The continued increase in the incidence of type 2 diabetes (T2D) has elicited the need to investigate in more depth and breadth the risk factors that contribute to this disease in order to have a more comprehensive vision of the process to illuminate the path towards more precise and effective preventive interventions [1–3]. In addition, and given the impact of T2D on cardiovascular diseases (CVD) [2, 4–8] we need to gain further knowledge about the factors driving the connection between T2D and CVD. In this regard, there is increasing evidence linking chronodisruption with metabolic disorders [9–12].

All living things have acquired, in the course of evolution, an internal circadian timing system so as to adapt to the rhythmically occurring daily changes in their environment [13, 14]. In humans, this circadian timing coordinates virtually all physiological processes encompassing the states of sleep and wakefulness, endocrine functions and cardiovascular activity [15, 17]. It has been shown that several aspects of the cardiovascular physiopathology and incidence of CVD events, such as myocardial infarction, ischemic and hemorrhagic stroke, have diurnal variation, peaking in the early morning hours [18–23]. Moreover, a higher risk of obesity, metabolic alterations and T2D has been found in shift workers [24–27] and adverse cardiometabolic effects (including postprandial glucose in a range typical of a prediabetes state) have been detected in subjects who underwent forced circadian misalignment [28].

The endogenous circadian timing system includes the suprachiasmatic nucleus (SNC) in the hypothalamus, as well as peripheral oscillators in the different organs [29]. The core molecular clock is composed of a transcription-translation feedback loop that oscillates with 24-h rhythmicity [30]. Several core clock genes are involved in the regulation of this system [14, 30]. Among them, the CLOCK (Circadian Locomotor Output Cycles Kaput) gene, a transcription factors from the positive limb of the molecular clock that forms a complex with BMAL1 (Brain and Muscle ARNT-Like 1), is one of the most relevant [14, 30, 31]. In murine models, mutations in the clock gene were first associated with glucose homeostasis [32] and later with obesity, hyperglycemia and hyperinsulinemia [33]. In humans, various CLOCK single nucleotide polymorphisms (SNPs) (i.e. rs1464490, rs3749474, rs4864584, rs4580704 and rs18012602) have also been associated with obesity [34–36], hyperglycemia and greater prevalence of T2D in cross-sectional studies [36, 37]. However, there has been no longitudinal study to analyze the influence of CLOCK SNPs on T2D incidence. The CLOCK-rs4580704 C > G, a tag SNP in linkage disequilibrium with other CLOCK SNPs, is one of the most relevant SNPs for these associations [35, 36]. In our previous work in the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) population, we observed that carriers of the minor allele (G) for the CLOCK-rs4580704 SNP presented lower weight, fasting-insulin and hyperglycemia risk [36].

Moreover, we detected a gene-diet interaction between the CLOCK-rs4580704 SNP and monounsaturated fatty acids (MUFA) intake in determining fasting glucose and insulin resistance in GOLDN [36]. When the MUFA intake was low, no differences were found for these parameters. However, when MUFA intake was high, minor allele carriers had significantly lower levels. We also suggested that this SNP could be relevant in modulating CVD risk [36]. However, the latter connection has not been specifically investigated. Bearing in mind that there may be a greater alteration in circadian rhythms in T2D [9–11, 16], we hypothesized that the effect of variation in clock genes on CVD risk might be greater in T2D subjects. Our aims, therefore, were as follows: (1) to analyze the association between the tag SNP CLOCK-rs4580704 and the incidence of T2D; (2) to analyze the association between the CLOCK-rs4580704 SNP and incidence of clinical CVD events depending on T2D status and; (3) to study whether a dietary intervention with the Mediterranean diet (MedDiet), very high in MUFA, within the framework of the PREDIMED trial, modulates these associations.

The present study was conducted within the framework of the PREvención con DIeta MEditerránea (PREDIMED) trial, whose design has been described in detail elsewhere [38]. Briefly, the PREDIMED study is a large, multicenter, randomized and controlled clinical trial aimed at assessing the effects of the Mediterranean diet (MedDiet) on primary cardiovascular prevention [38, 39]. This study was registered at controlled-trials.com (http://www.controlledtrials.com/ISRCTN35739639). Participants were randomized to one of three interventions: a MedDiet supplemented with extra virgin olive oil (EVOO), a MedDiet supplemented with mixed nuts, or advice on a low-fat diet (control diet). Here we included 7098 participants from whom DNA was isolated, the polymorphism determined, and who had valid data for the main clinical variables analyzed. The completion date of this study was December 2010 and the total number of randomized subjects was 7447. The 7098 participants included in the present study did not differ in the main characteristics from those of the total cohort. Additional file 1: Figure S1 shows the CONSORT flowchart of the trial for the primary outcomes. Details of the PREDIMED trial including sample size calculations and interim analysis have been fully described elsewhere [38, 39]. Briefly, From October 2003 physicians in Primary Care Centers selected high cardiovascular risk participants. Eligible were community-dwelling persons (55–80 years for men; 60–80 years for women) who met at least one of two criteria: T2D or 3 or more cardiovascular risk factors as previously detailed [38, 39]. Taking into account that no significant differences in CVD incidence were found between the MedDiet groups [39], data were analyzed pooling together the MedDiet intervention groups versus the control group.

A detailed description of the nutritional interventions has been provided elsewhere [39]. The Institutional Review Board/Ethics Committee of each participating center approved the study protocol. All participants provided written informed consent. Participants were followed up for a median of 4.8 years (interquartile range 2.8–5.8 years). Incidence of CVD as primary outcome and incidence of T2D in non-T2D participants (as secondary study outcome) were assessed.

In this longitudinal study, undertaken on 7098 participants of the PREDIMED trial, we have been able to demonstrate the relevance of the CLOCK-rs4580704 SNP in the incidence of T2D and related processes, which finally lead to a higher incidence of CVD in T2D subjects, strengthening the connection between genetic variants in core clock genes, metabolic alterations and CVD risk. In agreement with our results, previous studies in humans analyzing this or other SNPs in linkage disequilibrium (rs1801260, rs3736544, rs4864548 and rs3749474) had already reported that the variant allele is associated with lower hyperglycemia and prevalence of T2D [36, 37]. However, given that cross-sectional studies may be confounded by other factors related with obesity and other phenotypes concurrently associated with T2D, our longitudinal study provides a higher evidence level for the contribution of the CLOCK gene to T2D incidence. Moreover, our results are supported by observations on murine models, given that mice with mutations resulting in under-expression of the clock gene presented obesity and hyperglycemia [32, 33]. This dual phenotype has been a limiting factor in establishing causality as it was not possible to tease apart whether the greater risk of diabetes in those mice [33] was directly related to the functionality of the clock gene or was mediated by obesity. Our study and previous epidemiological studies [34–36, 44] show that the CLOCK-rs4580704 SNP was also significantly associated with body-weight. However, through multivariable adjustment, we have been able to show that the associations of the CLOCK-rs4580704 SNP with fasting glucose and T2D incidence remain statistically significant after adjustment for BMI, supporting the notion that the observed effect may be additional to that of the association with obesity.

Furthermore, we have shown that the association between the CLOCK-rs4580704 SNP and T2D incidence is modulated by diet. Thus, in the stratified analysis by dietary intervention (MedDiet vs control diet), the protective association between the minor G-allele and T2D risk was enhanced in subjects allocated to the MedDiet intervention group being highly statistically significant. However, no significant protective effect in carriers of the G-allele was observed in subjects in the control (low-fat diet) group. Although, in our study, the gene-diet interaction term between the SNP and dietary intervention on T2D incidence was borderline significant (P = 0.052) at P < 0.05, taking into account that we can consider a P < 0.1 for the significance level of the interaction term as suggestive of interaction when some criteria of reliability are accomplished [45], our study fulfill these criteria to avoid false negative results and we considered this significance level (P < 0.1) to conduct the stratified analyses by MedDiet intervention groups. These criteria require that [45] the subgroup (dietary intervention) variable was measured at baseline and randomly distributed; we had an a priori hypothesis of the higher protective effect of the variant allele in the MedDiet intervention group (high MUFA); the direction of the subgroup effect was specified; a small number of hypotheses were tested and the size of the interaction effect was large. Our hypothesis of a higher protective effect against T2D of the variant allele (G) was based on our previous results obtained in the GOLDN study [36]. Therefore, our results support that the association between the CLOCK-rs4580704 SNP and T2D can be modulated by dietary intake, having the MedDiet pattern an increasing protective effect of the G-allele against the disease.

The dietary modulation of the CLOCK gene expression effects on metabolic phenotypes has been previously shown is animal models [11, 32, 33]. Moreover, in our previous study in a US white population we specifically reported a gene-diet interaction between the CLOCK-rs4580704 SNP and the contribution of MUFAs in the diet in determining fasting glucose concentrations and insulin resistance (HOMA-IR). Hence, when the MUFA intake (% of energy) was below the median (<13.2 %), no significant differences were found for plasma glucose concentrations and HOMA-IR between carriers and non-carriers of the G-allele. However, when MUFA intake was high (≥13.2 %), G-allele carriers had significantly lower plasma glucose concentrations and lower HOMA-IR values than did non-carriers. The MedDiet is precisely characterized by a high contribution of MUFAs (about 20 %), coming mainly from olive oil. In the intervention with MedDiet group, the MUFA content in the diet was significantly higher than in the low fat control group, as we described previously [39]. This high MUFA intake in the MedDiet intervention group could be one of the factors that may contribute to explain the higher protective effect against T2D incidence observed for G-carriers in the MedDiet intervention group but not in the control group in in agreement with our previous results obtained in the cross-sectional study carried out in the US population [36]. Furthermore, we can even hypothesize that the interaction effect between groups could be even higher and more significant if more differences in MUFA intake were reached (being a Mediterranean population, MUFA intake in the PREDIMED study is generally higher than in the US population). Although the mechanisms explaining the observed gene-diet modulation on T2D remain unknown, recent studies have suggested that DNA methylation may be an important mechanism to drive circadian clock plasticity [46, 47]. In support of this argument, we have shown, in Mediterranean subjects [48], that the percentage of methylation of certain CpG islands in the CLOCK gene were significantly associated with MUFA and PUFA intakes, adding preliminary evidence to the potential epigenetic dietary modulation, although this requires further additional work to be confirmed.

The polymorphism in the CLOCK-rs4580704 gene is located in an intron and its function is still unknown. However, it involves a tag SNP that is in linkage disequilibrium with other SNPs in the 3’UTR region that could be modulated by microRNAs. In our previous work focusing on this SNP [36], a specific bioinformatics analysis for CLOCK-rs4580704 SNP allowed us to affirm that SNP rs4580704 was predicted to produce an allele-specific CREM (cAMP responsive element modulator) binding site (C allele on forward strand binds CREM, G allele does not). In this bioinformatics analysis we used MAPPER [49] to identify potential allele-specific transcription factor binding sites and RNAfold [50] within the Vienna RNA package as previously detailed [36]. CREM has been shown to be responsible for circadian expression in the mouse of many genes that could also be implicated in T2D and CVD risk [51–53]. Either by itself, or as an indicator of another functional SNP, the variant G-allele of the CLOCK-rs4580704 SNP in our study is associated with lower fasting glucose concentrations in non-diabetics, lower T2D incidence and less CVD risk in T2D subjects. Bearing in mind that in murine clock mutant models [11, 32, 33], the reduction in clock gene expression is precisely associated with an opposite phenotype characterized by hyperglycemia, hyperinsulinemia and metabolic syndrome and considering that β cell clock gene ablation in mice caused severe glucose intolerance [10], the minor CLOCK-rs4580704 G-allele in humans would be associated with a higher conservation of gene functionality, i.e. in this case it would be the major allele which would provoke a situation of higher vulnerability to circadian adaptation and higher likelihood of chronodisruption. Carriers of the variant G-allele would be subjects that could better adapt to circadian disruption (i.e., deficient sleep, inadequate time of meals, shift work), or who maintain their rhythmicity of its processes with less alterations, presenting greater protection against obesity, T2D and future CVD.

Other SNPs in core circadian genes, such as BMAL1, that forms a complex with CLOCK, have been associated with T2D prevalence in some human studies and in murine models [11, 32, 54, 55] reinforcing the contribution of the circadian system to the etiology of T2D; however, no human study has reported an association between SNPs in core circadian genes and incidence of stroke as demonstrated in the current study. It is interesting to note that the association of the CLOCK-rs4580704 SNPs with stroke was only present in T2D subjects. This is compatible with a higher chronodisruption in these subjects, resulting in a higher CVD risk in a relatively short follow-up period (median follow-up approx. 5 years). Circadian variations in relevant risk factors for CVD such as insulin sensitivity, blood pressure, renal function, heart rate, platelet aggregability, fibrinolytic markers and levels, hormone concentrations etc. [9, 16, 17, 56–58] may explain the morning onset of myocardial infarction, stroke and other CVD clinical events [17, 18, 23]. These circadian variations may be more important in T2D patients and their dysregulation contributing to a higher CVD (mainly stroke) incidence in more susceptible individuals such as CLOCK-rs4580704 CC homozygous subjects (for whom our results suggest a lower flexibility). Although it is well known that T2D subjects have higher CVD risk than non-diabetic subjects [7, 8], a great heterogeneity has been described [59]. This heterogeneity in CVD risk is only partially understood but is a key consideration for our understanding of the nexus of T2D and CVD and for the development of individualized CVD risk reduction strategies. According to our results, we suggest circadian alterations due to functional genetic variants in core clock genes such as CLOCK as another factor that may contribute to this heterogeneity and higher risk of CVD (mainly, stroke) in those subjects. As measuring the processes of circadian dysregulation is complex, the CLOCK-rs4580704 SNP can be considered as a proxy for higher chronodisruption risk in CC subjects. Thus, this SNP will be of interest as an instrumental variant in future Mendelian Randomization in other cohorts analyzing CVD risk in T2D subjects.

In conclusion, we have described for the first time an association between the CLOCK-rs4580704 SNP and incidence of T2D. These results are in agreement with previous knowledge obtained from animal and human cross-sectional studies. In addition we have found that this association with T2D can be modulated by dietary intake in the framework of a randomized controlled trial, with the MedDiet increasing the protective effects of the G-allele against T2D. Moreover, we have extended our finding also showing a novel association between the G-allele and protection against stroke in T2D subjects. Although our results highlight the contribution of the circadian system both on the incidence of T2D and CVD risk in T2D subjects it is presumable that the functional effects of CLOCK gene could be either circadian rhythm-dependent or independent. In the present study, as there is no assessments regarding the circadian rhythm of the study subjects, the suggested link between the CLOCK SNP and chronodisruption remains speculative, and more studies are needed to reveal the mechanisms behind these epidemiological associations.