Heritability of depression

For many years twin and family studies have highlighted the importance of genetic and environmental factors in the mediation of vulnerability to depression (Cohen-Woods et al. Reference Cohen-Woods, Craig and McGuffin2013). In the first meta-analysis of epidemiological studies in depression, Sullivan et al. (Reference Sullivan, Neale and Kendler2000) compared susceptibility in monozygotic and dizygotic twins. The majority of the variance in liability was attributed to environmental effects specific to the individual (63%), whilst genetic effects accounted for 37%. Subsequent studies have produced similar results (Kendler et al. Reference Kendler, Gatz, Gardner and Pedersen2006; Franz et al. Reference Franz, Lyons, O'Brien, Panizzon, Kim, Bhat, Grant, Toomey, Eisen, Xian and Kremen2011; Nivard et al. Reference Nivard, Dolan, Kendler, Kan, Willemsen, van Beijsterveldt, Lindauer, van Beek, Geels, Bartels, Middeldorp and Boomsma2015). This genetic contribution has been examined using linkage and association studies. Gene linkage studies for depression, as well as for other common complex disorders, have been perceived by some to be of only limited success (McGuffin et al. Reference McGuffin, Cohen and Knight2007; Nair & Howard, Reference Nair and Howard2013); hence the focus on GWAS. However, even for simple traits, genetic variants identified by GWAS are rarely shown to account for more than 20% of the heritability (Roberts et al. Reference Roberts, Vogelstein, Parmigiani, Kinzler, Vogelstein and Velculescu2012; Wood et al. Reference Wood, Esko, Yang, Vedantam, Pers, Gustafsson, Chu, Estrada, Luan, Kutalik, Amin, Buchkovich, Croteau-Chonka, Day, Duan, Fall, Fehrmann, Ferreira, Jackson, Karjalainen, Lo, Locke, Magi, Mihailov, Porcu, Randall, Scherag, Vinkhuyzen, Westra, Winkler, Workalemahu, Zhao, Absher, Albrecht, Anderson, Baron, Beekman, Demirkan, Ehret, Feenstra, Feitosa, Fischer, Fraser, Goel, Gong, Justice, Kanoni, Kleber, Kristiansson, Lim, Lotay, Lui, Mangino, Mateo Leach, Medina-Gomez, Nalls, Nyholt, Palmer, Pasko, Pechlivanis, Prokopenko, Ried, Ripke, Shungin, Stancakova, Strawbridge, Sung, Tanaka, Teumer, Trompet, van der Laan, van Setten, Van Vliet-Ostaptchouk, Wang, Yengo, Zhang, Afzal, Arnlov, Arscott, Bandinelli, Barrett, Bellis, Bennett, Berne, Bluher, Bolton, Bottcher, Boyd, Bruinenberg, Buckley, Buyske, Caspersen, Chines, Clarke, Claudi-Boehm, Cooper, Daw, De Jong, Deelen, Delgado, Denny, Dhonukshe-Rutten, Dimitriou, Doney, Dorr, Eklund, Eury, Folkersen, Garcia, Geller, Giedraitis, Go, Grallert, Grammer, Grassler, Gronberg, de Groot, Groves, Haessler, Hall, Haller, Hallmans, Hannemann, Hartman, Hassinen, Hayward, Heard-Costa, Helmer, Hemani, Henders, Hillege, Hlatky, Hoffmann, Hoffmann, Holmen, Houwing-Duistermaat, Illig, Isaacs, James, Jeff, Johansen, Johansson, Jolley, Juliusdottir, Junttila, Kho, Kinnunen, Klopp, Kocher, Kratzer, Lichtner, Lind, Lindstrom, Lobbens, Lorentzon, Lu, Lyssenko, Magnusson, Mahajan, Maillard, McArdle, McKenzie, McLachlan, McLaren, Menni, Merger, Milani, Moayyeri, Monda, Morken, Muller, Muller-Nurasyid, Musk, Narisu, Nauck, Nolte, Nothen, Oozageer, Pilz, Rayner, Renstrom, Robertson, Rose, Roussel, Sanna, Scharnagl, Scholtens, Schumacher, Schunkert, Scott, Sehmi, Seufferlein, Shi, Silventoinen, Smit, Smith, Smolonska, Stanton, Stirrups, Stott, Stringham, Sundstrom, Swertz, Syvanen, Tayo, Thorleifsson, Tyrer, van Dijk, van Schoor, van der Velde, van Heemst, van