3. Methodological Pitfalls and Future Challenges

An evaluation of studies included in the present review revealed some similarities but also great differences in study designs and methods. We noted that most studies had a cross-sectional design (65%), while only 13.9% and 20.9% were case-control or prospective studies, respectively. This did not allow us to understand the causal relationship between dietary factors and changes in DNA methylation, especially for studies on diseased patients. For this reason, further large-scale prospective research is encouraged to uncover the complex relationship of diet with DNA methylation among healthy people, but also to understand whether changes in DNA methylation could represent a molecular mechanism underlying the protective or risky effect of dietary factors against cancer and other diseases.

With respect to dietary assessment, almost all studies (90.7%) collected data using Food Frequency Questionnaire (FFQ), which currently represents the gold standard in nutritional epidemiology. Although this method does not preclude errors and inaccuracies of its measurements, other tools for dietary assessment (e.g., weighted records and 24-h recalls) are also prone to misreporting [75]. In fact, the development of novel dietary assessment tools could help to overcome limitations of traditional assessments [76], however, their costs and unresolved intrinsic problems related to self-reporting prevent their extensive use in epidemiological research. Moreover, only a few studies (25.6%) evaluated the complex mixture of foods into a dietary pattern rather than the consumption of specific foods or nutrients. Indeed, nutritional research is generally moving towards the study of dietary patterns using both a priori (e.g., predefined dietary indexes) or a posteriori (i.e., dimensionality reduction techniques) approaches [77,78,79,80,81,82,83,84,85,86].

Differences between studies also regarded methods used for DNA methylation analysis. Indeed, most studies applied methylation-specific PCR and pyrosequencing of bisulfite-treated DNA (37.2% and 34.9%, respectively), while 16.3% used alternative methods (e.g., enzyme-linked immunosorbent assay, Methylight assay, and mass spectrometry methods). The choice between these techniques followed the strides forward in DNA methylation profiling, but also depended on studies’ objectives, sample size and associated costs. Only in recent years, instead, some studies (11.6%) started applying DNA methylation microarrays based on the Illumina Beadchip technology to analyze more than 450,000 CpG sites. Each of the abovementioned techniques presented strengths and weaknesses that were systematically evaluated by Laird [87]. Nowadays, however, advances in DNA methylation microarrays along with reduction in their costs make them a versatile approach to explore the association between dietary factors and DNA methylation status, also in large-size studies [88].

Another issue to be considered when interpreting findings of this kind of study regards the significance of DNA methylation changes observed. While for many genes and genomic sequences it is clear their function and probable association with health and diseases, for others—especially for DMRs identified by epigenome-wide association studies—it is necessary to understand what they exactly entail. Indeed, the biological effect of a particular DNA methylation change is often unknown and unpredictable based on our current knowledge. In support of this, functional studies using in vivo models might help to investigate the significance of DNA methylation changes and to solve controversies on the effect of specific nutrients and foods. For instance, examining studies included in the current review, it was not clear if folate intake was positively or negatively associated with DNA methylation. In fact, it is likely that folates—as well as other methyl donors—are important for DNA methylation maintenance, while other nutrients and/or bioactive foods exert their influence directly on enzymes involved in the methylation process. In fact, several reviews properly summarized evidence of this influence coming from in vivo studies [89,90]. Thus, in addition to investigate the association between dietary factors and DNA methylation, a deeper insight into the consequences of epigenetic changes on physiological and pathological events should be recommended. Studies on LINE-1 methylation represented a prime example of this need. Indeed, while methylation status of LINE-1 sequences could be considered a surrogate marker of global DNA methylation [13], the impacts of both hypo-and hyper-methylation should be considered. Indeed, aberrant methylation of LINE-1 sequences was previously associated with aging, cancer, neurodegeneration, metabolic disorders, and obesity by affecting genome and chromosome stability [13,14,16,24,91,92,93,94,95]. Moreover, since DNA methylation is highly cell-and tissue-specific, differences in tissue samples or blood cell composition should be addressed when comparing results from different studies. Last but not least, DNA methylation is highly sensitive to environmental exposure-in its broadest and most comprehensive meaning–and genetic factors [17,20,56,96,97,98,99,100,101,102]. For this reason, future studies should comprehensively evaluate not only dietary factors and their relationship with DNA methylation, but also the effects of other lifestyles, social determinants, environmental exposures, and genetic variants that might affect the methylation process. Finally, investments are necessary to promote studies able to assess the potential applications of genome knowledge into clinical practice to improve public health [103].