Adiponectin-SOGA Dissociation in Type 1 Diabetes

Context:

Circulating adiponectin is elevated in human type 1 diabetes (T1D) and nonobese diabetic (NOD) mice without the expected indications of adiponectin action, consistent with tissue resistance.

Objective:

Adiponectin stimulates hepatocyte production of the suppressor of glucose from autophagy (SOGA), a protein that inhibits glucose production. We postulated that due to tissue resistance, the elevation of adiponectin in T1D should fail to increase the levels of a surrogate marker for liver SOGA, the circulating C-terminal SOGA fragment.

Main Outcome Measures:

Liver and plasma SOGA were measured in NOD mice (n = 12) by Western blot. Serum adiponectin and SOGA were measured in T1D and control (Ctrl) participants undergoing a three-stage insulin clamp for the Coronary Artery Calcification in T1D study (n = 20). Glucose turnover was measured using 6,6[²H₂]glucose (n = 12).

Results:

In diabetic NOD mice, the 13%–29% decrease of liver SOGA (P = .003) and the 30%–37% reduction of circulating SOGA (P < .001) were correlated (r = 0.826; P = .001). In T1D serum, adiponectin was 50%–60% higher than Ctrl, SOGA was 30%–50% lower and insulin was 3-fold higher (P < .05). At the low insulin infusion rate (4 mU/m²·min), the resulting glucose appearance correlated negatively with adiponectin in T1D (r = −0.985, P = .002) and SOGA in Ctrl and T1D (r = −0.837, P = .001). Glucose disappearance correlated with adiponectin in Ctrl (r = −0.757, P = .049) and SOGA in Ctrl and T1D (r = −0.709, P = .010). At 40 mU/m²·min, the lowered glucose appearance was similar in Ctrl and T1D. Glucose disappearance increased only in Ctrl (P = .005), requiring greater glucose infusion to maintain euglycemia (8.58 ± 1.29 vs 3.09 ± 0.87 mg/kg·min; P = .009).

Conclusions:

The correlation between liver and plasma SOGA in NOD mice supports the use of the latter as surrogate marker for liver concentration. Reduced SOGA in diabetic NOD mice suggests resistance to adiponectin. The dissociation between adiponectin and SOGA in T1D raises the possibility that restoring adiponectin signaling and SOGA might improve the metabolic response to insulin therapy.

Treatment with the goal of achieving metabolic normalization in human type 1 diabetes (T1D) by peripheral injection or infusion of insulin requires hyperinsulinemia (1–3). Despite peripheral hyperinsulinemia, liver metabolism may not be normalized. Additional factors are likely to be involved, including adiponectin. Clinical studies indicate that adiponectin plays an important role in glucoregulation (4, 5). Thus, the widely reported increase of circulating adiponectin in T1D (Supplemental Table 1) suggests tissue resistance to adiponectin, as previously proposed (6–8).

Human adiponectin is a 244-amino acid protein that circulates in the following: 1) two high-molecular-weight (HMW) multimers composed of either 18 or 12 adiponectin molecules, 2) a low-molecular-weight (LMW) hexamer composed of six adiponectin molecules, and 3) a trimer composed of three adiponectin molecules (7). HMW and LMW adiponectin, comprising approximately 90% of the total circulating adiponectin in humans, bind to cells through T-cadherin, a peripheral membrane protein that colocalizes with caveolin-rich lipid rafts (9, 10). HMW and LMW adiponectin are composed of adiponectin trimers. Free adiponectin trimer, comprising approximately 10% of the total circulating adiponectin in humans, binds to cells through adiponectin receptor 2 (AdipoR2) (11, 12). An acidic environment promotes trimer dissociation from HMW and LMW adiponectin and the proteolytic cleavage of the collagenous domain, on the N-terminal end of adiponectin, generating the globular trimer (13, 14). Free globular adiponectin trimers bind to adiponectin receptor 1 (AdipoR1) (11, 12).

Preclinical evidence in mice suggests that circulating free full-length and globular adiponectin trimers stimulate glucose uptake, whereas circulating HMW and LMW adiponectin suppress endogenous glucose production (15–19). This suppression results from decreases in rate-limiting enzymes and gluconeogenic substrates (7). AMP-activated protein kinase is a multimeric protein in the adiponectin signaling pathway that stimulates protein turnover, thereby increasing substrate availability (20). The cellular mediators that restrict substrate availability could include T-cadherin, AdipoR1, and AdipoR2 and the recently described suppressor of glucose from autophagy (SOGA).