Oort, Vermeulen, Verweij, Vonk, Waite, Waldenberger, Wennauer, Wilkens, Willenborg, Wilsgaard, Wojczynski, Wong, Wright, Zhang, Arveiler, Bakker, Beilby, Bergman, Bergmann, Biffar, Blangero, Boomsma, Bornstein, Bovet, Brambilla, Brown, Campbell, Caulfield, Chakravarti, Collins, Collins, Crawford, Cupples, Danesh, de Faire, den Ruijter, Erbel, Erdmann, Eriksson, Farrall, Ferrannini, Ferrieres, Ford, Forouhi, Forrester, Gansevoort, Gejman, Gieger, Golay, Gottesman, Gudnason, Gyllensten, Haas, Hall, Harris, Hattersley, Heath, Hengstenberg, Hicks, Hindorff, Hingorani, Hofman, Hovingh, Humphries, Hunt, Hypponen, Jacobs, Jarvelin, Jousilahti, Jula, Kaprio, Kastelein, Kayser, Kee, Keinanen-Kiukaanniemi, Kiemeney, Kooner, Kooperberg, Koskinen, Kovacs, Kraja, Kumari, Kuusisto, Lakka, Langenberg, Le Marchand, Lehtimaki, Lupoli, Madden, Mannisto, Manunta, Marette, Matise, McKnight, Meitinger, Moll, Montgomery, Morris, Morris, Murray, Nelis, Ohlsson, Oldehinkel, Ong, Ouwehand, Pasterkamp, Peters, Pramstaller, Price, Qi, Raitakari, Rankinen, Rao, Rice, Ritchie, Rudan, Salomaa, Samani, Saramies, Sarzynski, Schwarz, Sebert, Sever, Shuldiner, Sinisalo, Steinthorsdottir, Stolk, Tardif, Tonjes, Tremblay, Tremoli, Virtamo, Vohl, Electronic Medical, Genomics, Consortium, Consortium, LifeLines Cohort, Amouyel, Asselbergs, Assimes, Bochud, Boehm, Boerwinkle, Bottinger, Bouchard, Cauchi, Chambers, Chanock, Cooper, de Bakker, Dedoussis, Ferrucci, Franks, Froguel, Groop, Haiman, Hamsten, Hayes, Hui, Hunter, Hveem, Jukema, Kaplan, Kivimaki, Kuh, Laakso, Liu, Martin, Marz, Melbye, Moebus, Munroe, Njolstad, Oostra, Palmer, Pedersen, Perola, Perusse, Peters, Powell, Power, Quertermous, Rauramaa, Reinmaa, Ridker, Rivadeneira, Rotter, Saaristo, Saleheen, Schlessinger, Slagboom, Snieder, Spector, Strauch, Stumvoll, Tuomilehto, Uusitupa, van der Harst, Volzke, Walker, Wareham, Watkins, Wichmann, Wilson, Zanen, Deloukas, Heid, Lindgren, Mohlke, Speliotes, Thorsteinsdottir, Barroso, Fox, North, Strachan, Beckmann, Berndt, Boehnke, Borecki, McCarthy, Metspalu, Stefansson, Uitterlinden, van Duijn, Franke, Willer, Price, Lettre, Loos, Weedon, Ingelsson, O'Connell, Abecasis, Chasman, Goddard, Visscher, Hirschhorn and Frayling2014). GWAS has had limited success for depression in finding significant associations with individual genetic variations and there has been no evidence for a recessive model (Cohen-Woods et al. Reference Cohen-Woods, Craig and McGuffin2013; Chang et al. Reference Chang, Glymour, Walter, Liang, Koenen, Tchetgen, Cornelis, Kawachi, Rimm and Kubzansky2014; Flint & Kendler, Reference Flint and Kendler2014; Levinson et al. Reference Levinson, Mostafavi, Milaneschi, Rivera, Ripke, Wray and Sullivan2014; Power et al. Reference Power, Keller, Ripke, Abdellaoui, Wray, Sullivan and Breen2014; Schneider et al. Reference Schneider, Engel, Fasching, Haberle, Binder, Voigt, Grimm, Faschingbauer, Eichler, Dammer, Rebhan, Amann, Raabe, Goecke, Quast, Beckmann, Kornhuber, Seifert and Burghaus2014). Even when technical and statistical aspects of GWAS have been taken into account (Gusev et al. Reference Gusev, Bhatia, Zaitlen, Vilhjalmsson, Diogo, Stahl, Gregersen, Worthington, Klareskog, Raychaudhuri, Plenge, Pasaniuc and Price2013) the extent of this disparity between expected and verified genetic components in depression has remained considerable (Castillo-Fernandez et al. Reference Castillo-Fernandez, Spector and Bell2014) and family history continues to be the most effective method of predicting risk (Maher, Reference Maher2008). Nevertheless, trying to dovetail epidemiological studies which suggest a considerable genetic component for depression with genetic studies providing a dearth of single nucleotide polymorphism (SNP) associations remains a significant problem.

