The major strength of this

The major strength of this study is the large sample size which increases the statistical power to detect the likely subtle differences in peripheral methylation observed in psychiatric disorders (Olsson et al., 2010). Larger studies help provide more robust estimates of true associations, as small underpowered studies are prone to false positives, or type 1 errors (Ioannidis, 2005). Another strength of our study is the simultaneous investigation of both ACE promoter methylation and genetic variation, and their association with depression. Other strengths of our study include the diagnosis of depression which was based on a clinical assessment (MINI, DSM-IV) of MDD, as well as a validated questionnaire of severe depressive symptoms (CES-D≥16). This study was conducted on an older French Caucasian population and while applying validated assessments, did not take into account age at onset of first depressive disorder. The extent to which these results can be applied to earlier onset depression with possible different etiologies in younger age groups remains to be determined (Blazer, 2003). DNA methylation patterns vary with age and ethnicity (Zhang et al., 2011, Johnson et al., 2012). The association between methylation and depression can be confounded by many lifestyle and health factors that affect methylation and depression status and we controlled for these in our analysis, however further analysis is still needed. Of particular importance are functional impairment, a characteristic of depressive disorders as listed in the DSM criteria (American Psychiatric Association, 1994); antidepressant use, which has been shown to modify methylation levels (Menke and Binder, 2014); and ACE inhibitors, which may have antidepressant-life effects (Vuckovic et al., 1991). Given our cross-sectional analysis, our findings may result from an association with one or more of these confounding factors. While the present study was able to consider a much wider range of potential confounders than the previous study, the potential for residual confounding from factors where information was not available or collected, remains possible. For example, factors known to influence DNA methylation include stress (beyond the Fomepizole receptor marker) (Booij et al., 2013), methyl donors and nutrition (Anderson et al., 2012). There may also be unknown factors that influence DNA methylation and/or depression that have yet to be determined. Finally, no adjustment for multiple testing was made, and none of the associations would remain significant at the stringent Bonferroni corrected p value of 0.00045 (accounting for 16 CpG units and 7 SNPs, thus 112). However, such an adjustment would be overly conservative, as Bonferroni correction assumes independence of tests, and the tests we performed were not independent. Three of the SNPs for example, are in high linkage disequilibrium as stated in the text (page 7, section 2.3) and methylation levels at all but one of the CpG units was significantly correlated with other CpG units (p<0.05). Our study focuses on examining potential peripheral biomarkers of depression. Depression is a brain-based disorder, however it is becoming increasingly recognised as a systemic disease, with physiological changes observed in peripheral tissue (O'Donovan et al., 2010, Thomson et al., 2014). DNA methylation levels between blood and post-mortem brain tissues may be correlated, at least for some genes (Horvath et al., 2012). Furthermore, to be a useful peripheral biomarker, blood methylation levels do not need to reflect brain methylation levels or underlying mechanisms of depression, as long as they can be shown to be able to consistently distinguish depressed from non-depressed individuals across multiple independent studies. Another limitation of our study is the candidate gene approach although there is evidence for the involvement of ACE in depression. This candidate approach requires a strong a priori hypothesis, and being a complex disorder, may however, only have captured a small part of depression pathogenesis. For example, ACE levels or methylation may only be relevant to manifestations of depression that involve dysregulation of the HPA axis. Epigenome-wide association studies (EWAS) could be an alternative approach, enabling interrogation of large proportions of genomic locations without an a priori hypothesis. EWAS studies have the added advantage of being able to adjust for cellular heterogeneity through computational estimation of cell composition, which may have distinct methylation profiles (Jaffe and Irizarry, 2014). This was not achievable in our study, potentially confounding our methylation analyses and may account for the lack of associations observed between depression and ACE methylation. Unfortunately costs remain very high for large cohort EWAS studies such as our study cohort (Rakyan et al., 2011).