Adiponectin Increases Skeletal Muscle Mitochondrial Biogenesis by Suppressing Mitogen-Activated Protein Kinase Phosphatase-1
Adiponectin enhances mitochondrial biogenesis and oxidative metabolism in skeletal muscle. This study aimed to investigate the underlying mechanisms through which adiponectin induces mitochondrial biogenesis in skeletal muscle. Mitochondrial contents, expression, and activation status of p38 mitogen-activated protein kinase (MAPK) and PPARγ coactivator 1α (PGC-1α) were compared between skeletal muscle samples from adiponectin gene knockout, adiponectin-reconstituted, and control mice. Adenovirus-mediated adiponectin and MAPK phosphatase-1 (MKP1) overexpression were used to verify the relationship of MKP1 and PGC-1α in adiponectin-enhanced mitochondrial biogenesis using cultured C2C12 myotubes and PGC-1α knockout mice. An inhibitory effect of adiponectin on MKP1 gene expression was observed in mouse skeletal muscle and cultured C2C12 myotubes. Overexpression of MKP1 attenuated adiponectin-enhanced mitochondrial biogenesis, with significantly decreased PGC-1α expression and p38 MAPK phosphorylation. Although in vivo adiponectin overexpression reduced MKP1 protein levels, the stimulative effects of adiponectin on mitochondrial biogenesis vanished in skeletal muscle of PGC-1α knockout mice. Therefore, our study indicates that adiponectin enhances p38 MAPK/PGC-1α signaling and mitochondrial biogenesis in skeletal muscle by suppressing MKP1 expression.
Owing to huge tissue mass and relatively high energy demand, skeletal muscle plays a critical role in energy expenditure in humans and most rodents (1,2). The mitochondrion is an organelle in mammalian cells that converts nutritional metabolites into adenosine triphosphate for cellular energy supply. It has been well documented that mitochondrial dysfunction, particularly impaired oxidative metabolism, in skeletal muscle is closely associated with obesity and insulin resistance (3–6).
Adiponectin was initially identified as an adipocyte-derived hormone that regulates energy homeostasis by increasing insulin sensitivity (7,8). Several recent studies have demonstrated that skeletal muscle also expresses adiponectin (9–12). Most importantly, regardless of the origin of adiponectin, significant evidence suggests that adiponectin improves skeletal muscle oxidative metabolism through its own receptors and downstream signaling (7,13–15). Adiponectin mRNA and blood protein levels are inversely associated with obesity (7), which is a common cause of insulin resistance in humans. Therefore, it has been proposed that hypoadiponectinemia and impaired adiponectin signaling may contribute to the decreased skeletal muscle oxidative metabolism in obesity-associated insulin resistance and even type 2 diabetes.
Using adiponectin gene knockout and transgenic mouse models, studies have demonstrated that adiponectin increases mitochondrial biogenesis and oxidative capacity in skeletal muscle (16,17). Activated AMP-activated protein kinase and increased peroxisome proliferator–activated receptor γ coactivator 1α (PGC-1α) have been suggested to mediate the regulatory effects of adiponectin on mitochondrial biogenesis and function (13,16–18). However, the signaling pathway from adiponectin receptors to PGC-1α gene expression is still largely unknown.
p38 is a member of the mitogen-activated protein kinase (MAPK) family and has been identified as a downstream molecule in the adiponectin-signaling pathway (15,18,19). There are four isoforms of p38 MAPK (α, β, γ, and δ), and p38α and p38γ are most abundantly expressed in skeletal muscle (20,21). p38 MAPK plays an important role in maintaining skeletal muscle energy homeostasis, myotube differentiation, and skeletal muscle tissue mass (20–23). Activation of p38 MAPK increases not only PGC-1α gene expression but also PGC-1α activity (24–27). Inhibition of p38 MAPK completely blocks Ca2+-induced PGC-1α gene expression (28). Studies have also reported that adiponectin increases PGC-1α expression followed by mitochondrial biogenesis and fatty acid oxidation (16–18). Therefore, p38 MAPK may play a pivotal role in mediating adiponectin-stimulated PGC-1α gene expression, activation, and mitochondrial biogenesis in skeletal muscle. Identifying how adiponectin activates p38 MAPK will provide important information regarding the underlying mechanism through which adiponectin induces PGC-1α expression and mitochondrial biogenesis in skeletal muscle.