SOGA is a protein produced in liver cells that inhibits autophagy, thereby decreasing the availability of gluconeogenic substrates (21). Antisense oligonucleotide knockdown of SOGA in isolated hepatocytes prevents HMW and LMW adiponectin from inhibiting glucose production (21). Isolated hepatocytes release a C-terminal 80-kDa SOGA fragment that can be further cleaved to produce the circulating C-terminal 28-kDa fragment in humans and 25-kDa fragment in mice (21). These circulating SOGA fragments could provide a surrogate marker of adiponectin signaling. In support of this theory, serum adiponectin and 28-kDa SOGA show a linear correlation in healthy women (21). Furthermore, genetic modification, calorie restriction, and pioglitazone-mediated increases of circulating adiponectin in mice are associated with increases in circulating 25-kDa SOGA (21).

In this study, we set out to determine whether the widely reported elevation of adiponectin in T1D is accompanied by a decrease in SOGA. This would be compatible with tissue resistance to adiponectin in T1D as previously proposed. An increase in circulating adiponectin and a decrease in SOGA were observed in diabetic nonobese diabetic (NOD) mice that develop spontaneous autoimmune T1D (21, 22). This and other indications of adiponectin resistance in T1D led us to propose that the widely reported elevation of adiponectin in T1D may fail to produce the expected increase in SOGA. Our results have been presented in part in abstract form (23).

Materials and Methods

Human participants

Serum samples from T1D patients and nondiabetic control volunteers (Table 1) were available from the Coronary Artery Calcification in T1D study at the Barbara Davis Center for Childhood Diabetes (University of Colorado, Anschutz Medical Campus, Denver, Colorado) (24). Enrollment criteria for T1D patients included diagnosis before the age of 30 years, or a clinical course consistent with T1D, the initiation of insulin therapy within a year of diagnosis, 4 or more years of duration of diabetes, the absence of macroalbuminuria and diagnosed coronary artery disease. They all received recombinant human insulin. Age- and sex-matched nondiabetic control participants were also without diagnosed coronary artery disease. Premenopausal women were studied between menstrual cycle days 2 and 10. Informed consent was obtained as required by the Colorado Combined Institutional Review Board.

Hyperinsulinemic clamp

Glucose kinetics were measured using a primed-continuous infusion of 6,6[²H₂]glucose starting at t = 0 minutes and calculated as previously (6). The 2-hour baseline period (t = 0–120 min) was followed by a primed and then continuous infusion of insulin in three steps: at 4 mU/m² · min from 120 to 210 minutes, 8 mU/m² · min from 210 to 300 minutes, and 40 mU/m² · min from 300 to 390 minutes (3, 6, 25). Glycemia was maintained constant at approximately 90 mg/dL by a readout adjusted infusion of 20% dextrose. The isotope was also added to the dextrose infused. The concentration of insulin required for 50% inhibition of glucose appearance (Ra) (IC₅₀) was calculated as previously described (3).

Human serum collection

Volunteers were maintained on a standardized diet (50% carbohydrate; 20% protein; 30% fat) and asked to refrain from vigorous physical activity for 3 days before the clamp. All subjects with T1D underwent continuous glucose monitoring using a MiniMed Gold System (Medtronic) for three days. A standard meal was provided at the study unit on the night before the clamp. T1D subjects received their usual insulin dose with dinner. Beginning at 8:00 pm on the night before the clamp, insulin was administered by iv infusion using a sliding scale adjustment to achieve a glucose target range of 80–110 mg/dL. Sera from arterialized venous blood samples collected before and during the clamp were stored at −80°C and shipped to the University of North Carolina at Chapel Hill on dry ice for the measurement of adiponectin and SOGA. Insulin was measured using a human specific RIA (catalog number HI-14K; Millipore) at the study site (6). Sera from T1D subjects were not tested for the presence of endogenous insulin antibodies prior to analysis.

