Circulating long noncoding RNA GAS5 levels are correlated to prevalence of type 2 diabetes mellitus

Abstract

Background: Diabetes mellitus (DM), a metabolic disease, is characterized by impaired fasting glucose levels. Type 2 DM is adult onset diabetes. Long non-coding RNAs (lncRNAs) regulate gene expression and multiple studies have linked lncRNAs to human diseases.

Methods: Serum samples obtained from 96 participating veterans at JAH VA were deposited in the Research Biospecimen Repository. We used a two-stage strategy to identify an lncRNA whose levels correlated with T2DM. Initially we screened five serum samples from diabetic and non-diabetic individuals using lncRNA arrays. Next, GAS5 lncRNA levels were analyzed in 96 serum samples using quantitative PCR. Receiver operating characteristic (ROC) analysis was performed to determine the optimal cutoff GAS5 for diagnosis of DM.

Results: Our results demonstrate that decreased GAS5 levels in serum were associated with diabetes in a cohort of US military veterans. The ROC analysis revealed an optimal cutoff GAS5 value of less than or equal to 10. qPCR results indicated that individuals with absolute GAS5 < 10 ng/μl have almost twelve times higher odds of having diabetes (Exact Odds Ratio [OR] = 11.79 (95% CI: 3.97, 37.26), p < 0.001). Analysis indicated area under curve (AUC) of ROC of 0.81 with 85.1% sensitivity and 67.3% specificity in distinguishing non-diabetic from diabetic subjects. The positive predictive value is 71.4%.

Conclusion: lncRNA GAS5 levels are correlated to prevalence of T2DM.

General significance: Assessment of GAS5 in serum along with other parameters offers greater accuracy in identifying individuals at-risk for diabetes.

Keywords: AUC, area under curve; BMI, body mass index; CI, confidence interval; DM, diabetes mellitus; Diabetes; GAS5; GAS5, growth-arrest specific transcript 5; NMD, nonsense mediated decay; ROC, receiver operating characteristics; Serum; Veterans; lncRNA; lncRNA, long noncoding RNA.

Fig. 1

LncRNA profiler array was used to determine expression of lncRNAs in serum from diabetic (n = 5) and non-diabetic (n = 5) samples. RNA was extracted from serum samples, reverse transcribed and cDNA used in lncRNA Profiler (SBI). Average Ct was calculated for diabetic and non-diabetic samples, values normalized to hoU6 snRNA. Log graph shows fold change calculated as 2^ − (Ct_(avDM) − Ct_(avNDM) × ΔCt_{(geo mean)}). Analysis was performed by software from SBI.

Fig. 2

(a) Serum from non-diabetic (n = 49) and diabetic (n = 47) were analyzed for expression of GAS5 using qPCR. A standard curve was generated using 100–0.4 ng/μl GAS5 plasmid. Samples were normalized to U6snRNA. Graph shows absolute quantification of GAS5 (ng/μl) vs HbA1c. (b) Plot of mean (95% confidence interval) GAS5 and HbA1c levels by diabetic status. The confidence limits are mean ± 2 ∗ standard error.

Fig. 3

Receiver operating curve (ROC) was performed on GAS5 levels in serum from diabetic patients (n = 47) and non-diabetic patients (n = 49) to determine the optimal cutoff values. Area under curve (AUC) for GAS5 is 0.81 (95% CI: 0.72, 0.90); Sensitivity of 85.1% (95% CI: 72.3%, 92.6%), specificity of 67.3% (95% CI: 53.4%, 78.8%), positive predictive value PPV = 71.4% (95% CI:57.8%, 82.7%).

Fig. 4

The waterfall plot shows the classification accuracy of the optimal cutoff point for all 96 patients (diabetic (n = 47) + non-diabetic (n = 49) = 96), as well as the overall odds ratio for GAS5 as a predictor of diabetes (OR = 0.75 (95%CI: 0.64, 0.87), p < 0.001).