Central and peripheral pathogenetic forms of type 2 diabetes: a proof-of-concept study

Hypothesis Previous studies provide evidence that glycated haemoglobin (HbA1c) and fasting plasma glucose (FPG) should not be considered as interchangeable alternatives in the diagnosis of the same type 2 diabetes, but as indicators of its different pathogenetic subtypes. This study was conducted to determine whether a particularly high amount of glucose in either HbA1c form or in fasting plasma would be found in diabetic patients genetically predisposed for either intensive cognitive or intensive muscle metabolic activity, respectively. Methods HbA1c and FPG levels, polymorphisms of genes indicating the predisposition to different cognitive activity (the dopamine D2 receptor (DRD2/ANKK1)), muscle activity (peroxisome proliferator-activated receptor gamma, coactivator 1 alpha (PGC1A(PPARGC1A))), and vascular regulation of general metabolic activity (the angiotensin 1 converting enzyme (ACE)) were assessed in diabetic patients and nondiabetic controls. Results DRD2/ANKK1 polymorphism that affects baseline central arousal determined HbA1c variations uncorrelated with FPG in total and clinical groups. The mutation of PGC1A mainly affecting peripheral glucose metabolism had an effect on FPG correlated or uncorrelated with HbA1c depending on the effect assessment in the total sample or in the nondiabetic group, respectively. ACE insertion/deletion (I/D) gene polymorphism was associated with both HbA1c and FPG fluctuations, but only in diabetic patients. Conclusion The findings provide evidence that the HbA1c and FPG may predict the risks for different subtypes of type 2 diabetes associated with either brain or muscle metabolic activity in genetically vulnerable people.