MAPK phosphatases (MKPs) are a family of protein phosphatases that specifically dephosphorylate the Thr and Tyr residues, also known as TXY motif, of MAPKs, leading to the deactivation of MAPKs (29,30). Therefore, MKPs play a critical role in regulating the activity of MAPKs. Similar to other MAPKs, the level of phosphorylation of p38 MAPK at Thr180 and Tyr182 is determined by the balance of the activities of MAPK kinases (MKKs) and MKPs.
Here, we provide evidence that MKP1 plays an important role in mediating adiponectin-enhanced mitochondrial biogenesis and oxidative metabolism. Our study demonstrated that adiponectin suppresses MKP1 protein expression in skeletal muscle, thereby leading to p38 MAPK activation and PGC-1α gene expression.
RESEARCH DESIGN AND METHODS
Acetyl-CoA, oxaloacetate, 5,5′-dithiobis (2-nitrobenzoic acid), triethanolamine, and insulin were purchased from Sigma (St. Louis, MO). Anti-MKP1 antibody was from Millipore (Billerica, MA). Antibodies for PGC-1α and phospho– and total p38 MAPK (all isoforms) were from Cell Signaling (Danvers, MA). Horse serum, penicillin-streptomycin, Dulbecco’s modified Eagle’s medium (DMEM), and MitoTracker Green FM were from Invitrogen (Carlsbad, CA). Recombinant mouse adiponectin protein was purchased from Biovendor (Candier, NC).
Adiponectin gene knockout (Adipoq−/−) mice were created in Dr. Philipp Scherer’s laboratory (31) on a 129/SvEv genetic background and were backcrossed to C57BL/6 for six generations. Wild-type (WT) littermates were used as controls for Adipoq−/− mice. PGC-1α gene knockout mice (PGC-1α−/−) were provided by Dr. Daniel P. Kelly (Washington University, St. Louis, MO) (32). All mice were maintained under standardized conditions with a 12-h/12-h light/dark cycle. The experiments using mouse models were carried out under the Association for Assessment and Accreditation of Laboratory Animal Care guidelines with approval of the Animal Care and Use Committee. To overexpress or reconstitute adiponectin, we used adenovirus-Acrp30 (Ad-Acrp30) vector–mediated intravenous gene transduction, which leads to transduced gene expression primarily in hepatocytes (33–35). For these studies, purified adenovirus (1 × 109 pfu per mouse) vector was injected into the mouse through the tail vein (35). Tissue samples were collected 3 days after viral injection. Consistent with our previous report (34), mouse blood adiponectin levels were increased ∼10- to 12-fold compared with WT mice that received adenovirus encoding green fluorescent protein (GFP) (Supplementary Fig. 1A). In addition, blood multimeric adiponectin levels were proportionally elevated in Ad-Acrp30–injected mice (Supplementary Fig. 1B) (34).
C2C12 myoblasts were purchased from American Type Culture Collection. Immortalized mouse embryonic fibroblasts (MEFs) from MKK3 and -6 double knockout or WT mice were established by the laboratory of Dr. Roger Davis (University of Massachusetts Medical School, Worcester, MA). C2C12 cells were maintained in DMEM supplemented with 10% (Gemini Bio-Products, Woodland, CA), 200 mmol/L l-glutamine, and 200 units/mL penicillin and 50 ug/mL streptomycin in an atmosphere of 95% air and 5% CO2. When C2C12 cells reached 90% confluence, myotube differentiation was induced by switching the medium to DMEM supplemented with 2% horse serum. The culture medium was changed daily. Polynucleated myotubes were obtained within 4 days. A coculture system was used for adiponectin treatment (33). After overnight coculture, adiponectin in the medium reached a level of 40–50% of that in C57BL/6 mouse serum (Supplementary Fig. 1C). All multimeric adiponectins were secreted from Ad-Acrp30–transduced FAO cells and were present in the coculture medium (33) (Supplementary Fig. 1B). Differentiated C2C12 myotubes were cocultured overnight in an insert well with Ad-Acrp30– or Ad-GFP–transduced FAO cells.
Adenoviral vectors and plasmid construction.