The hypothalamic–pituitary–adrenal (HPA) axis

One major endocrinological finding in depression is of a dysregulation of the HPA axis – the foremost neuroendocrine stress response system. Stress has profound effects upon a broad range of physiological systems and is an established trigger for mental illness (Meaney, Reference Meaney2001; Gallagher et al. Reference Gallagher, Watson, Smith, Young and Ferrier2007; Binder et al. Reference Binder, Bradley, Liu, Epstein, Deveau, Mercer, Tang, Gillespie, Heim, Nemeroff, Schwartz, Cubells and Ressler2008; Turner et al. Reference Turner, Alt, Cao, Vernocchi, Trifonova, Battello and Muller2010; Klengel et al. Reference Klengel, Pape, Binder and Mehta2014). Dysregulation of the HPA axis in unipolar depression has been consistently reported since the 1960s (Gibbons, Reference Gibbons1964; O'Toole et al. Reference O'Toole, Sekula and Rubin1997; McAllister-Williams et al. Reference McAllister-Williams, Ferrier and Young1998; Holsboer, Reference Holsboer2000; Young, Reference Young2004; Moser et al. Reference Moser, Molitor, Kumsta, Tatschner, Riederer and Meyer2007; Pariante & Lightman, Reference Pariante and Lightman2008). Neuroendocrine studies have shown increased basal and/or activated levels of the HPA axis hormones – corticotropin-releasing hormone (CRH), vasopressin (arginine vasopressin; AVP), adrenocorticotropic hormone (ACTH) and cortisol – in plasma, saliva and cerebrospinal fluid (Herbert, Reference Herbert2013; Belvederi Murri et al. Reference Belvederi Murri, Pariante, Mondelli, Masotti, Atti, Mellacqua, Antonioli, Ghio, Menchetti, Zanetidou, Innamorati and Amore2014). Structural changes have also been seen in post-mortem and imaging studies of depressed patients including increased numbers of CRH-secreting neurones in the hypothalamus (Raadsheer et al. Reference Raadsheer, Hoogendijk, Stam, Tilders and Swaab1994) and enlarged pituitary and adrenal gland volumes (Kessing et al. Reference Kessing, Willer and Knorr2011). Meanwhile non-suppression in the dexamethasone/CRH test has been associated with inferior treatment response and increased relapse rates in depression (Aubry et al. Reference Aubry, Gervasoni, Osiek, Perret, Rossier, Bertschy and Bondolfi2007; Ising et al. Reference Ising, Horstmann, Kloiber, Lucae, Binder, Kern, Kunzel, Pfennig, Uhr and Holsboer2007; Medina et al. Reference Medina, Seasholtz, Sharma, Burke, Bunney, Myers, Schatzberg, Akil and Watson2013).

