This study revealed that the mean%Meth levels of CpGs located within and near the second exon of the DPP4 gene (CpG94 to CpG102) were comparable in VAT of non-diabetic severely obese MetS− and MetS+. Subjects classified into quartiles based on their CpG94-102 %Meth levels had similar characteristics regarding MetS phenotypes, but were discordant for their plasma total-cholesterol levels. In a sub-sample of non-diabetic severely obese premenopausal women, CpG94-102 %Meth levels in VAT tended to be correlated with their %Meth levels in WBCs, and %Meth levels of four individual CpGs (CpG95, CpG99, CpG100 and CpG101) were significantly correlated.
The presence of similar DPP4 %Meth levels in VAT between MetS− and MetS+ subjects rejects the previous hypothesis of a significant difference between the two groups [3, 4]. This observation may potentially be explained by the absence of differential DPP4 gene expression in VAT of the studied men and women, which contrasts with the differential expression observed in our previous studies on severely obese men MetS− (n=7) and MetS+ (n=7) [3, 4]. It is actually difficult to clearly explain this discrepancy. Beside sex, the same subject selection criteria were applied and the characteristics of male subjects were quite similar as those from the microarray study . Potential predictors of DPP4 gene expression in VAT were derived from a complimentary stepwise regression analysis (predictive variables: age, sex, waist circumference and smoking; data not shown). Only smoking was significantly associated with DPP4 mRNA levels in combined men and women (r2=0.05, p=0.02). The proportion of smokers among the study subjects was lower as compared to the microarray experiment where about half of subjects in each group were smokers . Since frequencies of smokers were not different between MetS groups in the actual and the microarray studies, it would thus potentially anneal the effect of smoking when performing DPP4 mRNA comparisons between MetS groups. However, it may not reveal potential interactions with smoking which would have induced a greater DPP4 expression among MetS+ men taking part of the microarray experiment as compared to male subjects herein. However, to the best of our knowledge, no published studies reported that smoking was related with differential DPP4 gene expression. Added to this explanation, other subjects’ characteristics which were not evaluated in our present and previous studies may have confounded DPP4 gene expression, such as other DPP4 locus regions under epigenetic regulation [19, 20] via stimulation by cytokines [21, 22], the type of cells (e.g. adipocytes vs. stromo-vascular cells) and their differentiation state  within VAT, or even the adipocyte volume . Further epigenetic and expression studies of the DPP4 gene in cell-specific analyses and with some VAT structural characterization may potentially help to better understand the previous DPP4 differential expression observed in VAT of non-diabetic severely obese men MetS− and MetS+ [3, 4].
This study aimed also to test whether the metabolic and plasma lipid profiles are variable between DPP4%Meth quartiles in VAT. Apart from different plasma triglyceride concentrations between CpG99 %Meth quartiles, similar MetS phenotypes were observed between individual CpGs and combined CpG94-102 %Meth quartiles. In a previous study, positive correlation between plasma HDL-cholesterol levels and CpG94-102 %Meth levels in VAT of non-diabetic severely obese women were observed . However, the actual study did not reveal any association between HDL-cholesterol concentrations and CpG94-102 %Meth levels (Pearson correlation analysis; data not shown) or CpG94-102 %Meth quartiles, even in sex-specific analyses (data not shown). Even though the actual and the previous studies  revealed discordant associations between DPP4 %Meth in VAT and MetS phenotypes, they both underlined an association between the DPP4 gene and plasma lipid profile. This observation is further supported by the presence of significant differences in total-cholesterol concentrations between CpG94-102 %Meth quartiles in this study, which are greater in subjects within Q1 as compared to those within Q2 and Q3, but without any difference between quartiles in DPP4 mRNA abundance in VAT. A previous genetic investigation at the DPP4 locus also revealed that two common single nucleotide polymorphisms were inconsistently associated with the risk of high triglyceride and cholesterol levels in a multi-stage study design conducted in severely obese individuals . These results suggest that both DPP4 %Meth levels (in VAT) and polymorphisms may influence the association between the DPP4 gene and the plasma lipid profile, such as triglycerides and total-cholesterol levels. This hypothesis may partly explain the inconsistency seen in the relation between DPP4 genetic and epigenetic variations with the plasma lipid profile and with DPP4 mRNA levels in VAT. Larger studies which would include both types of variations may increase the chance to observe genetic associations in future studies. Their functional impacts on DPP4 gene expression and function (protein and activity), as well as on DPP4-cleaved protein levels, would then merit further investigations.
