig. 2D, apoA-I + block vs. NaCl + block). In the heart, apoA-I–treated animals showed a trend toward a 3.6-fold increase in glucose/18F-FDG content compared with NaCl control animals (Fig. 2E, apoA-I vs. NaCl), whereas insulin blockade in NaCl control animals showed a trend toward reduced glucose/18F-FDG content (Fig. 2E, 44% less for NaCl + block vs. NaCl). The apoA-I had the largest effect on the insulin blockade animals, with a 14-fold increase in glucose/18F-FDG accumulation compared with the insulin-blocked NaCl control group (Fig. 2E, apoA-I + block vs. NaCl + block). Unexpectedly, the observed increase exceeded by 2.1-fold that of apoA-I–treated animals, suggesting that the lack of insulin and/or increased blood glucose availability enhanced the apoA-I effect on the heart. The finding was supported by visualization of the time-dependent accumulation of glucose/18F-FDG in the heart, as integrated over 15-min time intervals (Fig. 2F).
Glucose Disposal in Lean Mice
As shown above, short-term apoA-I protein treatment of obese and insulin-resistant mice leads to increased glucose tolerance. We next asked whether the obese and insulin-resistant state is a prerequisite for the positive effects of the apoA-I protein, or whether mice with a more balanced metabolic state also responded in a similar manner. To investigate this, we monitored the effects of apoA-I treatment in lean mice. Animals were treated as in Fig. 1A, but with the HFD replaced with a normal chow diet, followed by GTT and PET-CT analyses. The apoA-I treatment resulted in a reduction in basal blood glucose levels (∼30% lower compared with the NaCl control) (Fig. 3A) but no significant change in basal insulin levels (Fig. 3B). The glucose clearance capacity in the lean animals was more efficient than in the DIO mice (AUC glucose levels were 19–34% lower in the respective treatment groups in the lean vs. DIO mice) but followed the same relative pattern as for the DIO animals (compare Fig. 3C and D with Fig. 1C and D, respectively). The AUC insulin values were low (<20,000 pmol/L × min) and comparable among the four groups (data not shown).
Next, we used the glucose tracer molecule 18F-FDG to study tissue-specific accumulation of glucose in lean animals. Lean mice were treated in a manner similar to that described for the DIO mice in Fig. 2. No significant differences in glucose accumulation were observed in liver and epididymal fat among the four groups (data not shown). Soleus muscle tissue showed a trend toward increased uptake in the two groups of apoA-I–treated animals (with or without insulin block) compared with that in the NaCl control group (Fig. 3E). Glucose uptake in the heart was significantly affected with a twofold to threefold increase in the apoA-I–treated groups compared with control groups (Fig. 3F and G).
Titration of Glucose: In Vivo Glucose Clearance and Uptake by Heart
To further investigate how changes in blood glucose concentration affect the glucose uptake in the heart, mice were treated as in Fig. 1A, followed by GTT analysis and γ-quantification of the radioactive signal (18F-FDG tracer) in dissected hearts, using decreasing amounts of injected glucose (40, 30, and 20 mg/mouse i.p.) (Fig. 4). All apoA-I–treated groups had significantly lower AUC glucose values compared with the NaCl control group (Fig. 4A and B), whereas there was no statistically significant difference in AUC insulin values between the treatment groups (Fig. 4C and D). However, a trend of a 1.3-fold increase in the AUC insulin value in the apoA-I–treated animals compared with NaCl control was observed. Moreover, the data confirmed the results in Fig. 1D, with a relative contribution of 44% and 56%, respectively, for the insulin-independent effect of apoA-I protein and the increased levels of insulin in circulation. Titration of glucose (40, 30, and 20 mg/mouse) in the insulin-blocked animals showed a trend toward reduced AUC glucose values in the 20 mg/mouse group (Fig. 4B), and significant decreases in 18F-FDG/glucose accumulation in the heart for both the 20 and 30 mg/mouse groups compared with the 40 mg/mouse group (Fig. 4E). Interestingly, the 18F-FDG signals in the 30 mg/mouse group were reduced by 45% compared with those in the 40 mg/mouse group (both insulin blocked) in the presence of comparable levels of glucose in circulation (Fig. 4A and B), yet were twofold to fivefold higher for the apoA-I groups than for the NaCl control group (Fig. 4E), suggesting that the apoA-I–stimulated increase in glucose uptake to the heart is particularly important at a high glucose load in the insulin-devoid state but clearly is still active even at low glucose levels.
