Send Orders for Reprints to reprints@benthamscience.net Current Diabetes Reviews, 2022, 18, e010321189862 RESEARCH ARTICLE "<@>A/ABBC "<C>@/?D<> Growth Differentiation Factor-15 as a Biomarker of Obese Pre-diabetes and Type 2 Diabetes Mellitus in Indian Subjects: A Case-control Study BENTHAM SCIENCE 1,* 1,* 1 1 Dipayan Roy , Purvi Purohit , Anupama Modi , Manoj Khokhar , Ravindra Kumar Gayaprasad 2 3 1 1 Shukla , Ramkaran Chaudhary , Shrimanjunath Sankanagoudar and Praveen Sharma 1 Department of Biochemistry, AIIMS, Jodhpur, Rajasthan, India; 2Department of Endocrinology and Metabolism, AIIMS, Jodhpur, Rajasthan, India; 3Department of General Surgery, AIIMS, Jodhpur, Rajasthan, India Abstract: Background: Type 2 diabetes mellitus (T2DM) is an ever-growing epidemic in India and poses significant morbidity, mortality, and socioeconomic burden. Introduction: Growth differentiation factor-15 (GDF15) is a stress-responsive cytokine, increased in T2DM patients compared to control subjects without the disease. We aimed to assess whether serum GDF15 and adipose tissue GDF15 expression can differentiate between obese pre-diabetes and T2DM and control populations. ARTICLE HISTORY Received: July 10, 2020 Revised: October 25, 2020 Accepted: November 02, 2020 Current Diabetes Reviews DOI: 10.2174/1573399817666210104101739 Methodology: We recruited 156 individuals including 73 type 2 diabetes, 30 pre-diabetes, and 53 healthy controls. Clinical history, anthropometric measurements and biochemical profiling were taken. Insulin resistance indices were calculated following HOMA models. Serum GDF15 was measured by sandwich ELISA. Visceral adipose tissue (VAT) expression of GDF15 was observed in 17 T2DM patients and 29 controls using SYBR Green chemistry in RT-PCR using GAPDH as the housekeeping gene. The data were analyzed on R programming platform using RStudio. Results: Serum GDF15 was significantly higher (p<0.001) in T2DM subjects (median 1445.47 pg/mL) compared to pre-diabetes (627.85 pg/mL) and healthy controls (609.01 pg/mL). Using the ΔΔCt method, the VAT GDF15 expression was 1.54 fold and 1.57 fold upregulated in T2DM (n=17) compared to control subjects (n=29), and obese (n=12) compared to non-obese (n=34)subjects, respectively. The optimal cut-off point following Youden’s index method was found to be 868.09 pg/mL. ROC curve analysis revealed that serum GDF15 had a sensitivity, specificity, and area under the curve (AUC) of 90.41%, 79.52%, and 0.892 respectively. GDF15 levels were significantly associated with age, BMI, HbA1c, fasting blood sugar, and insulin resistance indices. Conclusion: Hence, serum GDF15 is a biomarker for T2DM patients in our study population from Western India. However, larger prospective cohorts are necessary to validate this claim. Keywords: GDF15, MIC-1, biomarker, visceral adipose tissue, cutpointr, type 2 diabetes, obesity. 1. INTRODUCTION Growth Differentiation Factor (GDF15) is an anti-inflammatory cytokine belonging to the transforming growth factor-β (TGF-β) superfamily [1]. Initially thought to be an inhibitor of macrophage activation, it is now known to control various cellular processes like cell cycle, cell proliferation and differentiation [2]. GDF15 is secreted mostly from liver, kidney, lung and placenta, and increased serum levels are observed in response to stress and inflammation [3]. It has been proposed as an adipokine and shown to reduce body *Address correspondence to these authors at the Department of Biochemistry. AIIMS, Jodhpur, Rajasthan-342005, India; Tel: +91 9928388223; +91 9358035259; E-mails: dr.purvipurohit@gmail.com, d.roy12092009@gmail.com 1875-6417/22 $65.00+.00 weight and fat mass, increase thermogenesis and lipolysis and improve insulin sensitivity and glucose tolerance in animal models [2, 4, 5]. Hence, GDF15 may act as a metabolic regulator of insulin signaling, making it a useful marker for insulin resistance (IR) and type 2 diabetes mellitus (T2DM) [6]. In humans, increased circulating GDF15 levels have been positively associated with age, body fat, serum triglyceride, fasting glucose, fasting insulin, homeostatic model assessment of insulin resistance (HOMA-IR) and HbA1c in diabetic and obese populations [7, 8]. In an obese Swedish population, GDF15 was an independent predictor of future IR and abnormal glucose control [9]. Circulating GDF15 levels were positively associated with risk of incident diabetes in the Malmö Diet and Cancer-Cardiovascular Cohort and the Whitehall II study [10, 11]. It has also been claimed to be a novel biomarker for detecting impaired fasting glucose © 2022 Bentham Science Publishers 1 Current Diabetes Reviews, 2022, Vol. 18, No. 1 e010321189862 in non-diabetic patients [12]. Type 2 diabetes is the most important driver of the diabetes epidemic in India [13]. The development of T2DM, as well as IR, is attributed to a lowgrade inflammatory response in adipose tissue [14]. Current pieces of evidence on several human studies show significant differences of GDF15 levels between T2DM and control subjects; however, the sera levels considerably varied across studies. Furthermore, there is yet no cut-off point above which GDF15 levels can be used as a marker in T2DM. Also, to the best of our knowledge, there is a dearth of data on serum levels of GDF15 in obese pre-diabetes subjects as compared to obese T2DM subjects and controls in Indian population. In this study, we aimed at assessing the serum GDF15 levels and its optimal cut-off values for discriminating between obese diabetes and control subjects and demonstrated its association with anthropometric parameters and insulin indices in T2DM of a Western Indian population. We have also studied the expression of GDF15 in the visceral adipose tissue (VAT) of a selected number of subjects from this population to ascertain VAT as a source of GDF15. 