The HPA axis is regulated by negative feedback loops incorporating glucocorticoids (GCs) and GC receptors (GRs) (Alt et al. Reference Alt, Turner, Klok, Meijer, Lakke, Derijk and Muller2010). Thus, studies exploring the mechanism underlying the HPA axis dysregulation that has been documented in depression have focused upon abnormal GR expression and function (Pariante, Reference Pariante2006, Reference Pariante2009; Cowen, Reference Cowen2010; Anacker et al. Reference Anacker, Zunszain, Carvalho and Pariante2011). Indeed, preclinical and clinical investigations have implicated a significant role for GR abnormalities in depression. Using knock-out mice Boyle et al. (Reference Boyle, Brewer, Funatsu, Wozniak, Tsien, Izumi and Muglia2005) demonstrated that reduced GR function led to disrupted negative feedback inhibition of the HPA axis and depression-like behaviour. Depressed patients have shown reduced peripheral GR levels and increased 24-h cortisol levels (Yehuda et al. Reference Yehuda, Boisoneau, Mason and Giller1993). In line with this, decreased GR mRNA has been demonstrated in post-mortem frontal cortices from depressed patients (Webster et al. Reference Webster, Knable, O'Grady, Orthmann and Weickert2002), although a more recent study has found increased GR expression in amygdala samples from depressed patients (Wang et al. Reference Wang, Verweij, Krugers, Joels, Swaab and Lucassen2014). These findings have been bolstered by studies implicating improved GR function and increased GR expression in the mechanism of action of certain antidepressants (Pariante & Miller, Reference Pariante and Miller2001).

Early life adversity, depression and the HPA axis

The impact of maternal care and early life adversity has been investigated across several species. Rodents temporarily separated from their mothers in the first few months of life, and those whose mothers provide low levels of licking and grooming (LG) and arched back nursing (ABN) care (Meaney, Reference Meaney2001; Daskalakis et al. Reference Daskalakis, Bagot, Parker, Vinkers and de Kloet2013), have been shown to exhibit depression-like phenotypes (Caldji et al. Reference Caldji, Diorio and Meaney2000; Murgatroyd et al. Reference Murgatroyd, Patchev, Wu, Micale, Bockmuhl, Fischer, Holsboer, Wotjak, Almeida and Spengler2009). Early adversity has also been associated with abnormal HPA axis function (Fish et al. Reference Fish, Shahrokh, Bagot, Caldji, Bredy, Szyf and Meaney2004; Archer et al. Reference Archer, Oscar-Berman, Blum and Gold2013). Liu et al. (Reference Liu, Diorio, Tannenbaum, Caldji, Francis, Freedman, Sharma, Pearson, Plotsky and Meaney1997) demonstrated that high-LG-ABN maternal conditions, triggered by brief human handling, were associated with lower levels of plasma ACTH and corticosterone in response to stress in rat offspring up to 100 days old. These offspring also had greater GC sensitivity, increased levels of hippocampal GR mRNA and lower levels of hypothalamic CRH mRNA. In an associated study, Francis et al. (Reference Francis, Diorio, Liu and Meaney1999) showed that rat pups cross-fostered from low- to high-LG-ABN mothers had dampened behavioural responses to stress. More interesting still, these offspring emulated their high-LG-ABN foster mothers when caring for their own offspring, while low-LG-ABN foster mothers rearing pups from high-LG-ABN biological mothers produced offspring with heightened stress responses and low-LG-ABN maternal behaviour. Further studies (Murgatroyd & Nephew, Reference Murgatroyd and Nephew2013; Murgatroyd et al. Reference Murgatroyd, Quinn, Sharp, Pickles and Hill2015) have been able to show that exposing rat mothers to chronic stress during lactation leads to reduced levels of maternal care as well as altered neuropeptide regulation and GR expression. Moreover, pups whose mothers were exposed to chronic stress tend themselves to exhibit reduced maternal care in adulthood. These results supported prior studies (Denenberg & Rosenberg, Reference Denenberg and Rosenberg1967; Danchin et al. Reference Danchin, Charmantier, Champagne, Mesoudi, Pujol and Blanchet2011) demonstrating that differences in gene expression could be passed from one generation to the next by non-genomic means. Further work has been carried out on non-genomic transgenerational inheritance and it continues to attract much attention (Bohacek et al. Reference Bohacek, Gapp, Saab and Mansuy2013; Crean et al. Reference Crean, Kopps and Bonduriansky2014; Babenko et al. Reference Babenko, Kovalchuk and Metz2015).