Introduction
Despite the known limitations of diagnostic criteria and early risk for type 2 diabetes derived from fasting (FPG) and/or 2-h plasma glucose levels, they are still widely accepted. Glycated hemoglobin A1 or the percentage of total hemoglobin that is bound by glucose (HbA 1c ) reflects an average level of plasma glucose over the previous 2-4 months and is also widely used as a marker of chronic glycemia. It was concluded that HbA 1c and FPG display a different sensitivity versus specificity in identifying people at risk for later development of diabetes, and Expert Committees in diabetes consider that a better biochemical marker should be selected as a more reliable diagnostic tool (1,2). However, this discordance may occur because HbA 1c and FPG each measure different physiological processes associated with the risk for type 2 diabetes. In this case, these glycemic parameters should be considered as partly overlapped, but not as interchangeable alternatives.
Previous studies showed that HbA 1c is not affected by plasma glucose instability and has a close relationship with mean or average plasma glucose (MPG) (3). However, correlations between HbA 1c and FPG are low. Twin, family, and genome-wide association studies show that HbA 1c and FPG levels are heritable with substantial genetic effects on their inter-individual variations, and that genetic factors influencing them do not overlap (4). Although some heritable HbA 1c variations are unrelated to glycemic status in individuals with and without type 2 diabetes (e.g., altered erythrocyte physiology), individual variance in both glycemic parameters can reflect different aspects of energy metabolism that are modulated by different gene polymorphisms. However, hypothesis-free, genome-wide association studies with multiple testing do not seem very helpful in exploring this challenge because they may miss some important physiological relationships (5). The main analytical approach of genome-wide association studies is focused on minimizing the false positive rate at the expense of controlling the false negative rate, and this bias determines the 'missing heritability' of type 2 diabetes susceptibility (6,7,8). This gives an additional warning that results of data-driven studies should not exclusively dictate or restrict scientific development from using the conventional approach based on prior hypotheses (9,10).
The hypothesis-driven candidate gene approach based on an inferred physiological role of previously discovered genes may investigate the subject more effectively. For example, physiological mechanisms between feeding and fasting are differently regulated for keeping glucose levels sufficient: (i) for permanent functioning of the brain and other glucose-dependent tissues (switching between idled and non-idled brain conditions; neutral, positive, and negative moods; and various cognitive activities) and (ii) for an intermittent energy supply to all other organs (e.g., switching between glucose and fatty acid oxidation in active and relaxed muscles) (11). Previous findings allow us to consider that HbA 1c level may be associated with genetic factors influencing more stable central mechanisms (habitual cognitive and dietary patterns), but FPG level may be determined by genetic factors influencing more flexible peripheral mechanisms which regulate energy metabolism (cardiovascular and metabolic activity patterns). For example, a study of children with and without attentiondeficit/hyperactivity disorder (ADHD) found that FPG did not differ between the groups, but HbA 1c was higher in the ADHD-group (12). This HbA 1c -related glucose regulation may be modulated by a central dopaminergic system in general and by polymorphism of the dopamine D2 receptor (DRD2) gene in particular (e.g., rs1800497 (TaqIA) single-nucleotide polymorphism (SNP), later identified within the exon 8 of ankyrin repeat and kinase domain containing 1 (ANKK1)). Mutations in this gene were found to be associated with ADHD and appear to be relevant to cognitive skills, as well as type 2 diabetes, dietary, and other motivated and compulsive/impulsive behaviors (13,14,15,16). Other studies support that MPG-or HbA 1c -, but not FPG, -related glucose regulation is associated with cognitive performance and dietary preferences.
For example, only higher HbA 1c or MPG was associated with cognitive decline and cognitive impairments (e.g., lower scores of visual motor speed, attention, learning, and verbal memory) in people with and without type 2 diabetes (17,18). These cross-sectional and prospective relationships were not mediated by cardiovascular (CV) or other metabolic complications and may be determined by more fat, carbohydrate, or generally higher caloric food preference as habitual dietary behavior in people with higher HbA 1c (19,20,21). It seems evolutionary justified, that meal restriction claims better performance skills for obtaining food, but an easy access to plenty of meal results in a 'lazy' or 'idled' brain (i.e., in the brain with low cognitive performance). In these cases, the levels of HbA 1c may display an amount of glucose unclaimed by the brain and irrevocably removed from energy metabolism. A particularly high amount of this unclaimed glucose in HbA 1c form may be predicted in people with the phenotypically 'idled' or 'lazy' brains that are genetically predisposed for hard cognitive working (e.g., individuals with A1-T allele of TaqIA). Fewer studies found that higher FPG values were also associated with lower cognitive skills, but only in patients with CV diseases or metabolic syndrome (22,23). Compared with HbA 1c , the FPG relationship with cognitive functions may be driven indirectly by peripheral hyperglycemia-related processes resulting in severe cardiovascular and metabolic impairments (24).
In some studies, the presence of D allele in case of the angiotensin 1 converting enzyme (ACE) insertion/ deletion (I/D) gene polymorphism (determining higher ACE levels and thus higher vascular resistance) conferred an increased risk of glucose intolerance (detected by higher 2-h glycemia) in a nondiabetic population (25). However, this association may be moderated by other genetic or nongenetic factors, as another study found that overweight women who are homozygous for the D allele of this gene had more intensive glucose utilization (detected by hyperinsulinemic-euglycemic clamps) compared with those who are homozygous for the I allele (26). ACE gene polymorphism, as well as gene polymorphisms affecting other metabolic factors like peroxisome proliferatoractivated receptor gamma (PPARG2) and its coactivator (PGC1A (PPARGC1A)), may have opposite effects on glucose metabolism in different (e.g., clinical and nonclinical) populations (27,28,29,30,31). Hemodynamic (ACE) and metabolic (PPARG2 and PGC1A) gene polymorphisms were found to have a weak impact on HbA 1c level and cognitive functions, but have a strong impact on body composition and physical activity in the general population and diabetic patients (32,33,34,35,36). FPG, but not HbA 1c , is mainly related to body composition (e.g., body weight, visceral fat area, body mass index, body fat percentage, body muscle percentage, waist circumference) and exercise (walking, steps/day and physical activity, min/day) indicators (20). It seems physiologically justified, when 'lazy' or 'idled' muscles with low metabolism do not claim the glucose intended for them, its unutilized and undeposited excess is detected in fasting and 2-h plasma. A particularly high amount of unclaimed glucose in fasting plasma may be predicted in persons with the 'lazy' muscles that are genetically predisposed for intensive or hard physical work (e.g., individuals with D allele of ACE I/D gene polymorphism or G allele of rs8192678 (Gly482Ser) polymorphism of PGC1A gene).
This research addresses two above-mentioned predictions of central and peripheral physiological mechanisms associated with HbA 1c and FPG and thus different central and peripheral pathophysiological forms of type 2 diabetes by studying the relationships between polymorphisms of central (DRD2), hemodynamic (ACE), and metabolic (PGC1A) genes and these two glycemic indicators in diabetic patients and nondiabetic control group.