2. MATERIALS AND METHODS 2.1. Ethics This study conforms to the guidelines of the Institutional Ethics Committee (IEC) of AIIMS, Jodhpur. Written informed consent was taken from each participant before enrolling in the study. 2.2. Selection and Description of Participants The sample size was primarily calculated based on the mean difference of serum GDF15 in T2DM population compared to normal glucose tolerant individuals, according to the study of Hong et al. [12] using OpenEpi software (https://www.openepi.com/SampleSize/SSMean.htm). A two-sided confidence interval of 95% and a power of 80% were taken. The calculated sample size came to be 26 for each group. Considering a 10% fallout rate, the total sample size for each group came to be 29. Participants were recruited from July 2018 to September 2019 from the Department of Endocrinology and Metabolism, the Department of General Surgery, and the Department of Biochemistry at AIIMS, Jodhpur. The patients who would undergo visceral or abdominal surgery were selected for VAT collection. A total of 156 participants, aged between 18-60 years, were included in this study, of which 73 were T2DM patients, diagnosed according to ADA criteria [15]. Patients with an existing or previous history of diabetic complications, chronic liver diseases, thyroid disorders, cancer, acute inflammatory conditions, hypolipidemic medications, or pregnancy were excluded.30 pre-diabetic subjects were included according to either HbA1c cut-off levels (5.7-6.4%) or fasting plasma glucose levels (101-125 mg/dL).53 age and sex-matched metabolically healthy individuals were taken as controls. For the VAT samples, only those patients (n=17) who fulfilled the selection criteria and consented for donating 2 Roy et al. their VAT participated in the study. They underwent minimally invasive laparoscopic surgery for chronic cholelithiasis and hernia, and additionally, did not have any acute inflammatory condition, malignancy, or associated diabetic complications, and were not over 60 years old. 29 control patients who did not have diabetes or pre-diabetes, were recruited for VAT collection following the same inclusion and exclusion criteria. 2.3. Anthropometric Measurements Weight was taken on a digital weighing machine to the nearest 0.1 kg, and height using a stadiometer. Measurements were taken while patients were barefoot and lightly clothed without ornaments, looking straight ahead with a line of sight parallel with the floor. BMI was calculated by the formula, BMI = weight/height2 (in kg/m2). The cases were divided into normal BMI and high BMI (BMI <25.0 kg/m2 and BMI ≥25.0 kg/m2). Waist circumference was measured at the level parallel to the floor, the midpoint between the lower margin of the last palpable rib in the midaxillary line and the top of the iliac crest at the end of expiration [16]. Hip circumference was measured at the level parallel to the floor at the widest circumference of the buttocks. The waist-to-hip ratio (WHR) was calculated by dividing waist circumference by hip circumference. 2.4. Biochemical Parameters Venous blood samples were drawn after overnight fasting in 3 vacutainers, i.e., clot activator (for serum separation), EDTA-coated (for plasma separation) and NaF-coated (for glucose estimation). Serum was separated within half an hour, and routine parameters were run immediately, including fasting blood glucose (FBG), total cholesterol (TC), triglycerides (TG), HDL-cholesterol and LDL-cholesterol in fully-automated AU480 and AU680 Clinical Chemistry Analyzer (Beckman Coulter). HbA1c was estimated immediately from EDTA vial samples by latex agglutination inhibition assay. The concentrations of both HbA1c and total hemoglobin were determined and the HbA1c/total Hb ratio was expressed as percentage HbA1c. Fasting serum insulin was measured by chemiluminescence immunoassay (ADVIA Centaur XP). All analyses were performed according to the manufacturer’s instructions and after method verification. All serum samples were stored at -80°C for further processing. Serum GDF15 was measured in batches by GDF15 Human ELISA Kit (Invitrogen, Thermo Fisher Scientific, #EHGDF15) following the manufacturer’s protocol. HOMA was used to calculate insulin resistance (HOMA-IR), insulin sensitivity (HOMA-S), and β-cell function (HOMA%β) [17]. HOMA-S was calculated as reciprocal of HOMA-IR. For a model-derived estimation of % β-cell function (%B) and % sensitivity (%S), we used the HOMA2 calculator, downloaded from https://www.dtu.ox.ac.uk/homacalculator/download.php. Further, other IR indices were calculated, which include TC/HDL-cholesterol (TC/HDL) ratio [18], TG/HDL-cholesterol (TG/HDL) ratio [19], and triglyceride glucose (TyG) index [20]. GDF15 as a Biomarker in Obese, type 2 Diabetic Indians e010321189862 2.5. Tissue RNA Isolation, Reverse Transcription, and Real- time PCR We used TRIzol (HiMedia RNA-XPressTM Reagent), a commercially available RNA extraction reagent (http://himedialabs.com/TD/MB601.pdf) for extracting total RNA from the VAT samples, based on the previously described optimisation [21] of the single-step acid-phenol-chloroform extraction method [22]. The VAT was drawn from each patient during surgery by the operating surgeon and was immediately transferred to the laboratory in 1% phosphate buffer saline (PBS) in a sterile container. 