In humans early life adversity is acknowledged as a significant risk factor for many psychiatric and non-psychiatric illnesses (Rutter, Reference Rutter1985; Meaney, Reference Meaney2001; Maniglio, Reference Maniglio2009). Childhood maltreatment – incorporating physical, emotional and sexual abuse and physical and emotional neglect – is a significant source of early life adversity in human populations. In Britain, childhood maltreatment has been estimated to occur in 15–25% of the population (May-Chahal & Cawson, Reference May-Chahal and Cawson2005; Radford et al. Reference Radford, Corral, Bradley and Fisher2013) with comparable figures quoted internationally (Ishida et al. Reference Ishida, Klevens, Rivera-Garcia and Mirabal2013; Barbosa et al. Reference Barbosa, Quevedo, da Silva Gdel, Jansen, Pinheiro, Branco, Lara, Oses and da Silva2014; Finkelhor et al. Reference Finkelhor, Shattuck, Turner and Hamby2014; Wildeman et al. Reference Wildeman, Emanuel, Leventhal, Putnam-Hornstein, Waldfogel and Lee2014). Those who are subjected to childhood maltreatment are thought to be at greater risk of later life depression (Klengel et al. Reference Klengel, Pape, Binder and Mehta2014). More specifically, Bifulco et al. (Reference Bifulco, Brown and Adler1991) demonstrated an increased risk of depression in women who had been abused as children and other subsequent studies have supported these findings (Kendler et al. Reference Kendler, Bulik, Silberg, Hettema, Myers and Prescott2000; Widom et al. Reference Widom, DuMont and Czaja2007; Alt et al. Reference Alt, Turner, Klok, Meijer, Lakke, Derijk and Muller2010; Bale et al. Reference Bale, Baram, Brown, Goldstein, Insel, McCarthy, Nemeroff, Reyes, Simerly, Susser and Nestler2010; Heim et al. Reference Heim, Shugart, Craighead and Nemeroff2010). Such individuals can exhibit persistent neuroendocrine and anatomical changes including: GC insensitivity, increased central CRH activity, immune up-regulation and reduced hippocampal volume (Heim & Nemeroff, Reference Heim and Nemeroff2001; Heim et al. Reference Heim, Newport, Mletzko, Miller and Nemeroff2008; Hornung & Heim, Reference Hornung and Heim2014). Similar abnormalities have been seen in parentally bereaved children in the form of elevated 24-h salivary cortisol concentrations (Nicolson, Reference Nicolson2004) and elevated cortisol in the dexamethasone/CRH test (Tyrka et al. Reference Tyrka, Wier, Price, Ross, Anderson, Wilkinson and Carpenter2008). However, stress diathesis remains a complex phenomenon with no absolute demarcation between brief (potentially beneficial) and persistent (potentially damaging) stress. In some instances early adversity has appeared to prime, or ‘stress inoculate’, the individual to later adversity (Anisman et al. Reference Anisman, Zaharia, Meaney and Merali1998; Carpenter et al. Reference Carpenter, Carvalho, Tyrka, Wier, Mello, Mello, Anderson, Wilkinson and Price2007; Watson et al. Reference Watson, Owen, Gallagher, Hearn, Young and Ferrier2007; Elzinga et al. Reference Elzinga, Roelofs, Tollenaar, Bakvis, van Pelt and Spinhoven2008; Daskalakis et al. Reference Daskalakis, Bagot, Parker, Vinkers and de Kloet2013). Thus, a relationship clearly exists between early adversity, depression and HPA axis function. The potential for individual genes to exert a mediating role in this relationship is the subject of much current study. One such gene is the GR gene; NR3C1.