Ethical statement
Institutional and governmental regulations concerning the ethical use of human volunteers were followed. The study was approved by the Ethics Committee of the Sholokhov Moscow State University for the Humanities and performed in compliance with the recommendations of Helsinki Declaration II.

Study populations
This study was based on a larger cohort of outpatients diagnosed with type 2 diabetes, living in Moscow (Russia), and listed in a local medical register. For the study, 103 participants (73 females) aged 23 -83 years (mean ± s.d., 65.7 ± 11.2 years) with undefined ethnicity (due to ethical regulations of the local medical register) were randomly selected from a sub-group of patients with uncomplicated type 2 diabetes for no less than 2 years. Classification of type 2 diabetes was done according to the WHO diagnostic criteria (37). For inclusion in the study, the patients needed to be a noncompliant with the treatment regimen in order to decrease the confounding effect of the treatment on expected relationships. Of the total sample of diabetic patients listed in the local medical register, 50% were identified as missing medication in the previous week. The recommended treatment modalities in the selected sample were metformin alone (20%); metformin in combination with dipeptidyl peptidase IV inhibitor (5%); glucagonlike peptide 1 alone (5%); and insulin alone (10%) or in combination with oral antidiabetic agents (metformin and sulfonylurea; 60%). Sixty subjects (42 females) without diabetes and undefined ethnicity (due to the same ethical regulations) aged 19-92 years (53.8 ± 16.2 years) were enrolled in the control group. They visited primary care centers for a regular annual medical examination and did not either have hypertension, coronary heart disease, or any other neuropsychiatric, cardiovascular, or metabolic diseases. All selected patients and healthy subjects signed a written consent for participation in the study.
An effective sample size for genetic model with continuous outcomes was calculated using Quanto (38). Sample size = 60 was found effective for α = 0.05 and power = 0.80 with β = 0.60, dominant allele frequency = 0.54 and mean (s.d.) of FPG = 5.3 (0.7) mmol/L obtained from a previous study and accepted for a group of subjects without diabetes (39). Additional 43 subjects were included in the patient group to cover for possible allele frequency deviation associated with the disease (40).

Laboratory assays
Blood samples of the people with and without diabetes were collected during 2012-2015 on a regular weekday morning after an overnight fast during an outpatient visit to local hospital laboratories or primary care centers in Moscow (Russia) and immediately delivered to the Moscow Diagnostic and Clinical Center #1. The samples were analyzed for fasting plasma glucose (FPG), glycated hemoglobin A1 (HbA 1c ), total cholesterol, low-and highdensity lipoprotein-cholesterol (direct measurements), triglycerides, plasma creatinine, total bilirubin, total protein, aspartate transaminase, and alanine transaminase

Genetic analysis
Total genomic DNA was extracted from peripheral blood leukocytes and assessed as described in detail elsewhere (41). In particular, SNP genotyping of PGC1A (Gly482Ser, rs8192678, G > A nucleotide change) and DRD2/ANKK1 (TaqIA, rs18000497, A1-T, and A2-C nucleotide change) was performed by the real-time PCR method using a TaqMan (hydrolysis) probe-based assay (Thermocycler DT-322; DNA-Technology, Russia) or using 2% agarose gels with a 100-bp DNA Ladder (SibEnzyme Ltd, Russia) and a gel imaging system for genotyping ACE (insertion/deletion, I/D) polymorphisms in intron 16 (42). The end-point readings were analyzed according to the manufacturer's instructions.
One-third of the samples of patients and healthy subjects was randomly processed in duplicate for each genotype analysis to protect the study against genotyping errors. Genotype data were obtained in 98.7-100% of the DNA samples with a genotype error rate of 0.0%.