Adenoviral vectors encoding adiponectin, MKP1, LacZ, and GFP were created using the ViraPower Adenoviral Expression system (Invitrogen). Adenovirus was produced in HEK293 cells and purified by cesium chloride gradient centrifugation as previously described (34). Plasmids pcDNA-MKP1, pcDNA-p38, and pcDNA-PGC-1α were created by subcloning the cDNAs into pcDNA3.1 vector. Plasmid vector encoding 262A/265A/298A triple-mutated PGC-1α (PGC-1α 3A) was provided by Dr. Pere Puigserver (Dana-Farber Cancer Institute, Boston, MA). The luciferase reporter construct was generated by inserting the 2-kb mouse Cidea promoter into pGL3 basic vector (pGL3-Cidea-Luc).
Mitochondrial content and enzyme activity assay.
Mitochondrial content was determined by measuring mitochondrial DNA (mtDNA) levels, mitochondrial mass with MitoTracker Green probe, and citrate synthase activity. DNA was prepared using a Qiagen DNeasy blood and tissue kit. mtDNA was quantified using real-time PCR using a pair of primers for the Cox2 mtDNA region: forward, 5′-TTTTCAGGCTTCACCCTAGATGA-3′; reverse, 5′-GAAGAATGTTATGTTTACTCCTACGAATATG-3′ (36). MitoTracker Green is a cell-permeable mitochondrion-selective dye and has been used for mitochondrial mass measurement (16). Briefly, after adiponectin treatment, C2C12 myotubes were washed twice with PBS and stained with 100 nmol/L MitoTracker Green FM for 30 min at 37°C. Fluorescence intensity was detected following the protocol provided by the manufacturer, and values were normalized to total protein levels.
Citrate synthase activity was measured spectrophotometrically using cell or tissue homogenates (16). Briefly, citrate synthase activity was measured at 37°C in 0.1 mol/L Tris-HCl (pH 8.3) assay buffer containing 0.1 mmol/L 5,5′-dithiobis (2-nitrobenzoic acid), 0.5 mmol/L oxaloacetate, 50 μmol/L EDTA, and 5 mmol/L triethanolamine hydrochloride. After an initial 2-min absorbance reading taken at 412 nm, the reaction was initiated with the addition of 3.0 mmol/L acetyl-CoA, and the change in absorbance was measured every 10 s for 2 min.
Data are expressed as means ± SEM. Statistical analysis was performed using Student t test or ANOVA, followed by contrast test with Tukey or Dunnett error protection. Differences were considered significant at P < 0.05.
Adiponectin induces p38 MAPK activation, PGC-1α expression, and mitochondrial biogenesis in skeletal muscle.
Using Adipoq−/− mice and adenovirus-mediated in vivo gene transduction, our results show that phosphorylation of p38 MAPK at Thr180 and Tyr182, which represents its activation state, was significantly reduced in skeletal muscle of Adipoq−/− mice (Fig. 1A). Consistent with previous reports (16–18), mitochondrial content and expression levels of PGC-1α and mitochondrial marker Cox5a were significantly reduced in skeletal muscle of Adipoq−/− mice (Supplementary Fig. 2A–C). In contrast, adiponectin reconstitution increased p38 MAPK phosphorylation in skeletal muscle from mice that received Ad-Acrp30 injection (Fig. 1A). Furthermore, overnight adiponectin treatment also significantly increased p38 MAPK phosphorylation and PGC-1α protein levels in cultured C2C12 myotubes (Fig. 1B). The levels of mtDNA and Cox5α mRNA were also significantly increased in adiponectin-treated C2C12 myotubes (Supplementary Fig. 2D and E). These results clearly indicate that adiponectin increases p38 MAPK phosphorylation, PGC-1α expression, and mitochondrial biogenesis in skeletal muscle.
Adiponectin suppresses MKP1 gene expression in skeletal muscle.
