Type 2 diabetes (T2D) is a worldwide disease and it is expected to affect more than 629 million people by 2045 [1]. Type 2 diabetes is a complex disease, characterized by the interplay between environmental and genetic factors, and it is mainly determined by impaired beta cell function and/or insulin sensitivity [2].

In the past years, many studies have been conducted in order to unravel the genetic background of type 2 diabetes. Before 2007, candidate and linkage approaches lead to identify a small number of associated loci (PPARG, KCNJ11, CAPN10, WFS1,IRS1 and TCF7L2) [3] [4]. From 2007 to present day, the advent of high-throughput genotyping technology, such as Genome Wide Association (GWA) Scan, increased the number of T2D associated loci, identified in different ethnic groups, up to over 240 genetic risk loci with about 400 independent association signals [5] [6] [7]. Initially, the first GWA studies lead to replicate the already known associated SNPs and to identify at least six novel T2D loci (CDKN2A/2B, IGF2BP2, CDKAL1, HEEX, SLC30A8) [8] [9] [10] [11]. The number of candidate polymorphisms further increased in the following years as a result of other GWA and meta-analysis studies that added KCNQ1, CDC123, CAMK1, JAZF1, THADA, TSPAN8, ADAMTS9, NOTCH2/ADAM30 to the list [12] [13] [14] [15]. In 2010, Voight and collaborators discovered other T2D associated loci (ARAP1, BCL11A,ZBED3,KLF14,TP53INP1,CHCHD9,CENTD2-ARAP1,HMGA2,ZFAND6,PRC1,TLE4 and DUSP9) and interestingly, among them, there were also two MODY genes (HNF1A and HNF4A) [16]. Previously published papers already reported an effect of SNPs located into MODY genes (HNF4A,PDX1,NEUROD1,KLF11,GCK,CEL)on type 2 diabetes, suggesting an influence of these monogenic diabetes loci also on the common form of the disease [17] [18] [19] [20]. Subsequent studies expanded their investigation to the genetic background of type 2 diabetes related traits, such as fasting plasma glucose and obesity, and they discovered a number of associated loci (ADCY5,DGKB,GCK,GCKR,MTNR1B,PROX1,BCDIN3D/FAIM2,AIF1/NCR3,FTO,GNPDA2,MCR4,NEGR1,SFRS10,SH2B1 and TMEM18) [21] - [26]. More recently many other polymorphisms have been added to the list of T2D associated loci reaching a number of over 400 common variants involved in this disease susceptibility [6] [7] [27].

Despite the high number of associated loci, genetics can explain less than 20% of type 2 diabetes heritability [28]. Moreover, investigating the role of genetics on T2D pathophysiological phenotypes, such as impaired beta cell function and insulin sensitivity, researchers found that insulin secretion was much more genetically determined than insulin action [16] [29] [30] [31] [32] [33]. Most of the GWA and meta-analysis studies have been conducted in Caucasians or Asian ancestry, however, as far as we know, there are no data regarding a quite large Italian sample. Thus, in this study, we aimed to genotyped 143 polymorphisms on 1875 Italian subjects (927 controls and 948 diabetic patients) in order to investigate their association with type 2 diabetes.

In this study we tested the association of 133 polymorphisms, in part already identified as T2D associated loci and in part selected as tag SNP within loci of interest, with type 2 diabetes on 1875 Italian subjects (927 controls and 948 cases).

The novelty of the study is that, as far as we know, this is the first study that investigated a number of polymorphisms, already associated to type 2 diabetes or located into/close to genes with a possible biological role in the development of the disease, in a quite large Italian population, thus giving information about the allelic frequency of these SNPs in Italians subjects.

Not surprisingly, TCF7L2 variants displayed the strongest association signals, and this was consistent with previously published data, reporting for rs7903146 the highest odd ratio (OR: 1.37) of all type 2 diabetes associated polymorphisms, in different ethnic groups [51] [52] [53]. Different mechanisms have been hypothesized in order to explain the biological role of TCF7L2 in determining type 2 diabetes, particularly it could: 1) decrease beta cell proliferation and insulin secretion; 2) alter insulin process or release; 3) affect GLP1 signalling in beta cell [54]; 4) alter local chromatin structure in islet cells [55].

Three loci, reported in this study, have been previously associated to type 2 diabetes through candidate-gene approach: KCNJ11 and ABCC8, both encoding two subunits of the potassium channel, an important protein involved in the insulin release process [56] [57]; and WFS1, which is involved in the Wolfram Syndrome, a monogenic disease characterized also by diabetes mellitus [58]. Particularly, variants in WFS1 have been linked to gut incretin hormones such as pancreatic glucose-dependent insulinotropic peptide sensitivity. Because endoplasmic reticulum stress has been recognized as contributing to insulin resistance in type 2 diabetes [59], the dissection of pathogenic mechanisms in monogenic ER stress-mediated β-cell loss may also contribute to the understanding of type 2 diabetes pathogenesis [60].

All other associated variants have been previously identified through GWA studies [3] [4] [5] [6] [7]. Many of them have a biological role that could explain their involvement in type 2 diabetes, even if the molecular mechanisms are not yet clear and further functional studies are needed: IGF2BP2, which binds IGF2 mRNA and influences pancreas development [3] [61]; FTO, the strongest obesity gene so far identified, which probably promotes diabetes onset through an effect on BMI [62]; HNF1B, which regulates gene transcription of insulin and other proteins involved in glucose metabolism [63]; finally G6PC2, belonging to the glucose-6-phosphatase catalytic subunit family, and GCKR,theglucokinase regulatory protein, both involved in glucose metabolism, previously associated to fasting glucose levels and type 2 diabetes [64]. Glucokinase and glucose-6-phosphatase are the molecular bases of the glucose/glucose-6-phosphate cycle, which directly affects the glucose homeostasis through insulin secretion. Notably, the association between G6PC2 SNPs and FPG was reported only in non-diabetic population, but not in T2D cohorts [65].

Moreover, we previously demonstrated that GCKR variability may play a pathogenetic role in type 2 diabetes affecting insulin secretion [40]. These three genes could be considered as determinants of glucose homeostasis that finely regulate beta cell function [40] [64].

The biological function of the last two loci that we found associated to type 2 diabetes in our Italian population, JAZF1 and ADAMTS9, and that have been already identified in previous GWAS, is not yet well known, thus it is difficult to hypothesize their possible functional mechanisms in type 2 diabetes [12] [13], even because contrasting results have been found [66]. We previously demonstrated that variations in ADAMTS9 were related to insulin secretion and a recent study of Cheng and colleagues indicated a positive regulatory role of TCF7L2 on endogenous ADAMTS9 expression [63] [67]. JAZF1 has been implicated in glucose and metabolism, and its overexpression could result in enhanced glucose tolerance and insulin sensitivity, reduced body weight gain, inhibited the expression of genes related to lipid metabolism and lipid accumulation in adipocytes [68] [69] [70] [71]. A recent study of Kobiita A. et al., shows that impaired JAZF1 function in T2D contributes to ribosomal and ER stress, activation of apoptosis pathways, and ultimately β-cell demise [72].

In conclusion, we replicated the association of 11 loci with type 2 diabetes in an Italian population, further genetic and functional studies will be necessary to identify the causal variants, if any, located into these T2D-associated loci.

The authors have no potential conflicts of interest to disclose.

Bonetti, S., Zusi, C., Rinaldi, E., Boselli, M.L., Patuzzo, C., Trabetti, E., Malerba, G., Bianchi, C., Lombardi, G., Zaza, G., Del Prato, S., Gambaro, G., Bonora, E., Bonadonna, R.C. and Trombetta, M. (2021) Common Variants Associated to Type 2 Diabetes in the Italian Population. Open Journal of Endocrine and Metabolic Diseases, 11, 24-42. https://doi.org/10.4236/ojemd.2021.111003

Common variants associated to type 2 diabetes in the Italian population.

Mathematical Modeling of Beta Cell Function

The analysis of the glucose and C-peptide curves during the OGTT follows the general strategy described in previous publications [41] [42] with some modifications and builds upon previous works from other laboratories [43] [44]. The kinetics of C-peptide is described with a two-compartment model, in which the two pools (1 and 2) exchange with each other and the irreversible loss of the hormone is from pool 1, the same where C-peptide concentration is measured. C-peptide kinetic parameters are computed according to the equations by Van Cauter et al. [45].

Herein are the equations describing the model of glucose induced insulin secretion during an OGTT:

d c p 1 ( t ) / d t = ISR ( t ) + c p 2 ⋅ k 12 − ( k 01 + k 21 ) ⋅ c p 1 (1)

where ISR = insulin secretion rate, cp₁ = C-peptide mass in the sampling (accessible) compartment, cp₂ = C-peptide mass in the remote compartment, k₁₂ and k₂₁ = rate constants of the exchange between the two C-peptide compartments, and k₀₁ = rate constant of the irreversible loss of C-peptide from the accessible compartment. Note that the values of the volume of distribution of C-peptide pool 1 (accessible compartment), k₁₂,k₂₁, and k₀₁ are computed according to the equations by Van Cauter et al. [45].

ISR ( t ) = BSR + DSR ( t ) + PSR ( t ) (2)

where BSR = basal insulin secretion rate, DSR = insulin secretion rate due to the derivative (or dynamic) component, and PSR = insulin secretion rate due the proportional (or static) component.

BSR = C P s s ⋅ V 1 ⋅ k 01 (3)

where CP_ss is basal C-peptide concentration and V₁ is the volume of the accessible compartment of C-peptide.

From the modeling viewpoint, DSR(t) and PSR(t) are the components which in intravenous glucose tolerance tests or hyperglycemic clamps describe classical first phase insulin secretion and second phase insulin secretion, respectively. Furthermore, from a physiological viewpoint, the sum of BSR and PSR(t) describes the relationship linking glucose concentration and insulin secretion rate, in the absence of the derivative component (DSR).

DSR(t) and PSR(t) are mathematically defined as follows:

DSR ( t ) = X 1 ( t ) ⋅ τ − 1 (4)

d X 1 ( t ) / d t = σ 1 ⋅ [ d G ( t ) / d t ] / [ log ( 1.1 + t ) ] − X 1 ( t ) ⋅ τ − 1 if dG(t)/dt > 0 (5)

d X 1 ( t ) / d t = − X 1 ( t ) ⋅ τ − 1 if dG(t)/dt ≤ 0 (6)

where σ₁ = glucose sensitivity of derivative control of insulin secretion, G = plasma glucose concentration, X₁ = C-peptide (insulin) mass made available for the derivative component of insulin secretion, τ = time constant of the derivative component of insulin secretion, and the term log(1.1 + t) accomodates the time-associated decline of σ₁ documented in humans during a hyperglycemic stimulus [46].

PSR ( t ) = X 2 ( t ) ⋅ δ − 1 (7)

d X 2 ( t ) / d t = σ 2 ⋅ [ G ( t ) − θ ] − X 2 ( t ) ⋅ δ − 1 (8)

where σ₂ = glucose sensitivity of the proportional component of insulin secretion, X₂ = C-peptide (insulin) mass made available for the proportional component of insulin secretion, δ = time constant of the proportional component of insulin secretion, θ = glucose threshold above which the beta-cell responds with the proportional component of insulin secretion to plasma glucose concentration.

This model was implemented in the SAAM 1.2 software (SAAM Institute, Seattle, WA) [47] to estimate its unknown parameters. Numerical values of the unknown parameters were estimated by using nonlinear least squares. Weights were chosen optimally, i.e., equal to the inverse of the variance of the measurement errors, which were assumed to be additive, uncorrelated, with zero mean, and a coefficient of variation (CV) of 6% - 8%. The unknown parameters of the model are: CP_ss, σ₁, τ, σ₂, δ, and θ. They were estimated with good precision, as shown by their CVs (Supplemental TableS1). Since in the GENFIEV Study the data set was to the time window 0’ - 120’, CPss was assumed to be equal to baseline C-peptide concentration in all subjects. For the same reasons, in those subjects of the GENFIEV Study in whom glucose levels did not go back close to baseline levels, θ was assumed to be equal to baseline glucose concentration.

There are two main physiological outputs of the model:

1) derivative control (units: [pmol∙m⁻² BSA]∙[mmol∙l⁻¹∙min⁻¹]^–1): it is the amount of insulin secreted in response to a rate of glucose increase of 1 mmol/l per min which lasts for 1 minute;

2) stimulus-response curve linking glucose concentration (x axis) to insulin secretion rate (y axis): as explained above, it is the sum of BSR and PSR. With the purpose of avoiding artifactual increases in the power of statistical analyses, we used the stimulus-response curve at the pre-determined glucose concentrations of 5.5, 8.0, 11.0, 15.0 and 20.0 mmol/l.