Data analyses
Descriptive and inferential analyses were performed by SPSS (SPSS Science, Chicago, IL) using Pearson product-moment correlation and General Linear Models by Type III method (GLM) to evaluate significance of multivariate and univariate effects and bootstrapped confidential intervals (CIs) of regression estimates. Biascorrected and accelerated bootstrap procedure with 1000 bootstrap replications of regression coefficients was used to derive their nonparametric 95% CIs from empirical sampling distribution. The bootstrap procedure was considered as a robust alternative to inference based on parametric assumptions (such as normally distributed errors) to confirm or validate findings obtained by parametric analyses, and is recommended for reporting inferences in scientific reports (43). Differences at P < 0.05 were regarded as significant for parametric analyses. Parameter estimates are expressed as nonstandardized (B) regression coefficients and their standard errors (s.e.) or as means (M) and their standard deviations (s.d.). Where necessary, a partial η 2 was reported as a measure of strength of association (effect size), which is comparable to R 2 expressing the percentage of explained variance. Sex and age were included in all models to adjust effects for these factors. As the distribution of age was skewed, its common logarithm (log 10 ) was used. The GLM procedure was conducted with two measures of glycemia, FPG and HbA 1c , treated as continuous dependent variables representing multiple measurements of the same process to test the hypothesis that both together are affected by gene polymorphisms. All multivariate F values were obtained by Pillai's trace statistic, which is equivalent to partial η 2 and R 2  Table 1 lists the values of demographic and metabolic parameters separately for patients with diabetes and control participants. Patients in diabetes group were significantly older and had significantly higher values of FPG, HbA 1c , triglycerides, and alanine transaminase, but lower values of total and high-density lipoprotein cholesterols than controls. FPG weakly correlated with HbA 1c in diabetes and control groups (r = 0.35 and 0.22, P < 0.001 and 0.10, respectively). There were no significant differences among the groups in sex, creatinine, total protein, low-density lipoprotein cholesterol, aspartate transaminase, and total bilirubin.

Hardy-Weinberg equilibrium test
Distribution of DRD2/ANKK1 genotype in diabetes group, ACE and PGC1A genotypes in control group (P < 0.01, 0.005, and 0.05, respectively), and all three genetic loci in the total sample (P < 0.05) was significantly different from the expectations of HWE. Although, in principle, genotyping errors may be responsible for the observed statistically significant deviations from HWE, this was unlikely for these samples because double scoring and duplicate genotyping in randomly selected subsets of 30% individuals confirmed the results for each locus. Therefore, the probable reason for the departure of observed genotype frequencies for the three loci is not genotyping error but the  association of loci with disease and/or population stratification (e.g., intensive multiethnic migration, generations overlapping, and unequal sex ratio (M < F) in this region). More importantly, in contrast to casecontrol (i.e., gene-disease association) studies, the expected inferences on quantitative association of the polymorphisms with FPG and HbA 1c phenotypes or its lack should remain valid and be independent of the fact that these polymorphisms in the type 2 diabetes or control individuals are not in HWE (44). Significant simple and interaction effects of ACE and PGC1A polymorphisms and interaction effect of Sex*DRD2/ANKK1 genotype on age in the total sample confirmed the assumption of generations overlapping as a probable cause violating HWE in this study. The presence of the PGC1A gene GG genotype and the ACE I allele each separately increased the odds of surviving to older age (Fs(1,159) = 4.96 and 4.16, P = 0.027 and 0.043; η 2 = 0.03 and 0.03, respectively, after adjustment for Sex; confirmed by bootstrap 95% CIs: 0.01-0.08 and 0.004 -0.08, respectively). However, association of the presence of the PGC1A A allele with the ACE gene DD genotype significantly increased the risk of not surviving to older age (F(1,158) = 4.46, P = 0.036; η 2 = 0.03, after adjustment for Sex; confirmed by bootstrap 95% CI: 0.01 -0.15). Women with DRD2/ANKK1 gene A2A2 genotype and men with DRD2/ANKK1 A1 allele had higher odds to survive to older age (F(1,157) = 4.37, P = 0.038; η 2 = 0.03; confirmed by bootstrap 95% CI: 0.02 -0.18).