MKK3 and MKK6 are two key upstream kinases for p38 MAPK activation. To verify the role of MKK3 and MKK6 in adiponectin-induced p38 MAPK activation, MEFs from MKK3 and MKK6 double knockout (MKK3&6KO) mice were used. The results show that the magnitudes of adiponectin-induced p38 MAPK phosphorylation were significantly lower in MKK3&6KO MEFs compared with WT MEFs (Fig. 2), which suggests that MKK3 and MKK6 are required for maximal adiponectin-stimulated p38 MAPK phosphorylation. This result is in line with a recent study, which reported that MKK3 plays an important role in adiponectin-induced p38 MAPK phosphorylation (19). However, although there was a reduction of adiponectin-induced p38 MAPK phosphorylation in MKK3&6KO MEFs, adiponectin treatment still significantly increased p38 MAPK phosphorylation in MKK3&6KO MEFs (Fig. 2). In other words, MKK3 and -6 gene deletion did not completely eliminate the stimulatory effects of adiponectin on p38 MAPK activation. These results suggest that, in addition to MKK3 and MKK6, other molecules or signaling pathways are involved in adiponectin-induced p38 MAPK activation in skeletal muscle.
MKP1 is the prototype of the MKP family and is ubiquitously expressed. Our results show that the expression levels of MKP1 were significantly increased in skeletal muscle of Adipoq−/− mice, while adiponectin reconstitution reduced MKP1 protein levels in skeletal muscle (Fig. 1A). Similar inhibitory effects of adiponectin on MKP1 expression were observed in C2C12 myotubes (Fig. 1B). These results indicate that adiponectin inhibits MKP1 expression in skeletal muscle.
Adiponectin enhances p38 MAPK phosphorylation, PGC-1α expression, and mitochondrial biogenesis by suppressing MKP1 expression in C2C12 myotubes.
A previous in vivo study reported that MKP1-deficient mice expressed higher levels of PGC-1α in skeletal muscle with enhanced energy expenditure (30). Using adenovirus-mediated gene transduction and cultured C2C12 myotubes, we studied the effects of MKP1 on PGC-1α gene expression and mitochondrial biogenesis. As expected, MKP1 overexpression robustly decreased p38 MAPK phosphorylation (Fig. 3A). Expression of PGC-1α was also significantly reduced at the protein (Fig. 3A) and mRNA (Supplementary Fig. 3) levels. Mitochondrial content and citrate synthase activities were also significantly lower in Ad-MKP1–transduced myotubes compared with Ad-GFP–treated cells (Fig. 3B and C). Results from these in vitro studies further demonstrate that MKP1 plays an important role in PGC-1α gene expression and mitochondrial biogenesis in myotubes.
Posttranslational modification plays an important role in regulating PGC-1α activity. p38 MAPK phosphorylates Thr262, Ser265, and Thr298 residues of PGC-1α and increases its activity (25). MKP1 reduces p38 MAPK activity. We carried out a functional assay to determine whether MKP1 reduces PGC-1α activity through p38 MAPK–induced PGC-1α phosphorylation by using Cidea promoter–directed luciferase reporter construct and a plasmid encoding PGC-1α with 262A/265A/298A triple mutation (PGC-1α 3A) (25,37). Consistent with previous reports (25), p38 MAPK increased PGC-1α activity, while MKP1 displayed an opposing effect (Fig. 3D). However, MKP1 failed to reduce luciferase activities in cells expressing PGC-1α 3A. Together, these results indicate that MKP1 reduces PGC-1α activity by decreasing both its expression and p38 MAPK–induced phosphorylation.
For determination of the role of MKP1 in adiponectin-enhanced PGC-1α gene expression and mitochondrial biogenesis, MKP1 was overexpressed using an adenovirus vector to counteract the inhibitory effects of adiponectin on MKP1 expression. MKP1 protein levels were significantly elevated in C2C12 myotubes owing to the potent CMV promoter (Fig. 3A). Similar to the results presented in the section above, overnight adiponectin treatment significantly increased PGC-1α expression (Supplementary Fig. 4A) and levels of mitochondrial content and citrate synthase activity in control vector-transduced C2C12 myotubes (Fig. 3E and Supplementary Fig. 4B). However, expression of PGC-1α and mtDNA and citrate synthase activity were comparable between adiponectin-treated and control cells with MKP1 overexpression (Fig. 3E and Supplementary Fig. 4). These results indicate that MKP1 overexpression abolishes the stimulating effects of adiponectin on PGC-1α gene expression and mitochondrial biogenesis in myotubes.
MKP1 mediates adiponectin-enhanced mitochondrial biogenesis through PGC-1α.