Epigenetics

In addition to genetic factors, epigenetics is increasingly being investigated for its potential role in integrating environmental exposures and determining disease susceptibility. Epigenetic mechanisms are responsible for heritable changes in gene expression that are not a consequence of changes in the DNA sequence (Bird, Reference Bird2002, Reference Bird2007; Weaver et al. Reference Weaver, D'Alessio, Brown, Hellstrom, Dymov, Sharma, Szyf and Meaney2007; Akbarian & Huang, Reference Akbarian and Huang2009; Bale et al. Reference Bale, Baram, Brown, Goldstein, Insel, McCarthy, Nemeroff, Reyes, Simerly, Susser and Nestler2010; Murgatroyd & Spengler, Reference Murgatroyd and Spengler2011; Booij et al. Reference Booij, Wang, Levesque, Tremblay and Szyf2013). Factors influencing the patterns of epigenetic modification include cell type, developmental stage and the nature and severity of environmental stressors. The fact that epigenetic factors can be modulated by environmental stress and that these epigenetic changes can then be stably inherited within somatic cells means that epigenetic factors have the potential to modulate the effects of environment on health outcomes many years after initial exposure. Thus, epigenetics could explain the extent and nature of the risk of depression conferred by gene–environment interactions.

Epigenetic mechanisms include DNA methylation, histone post-translational modifications such as methylation, acetylation, phosphorylation, ubiquitylation and sumoylation along with non-coding RNAs comprising of micro, small interfering, Piwi-interacting and small nucleolar RNA. Alterations in DNA methylation in particular have been associated with the development of a number of psychiatric illnesses (Booij et al. Reference Booij, Wang, Levesque, Tremblay and Szyf2013; Aberg et al. Reference Aberg, McClay, Nerella, Clark, Kumar, Chen, Khachane, Xie, Hudson, Gao, Harada, Hultman, Sullivan, Magnusson and van den Oord2014; Klengel et al. Reference Klengel, Pape, Binder and Mehta2014; Zannas et al. Reference Zannas, Provençal and Binder2015).

DNA methylation was first documented in calf thymus tissue by Hotchkiss (Reference Hotchkiss1948) and has since become the most widely studied epigenetic mechanism. It involves the addition and maintenance of methyl groups on the 5-carbon position of cytosines at the 5′ end of cytosine–guanine dinucleotides (CpG sites), on one or both DNA strands, through the action of DNA methyltransferases. CpG sites are the least frequent dinucleotides in mammalian genomes, but many are present in clusters, so-called CpG ‘islands’. These islands have been defined as being over 200 bp in length, being more than 50% CG rich and having an expected:observed CpG ratio of more than 0.6. In humans there are thought to be approximately 29 000 CpG islands representing close to 1% of the human genome. Throughout the entire genome about 70% of the CpG sites are thought to be methylated yet those within CpG islands are often unmethylated. Furthermore, CpG islands are often located within gene promoter regions; if these regions become methylated it is usually associated with gene silencing (Turner & Muller, Reference Turner and Muller2005).