Genotype distribution
Compared with the control group, the frequency of the G allele polymorphism of PGC1A gene was significantly higher, but the frequency of the PGC1A AA genotype was lower in the diabetes group (Table 2). No evidence of differences between the groups was found in genotype or allele frequency of the other two genetic polymorphisms.

Levels of FPG and HbA 1c in different genotypes
Multivariate test found that DRD2/ANKK1 and PGC1A, but not ACE allele polymorphism, significantly determined the fluctuation of glycemic status in the total group  respectively). HbA 1c , but not FPG, levels were significantly higher in non-A2 allele carriers of DRD2/ANKK1 (A1A1 genotype) compared with subjects with A2A2 & A2A1 genotypes, but both HbA 1c and FPG levels were higher in D allele carriers of ACE (DD & ID genotypes) compared with subjects with II genotype ( Table 4). The multivariate test found that only the PGC1A allele polymorphism determined the fluctuation of glycemic status in the control group (Pillai's trace = 0.09 (0.10); F(2) = 2.53 (2.89); P = 0.089 (0.064)). FPG, but not HbA 1c , levels were significantly higher in non-G allele carriers of PGC1A (AA genotype) compared with subjects with

Discussion
Although the distributions of all three genetic loci in the total sample were significantly different from the expectations of HWE, our additional findings attribute this departure to the association of polymorphisms with generations overlapping and the rate of surviving to old age. This is consistent with the genetic models of morbidity and mortality protection effects of PGC1A gene GG genotype (due to chronic stress resilience effect of high PGC1A expression), ACE I allele (due to acute stress resilience effect of lower serum ACE activity), and DRD2/ANKK1 gene polymorphism depending on sex (due to stress resilience effects of dopamine binding in the striatum in females with lower and males with higher sensitivity to motivational/reward factors) (45,46,47,48,49,50). The main findings of quantitative associations of the polymorphisms with FPG and HbA 1c phenotypes or their lack are independent of these deviations (44).
This study confirmed that fasting plasma glucose and glycated hemoglobin are weakly correlated in diabetes and nondiabetes groups. DRD2/ANKK1 allele polymorphism determined an uncorrelated part of variations of glycated hemoglobin in the total and clinical groups. This corresponds to the hypothesis that people with high baseline levels of cortical arousal that claim extensive energy delivery for mental processes (e.g., ruminations) are at higher risk for the transition of glucose regulation mechanism from normal to abnormal and for the development of more severe type 2 diabetes (A1A1 genotype) compared with people with low baseline levels of cortical arousal prone to novelty-or sensation-seeking behaviors (A2 allele carriers). Indeed A1A1 genotype determines reduced D2 receptor density and binding, but increased baseline dopamine neuron activity associated behaviorally with high harm avoidance and low novelty seeking behaviors (51). It suggests that if cognitive activity decreases in these people for a long time, the oversupply of glucose is left unclaimed and this increases the risk for diabetes and its higher severity.
In this study, the frequencies of PGC1A GG and GA genotypes were significantly higher in the diabetes group compared with the control group and this difference was confirmed by the higher glycemic status in general (higher levels of both fasting plasma glucose and glycated hemoglobin) in PGC1A G-allele carriers compared with AA genotype in total group. Thus, PGC1A allele polymorphism determined a correlated part of variations of fasting plasma glucose and glycated hemoglobin in the total group. However, this PGC1A effect on total glycemic status was only related to the risk for the transition of glucose regulation mechanism from normal to abnormal, because PGC1A GG and GA genotypes determine a lower fasting plasma glucose level uncorrelated with glycated hemoglobin level in a nondiabetic control group. These opposite relationships between PGC1A polymorphism and fasting plasma glucose may potentially cancel each other out if assessed in mixed populations (52). This finding corresponds with a potential role of PGC1A as a protective factor or a mediator of disease progression depending on affected tissues. AA genotype was found to determine reduced activity of PGC1A compared with GG and GA genotypes (53). Thus, higher levels of PGC1A determined by PGC1A GG and GA genotypes were found to be associated with higher fasting plasma glucose and glycated hemoglobin levels in diabetic patients included in this study. Indeed, it is known that PGC1A activity is strongly activated in the liver and pancreatic β-cells of the subjects with obesity and type 2 diabetes (29). This could potentially contribute to increased hepatic glucose production which in turn would contribute to the higher hyperglycemia in patients with type 2 diabetes included in this study. However, overexpression of PGC1A in skeletal muscle results in increased glucose uptake by this tissue (29). This can contribute to reducing plasma glucose levels found in nondiabetic subjects with PGC1A GG and GA genotypes. Thus regularly active muscles due to exercises recommended in the STOP-NIDDM trial become a main protective or resilience factor against the development of insulin resistance and type 2 diabetes only in people with PGC1A GG genotype in contrast to those with PGC1A AA genotype shown sensitive in the same trial only to glucose-reducing treatment by acarbose (52). It suggests that if muscle activity decreases in the former group of people, the overexpression of PGC1A in the liver and the pancreatic β-cells becomes dominant in the effect on plasma glucose level and this increases the risk for diabetes.