Measurement of adiponectin and SOGA in human serum

Sera were prepared for Western blot analysis by heating 1 μL in 10 μL of loading buffer (0.2 M Tris; 50% glycerol, pH 6.8; 0.1% sodium dodecyl sulfate; 0.1 M dithiothreitol) at 95°C for 5 minutes. Samples were loaded on NuPAGE Novex 4%–12% gradient bis-Tris acetate gels (Invitrogen). After electrophoresis, serum protein was transferred to Whatman Protran BA83 nitrocellulose (GE Healthcare). Nonspecific binding sites were blocked in a Tris-buffered saline-Tween 20 solution comprised of 50 mM Tris-HCl, pH 7.5; 150 mM NaCl; 0.05% Tween 20; and 1% fat-free powdered milk. Blots were exposed to rabbit polyclonal antisera recognizing the human-specific sequence on the N-terminal end of adiponectin (DQETTTQGPGV). Antibody binding was detected using horseradish peroxidase conjugated goat antirabbit IgGs. Densitometry was performed using a Versa-Doc 400 MP digital image capturing apparatus (Bio-Rad Laboratories). SOGA was measured separately on the same blots using rabbit polyclonal antisera recognizing the human-specific sequence on the C-terminal end of human SOGA (STQSLTSC*FARSSRSAIRHSPSKC, where C* is acetamidomethyl cysteine). Rabbit polyclonal antisera used to measure circulating SOGA in human serum also cross-reacted with a peptide of 50 kDa. Human serum SOGA values in Figures 1 and 2 reflect the measurement of the 28-kDa peptide. Signal variability from identical samples was below 10% within the same gels and below 8% between gels. The production of rabbit polyclonal antisera for human adiponectin and SOGA was previously described (21, 26).

Mouse liver and plasma collection

Female NOD mice (12–18 wk of age) were housed in a barrier facility at the University of North Carolina at Chapel Hill (27). The diabetes status of the colony was monitored daily by urine test. The control group consisted of age-matched nondiabetic female NOD mice. Liver and plasma samples for the measurement of SOGA were collected within 24–48 hours of diabetes onset. Plasma was collected from the tail tip of unrestrained mice and stored at −20°C. The liver was dissected immediately after euthanasia by cervical dislocation, snap frozen in liquid nitrogen, and stored at −80°C. All procedures were approved by the Institutional Animal Care and Use Committee.

Western blot measurement of SOGA in liver homogenates and plasma

Frozen liver samples were homogenized in radioimmunoprecipitation assay buffer (50 mM Tris-HCl, pH 7.5; 150 mM NaCl; 1% Nonidet-P40; 0.1% sodium dodecyl sulfate; 1% phenylmethylsulfonyl fluoride; 1 mM Na₃VO₄; 0.1% aprotinin; and 0.1% phosphatase inhibitor cocktail-2; Sigma-Aldrich). Whole-liver homogenates were centrifuged at 12 000 × g for 15 minutes at 4°C. The bicinchoninic acid method was used to determine the protein concentration of the supernatant. Homogenates were prepared for Western blot analysis by heating 15 μg of protein in 10 μL of loading buffer at 95°C for 5 minutes, as described above. Plasma samples were prepared by heating 1 μL of plasma in 10 μL of sample buffer. Blots were exposed to rabbit polyclonal antisera recognizing the mouse-specific sequence on the C-terminal end of murine SOGA (CSAQSLASCFIRPSRN), prepared as previously described (21).

Statistical analyses

Analyses were performed using SPSS 17.0 software for Windows (SPSS). The Shapiro-Wilk test was used to assess normality in data distribution. An independent-samples t test was used to determine differences between two groups showing a normal distribution. Mann-Whitney test was used to calculate differences between groups that did not show a normal distribution. A two-way ANOVA was used to determine differences between groups showing a normal distribution when the dependent variable could be affected by more than one independent variable. A repeated-measures ANOVA was used to determine differences over time. A Wilcoxon signed rank test was used to determine differences between two related groups that lacked normal distribution. The bivariate Pearson correlation test was used to determine a linear relationship between two variables. Spearman's rank-order correlation test was used to identify a linear relationship between two variables lacking a normal distribution. All values of P < .05 were considered significant.