The results from the above studies suggest that adiponectin increases PGC-1α gene expression and mitochondrial biogenesis in skeletal muscle. Our studies and results from other studies (30) indicate that MKP1 inhibits mitochondrial biogenesis. For verification of the relationship of MKP1 and PGC-1α in adiponectin-induced mitochondrial biogenesis in skeletal muscle, the PGC-1α−/− mouse model was used. Blood adiponectin protein levels were significantly elevated after adenoviral vector–mediated in vivo gene transduction (34). Overexpression of adiponectin increased PGC-1α protein levels in skeletal muscle from WT control mice (Fig. 4A). The levels of citrate synthase activity (Fig. 4B), mitochondrial marker Cox5α, and mtDNA (Fig. 4C and D) were also significantly increased in Ad-Acrp30–treated WT mice. However, despite significantly reduced MKP1 protein levels in Ad-Acrp30–treated WT and PGC-1α−/− mice (Fig. 4E), adiponectin overexpression failed to increase Cox5α and citrate synthase activity in PGC-1α−/− mice (Fig. 4B–D). These results indicate that PGC-1α plays an essential role in adiponectin-enhanced mitochondrial biogenesis. These results further suggest that MKP1 mediates adiponectin-enhanced mitochondrial biogenesis through PGC-1α.
Adiponectin is an adipocyte-derived hormone that exhibits multiple favorable functions in glucose and lipid metabolism. In addition to its insulin-sensitizing effect, recent studies have demonstrated that adiponectin may regulate energy metabolism through pathways independent of insulin signaling. Previous studies and the current study have demonstrated that adiponectin enhances mitochondrial biogenesis. Through the use of a PGC-1α knockout mouse model, our study further confirms the key role of PGC-1α in adiponectin-enhanced mitochondrial biogenesis. Most importantly, our study identifies MKP1 as a new downstream molecule of adiponectin signaling. This study provides clear evidence that adiponectin reduces MKP1 protein levels leading to increased p38 MAPK and PGC-1α activation in skeletal muscle.
It has been well established that p38 MAPK plays a critical role in skeletal muscle biogenesis and energy metabolism (25–27,38). p38 MAPK is among the first group of signaling proteins that have been identified as downstream of the adiponectin signaling pathway (15,18,19). However, it was unknown for a long period how adiponectin induces p38 MAPK and its downstream activation. A recent study reported that the adaptor protein APPL1 (adaptor protein, phosphotyrosine interaction, PH domain and leucine zipper containing 1) selectively transduces signals from the adiponectin receptor to the TGF-β-activated kinase 1 (TAK1)-MKK3-p38 MAPK pathway (19). Consistent with this report, our study found that MKK3 and MKK6 gene deletion significantly attenuated adiponectin-induced p38 MAPK phosphorylation. These studies indicate that the canonical MKK/MAPK pathway is required in adiponectin-induced p38 MAPK activation. Although other posttranslational modifications may exist, at this point, phosphorylation of the TXY motif has been identified as the on-and-off switch for p38 MAPK and other MAPKs. MKK3 and MKK6 phosphorylate and activate p38 MAPK. In contrast, MKPs, including MKP1, specifically dephosphorylate the TXY motif and deactivate p38 MAPK (29). The counteraction of MKKs and MKPs determines the phosphorylation level and kinase activity of p38 MAPK. Therefore, MKPs are also important in controlling p38 MAPK activation. Using both in vivo and in vitro systems, our study demonstrates that adiponectin reduces MKP1 protein levels in skeletal muscle. Furthermore, MKP1 overexpression attenuated adiponectin-enhanced PGC-1α gene expression and mitochondrial biogenesis in C2C12 myotubes. Decreased MKP1 protein leads to enhanced p38 MAPK activation and then to higher PGC-1α expression and mitochondrial biogenesis in skeletal muscle.
PGC-1α is a transcriptional coactivator and plays an important role in mitochondrial biogenesis. Our study suggests that adiponectin enhances mitochondrial biogenesis through the MKP1/p38/PGC-1α pathway. p38 MAPK activates PGC-1α by increasing not only its gene expression but also phosphorylation (24–27). This raises the question of whether adiponectin increases PGC-1α activity by increasing p38-mediated phosphorylation of PGC-1α. Owing to the lack of antibody against phosphorylated PGC-1α at Thr262, Ser265, and Thr298, which are phosphorylated by p38, our study does not provide direct evidence indicating whether adiponectin increases PGC-1α phosphorylation at these three sites. However, the functional assay showed that expressing triple mutant PGC-1α 3A abolished p38 MAPK–enhanced promoter activation but not adiponectin-increased luciferase activities (Supplementary Fig. 5A). These results suggest that, in addition to p38-mediated PGC-1α phosphorylation, adiponectin enhances PGC-1α activity through other pathway(s).