Other findings indicated that ACE gene deletion polymorphism was associated with the glycemic status indicated by correlated glycated hemoglobin and plasma glucose levels with higher values in ACE DD and ID genotypes compared with subjects with II genotype only in diabetic patients. It corresponds with some reports that found no evidence of ACE polymorphism as a risk factor for the development of type 2 diabetes (42), but as a risk factor for more complicated or severe type 2 diabetes (54). The presence of ACE D allele determines higher ACE levels and thus higher vascular resistance with decreasing tissue delivery of glucose. The high vascular resistance together with poor glycemic control (i.e., both high plasma glucose and glycated hemoglobin levels) may increases the risk for atherosclerosis (55).
Therefore, the findings suggest that adopting a healthy lifestyle with more intensive cognitive or more intensive physical activity may have a protective (resilience) effect against the development of two different subtypes of type 2 diabetes in susceptible individuals, who are genetically predisposed to more intensive energy supply for respective brain or muscle activity. The interventions for increasing cognitive activity or decreasing vascular resistance may further protect against complications in patients with these two cortical or metabolic/vascular subtypes of type 2 diabetes. This further suggests that the metabolic/vascular pathway in the development of diabetes and its severity may be associated with 'lazy' muscles determining a concordant increase in glycated hemoglobin and fasting plasma glucose levels. 'Cortical arousal' mechanism of the development of type 2 diabetes may be associated with 'idled' brain determining an isolated increase in glycated hemoglobin level without a significant effect on fasting plasma glucose. Thus, both glycated hemoglobin and fasting plasma glucose should be evaluated to indicate principal pathogenetic differences in the development of type 2 diabetes for its correct treatment and prevention.
The main limitation of this study is the relatively small sample size. However, according to the power analysis, larger sample sizes would be relevant in the case of gene polymorphisms for which effects were not expected to show evidence of practical importance or statistical significance to avoid type II ('false-negative') errors. The study's results confirm that the present sample size had enough power to demonstrate the predicted effects in order to support the main hypothesis. However, further validation of these findings in a larger population involving other factors modifying these and other related gene expressions is required. Modification of detected significant effects by variations in glycemic control and DNA, mRNA and protein expressions as well as polymorphisms of other genes with the same effects on glycated hemoglobin and fasting plasma glucose levels can explain weak effect sizes obtained in the study. This study was conducted to obtain proof-of-concept where genome-wide association research is methodologically ineffective. Future research should support the concept with respect to other genetic, epigenetic, and proteomic factors. These factors should increase explained variance or effect size of both pathogenic mechanisms and thus demonstrate clinical importance of the two subtypes of type 2 diabetes and that both glycemic parameters are not interchangeable as indicators since they contribute unique information in its pathogenesis. This consideration should hold expert committees in diabetes from precipitate revisions of the diagnostic criteria for diabetes in favor of either indicator as more reliable (2). Clinical trials would demonstrate a beneficial effect of different cognitively or physically oriented lifestyles in individuals with different susceptibility to hyperglycemic effects of the decrease in brain or muscle activity.

Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

Funding
The study was supported by the Russian Ministry of Science and Education project No. 2934 led by M.K.N.

Author contribution statement
MKN designed and coordinated the study and reviewed/edited the manuscript, DMD conceived the study, researched data, contributed to the design and coordination, and wrote the manuscript. Both authors read and approved the final version.