Results

The T1D and control (Ctrl) groups did not differ by sex, age, duration of T1D, waist circumference, weight, or body mass index (BMI) (Table 1). Age at diagnosis ranged from 3 to 31 years, with four diagnosed in adulthood (aged 18+ y), three during teen years (14–17 y), and three during childhood (3–12 y). There were six lean, seven overweight, and seven obese participants. In these categories, there were two, five, and three in the Ctrl group and four, two, and four, respectively, in the T1D group. The distribution of lean, overweight, and obese subjects did not differ between Ctrl and T1D by χ² analysis (P = .351). Glycated hemoglobin was increased and serum triglycerides were decreased in T1D (P < .05; two way ANOVA).

Figure 1, A and B, show the baseline values, t = 0, for total circulating adiponectin (HMW, LMW, free full length trimer) and circulating SOGA in men and women with and without T1D. Serum adiponectin was 50%–60% higher in T1D than Ctrl (P < .001), and SOGA was 30%–50% lower (P < .001). Women had higher levels of adiponectin (P = .002) and SOGA (P < .001). Adiponectin and SOGA did not correlate when Ctrl and T1D subjects were analyzed as a single cohort. As shown in Figure 1D, serum adiponectin and SOGA correlated in male and female T1D (r = 0.769, P = .009). Neither adiponectin nor SOGA correlated with BMI when Ctrl and T1D were analyzed as a single cohort (n = 20). BMI and adiponectin correlated in Ctrl (n = 10; r = 0.875, P = .001) but not in T1D. There were no correlations between adiponectin or SOGA and waist circumference.

Figure 2A shows insulin infusion rates for the three-stage hyperinsulinemic clamp. Figure 2B shows that fasting insulin was higher in T1D at t = 0 and 120 minutes (P < .010; Mann-Whitney). Insulin values did not differ at maximal hyperinsulinemia, clamp stage 3. Figure 2 C and D, show adiponectin and SOGA in serum at t = 0 and at the end of each clamp stage (t = 210, 300, and 390 min). The 50%–60% greater adiponectin in T1D was maintained throughout the clamp (P = .002).

After an overnight fast, at t = 0, when T1D and Ctrl were analyzed as a single group, adiponectin correlated with serum insulin levels (r = 0.522; P = .018). Serum SOGA correlated inversely with insulin (r = −0.470; P = .037). There was no correlation (Spearman) between clamp infusion rates and adiponectin or SOGA. When T1D and Ctrl were analyzed as a single group, there was no clamp effect on adiponectin (P = .257; repeated measures ANOVA). However, there was a clamp effect on adiponectin in Ctrl largely due to its reduction at t = 390 minutes (P < .05, Wilcoxon). SOGA remained 30%–50% lower in T1D throughout the clamp (P = .009) and showed a slight decline when T1D and Ctrl were combined (P = .033, repeated measures ANOVA) but not within each group.

Clamp glucose kinetic data were available for seven Ctrl (three males, four females) and five T1D (three males, two females) participants. Figure 3 shows that glucose and fasting Ra and glucose disappearance (Rd) were similar between these subsets of Ctrl and T1D. As shown in Figure 3A, there was an elevation of glucose in T1D at clamp stage 1 (P = .003; Mann-Whitney) because all participants received the same rate of insulin infusion. This resulted in the repeated-measures ANOVA, showing a clamp effect on glucose in both Ctrl and T1D (P = .001), but the two groups responded differently, as indicated by a significant clamp × group interaction (P = .003). At clamp stage 1, serum SOGA in Ctrl and T1D correlated with the resulting serum glucose (r = −0.739, P = .006).

As shown in Figure 3B, a clamp effect on glucose Ra was seen in both Ctrl and T1D (P < .001, repeated measures ANOVA), with a higher Ra in T1D (P = .049), due to the elevation of Ra at clamp stage 1 (P = .021, t test). During clamp stage 3 hyperinsulinemia, Ra was suppressed in both groups to values, no longer significantly different (P > .05, t test). There was no significant clamp × group interaction. At clamp stage 1, adiponectin correlated with suppressed Ra in T1D (r = −0.985, P = .002), and SOGA in Ctrl and T1D correlated with Ra (r = −0.837, P = .001).