It has been reported that phosphorylation of Ser570 of PGC-1α inhibits its transactivity (39). Using a phospho-specific antibody, we found that phosphorylation levels of PGC-1α Ser570 were remarkably decreased in adiponectin-treated C2C12 myotubes and skeletal muscle of adiponectin-overexpressed mice (Supplementary Fig. 5B and C). Apparently, AKT is not responsible for adiponectin-decreased PGC-1α phosphorylation at Ser570 because increased AKT phosphorylation was observed. Although more studies are required to identify the underlying mechanism, our study suggests that adiponectin increases PGC-1α activity by stimulating both gene expression and posttranslational modification.
MKP1 is a key negative regulator of the MAPK pathway, which plays a critical role in multiple physiological processes in skeletal muscle including mitochondrial biogenesis and energy metabolism (29,30). Low-level mitochondrial content and activity in skeletal muscle have been reported in both type 2 diabetic patients and obese subjects (3–5). Through use of mouse model and biochemical approaches, a study from Bennett’s group convincingly demonstrated that high-fat diet induces MKP1 gene expression in skeletal muscle and results in inhibition of the p38 MAPK/PGC-1α pathway and loss of oxidative myofibers (30). The same group has also reported that MKP1 gene deletion increases skeletal muscle mitochondrial respiration (40). Hypoadiponectinemia occurs in obese human subjects and diet-induced obese mouse models (7). Our study indicates that adiponectin inhibits MKP1 gene expression in skeletal muscle. Therefore, we postulate that hyoadiponectinemia may contribute to high-fat diet–induced MKP1 gene expression and its adverse effects on oxidative metabolism in skeletal muscle. However, caution should be taken in applying this finding in mice to humans. More animal and human studies are required to support this possibility.
The underlying mechanism through which adiponectin suppresses MKP1 gene expression is not clear. Previous studies have suggested that activation of p38 MAPK and Jun NH2-terminal kinase induces MKP1 gene expression (29), which may serve as a feedback inhibitor to prevent prolonged MAPK activation. Clearly, this is not the case for adiponectin-induced downregulation of MKP1 gene expression in myotubes. Similar to other reports, our study showed that adiponectin stimulates p38 MAPK phosphorylation. If p38 MAPK increases MKP1 gene expression, we should expect an increase of MKP1 expression in adiponectin-treated myotubes. However, the opposite effect was observed in our study. On the other hand, the inverse changes of p38 MAPK activity and MKP1 expression in adiponectin-treated myotubes suggest that adiponectin-induced p38 MAPK activation is most likely mediated by decreased MKP1 gene expression.
In summary, our results demonstrate that activation of the p38 MAPK/PGC-1α pathway is essential in adiponectin-induced mitochondrial biogenesis and oxidative metabolism in skeletal muscle. Inhibition of MKP1 gene expression plays an important role in adiponectin-stimulated p38 MAPK activation.
This work was supported by National Institutes of Health grants DK-080418 (to J.Sh.) and DK077643 (to J.Sh.), and American Diabetes Association Grant 7-07-CD23 (to J.Sh.).
No potential conflicts of interest relevant to this article were reported.
L.Q., B.K., H.s.Y., and B.L. contributed research data. J.Sc. contributed to viral vectors and discussion. J.Sh. conceived the study and wrote the manuscript. J.Sh. 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.
The authors thank Dr. Philipp Scherer (University of Texas Southwestern Medical Center), Dr. Daniel P. Kelly (Washington University in St. Louis), Dr. Roger Davis (University of Massachusetts Medical School), and Dr. Pere Puigserver (Dana-Farber Cancer Institute, Boston, MA) for sharing gene knockout mice and research reagents.
This article contains Supplementary Data online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db11-1475/-/DC1.
- Received October 20, 2011.
- Accepted February 7, 2012.
- © 2012 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. See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.