The link between DPP4 and the lipid profile contrasts with its well-known role on glucose homeostasis. DPP4 has recently been identified in VAT  and its function within this tissue is unknown. A potential link with plasma triglycerides could be made by the assumption that DPP4 may inhibits the effect of the incretin hormone glucose-dependent insulinotropic polypeptide on lipoprotein lipase synthesis and activity in adipocytes [23, 24], which would favor hypertriglyceridemia due to lower triglyceride hydrolysis as seen in lipoprotein lipase deficiency . The link between DPP4 in VAT and total-cholesterol concentrations is somewhat more difficult to delineate, but some assumptions could be made when considering DPP4 action within other tissues and plasma. Inhibition of DPP4 with vitaglipin has shown to increase the active form of glucagon-like peptide-1 along with a reduction in cholesterol content of chylomicrons after a fat-rich diet in type 2 diabetic subjects , which suggests an indirect link between DPP4 function and cholesterol absorption. Tahara et al. have also recently observed a dose response relationship between DPP4 protein levels and plasma total-cholesterol concentrations. However, in a multiple stepwise regression analysis HDL-cholesterol, but not total-cholesterol, was independently associated with DPP4 plasma levels in a Japanese population . DPP4 is also recognized as a potential peptidase involved in the truncation of the neuropeptide Y (NPY), which would modulate its receptor preference . The common SNP Leu(7)-to-Pro(7) (T1128C) in the NPY gene has previously been associated with higher synthesis and secretion of NPY , and also with higher total- and LDL-cholesterol concentrations in serum of obese Finnish and Dutch subjects . It is unknown how NPY may be related with serum total-cholesterol concentrations in this study , but these results suggest that an indirect link between DPP4 and total-cholesterol may exist via the modulation of NPY function. Furthermore, a recent study conducted by Zhang et al. also demonstrated an association between DPP4 expression and protein levels of two enzymes involved in cholesterol biosynthesis in melanoma cells , but it is unknown if this relation could be observed in other tissues, such as in the liver. Hence, these observations demonstrate few evidences relating the DPP4 enzyme with the lipid profile, which may be further clarified in other independent studies.
Currently, there is concern as to whether less-invasively obtainable human samples (i.e. blood and saliva) would be good surrogates for disease-associated epigenetic biomarkers, such concerns being raised because of the tissue specificity of both epigenetic patterns [31, 32] and disease-associated epigenetic variations . In the actual study, correlations were observed between the %Meth levels of some targeted CpGs in VAT and WBCs of a sub-sample of severely obese women. Although cell types expressing DPP4 mRNA in VAT and WBCs were not investigated in this study, lymphocytes are suspected to be a common link between the two compartments because they are DPP4-expressing cells (i.e. stimulated T-cells, B-cells and natural killer cells) [5–7] and they are present in both obese VAT [34–36] and peripheral WBCs. This hypothesis may thus partly explain the relationship seen between WBCs and VAT %Meth levels.
One study limitation needs to be outlined, which regards the heterogeneity of MetS definition to categorize subjects as being affected or not by obesity-related metabolic complications. To overcome this issue, the selection of non-diabetic individuals in the extremes of the MetS definition has been attempted in this study. However, it may not take into account other metabolic (e.g. insulin resistance) and VAT physiological (e.g. adipocyte size, immune cell infiltration) parameters that may better discriminate those expressing DPP4 at greater levels in their VAT [3, 4]. As previously underlined, further epigenetic and expression studies at the DPP4 locus would be needed to clarify what controls VAT DPP4 expression and whether its methylation levels influence gene expression in a cell-specific fashion.
In conclusion, this study demonstrated that %Meth of CpGs localized within and near the exon 2 of the DPP4 gene in VAT are not associated with MetS status. The actual study also revealed an association between the DPP4 %Meth with plasma total-cholesterol levels in severe obesity, which suggests a link between the DPP4 gene and plasma lipid metabolism. Finally, since the %Meth levels of some of the targeted DPP4 CpGs in VAT correlated with those observed in WBCs, peripheral WBCs may potentially be used as a surrogate measure for DPP4 methylation analysis in further epidemiological studies in relation with the plasma lipid profile.