As shown in Figure 3C, clamp elevated glucose Rd in Ctrl (P < .001; repeated measures ANOVA) but not in T1D, as indicated by a significant clamp × group interaction (P = .004). At clamp stage 1, the lowest insulin infusion rate, serum adiponectin correlated with reduced Rd in Ctrl (r = −0.757, P = .049). Serum SOGA in Ctrl and T1D correlated with reduced Rd (r = −0.709, P = .010) and IC₅₀ (r = −0.640, P = .034). At clamp stage 3, Ctrl required greater glucose infusion (8.58 ± 1.29 vs 3.09 ± 0.87 mg/kg·min; P = .009). SOGA showed a tendency to correlate with Rd at hyperinsulinemic clamp stage 3 (r = 0.526, P = .079), at which stage Rd was elevated in Ctrl but not in T1D.

Figure 4, A and B, show, in comparison with nondiabetic NOD mice, diabetic NOD mice (plasma glucose > 500 mg/dL) had 13%–29% less 80 kDa SOGA in whole-liver homogenates (P = .003, t test) and 30%–37% less circulating 25 kDa SOGA (P < .001, t test). Figure 4D shows a significant linear correlation between liver and plasma SOGA (r = 0.826; P = .001).

Discussion

This study provides novel evidence of tissue resistance to adiponectin in T1D. Our results suggest that SOGA is low in T1D because adiponectin fails to produce a normal level of signaling, which may play a role in the pathophysiology of hyperglycemia. It suggests that the C-terminal circulating SOGA fragment may be a marker of hepatic adiponectin action, linked to insulin sensitivity.

The linear correlation between SOGA levels in the liver and the circulation of NOD mice suggests the reduction of SOGA in T1D is not due to a defect in the secretory pathway that results in the accumulation of SOGA in the liver. This does not exclude the possibility that rates of SOGA exit from the circulation are elevated in T1D or that contributions from other tissues exist. In this regard, SOGA was previously identified by mass spectrometry in human embryonic kidney cells (28).

Tissue resistance to adiponectin may contribute to the SOGA deficit in T1D patients and NOD mice. The hypothesis that liver cells secrete a fragment of SOGA is consistent with reduced SOGA in the liver of NOD mice. This hypothesis is also consistent with the elevation of liver SOGA mRNA and circulating SOGA in pioglitazone treated ob/ob mice and the correlation between SOGA in hepatocytes and hepatocyte conditioned media (21).

Both the present and previous studies show that T1D patients require exogenous insulin to reach higher than normal levels to achieve euglycemia (1–3). At serum glucose levels attained by overnight insulin infusion, fasting levels of insulin were 3-fold higher in T1D patients. The increased insulin requirement for Ra suppression in T1D may be due to not only impaired insulin signaling, but also that of adiponectin, in the liver. When Ctrl and T1D participants were grouped, higher fasting serum SOGA correlated with lower insulin. Marked resistance to insulin suppression of Ra was previously reported at all clamp stages (3). In the present study, circulating SOGA correlated with suppressed glucose Ra at low insulin levels. SOGA may control Ra by reducing autophagy, a highly regulated process that leads to the hydrolysis of protein, glycogen, and lipids in acidic vacuoles (29). Small interfering RNA knockdown of SOGA in hepatocytes increases the number of acidic vacuoles, proteolytic activity, and glucose release into the media (21). With comparable hyperinsulinemia, at clamp stage 3, T1D required half the glucose infusion rate of Ctrl, the gold standard measure of insulin resistance.

Adiponectin regulation of SOGA does not appear to be completely absent in T1D. Gender differences in the circulating levels of adiponectin in T1D were associated with corresponding changes in the circulating levels of SOGA. In previous studies, a single injection of recombinant adiponectin had a delayed but long-lasting, glucose-lowering effect in diabetic NOD mice (17). The essential role of adiponectin in whole-body and β-cell metabolism and regeneration is particularly clear in the near absence of insulin (30, 31). At the lowest insulin infusion rate, adiponectin in Ctrl and SOGA in Ctrl and T1D correlated negatively with Rd. Hence, metabolic effect studies at different circulating concentrations are needed, in larger study cohorts.

Clamp hyperinsulinemia had a small but significant suppressive effect on serum SOGA. Previous studies suggest that insulin signaling can also increase SOGA. Circulating levels of SOGA increase 2 hours after feeding in healthy adults (21). Twenty-four hours after a single injection of insulin, diabetic NOD mice showed a partial increase in circulating SOGA (21). LY294002 inhibition of phosphatidylinositol 3-kinase, a proximal signaling intermediate in the insulin signaling pathway, lowers SOGA in hepatocytes (21). Rapamycin inhibition of mammalian target of rapamycin, an Akt/protein kinase B target in the insulin signaling pathway, also lowers SOGA in mice and in hepatocytes (AbdelBaky, O., and Combs, T.P., unpublished observations). Thus, the dose effects of insulin on SOGA also require further study.