Cellular Signaling in the Heart
Cellular signaling pathways in the hearts were analyzed by immunological detection. HFD-fed mice were treated with apoA-I or NaCl control intraperitoneal injections. After 3 h, half of the mice received a glucose load (intraperitoneal), and all mice were sacrificed after an additional 15 min and dissected to obtain their hearts. Protein extracts were separated by SDS-PAGE and analyzed for immunologic detection using p-specific antibodies for AKT, GSK3, and AMPK, as well as their total protein content. The apoA-I treatment resulted in an increase in AKT phosphorylation in the basal state, which was further amplified after the glucose challenge (Fig. 5A). Similarly, the signal ratios of GSK, which is a downstream target of AKT, were increased in the apoA-I–treated mice in both the basal and the stimulated states (Fig. 5B), whereas the effect of apoA-I on AMPK activation was less clear (Fig. 5C).
Lipid Analyses of Heart and Blood Serum
Finally, the levels of FFA and TG in the blood serum and the heart were determined. Data from serum and heart tissue collected from fasted DIO mice 3 h after apoA-I and NaCl injections are shown in Fig. 6. The FFA and TG content in the heart was similar between the two groups, whereas a trend toward a reduction in the levels of both TG (30% reduction) and FFA (26% reduction) is seen in serum.
Our study demonstrates that apoA-I potentiates an increased GSIS and exerts a direct insulin-independent stimulation of glucose disposal in peripheral tissues with quantitatively similar effects on the clearance of blood glucose. The data also verify muscle as an important in vivo target for the apoA-I effect, with more than a doubling of the glucose uptake in the insulin-blocked state. On the contrary, liver and fat glucose uptake appeared not to be influenced by the acute apoA-I treatment, and was only modestly affected by the insulin blockade, with the latter potentially being explained by the severe insulin resistance already occurring in these tissues after short-term HFD treatment (20). However, the fact that the effects of apoA-I on improved blood glucose control are also seen in the lean metabolically healthy animals suggests that the mechanism of action of the apoA-I protein is not to repair the metabolic defects occurring in the obese and insulin-resistant state. Rather, the data suggest a unique mechanism pathway with which to increase glucose tolerance that can be triggered that is independent of the metabolic state of the animals. From a therapeutic perspective, this is an advantage, because a potential antiglycemic drug based on the apoA-I biology could be effective in insulin-resistant patient groups irrespective of the nature of their reduced insulin sensitivity. Further studies will be needed to understand the mechanistic details.
The most striking effect of our study is the considerable apoA-I–induced glucose uptake in the heart, which is particularly pronounced after the blockade of insulin secretion in the DIO mice. The apoA-I–stimulated uptake in the heart also occurs in the lean mice. However, we observed no difference in the level of glucose uptake in the heart between the two groups of apoA-I–treated animals despite the fact that the insulin-blocked lean mice had a reduction in glucose clearance (Fig. 3C and D, apoA-I vs. apoA-I + block). The fact that this compensatory effect occurs only in the obese and insulin-resistant state suggests that the increased uptake in the heart occurs via an actively regulated mechanism and is not caused merely by the higher glucose availability. The finding that the high glucose levels in the NaCl control–treated animals (Fig. 1C and D) did not per se lead to an increase in heart glucose uptake (Fig. 2E) further supports this notion. However, a partial contribution of higher glucose availability to the increased glucose accumulation in the hearts of the apoA-I–treated and insulin-blocked mice can be expected, and was seen as a gradual decrease of injected glucose led to lower levels of FDG uptake (Fig. 4E).