500 mg of VAT was measured using an analytical balance and 250 μLTRIzol was added to it, and RNA was isolated.The total RNA was quantified using a microplate reader (BioTek Instruments, Inc.). Only those RNA samples with a 260/280 and 260/230 ratio of ≥1.8 were considered as suitable for Real time-polymerase chain reaction (RT-PCR). We used the miScript II RT Kit (QIAGEN) to yield complementary DNA (cDNA) from the samples using the 5x miScript HiFlex Buffer, following manufacturer’s protocol (https://www.qiagen.com/us/resources/download.aspx?id=0948c3fd-c643-4daaa777-11425991ba3e&lang= en). Each tube contained a reaction mixture of 10 μL, which were incubated at 37°C for 60 min and then at 95°C for 5 min. We then amplified the cDNA using miScript® SYBR® Green PCR Kit (QIAGEN) and RT2 qPCR Primer Assay (QIAGEN) for GDF15 (Catalog no: 330001 PPH01935C, RefSeq Accession no: NM_004864.2, Reference position: 443) on a BioRad CFX96 Real-Time system. A 10 μL volume reaction mixture was prepared for each sample, and the samples were run on a 96-well plate under the following conditions: initial activation step at 95°C for 15 minutes, followed by 40 cycles of 15 s at 94°C, 30 s at 55°C, and 30 s at 70°C. The gene expression was normalized using the housekeeping gene GAPDH (Catalog no: 330001 PPH00150F, RefSeq Accession no: NM_002046.5, Reference position: 842). 2.6. Statistics Data were analysed in R (version 3.5.3) using RStudio [23]. Variables were tested for normality by density plot, QQ plot, and Shapiro-Wilk significance test. Continuous variables were expressed as mean and standard deviations (SD) or median and interquartile range (IQR) according to their distributions. Categorical variables were expressed as percentages. Differences between groups were tested using oneway analysis of variance (ANOVA) for parametric data and Kruskal-Wallis test for non-parametric data for continuous variables, and Pearson’s chi-squared test for categorical variables. If significant differences between group means were detected, multiple pairwise comparisons were carried out for differences between each group pair, and P-values adjusted accordingly. For correlation analysis, Pearson’s correlation coefficient was used if both variables were normally distributed. Otherwise, Spearman’s rank correlation test was used. The optimal cut-off was determined, keeping in mind the maximization of the sum of sensitivity and specificity. Another approach entailed that the relative weights of sensitivity and specificity in their sum would be equal. The serum Current Diabetes Reviews, 2022, Vol. 18, No. 1 GDF15 cut-off was demonstrated through receiver operating characteristic (ROC) curve for the study population as well as separately for subgroups of BMI and age. Calculations for subgroups were carried out on the full sample and bootstrap samples to optimize the performance of the estimation method. To quantify and compare the VAT GDF15 expression between T2DM and control groups, we employed the 2ΔΔCt method [24]. For all analytical purposes, a two- tailed Pvalue less than 0.05 was considered to be statistically significant. 3. RESULTS 3.1. Study Characteristics The median age for T2DM patients was 48 years, with a male: female ratio of 1.92:1. The median age was 50 years for males and 47 years for females. For the healthy control and the pre-diabetes populations, the male: female ratio was 1.94:1 and 1.14:1, respectively. Basic clinical and biochemical characteristics of the study subjects are presented in Table 1. Significantly higher BMI and WHR, and higher levels of FBG, HbA1c, TC, TG, and LDL-cholesterol levels were observed in the diabetic group compared to controls. The between-group differences except for fasting insulin, HDL-cholesterol, and LDL-cholesterol, were statistically significant. HbA1c levels were significantly different across all three group pairs, i.e. controls, pre-diabetes and T2DM. WHR, FBG, and TG were significantly different in people in T2DM compared to both pre-diabetes and healthy controls, whereas BMI and total cholesterol levels were significantly higher in cases compared only to healthy controls. 3.2. Serum GDF15 Levels The median serum GDF15 concentration in cases was 1445.47 pg/mL (range 454.83-4627.35 pg/mL), and it was significantly higher than both pre-diabetes cases and healthy controls (in both comparisons, adjusted P<0.001) (Fig. 1a). However, the difference was not significant between the healthy control and pre-diabetes groups, despite a numerically higher GDF15 in the latter group. Median serum GDF15 levels were significantly higher in high BMI subjects (median 1030.85 pg/mL, n=94) compared to normal BMI subjects (median 670.78 pg/mL, n=62, P=0.036) (Fig. 1b). Agewise distribution of serum GDF15 levels in subjects aged 40 years and above were significantly higher (median 1299.09 pg/mL, n=96) compared to the population less than 40 years of age (median 599.64, pg/mL, n=60, P<0.001) (Fig. 1c). No significant gender-based differences were observed. To explore whether the differences between diabetic and control subject were confounded by BMI or age, we further divided our subjects into groups, once with pre-diabetes and healthy controls separately, and again, merging the two into a common group (since serum GDF15 did not significantly differ between healthy controls and pre-diabetes). Our analysis revealed that in both comparisons, serum GDF15 levels in each of the T2DM groups were significantly higher than the control groups, independent of BMI and age (Table 2, Supplementary tables S1-S4). 3 Current Diabetes Reviews, 2022, Vol. 18, No. 1 e010321189862 3.3. Correlation Analyses Spearman’s rank correlation test revealed significant associations between GDF15 levels and most of the clinical and biochemical parameters (Table 3). 3.4. GDF15 as a Marker in T2DM To establish serum GDF15 as a marker for T2DM, we used the cutpointr package for calculation of optimal cut-off values. We found the serum GDF15 cut-point for our study population to be 868.09 pg/mL, by maximizing the sum of sensitivity and specificity. The ROC curve analysis provided a sensitivity and specificity of 90.41% and 79.52% respectively, with an area under the curve (AUC) to be 0.892 (Fig. 2). Several optimal cut-off points obtained from various methods, including the method scores, have been listed in Table 4. Cut-offs were separately estimated on subgroups defined by BMI and Age (Supplementary figures S1, S2). For the normal BMI group, an optimal GDF15 cut- point of 900.39 pg/mL was obtained. The associated sensitivity, spe- Roy et al. cificity, and AUC were 94.12%, 82.22%, and 0.927, respectively. For the obese population, the cut-off was 868.09 pg/mL, with a sensitivity of 89.29%, a specificity of 76.32%, and an AUC of 0.882 (Fig. 3). For the <40 years old population, the serum GDF15 cut-off was 776.39 pg/mL, whereas it was 868.09 pg/mL in the subjects aged ≥40 years. The ROC analysis (Fig. 4) showed a sensitivity, specificity, and AUC of 87.50%, 82.69%, and 0.868 respectively for the former group, and 92.31%, 70.97%, and 0.862 for the latter. We further performed a logistic regression analysis, which showed that serum GDF15 significantly contributed to differentiating T2DM cases from controls (OR 15.07, 95% CI 6.62-38.21,P<0.001) (Table 5). In the unadjusted model, compared to the null deviance (199.95 on 155 degrees of freedom), GDF15 showed an improved deviance residual by 149.84 at the cost of 1 degree of freedom with an AIC of 153.84. Further, upon adjusting for BMI and metformin intake, serum GDF15 could still discriminate between T2DM and controls (OR 3.31, 95% CI 1.20-9.66, P=0.023). Fig. (1). (a): Boxplot showing significantly higher GDF15 levels in T2DM compared to pre-diabetes and healthy controls, (b) Significantly higher GDF15 levels (P=0.036) in high BMI subjects compared to normal BMI subjects, and (c) in subjects aged ≥40 years (P<0.001) compared to subjects aged <40 years. (A higher resolution / colour version of this figure is available in the electronic copy of the article). 4 GDF15 as a Biomarker in Obese, type 2 Diabetic Indians e010321189862 Current Diabetes Reviews, 2022, Vol. 18, No. 1 Table 1. Study characteristics; physical measurements and laboratory parameters in control population, pre-diabetic population, and type 2 diabetic cases (values are expressed in median and interquartile range). Parameters Controls (n=53) Pre-diabetes (n=30) T2DM (n=73) P-value Age (years) 30.00 (15.00) 38.00 (14.75) 48.00 (12.00) 0.000* BMI (kg/m ) 24.60 (3.57) 24.87 (5.57) 27.23 (4.73) 0.000* 2 WHR 0.90 (0.06) 0.88 (0.09) 0.95 (0.13) 0.000* Age ≥40 years (%) 10.90 8.97 41.67 0.000*† Obese (%) 7.55 20.00 24.66 0.001*† FBG (mg/dL) 96.00 (8.00) 100.50 (12.25) 172.00 (125.00) 0.000* HbA1c (%) 5.30 (0.30) 5.90 (0.38) 8.30 (3.10) 0.000* Fasting insulin (mIU/L) 9.60 (8.20) 11.60 (9.50) 9.9 (11.07) 0.218 HOMA-IR 2.30 (1.97) 2.97 (2.22) 3.93 (5.62) 0.016* HOMA %B 108.86 (84.25) 103.25 (66.53) 32.87 (47.38) 0.000* HOMA %S 43.49 (39.62) 33.69 (26.80) 25.47 (53.93) 0.016* HOMA2-IR 1.27 (1.08) 1.51 (1.21) 1.71 (1.46) 0.250 HOMA2%B 100.50 (51.40) 92.40 (38.50) 40.50 (39.45) 0.000* HOMA2%S 78.20 (64.20) 66.20 (57.90) 58.40 (67.65) 0.299 Total cholesterol (mg/dL) 161.00 (34.00) 180.00 (53.00) 193.00 (49.00) 0.000* Triglycerides (mg/dL) 97.00 (64.00) 118.50 (68.25) 161.00 (125.00) 0.000* HDL-C (mg/dL) 39.00 (11.00) 42.00 (15.75) 40.00 (9.00) 0.414 LDL-C (mg/dL) 103.00 (41.00) 115.50 (45.00) 121.00 (47.00) 0.055 TC/HDL-C ratio 4.17 (1.27) 4.56 (1.21) 5.08 (1.61) 0.014* TG/HDL-C ratio 2.48 (1.76) 2.94 (1.29) 4.18 (3.20) 0.000* TyG index 8.44 (0.61) 8.71 (0.69) 9.44 (0.97) 0.000* GDF15 (pg/mL) 609.01 (356.43) 627.85 (553.885) 1445.47 (1071.44) 0.000* * Statistically significant at P<0.05; kruskal-wallis test for between-group differences † Pearson’s Chi-squared test Fig. (2). Receiver operating characteristic (ROC) curve for T2DM using serum GDF15 cut-off value of 868.09 pg/mL. 5 Current Diabetes Reviews, 2022, Vol. 18, No. 1 e010321189862 Roy et al. 2 Fig. (3). Receiver operating characteristic (ROC) curve for T2DM using serum GDF15 cut-off values for BMI subgroups <25.0 kg/m and 2 ≥25.0 kg/m . (A higher resolution / colour version of this figure is available in the electronic copy of the article). 6 GDF15 as a Biomarker in Obese, type 2 Diabetic Indians e010321189862 Current Diabetes Reviews, 2022, Vol. 18, No. 1 Fig. (4). Receiver operating characteristic (ROC) curve for T2DM using serum GDF15 cut-off values for age subgroups <40 years and ≥40 years. (A higher resolution / colour version of this figure is available in the electronic copy of the article). 7 Current Diabetes Reviews, 2022, Vol. 18, No. 1 e010321189862 Roy et al. Table 2. Subgroup analysis for serum GDF15 levels in the study population (values are expressed in median and interquartile range). - Normal BMI High BMI P-value - T2DM Pre-diabetics Controls T2DM Pre-diabetics Controls - n 17 15 30 56 15 23 - GDF15 (pg/mL) 1890.45 (1204.83) 659.86 (762.11) 557.17 (279.25) 1354.08 (988.09) 493.53 (327.28) 651.13 (471.69) 0.000* T2DM Non-diabetics T2DM Non-diabetics - n 17 45 56 38 - GDF15 (pg/mL) 1890.45 (1204.83) 638.74 (328.10) 1354.08 (988.09) 607.53 (416.00) 0.000* Age <40 years n GDF15 (pg/mL) Age ≥40 years - T2DM Pre-diabetics Controls T2DM Pre-diabetics Controls - 8 16 36 65 14 17 - 1111.28 (392.25) 512.755 (355.06) 523.57 (294.97) 1515.69 (1114.22) 820.90 (804.83) 742.93 (251.39) 0.000* T2DM Non-diabetics T2DM Non-diabetics - n 8 52 65 31 - GDF15 (pg/mL) 1111.28 (392.25) 523.57 (319.48) 1515.69 (1114.22) 742.93 (498.14) 0.000* * Statistically significant at P<0.05; kruskal-wallis test for between-group differences For differences between each group pairs, Dunn Kruskal-Wallis multiple comparison test with Benjamini-Hochberg correction was carried out. No significant differences were observed between the BMI and age groups in either case. Table 3. Correlation analyses between serum GDF15 and anthropometric and biochemical parameters and insulin resistance indices. Variable * Spearman’s Rank Correlation Rho (ρ) P-value Age (years) 0.525 0.000* 2 BMI (kg/m ) 0.199 0.013* WHR 0.262 0.000* FBG (mg/dL) 0.530 0.000* HbA1c (%) 0.599 0.000* Fasting insulin (mIU/L) -0.077 0.338 HOMA-IR 0.156 0.052 HOMA %B -0.502 0.000* HOMA %S -0.156 0.051 HOMA2-IR 0.089 0.285 HOMA2%B -0.515 0.000* HOMA2%S -0.093 0.265 Total cholesterol (mg/dL) 0.191 0.017* Triglycerides (mg/dL) 0.328 0.000* HDL-C (mg/dL) -0.038 0.640 LDL-C (mg/dL) 0.082 0.308 TC/HDL-C ratio 0.201 0.012* TG/HDL-C ratio 0.284 0.000* TyG index 0.547 0.000* Statistically significant at P<0.05. 8 GDF15 as a Biomarker in Obese, type 2 Diabetic Indians e010321189862 Current Diabetes Reviews, 2022, Vol. 18, No. 1 Table 4. Optimal cut-off values along-with sensitivity, specificity, and accuracy of serum GDF15 for classifying diabetic from controls. Different methods usually employed for cut-off estimation have been tested with their minimization or maximization functions. Method Optimal Cut-off for GDF15 (pg/mL) Youden’s index 868.09 0.6993 90.41 79.52 84.62 Accuracy 868.09 0.8462 90.41 79.52 84.62 F1 score 868.09 0.8462 90.41 79.52 84.62 Cohen’s kappa 868.09 0.6936 90.41 79.52 84.62 Product of sensitivity and specificity 868.09 0.7189 90.41 79.52 84.62 Absolute difference of sensitivity and specificity 951.44 0.001 80.82 80.72 80.77 Odds ratio 951.17 17.65 80.82 80.72 80.77 Risk ratio 951.17 4.19 80.82 80.72 80.77 Absolute difference of PPV and NPV 1041.94 0.004 75.34 83.13 79.49 Method Score Sensitivity (%) Specificity (%) Accuracy (%) Youden’s index=sensitivity+specificity-1; F1 score=weighted average of precision and recall; PPV=positive predictive value; NPV=negative predictive value. Table 5. Multiple regression analysis of serum GDF15 as a marker in type 2 diabetes mellitus. Predictor Unadjusted Adjusted Coefficients P-value Coefficients P-value Intercept -0.4745 0.049* -3.6414 0.019* GDF15 2.7125 0.000* 1.1985 0.023* BMI - - 0.1179 0.052 Metformin - - 18.9084 0.988 * Statistically significant at P<0.05. Columns 2 and 3 represent the results from the unadjusted model; columns 4 and 5 resulted after adjusting for BMI and metformin intake. Fig. (5). Barplot showing higher VAT GDF15 expression in (a) T2DM cases compared to controls, and (b) obese compared to non-obese population. (A higher resolution / colour version of this figure is available in the electronic copy of the article). 9 Current Diabetes Reviews, 2022, Vol. 18, No. 1 e010321189862 3.5. Tissue GDF15 Expression The average ΔCT value for tissue GDF15 in T2DM patients was 1.26 compared to 1.88 for the control population, which meant that GDF15 expression in VAT was 1.54 times upregulated in T2DM patients. Again, the upregulation was 1.57 times in the obese population compared to the nonobese population. Hence, our analysis showed that GDF15 expression in VAT was increased in T2DM cases compared to control population, and obese compared to non-obese subjects (Fig. 5). 4. DISCUSSION Our study revealed that in a Western Indian population of Rajasthan, serum GDF15 could be a useful candidate biomarker in the presence of T2DM in patients without complications or comorbidities. This likelihood of detecting T2DM increases if the patient is obese and aged ≥40 years. Based on existing statistical methods, we further suggest serum GDF15 cut-off values to discriminate T2DM and healthy individuals. We also validated the results by showing that GDF15 expression in adipose tissue, a key regulator in the development of IR, was upregulated in T2DM subjects compared to controls. T2DM is a complex, multifactorial, and mostly preventable disease. It is often associated with obesity and an array of other secondary metabolic abnormalities such as IR, hyperlipidemia, and oxidative stress [14, 25]. To identify a marker that gets deranged early on in the disease course could benefit the patients in the long run. GDF15 is an anti-inflammatory protein from the TGF-β superfamily, which inhibits macrophage activation. Like other anti-inflammatory markers, its expression is increased at the onset of diabetes [26]. This increase occurs in response to the underlying inflammation in diabetes but is not sufficient to overcome chronic low-grade inflammation that ultimately leads to IR and β-cell dysfunction [27, 28]. It has also been suggested that GDF15 may be increased as a surrogate marker of p53 activation. Obesity-induced p53 activation plays a crucial role in developing IR [29]. GDF15 directly targets p53, which is induced by oxidative stress and has anti-apoptotic effects on target cells [30]. Further, Mullican et al. experimented on mice models and showed that GDF15 acts through GDNF family receptor α-like (GFRAL), a distant member of the TGF-β superfamily ligands and expressed in area postrema and nucleus tractus solitarius neurons [31]. GDF15 potentially exerts its effects on glucose homeostasis through the AKT-ERK crosstalk. They also showed that GFRAL-deficient mice did not respond to the anorectic effects of GDF15 [31]. In the Malmo Diet and Cancer (MDC) cohort study, GDF15 was shown to be useful for identifying risk population for T2DM [10]. They found this association to be stronger in the age group 60 years or less. However, in the current study, we could better differentiate controls from T2DM at age of ≥40 years and BMI ≥25 kg/m2. Carstensen et al. 2010 [11] showed GDF15 to be significantly higher in individuals who developed T2DM; while the XENDOS trial showed that GDF15 is significantly related to IR, and is an 10 Roy et al. independent predictor of abnormal glucose control [9]. Further, elevated GDF15 concentration at baseline was significantly associated with HOMA-IR after four years [9]. In our study, serum GDF15 was borderline significantly associated (P=0.052) with HOMA-IR. In a Czech population study, GDF15 levels were significantly higher in the T2DM group compared to the obese group without T2DM [7]. They also showed that this increase was further augmented by age and BMI. This suggests that with increasing obesity, an underlying inflammatory environment enhanced the expression of GDF15 that increases its secretion. Further, a highly significant negative association between GDF15 and β-cell function in the form of HOMA%β (ρ=-0.50, P<0.001) and HOMA2%β (ρ=-0.52, P<0.001) was observed. There was a borderline negative association between GDF15 and HOMA-%S (P=0.051). This makes GDF15 a surrogate marker of IR and insulin sensitivity and follows a 2012 study on obese women that reports an inverse correlation between insulin sensitivity and serum GDF15 during a clamp examination [32]. In a cross-sectional study from 2019, which included 160 obese subjects, GDF15 was significantly correlated with insulin sensitivity indices and estimated β-cell function [33]. The mean GDF15 levels in the T2DM groups in both Kempf et al. 2012 and Dostalova et al. 2009 were 1136 pg/mL and 1100 pg/mL. We found the median value of serum GDF15 in T2DM cases to be 1445.47 pg/mL. A 2014 study claimed GDF15 to be a novel biomarker for impaired fasting glucose [12]. They also reported a difference by sex, where males had a more elevated GDF15. But neither of these findings were there in our study. This difference may be attributed to the difference in ethnicities of the two populations. Lu et al. 2019 [34] pooled the data from eleven Gene Expression Omnibus (GEO) datasets and three articles and found that GDF15 levels, in differentiating T2DM cases than controls, had a sensitivity, specificity, and AUC of 83%, 59%, and 0.81, respectively, with an OR of 1.64 (1.35-1.99). The diagnostic accuracy measures in our study (sensitivity 90.41%, specificity 79.52% and AUC 0.892, with an adjusted OR of 3.31) with the optimal cut-point are comparable, albeit much higher in terms of specificity. Further, GDF15 expression in adipose tissue was upregulated more than 1.5 times in both T2DM and obese subjects compared to controls. GDF15 has the advantages of being used as a biomarker in diabetes. It is a protein, and hence only one fasting blood sample will be required for its measurement. Serum GDF15 was found to be stable at room temperature for 48 hours despite repeated freeze/thaw cycles. Also, the anticoagulant used in the sampling vials does not affect the measurement as observed in other studies [12, 35]. One of the strengths of our study was that serum GDF15 levels were measured in adult T2DM patients within 60 years of age without any complications or comorbidities, and therefore, reflects an unbiased quantification in these patients, which might get confounded by the presence of co-morbidities. However, the limitations of the current study have to be considered. Firstly, we recruited all patients who were previously diagnosed with T2DM, which means the clinical history was considerably variable in the disease population. Therefore, future studies may be planned, where GDF15 is GDF15 as a Biomarker in Obese, type 2 Diabetic Indians e010321189862 used as a population screening tool for obese individuals to identify undiagnosed diabetes population. Secondly, being an anti-inflammatory cytokine, GDF15 is increased in response to any inflammation, probably as a protective mechanism [36]. Therefore, the existence of comorbid conditions or complications, unbeknownst to the patient or not phenotypically expressed, will affect the circulating levels. It is also worth mentioning that the GDF15 levels quantified as part of this study, or the previous large-scale cohorts, was not immediately after collection but from frozen serum samples at -80°C. The variability of GDF15 in such samples over time needs evaluation. Lastly, many patients were on treatment, mostly on metformin, the most commonly prescribed glucose-lowering agent in T2DM. Recent reports from mice model and the Outcome Reduction with Initial Glargine Intervention (ORIGIN) trial identified GDF15 as a novel and dose-dependent biomarker for metformin [37-39]. Coll et al. 2019 [39] have suggested that metformin imparts its beneficial effects through elevating the circulatory levels of GDF15. Our multiple regression analysis showed that GDF15 is an independent predictor of T2DM after adjusting for metformin intake. However, further studies exploring the effects of other hypoglycemic drugs on GDF15 will be crucial. T2DM is asymptomatic in its initial stages. But like any disease, its underlying pathology sets up the mechanistic processes much earlier, and consequently, the compensatory mechanisms come into play. GDF15 is one such molecule that is increased as part of the body’s anti-inflammatory response and has multiple sites of production in humans. Moreover, being an inflammatory marker, its relative significance and underlying mechanistic pathways under various pathophysiological conditions are still mostly unknown. Until the pathways directly controlling GDF15 are entirely understood, its clinical utility may remain limited. CONCLUSION To conclude, we identified GDF15 levels as a candidate marker of T2DM in obese patients from a Western Indian population of Rajasthan and optimised its cut-off value, which can discriminate the cases from the controls and with high sensitivity and specificity, GDF15 seems to have good diagnostic accuracy, that is further supported by raised VAT expression of GDF15. Hence, we suggest that it can be an appropriate biomarker for diagnosing T2DM from the control population. However, large, prospective cohort studies are necessary to validate this claim. ETHICS APPROVAL AND CONSENT TO PARTICIPATE Current Diabetes Reviews, 2022, Vol. 18, No. 1 (www.icmje.org) with regard to the patient’s consent for research or participation in a study. CONSENT FOR PUBLICATION Informed consent was obtained from all patients prior to the study. AVAILABILITY OF DATA AND MATERIALS Not applicable. FUNDING This study was funded by AIIMS, JODHPUR Intramural funding, AIIMS/RES/2018/1923,and MD thesis grant 2018 No. AIIMS/IEC/2018/636. CONFLICT OF INTEREST The authors declare no conflicts of interest, financial or otherwise. ACKNOWLEDGEMENTS We acknowledge the support of all the technical staff at IPD Biochemistry Laboratory, AIIMS, Jodhpur. SUPPLEMENTARY MATERIAL Supplementary table S1:Group-wise comparison of serum GDF15 in T2DM, pre-diabetic and healthy individuals according to high BMI. Supplementary table S2:Group-wise comparison of serum GDF15 in T2DM and control individuals according to high BMI. Supplementary table S3:Group-wise comparison of serum GDF15 in T2DM, pre-diabetic and healthy individuals according to age. Supplementary table S4:Group-wise comparison of serum GDF15 in T2DM and control individuals according to age. Supplementary figure S1: Sensitivity and specificity plot for serum GDF15 in T2DM patients, grouped according to BMI <25.0 kg/m2 (Code 0) and BMI≥25.0 kg/m2 (Code 1). Supplementary figure S2: Sensitivity and specificity plot for serum GDF15 in T2DM patients, grouped according to age group <40 years (Code 0) and ≥40 years (Code 1). REFERENCES [1] This study has been approved by the Institutional Ethics Committee (IEC), AIIMS, Jodhpur, India. HUMAN AND ANIMAL RIGHTS All clinical investigations were conducted according to the Declaration of Helsinki principles, 1975, as revised in 1983. The study has complied with the guidelines of the International Committee of Medical Journal Editors [2] [3] Bootcov MR, Bauskin AR, Valenzuela SM, et al. MIC-1, a novel macrophage inhibitory cytokine, is a divergent member of the TGF-beta superfamily. Proc Natl Acad Sci USA 1997; 94(21): 11514-9. http://dx.doi.org/10.1073/pnas.94.21.11514 PMID: 9326641 Ding Q, Mracek T, Gonzalez-Muniesa P, et al. Identification of macrophage inhibitory cytokine-1 in adipose tissue and its secretion as an adipokine by human adipocytes. Endocrinology 2009; 150(4): 1688-96. http://dx.doi.org/10.1210/en.2008-0952 PMID: 19074584 Patel S, Alvarez-Guaita A, Melvin A, et al. GDF15 provides an endocrine signal of nutritional stress in mice and humans. Cell 11 Current Diabetes Reviews, 2022, Vol. 18, No. 1 [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] 12 e010321189862 Metab 2019; 29(3): 707-718.e8. http://dx.doi.org/10.1016/j.cmet.2018.12.016 PMID: 30639358 Macia L, Tsai VW, Nguyen AD, et al. Macrophage inhibitory cytokine 1 (MIC-1/GDF15) decreases food intake, body weight and improves glucose tolerance in mice on normal & obesogenic diets. PLoS One 2012; 7(4) http://dx.doi.org/10.1371/journal.pone.0034868 PMID: 22514681 Chrysovergis K, Wang X, Kosak J, et al. NAG-1/GDF-15 prevents obesity by increasing thermogenesis, lipolysis and oxidative metabolism. Int J Obes 2014; 38(12): 1555-64. http://dx.doi.org/10.1038/ijo.2014.27 PMID: 24531647 Adela R, Banerjee SK. GDF-15 as a target and biomarker for diabetes and cardiovascular diseases:a translational prospective. J Diabetes Res 2015; 2015:490842. http://dx.doi.org/10.1155/2015/490842 PMID: 26273671 Dostálová I, Roubícek T, Bártlová M, et al. Increased serum concentrations of macrophage inhibitory cytokine-1 in patients with obesity and type 2 diabetes mellitus: the influence of very low calorie diet. Eur J Endocrinol 2009; 161(3): 397-404. http://dx.doi.org/10.1530/EJE-09-0417 PMID: 19515791 Vila G, Riedl M, Anderwald C, et al. The relationship between insulin resistance and the cardiovascular biomarker growth differentiation factor-15 in obese patients. Clin Chem 2011; 57(2): 309-16. http://dx.doi.org/10.1373/clinchem.2010.153726 PMID: 21164037 Kempf T, Guba-Quint A, Torgerson J, et al. Growth differentiation factor 15 predicts future insulin resistance and impaired glucose control in obese nondiabetic individuals: results from the XENDOS trial. Eur J Endocrinol 2012; 167(5): 671-8. http://dx.doi.org/10.1530/EJE-12-0466 PMID: 22918303 Bao X, Borné Y, Muhammad IF, et al. Growth differentiation factor 15 is positively associated with incidence of diabetes mellitus: the Malmö Diet and Cancer-Cardiovascular Cohort. Diabetologia 2019; 62(1): 78-86. http://dx.doi.org/10.1007/s00125-018-4751-7 PMID: 30350239 Carstensen M, Herder C, Brunner EJ, et al. Macrophage inhibitory cytokine-1 is increased in individuals before type 2 diabetes diagnosis but is not an independent predictor of type 2 diabetes: the Whitehall II study. Eur J Endocrinol 2010; 162(5): 913-7. http://dx.doi.org/10.1530/EJE-09-1066 PMID: 20167682 Hong JH, Chung HK, Park HY, et al. GDF15 Is a Novel Biomarker for Impaired Fasting Glucose. Diabetes Metab J 2014; 38(6): 472-9. http://dx.doi.org/10.4093/dmj.2014.38.6.472 PMID: 25541611 Pradeepa R, Mohan V. Prevalence of type 2 diabetes and its complications in India and economic costs to the nation. Eur J Clin Nutr 2017; 71(7): 816-24. http://dx.doi.org/10.1038/ejcn.2017.40 PMID: 28422124 Hruby A, Frank B. The epidemiology of obesity: A big picture. Pharmacoeconomics 2015. Jul;33(7):673-89. http://dx.doi.org/10.1007/s40273-014-0243-x PMID: 25471927 American Diabetes Association. 2. Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes-2020. Diabetes Care 2020; 43 (Suppl. 1): S14-31. http://dx.doi.org/10.2337/dc20-S002 PMID: 31862745 World Health Organization. (2011). Waist circumference and waist-hip ratio : report of a WHO expert consultation, Geneva, 8-11 December 2008. World Health Organization. https://apps.who.int/iris/handle/10665/44583 Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985; 28(7): 412-9. http://dx.doi.org/10.1007/BF00280883 PMID: 3899825 Jeppesen J, Facchini FS, Reaven GM. Individuals with high total cholesterol/HDL cholesterol ratios are insulin resistant. J Intern Med 1998; 243(4): 293-8. http://dx.doi.org/10.1046/j.1365-2796.1998.00301.x PMID: 9627143 Zhang L, Chen S, Deng A, et al. Association between lipid ratios and insulin resistance in a Chinese population. PLoS One 2015; 10(1): e0116110. [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] Roy et al. http://dx.doi.org/10.1371/journal.pone.0116110 PMID: 25635876 Simental-Mendía LE, Rodríguez-Morán M, Guerrero-Romero F. The product of fasting glucose and triglycerides as surrogate for identifying insulin resistance in apparently healthy subjects. Metab Syndr Relat Disord 2008; 6(4): 299-304. http://dx.doi.org/10.1089/met.2008.0034 PMID: 19067533 Roy D, Tomo S, Modi A, Purohit P, Sharma P. Optimising total RNA quality and quantity by phenol-chloroform extraction method from human visceral adipose tissue: a standardisation study. MethodsX 2020; 101113 http://dx.doi.org/10.1016/j.mex.2020.101113 Chomczynski P, Sacchi N. The single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction: twenty-something years on. Nat Protoc 2006; 1(2): 581-5. http://dx.doi.org/10.1038/nprot.2006.83 PMID: 17406285 RStudio Team. RStudio: Integrated Development for R RStudio, Inc 2015.http://www.rstudio.com/ Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Δ Δ C(T)) Method. Methods 2001; 25(4): 402-8. http://dx.doi.org/10.1006/meth.2001.1262 PMID: 11846609 Tan CK, Chong HC, Tan EHP, Tan NS. Getting ‘Smad’ about obesity and diabetes. Nutr Diabetes 2012; 2(3) http://dx.doi.org/10.1038/nutd.2012.1 PMID: 23449528 Herder C, Zierer A, Koenig W, Roden M, Meisinger C, Thorand B. Transforming growth factor-beta1 and incident type 2 diabetes: results from the MONICA/KORA case-cohort study, 1984-2002. Diabetes Care 2009; 32(10): 1921-3. http://dx.doi.org/10.2337/dc09-0476 PMID: 19592635 Germain RN. Maintaining system homeostasis: the third law of Newtonian immunology. Nat Immunol 2012; 13(10): 902-6. http://dx.doi.org/10.1038/ni.2404 PMID: 22990887 Kolb H, Mandrup-Poulsen T. The global diabetes epidemic as a consequence of lifestyle-induced low-grade inflammation. Diabetologia 2010; 53(1): 10-20. http://dx.doi.org/10.1007/s00125-009-1573-7 PMID: 19890624 Minamino T, Orimo M, Shimizu I, et al. A crucial role for adipose tissue p53 in the regulation of insulin resistance. Nat Med 2009; 15(9): 1082-7. http://dx.doi.org/10.1038/nm.2014 PMID: 19718037 Berezin AE. Diabetes mellitus related biomarker: The predictive role of growth-differentiation factor-15. Diabetes Metab Syndr 2016; 10(1) (Suppl. 1): S154-7. http://dx.doi.org/10.1016/j.dsx.2015.09.016 PMID: 26482961 Mullican SE, Lin-Schmidt X, Chin CN, et al. GFRAL is the receptor for GDF15 and the ligand promotes weight loss in mice and nonhuman primates. Nat Med 2017; 23(10): 1150-7. http://dx.doi.org/10.1038/nm.4392 PMID: 28846097 Karczewska-Kupczewska M, Kowalska I, Nikolajuk A, et al. Hyperinsulinemia acutely increases serum macrophage inhibitory cytokine-1 concentration in anorexia nervosa and obesity. Clin Endocrinol (Oxf) 2012; 76(1): 46-50. http://dx.doi.org/10.1111/j.1365-2265.2011.04139.x PMID: 21645023 Schernthaner-Reiter MH, Itariu BK, Krebs M, et al. GDF15 reflects beta cell function in obese patients independently of the grade of impairment of glucose metabolism. Nutr Metab Cardiovasc Dis 2019; 29(4): 334-42. http://dx.doi.org/10.1016/j.numecd.2018.12.008 PMID: 30718144 Lu J, Zhang Y, Dong X, et al. Association between MIC-1 and Type 2 Diabetes: A Combined Analysis. Dis Markers 2019; 2019:7284691. http://dx.doi.org/10.1155/2019/7284691 PMID: 31827641 Kempf T, Horn-Wichmann R, Brabant G, et al. Circulating concentrations of growth-differentiation factor 15 in apparently healthy elderly individuals and patients with chronic heart failure as assessed by a new immunoradiometric sandwich assay. Clin Chem 2007; 53(2): 284-91. http://dx.doi.org/10.1373/clinchem.2006.076828 PMID: 17185363 Modi A, Dwivedi S, Roy D, et al. Growth differentiation factor 15 and its role in carcinogenesis: an update. Growth Factors 2019; 37(3-4): 190-207. GDF15 as a Biomarker in Obese, type 2 Diabetic Indians [37] [38] e010321189862 http://dx.doi.org/10.1080/08977194.2019.1685988 PMID: 31693861 Gerstein HC, Pare G, Hess S, et al. ORIGIN Investigators. Growth Differentiation Factor 15 as a Novel Biomarker for Metformin. Diabetes Care 2017; 40(2): 280-3. http://dx.doi.org/10.2337/dc16-1682 PMID: 27974345 Day EA, Ford RJ, Smith BK, et al. Metformin-induced increases [39] Current Diabetes Reviews, 2022, Vol. 18, No. 1 in GDF15 are important for suppressing appetite and promoting weight loss. Nature Metabolism 2019; 1: pp. 1202-8. http://dx.doi.org/10.1038/s42255-019-0146-4 Coll AP, Chen M, Taskar P, et al. GDF15 mediates the effects of metformin on body weight and energy balance. Nature 2019; 578: 444-8. http://dx.doi.org/10.1038/s41586-019-1911-y PMID: 31875646 13