A deregulated immune system may contribute to adiponectin resistance in T1D. In this scenario, antibody binding and proinflammatory cytokines associated with β-cell destruction could affect the function of adiponectin receptors and adiponectin signaling intermediates in the liver (32, 33). The first AdipoR1/R2 agonist, AdipoRon, restores glucose homeostasis in murine models of type 2 diabetes (T2D) (34). The therapeutic potential of AdipoR1/R2 activation in T1D is presently unknown.

The elevation of serum adiponectin in T1D patients has been extensively documented since it was first reported in 2002 (Supplemental Table 1). Eleven studies involving 2778 subjects, from 3 to 61 years of age, show that adiponectin is 28%–98% higher in T1D patients than Ctrl subjects. This contrasts markedly with human T2D, in which minor decreases in circulating adiponectin limit its effects (35). T1D and T2D represent clinical conditions that can be ameliorated by the elevation of adiponectin with thiazolidinedione medications (26, 36). Our study subjects were between 29 and 58 years of age. It is currently not known whether the adiponectin-SOGA dissociation occurs at a younger age.

The specific signals triggering the elevation of adiponectin in T1D are currently unknown. It is noteworthy that adiponectin increases with the duration of T1D. The average duration of T1D in the present study was nearly 2 decades. Adiponectin is higher in T1D patients without circulating C-peptide (37). The correlation between adiponectin and fasting insulin we found (with n = 20) is consistent with T1D adiponectin resistance, contributing to the need for higher circulating insulin. Elevation of adiponectin is also observed in other autoimmune and proinflammatory disease states including rheumatoid arthritis, lupus, inflammatory bowel disease, renal insufficiency, and cystic fibrosis (38). The elevation of adiponectin in T1D is probably related to tissue resistance rather than posttranslational modifications on adiponectin or the binding of other serum proteins to adiponectin (39, 40). It is noteworthy that these elevations of adiponectin in T1D have been observed whether serum was analyzed by a Western blot, RIA, or ELISA (Supplemental Table 1).

We have demonstrated a dissociation between adiponectin and SOGA in T1D. The concurrent elevation of adiponectin and reduction of SOGA suggests a deficiency in adiponectin signaling. Based on mouse and cell data, the reduction of SOGA in T1D patients can be linked to hyperglycemia, although the relative contribution of SOGA insufficiency remains to be determined. Presently the infusion and injection of insulin are insufficient to restore the metabolic effects of a fully functional pancreas. In this regard, any further correction of metabolic abnormalities via other steps that modulate hepatic glucose production may bring this ultimate goal closer. Agents that could demonstrably target the adiponectin signaling deficiency in T1D could potentially enhance insulin-mediated suppression of excessive hepatic glucose production.

Acknowledgments

This study was performed on the advice of the late Dr George Eisenbarth.

Each author made substantial contributions. LJ and OA did the Western blot analysis. Essential reagents were provided by T.P.C., J.K.S.-B., D.M.M., B.C.B., R.T., and P.E.S., E.B.M., T.P.C., and M.L. performed the statistical analysis and wrote the paper. Each author critically revised and approved the final version of the manuscript.

This work was supported by Grants DK075573 (to T.P.C.), HL061753 (to J.K.S.-B.), DK075360 (to D.M.M.), DK100256 (to R.T.), and DK055758 (to P.E.S.) from the US National Institutes of Health and Canadian Institutes of Health Research Grant MOP-62889 (to E.B.M.).

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• AdipoR1

  adiponectin receptor 1

 
• AdipoR2

  adiponectin receptor 2

 
• BMI

  body mass index

 
• Ctrl

  control

 
• HMW

  high-molecular-weight

 
• LMW

  low-molecular-weight

 
• NOD

  nonobese diabetic

 
• Ra

  glucose appearance

 
• Rd

  glucose disappearance

 
• SOGA

  suppressor of glucose from autophagy

 
• T1D

  type 1 diabetes

 
• T2D

  type 2 diabetes.

References