The relative FDG uptake per tissue mass observed here is comparable to what was previously described in wild-type and insulin receptor substrate knockout mice (19) as well as in human subjects with diabetes and control subjects (21). Considering that the heart mass is only about a few percent of the total muscle mass (0.15 g for heart and ∼8–10 g for skeletal muscle) (22), skeletal muscle is quantitatively more significant for glucose clearance, yet an altered cardiac energy metabolism profile may have implications in diabetic cardiomyopathy. It is known that glucose utilization as a source of energy by the cardiomyocyte is impaired in diabetes (23) and that the diabetic heart depends almost completely on fatty acid oxidation for ATP (24,25). This lack of energy flexibility is considered to be one of the factors that lead to diabetic cardiomyopathy (26,27). ApoA-I may thus provide a favorable shift in the balance of energy use toward increased glucose metabolism. Because both AKT and its downstream target GSK3 were activated in the hearts of the apoA-I–treated, HFD-fed animals, and this was particularly pronounced in the glucose-stimulated state, our data suggest that the increased uptake in the heart occurs via AKT signaling. However, because we also found that insulin blockade in the mice led to further amplification of the heart glucose uptake, the apoA-I–related AKT signaling appears to be regulated by other means than the classic insulin-signaling pathway (i.e., not activation/phosphorylation after binding of insulin to the insulin receptor). Importantly, the accelerated glucose uptake occurred without any net change in total FFA and TG levels in the heart, suggesting that the shift in the use of energy source was independent of the general lipid availability. However, the injected apoA-I has an effect on the circulating lipid profile, as shown by the lowered serum levels of FFA and TG already occurring within the 3-h time span, and alterations of specific lipids with potential regulatory roles in heart glucose uptake cannot be ruled out. For example, we and others (28–30) have previously shown that apoA-I–mediated cholesterol efflux and changes in the plasma membrane cholesterol content affect signaling pathways in adipocytes and pancreatic β-cells. Further studies focused on the detailed cellular regulatory mechanism will address this.
In summary, we have shown that apoA-I treatment of lean and obese mice leads to increased glucose tolerance via dual actions. This involves amplification of insulin release after glucose challenge as well as insulin-independent increase in glucose disposal. The data confirm skeletal muscle and identify the heart as targets for apoA-I–stimulated glucose uptake. The stimulated uptake in the heart is suggested to occur via a new regulatory mechanism that involves activation of the AKT pathway, and potentially also an altered cellular lipid profile, although total TG and FFA levels were not altered during the short treatment. Because patients with diabetes are known to have reduced apoA-I/HDL levels, one may speculate that a resulting reduced heart muscle glucose uptake contributes to the increased risk of CVD and diabetic cardiomyopathy for such patients. This is an exciting novel aspect of apoA-I/HDL biology that may provide a basis for new therapeutic agents in the treatment of diabetes and CVD.
Acknowledgments. The authors thank Drs. Gustav Grafström and Thuy Tran (Lund University Bioimaging Center) for valuable assistance in the PET-CT analyses, Ewa Krupinska and Dr. Wolfgang Knecht (Lund University Protein Production Platform [LP3]) for providing the proteins, and Anki Knutsson (Lund University) for assistance in the heart dissections.
Funding. This work was supported by grants from the Swedish Research Council (K2014-54X-22426-01-3), the Swedish Diabetes Foundation, the Albert Påhlsson Foundation, the Tage Blücher Foundation, and the Royal Physiographic Society in Lund, Sweden.
Duality of Interest. K.G.S. and J.O.L. have filed a patent application that partly relates to the findings of this study. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. J.D.-E. wrote the manuscript, contributed to the study design, and researched the data. M.L. and O.N.-W. contributed to the study design, researched the data, and reviewed and edited the manuscript. S.W.C and K.G.S contributed to the study design and discussion, and reviewed and edited the manuscript. J.O.L. contributed to the study design, contributed to the analysis and interpretation of the data, and wrote the manuscript. J.O.L 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.
- Received October 29, 2015.
- Accepted April 12, 2016.
- © 2016 by the American Diabetes Association. Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered.