Evaluation of PRKAA2 Genetic Variation on Metformin Ecacy as an Initial Therapy Among Drug-Naïve Patients With Type II Diabetes Mellitus

Background: Metformin is the most popular oral antidiabetic agent, which is recommended as initial monotherapy. AMPK is the pivotal target of metformin molecular mechanisms. AMPK subunit a2 (encoded by PRKAA2) is a gene contributable to increase type 2 diabetes mellitus (T2DM) risk. This study aimed to evaluate PRKAA2 rs2796498, rs2746342, and rs980799 genetic variations on metformin ecacy. Methods: This study enrolled 191 newly diagnosed Indonesia T2DM patients in primary health care. Patients who received metformin as monotherapy for at least 3 months were included for genotyping. Genotyping was performed using the Taqman assay. Results: Baseline characteristics showed that BMI was higher among AA than GG+AG (p=0.04). Patients with TT genotype showed a higher FBG and HbA1c than GG+GT (p=0.02 and p=0.02, respectively). There was no signicant difference in allele and genotype frequencies between responders and non-responders group in PRKAA2 rs2796498, rs9803799, and rs2746342. However, among PRKAA2 rs2796498, AG carrier had 0.32 times of responding in metformin ecacy after adjusting BMI, WC, blood pressure, lipid proles, and eGFR. Dominant model of rs2796498 showed a signicant association (OR=0.34, 95%CI=0.13 – 0.90) to metformin ecacy. Conclusions: Our ndings suggest that PRKAA2 rs2796498 genetic variation may affect metformin ecacy, especially AG carrier, in drug-naïve T2DM patients. and Applied Biosystems® qPCR 7500 Fast Real-Time PCR System. The nal reaction volume using in real-time PCR is 10mL, including 2.5 mL nucleotide-free water, 5mL TaqMan GTXpress mix, 0.5mL TaqMan SNP genotyping assay, and 2mL of genomic DNA. The thermal cycle for a reaction was as follows: 40 cycles at hold 95°C for the 20s, at denaturing 95°C for 3s, and then annealing 60°C for 30s. PRKAA2 SNPs were determined using the following primer sequences: PRKAA2 rs9803799 and rs2746342 might have no impact on metformin ecacy among Indonesian. Further study recruiting a larger sample size and engaging environment factors such as physical activities are required to conrm the role of PRKAA2 genetic variation, especially of rs2796498, rs9803799, and rs274634 on metformin ecacy among drug-naïve T2DM patients.


Introduction
Metformin is notable as the most prescribed medicine for T2DM patients as a recommended initial type 2 diabetes mellitus (T2DM) therapy by many international guidelines. Drug-naïve T2DM patients who have HbA1c 7-9% are suggested to consume metformin until it reaches the maximum doses before adding a second agent [1,2]. Metformin considerably has better safety and e cacy as initial T2DM therapy [3]. Metformin is a guanidine derivate which acts as an unchanged form. Metformin speci cally reduces blood glucose by enhancing glucose uptake in peripheral tissue and suppressing glucose hepatic production [4][5][6]. The underlying complex molecular mechanism of metformin has been observed in many studies, and AMP-activated protein kinase (AMPK) was mentioned to have a pivotal role in metformin action [7,8].
AMPK is a cellular energy sensor whose phosphorylation is stimulated by lower ATP concentration in cells. AMPK has three subunits consisting of a, b, and g. AMPK suppresses the anabolic process, which utilizes ATP [9]. Metformin stimulates phosphorylation of Thr-172 at a catalytic unit [10]. Some studies also demonstrated that metformin activates AMPK after inhibiting mitochondrial complex I respiratory chain. Thus it reduces the ATP level [8,11]. Furthermore, other studies reported that LKB-1 is also involved in AMPK activation by metformin [12]. Besides, AMPK via acetyl-CoA carboxylase 1 (ACC1) and ACC2 could inhibit malonyl CoA synthesis. As a result, it reduces lipogenesis and increases insulin sensitivity [13]. Therefore, metformin does not affect weight gain.
Nevertheless, there is a different e cacy of metformin among T2DM patients [14]. Metformin e cacy is affected by various factors, including age, lifestyle, baseline HbA1c, adherence [15,16], gene-drug interaction, and gene-environment interaction [17]. Inter-individual variability plays an important role in drug response affected by drug-gene interaction. Therefore, pharmacogenomic still be a promising study for discovering the relationship between gene and medicine's pharmacodynamic and pharmacokinetic, in this context especially metformin e cacy.
Many studies have explored the effect of genetic variation on metformin response in the pharmacokinetic area [18,19]. Otherwise, a little study conduct pharmacogenomic research focused on the pharmacodynamic of metformin. Accordingly, it is relevant to explore pharmacogenetic concerning the pharmacodynamic of metformin in achieving the HbA1c goal. Since AMPK is involved as the main target in the metformin molecular mechanism, genes coding AMPK may have contributable to metformin e cacy. A recent review declared that PRKAA2 coding AMPKa2 is one of a gene related to clinical outcomes after metformin therapy [20]. Nevertheless, that review has not mentioned an SNP of PRKAA2 contributing to metformin response yet.
Previous studies have reported an association of PRKAA2 with susceptibility of T2DM (rs2796498 and rs2746342) [21,22], even in metformin effectiveness (rs9803799) [23]. However, no studies observe the association between PRKAA2 rs2796498 and rs2746342 genetic variation and the effectiveness of metformin use. Notably, those three SNP have not been explored yet in Indonesia. Therefore, this study aimed to observe the genotype frequency among T2DM Indonesian patients receiving metformin. Furthermore, this study investigates the in uence of PRKAA2 rs2796498, rs9803799, and rs2746342 genetic variation on the e cacy of metformin in drug-naïve T2DM patients, determined by the reduction of HbA1c level.

Participants
This study recruited 191 participants from ten primary health care (PHC) in Yogyakarta Province, Indonesia. Physicians in PHC gave the rst prescription of metformin on newly diagnosed T2DM patients. Then participants were followed up for three months. The participant's selection following inclusion criteria: newly diagnosed T2DM patients with HbA1c>7%, 18-70 years old, do not consume other hyperglycemia agents, either oral or injectable dosage forms.
Participants were excluded when GFR<30 ml/min/m2, albumin serum 3.4 -4.8 g/dL, and diagnosed as diabetes gestational. Finally, only 106 newly diagnosed T2DM patients who had taken the rst prescription metformin as monotherapy and completed follow up for three consecutive months were enrolled in this study. Patients were classi ed into two groups: metformin responders (n=45) and metformin non-responders (n=61). All of our participants were of Indonesia origin. Metformin responders are patients who could achieve decreasing HbA1c>1.12% during follow up [24].
Anthropometric measurement, blood pressure, blood sample collecting for clinical chemistry and genotyping analysis, and medical record data were obtained for this study after patients signing informed consent. All participants were briefed about the study aim, procedure, duration, potential risk, and bene t. The study was performed in compliance, according to the Declaration of Helsinki. The study protocol was approved by the Medical and Health Research Ethics Committee (MHREC) Faculty of Medicine, Public Health, and Nursing Universitas Gadjah Mada -Dr. Sardjito General Hospital. Patients had the right not to participate in our study at any time.

Clinical measurement
Our study collected demographic data, anthropometric data, and laboratory results. Age and gender, as demographic data and metformin dose, were obtained from medical records. Nutritionists measured anthropometric measurements, including body weight, height, and waist circumference. Body Mass Index (BMI) was calculated by divided body weight (kg) by height (m2). Blood pressure was measured by the nurse in early registration. We examined laboratory results, including FBG, HbA1c, creatinine serum, HDL-c, Triglyceride-c, and total cholesterol. Laboratory data were collected by a PHC analyst that helped by a commercial laboratory after an overnight fast. FBG and creatinine serum was measured by the hexokinase method and enzymatic method, respectively.
HbA1c was calculated using high-performance liquid chromatography (Cobas D-10). Glomerular ltration rate (GFR) was calculated using CKD-EPI formulation using the creatinine level. Lipid pro les, including HDL-c, Triglyceride-c, and total cholesterol, were measured using Cobas C311. LDL was calculated using Friedewald formulation.

DNA extraction and genotyping
According to the kit protocol, genomic DNA was extracted from peripheral whole blood-EDTA using Geneaid® Blood DNA Mini Kit and stored at -20 until the genotyping procedure. Genotyping in rs2796498, rs9803799, and rs2746342 was performed using the TaqMan® genotyping assay and Applied Biosystems® qPCR 7500 Fast Real-Time PCR System. The nal reaction volume using in real-time PCR is 10mL, including 2.5 mL nucleotide-free water, 5mL TaqMan GTXpress mix, 0.5mL TaqMan SNP genotyping assay, and 2mL of genomic DNA. The thermal cycle for a reaction was as follows: 40 cycles at hold 95°C for the 20s, at denaturing 95°C for 3s, and then annealing 60°C for 30s. PRKAA2 SNPs were determined using the following primer sequences:

Statistical analysis
The baseline and follow-up participant's characteristics were analyzed and compared between the genotype model using an independent t-test or chi-square, as appropriate. We applied the dominant and recessive model in comparing clinical characteristics. Data are expressed in mean+SD for numerical data and n(%) for categorical data. Allele and genotype frequencies were evaluated by Hardy-Weinberg equilibrium (HWE) using chi-square. The association between responder status and genotypes was analyzed using multinominal logistic regression by adjusting for age, gender, BMI, WC, lipid pro les, glomerular ltration rate, and blood pressure. All statistical analysis was performed using SPSS version 25.0, and p<0.05 was considered statistically signi cant.

Results
Of the 191 participants who have classical T2DM signs, 62 were dropped out: 35 were HbA1c<7%, 3 were age>70 years old, 18 had eGFR<30 mL/min/m2, 20 did not comply with consuming metformin, and 23 needed the second agent. The baseline characteristics are shown in Table. 1. BMI in the AA group (29.00+5.34) was higher than the GG+AG group (25.01+4.21) in the rs2796498 recessive model (p=0.04). However, both have been classi ed as obesity. The differences between FBP and HbA1c before metformin treatment are only found in the rs2796498 recessive model (p=0.02 and p=0.04, respectively).
Speci cally, rs2796498 in the recessive model showed that the GG+GT group tended to have lower FBG and HbA1c before metformin therapy (174.39+58.08 mg/dL and 9.29+1.74%) than the TT group (211.67+74.21 mg/dL and 10.27+2.07%). Moreover, there were no signi cant differences in baseline characteristics in rs980377, either dominant or recessive models. Allele and genotype frequency of PRKAA2 genetic variation in responders and non-responder of metformin therapy are listed in Table 2. The HWE of rs2796498, rs9803799, and rs2746342 were 0.47, 0.03, and 0.92, respectively. It suggested that the PRKAA2 genotypes of rs2796498 and rs2746342 in our population are consistent with HWE, but not rs9803799 (p<0.05). Minor allele frequencies (MAFs) of rs2796498, rs9803799, and rs2746342 were 27.8% and 22.1%, 10.0% and 4.9%, 44,4% and 39.3%, respectively, in responders and non-responders group. As presented in Table 2., the frequency of allele G, T, and G of rs2796498, rs9803799, and rs2746342, respectively, tend to higher in metformin non-responders than responders group. Nonetheless, we could not nd the difference statistically of allele frequency between responders and non-responders (p=0.35 for rs2796498, p=0.15 for rs9803799, and p=0.46 for rs2746342). We could not nd any difference in blood pressure, FBG, HbA1c, HBA1c change, and lipid pro les after receiving metformin therapy based on PRKAA2 genetic variation (Table. 3). Furthermore, multinominal logistic regression models were applied to verify the effect of PRKAA2 genetic variation on metformin e cacy (Table 4). This study failed to detect the association between PRKAA2 genetic variation with metformin e cacy, even after adjusting age, gender, and adjusting for BMI, WC, and lipid pro les. Interestingly, we found some association after adjusting for BMI, WC, lipid pro les, blood pressure, and eGFR. As shown in
Accordingly, this current study evaluates the impact of PRKAA2 genetic variation on metformin e cacy. To the authors knowledge, this study is the rst pharmacogenomic research reporting allele frequency of rs2796498, rs9803799, and rs2746342 applied metformin therapy among Indonesia drug-naïve T2DM patients.
This recent study found that T2DM patients with AA genotype of rs2796498 had signi cantly higher BMI than GG+AG. However, this signi cant mean of BMI difference was not observed in rs9803799 and rs2746342. These ndings might imply that the AA genotype of rs2796498 is associated with obesity, then it should be correlated with lipid pro les. Our ndings contradict the data obtained by Jones et al. [25], which indicated that rs2796498 and rs2746342 genetic variations correlated lipid pro les. Furthermore, this study found that the TT genotype of rs2746342 had higher FBG and HbA1c levels signi cantly than GG+GT among study subjects before receiving metformin. Conversely, Shen et al. reported that the G allele had a higher FBG level and T2DM risk [22] . It could be caused by dissimilar ethnicity, where Shen et al. focused on Han Chinese, and this study focused on Indonesia. Nevertheless, the recessive model of rs2746342 did not in uence the difference in FBG, HbA1c, and HbA1c change after metformin therapy. It could be an early indicator that rs2746342 genetic variation does not alter metformin e cacy in our population.
AMPK, as the main target of metformin, regulates the function of hepatic glucose metabolism and pancreatic b-cell [8,11,26]. It has been widely agreed that the phosphorylation of AMPK is induced by metformin, although the speci c route is not clear yet. An animal study con rmed that 10 weeks of metformin treatment signi cantly escalated AMPK phosphorylation on the a2 subunit [27]. Several PRKAA2 genetic variations have been investigated to increase T2DM risk, including that's located in the intron region [21,22,28]. Nevertheless, only a few studies observe that SNP is related to metformin responses [23]. A review has been mentioned that PRKAA2 genetic variation is one of the genes contributing to the metformin mechanism [20]. However, it has not yet been discovered clearly in a clinical study.
Thus, this study investigated the association of PRKAA2 genetic variation and metformin e cacy among drug-naïve T2DM patients. Our results discovered that allele and genotype frequencies of rs2796498, rs9803799, and rs2746342 between responders and non-responders were not signi cantly different.
Interestingly, the wild type of rs2796498 (AA) and rs9803799 (GG) are detected in little number, either in the responder or non-responder group. It con rmed that those SNP has related to T2DM [21,29], and our study participants were T2DM, indeed.
Moreover, we found a signi cant association between rs2796498 in AG genotype and dominant model with metformin e cacy, after adjusting for BMI, WC, lipid pro les, blood pressure, and eGFR. AG carrier in rs2796498 had 0.32 times of metformin response compared with the GG carrier. Therefore, T2DM patients with AG carriers might have a poor response to metformin therapy. Apart from these, the dominant model found that AG+AA had a worse metformin e cacy compared with GG in drug-naïve T2DM patients. However, we could not detect that the A allele in uences metformin e cacy. This point emphasizes that AG carrier in rs2796498 is the predictive factors of metformin e cacy in our population. Although metformin is well-known as a better treatment for T2DM patients with obesity (ESC/EASD) [2], a different BMI in baseline could in uence the impact of AA genotype in metformin e cacy.
On the other hand, we could not identify any relationship between rs9803799 and metformin e cacy. Conversely, a study in the U.S population reported that rs9803799 had signi cant interaction with metformin [23]. In addition, rs9803799 deviated from HWE. Therefore, ndings related to rs980799 may be reported bias association [30].
This study still has several limitations. First, the epistatic mechanism may contribute to metformin e cacy. It is possible if other genes or SNPs also affect metformin response. Second, environmental factors such as diet, physical activities, and adherence might impact decreasing HbA1c. Third, an SNP was not in agreement with HWE. Finally, the sample size should be larger to detect the function of PRKAA2 in metformin e cacy. Further replication pharmacogenomic studies are needed to con rm the PRKAA2 genetic variation on metformin e cacy.

Conclusions
In summary, there were no signi cant differences between genotype and allele frequencies of PRKAA2 genetic variation with metformin e cacy. Only AG genotype and dominant model of PRKAA2 rs2796498 associated with metformin e cacy in drug-naïve T2DM patients treated with metformin as monotherapy, after adjusting for BMI, WC, blood pressure, eGFR, and lipid pro les. Nonetheless, PRKAA2 rs9803799 and rs2746342 might have no impact on metformin e cacy among Indonesian. Further study recruiting a larger sample size and engaging environment factors such as physical activities are required to con rm the role of PRKAA2 genetic variation, especially of rs2796498, rs9803799, and rs274634 on metformin e cacy among drug-naïve T2DM patients.