Crucially, DNA methylation patterns appear to be stable but also dynamic. This is evident in the maintenance of DNA methylation patterns throughout mitotic divisions but also in the overwriting of parental methylation patterns in meiosis and individual fetal methylation patterns emerging during development. The functional significance of this is thought to be that DNA methylation patterns have an important role in tissue differentiation by controlling cell type-specific patterns of gene expression (Holliday & Pugh, Reference Holliday and Pugh1975; Riggs, Reference Riggs1975; Razin & Riggs, Reference Razin and Riggs1980; Smith & Meissner, Reference Smith and Meissner2013). For example, DNA methylation-induced silencing of the POU5F1 (POU domain, class 5, transcription factor 1) gene (encoding the octamer-binding transcription factor 4; OCT4) appears to be crucial for embryonic stem cell differentiation (Reik, Reference Reik2007).

Although increased DNA methylation of gene-associated CpG islands has been clearly associated with loss of gene expression, the location and the extent of DNA methylation required to prevent gene expression remains unclear. DNA methylation is thought to exert a gene-silencing effect, in concert with other epigenetic modifications and signalling pathways (Day & Sweatt, Reference Day and Sweatt2011; Reul, Reference Reul2014), by inhibiting the binding of transcription factors to promoter regions. While the majority of studies have focused on the impact of promoter-associated changes in DNA methylation on gene expression, more recently methylation within distal enhancer elements has also been explored as an epigenetic mechanism that may have important impacts on gene activity (Plank & Dean, Reference Plank and Dean2014). Meanwhile, discordant monozygotic twin studies of patients with bipolar affective disorder, schizophrenia, autism spectrum disorder and depression have demonstrated differences in DNA methylation of less than 10% at specific CpG sites (Dempster et al. Reference Dempster, Pidsley, Schalkwyk, Owens, Georgiades, Kane, Kalidindi, Picchioni, Kravariti, Toulopoulou, Murray and Mill2011, Reference Dempster, Wong, Lester, Burrage, Gregory, Mill and Eley2014; Wong et al. Reference Wong, Meaburn, Ronald, Price, Jeffries, Schalkwyk, Plomin and Mill2014). Such localized DNA methylation changes have emerged as a potential mechanism underlying abnormalities in GR expression.

The NR3C1 gene

NR3C1 is found on chromosome 5q31–32 and is over 150 kb in length (Francke & Foellmer, Reference Francke and Foellmer1989; Turner et al. Reference Turner, Vernocchi, Schmitz and Muller2014). It has eight translated exons (numbered 2–9) and is thought to have up to 14 untranslated alternative first exons (termed 1a through to 1j, with 1a and 1c having six further subdivisions between them) at its 5′ end (Daskalakis & Yehuda, Reference Daskalakis and Yehuda2014). NR3C1 has a complex promoter structure with one promoter for each of its single alternative first exons (Turner et al. Reference Turner, Alt, Cao, Vernocchi, Trifonova, Battello and Muller2010). Alternative first exon transcripts are thought to be important for adjusting GR levels in accordance with cell or tissue type and dynamic environmental conditions (Turner et al. Reference Turner, Schote, Macedo, Pelascini and Muller2006) by differentially regulating translation efficiency and RNA stability (Bockmuhl et al. Reference Bockmuhl, Murgatroyd, Kuczynska, Adcock, Almeida and Spengler2011). Transcription factors known to modulate alternative first exon use include nerve growth factor inducible protein A (NGFIA) which binds to the exon 1f promoter (Weaver et al. Reference Weaver, Cervoni, Champagne, D'Alessio, Sharma, Seckl, Dymov, Szyf and Meaney2004). GR is a ligand-activated transcription factor crucial for the effective functioning of the HPA axis. It translocates to the nucleus after binding GCs to regulate the activity of specific target genes, including NR3C1 itself. NR3C1 splice variants and mRNA levels, GR isoforms, co-activators and co-repressors have all been associated with variations in GR activity (Binder, Reference Binder2009; Turner et al. Reference Turner, Alt, Cao, Vernocchi, Trifonova, Battello and Muller2010; Anacker et al. Reference Anacker, Zunszain, Carvalho and Pariante2011; Szczepankiewicz et al. Reference Szczepankiewicz, Leszczynska-Rodziewicz, Pawlak, Narozna, Rajewska-Rager, Wilkosc, Zaremba, Maciukiewicz and Twarowska-Hauser2014). Although GR is ubiquitously expressed, its levels are thought to be tightly controlled according to tissue, or even cell, type. Thus, within the brain, levels are higher in areas involved in the stress response such as the paraventricular nucleus, hippocampus and anterior pituitary (Fig. 1) (Karanth et al. Reference Karanth, Linthorst, Stalla, Barden, Holsboer and Reul1997; Uchida et al. Reference Uchida, Nishida, Hara, Kamemoto, Suetsugi, Fujimoto, Watanuki, Wakabayashi, Otsuki, McEwen and Watanabe2008; Booij et al. Reference Booij, Wang, Levesque, Tremblay and Szyf2013).

Fig. 1.

Portion of the human gene for glucocorticoid receptor (NR3C1) (above) and the cytosine–guanine (CpG)-rich region within exon 1f and its promoter (below). The 5′ end of the NR3C1 gene contains multiple alternative first exons labelled 1a to 1j. The sequenced region shows the exon 1f promoter and exon 1f itself (the latter being underlined) running in the 5′–3′ direction. Sequence numbering is in accordance with Daskalakis & Yehuda (Reference Daskalakis and Yehuda2014) with position +1 being the translational start site (ATG) 13 bp downstream from the start of exon 2. This region is equivalent to 36138 to 36527 in National Center for Biotechnology Information (NCBI) reference sequence NG_009062.1 (http://blast.ncbi.nlm.nih.gov/Blast.cgi). Boxes illustrate the canonical nerve growth factor (NGF) inducible protein A (NGFIA) binding sites, dotted boxes the putative NGF binding sites (Turner & Muller, Reference Turner and Muller2005; McGowan et al. Reference McGowan, Sasaki, D'Alessio, Dymov, Labonte, Szyf, Turecki and Meaney2009).

Early life adversity and methylation of NR3C1 in animals models

Weaver et al. (Reference Weaver, Cervoni, Champagne, D'Alessio, Sharma, Seckl, Dymov, Szyf and Meaney2004) demonstrated that in rat pups of low-LG-ABN mothers NR3C1 methylation was increased in hippocampal samples (to rates of 80–100%) within the NGFIA binding site of the GR gene's exon 1(7) promoter, the homologue of exon 1f in humans, and that this was associated with reduced GR expression. Conversely, they showed that high LG-ABN maternal care was associated with lower methylation rates (0–10%) of the exon 1(7) GR promoter in offspring. In agreement with a prior study (Meaney et al. Reference Meaney, Diorio, Francis, Widdowson, LaPlante, Caldji, Sharma, Seckl and Plotsky1996) examining GR expression, these differences in methylation rates persisted into the offspring's adulthood. Group differences were eliminated by trichostatin A, a histone deacetylase inhibitor thought to promote demethylation (Ou et al. Reference Ou, Torrisani, Unterberger, Provençal, Shikimi, Karimi, Ekstrom and Szyf2007), as well as by cross-fostering. This was the first study in the literature to show a clear link between mothering, long-term changes in DNA methylation patterns and subsequent gene expression. Further studies were conducted with, at times, conflicting results. Daniels et al. (Reference Daniels, Fairbairn, van Tilburg, McEvoy, Zigmond, Russell and Stein2009), for instance, investigated the impact of separating rat pups from their mothers between postnatal days 2 and 14. Assessment of hippocampi at postnatal day 21 revealed no significant change in methylation within exon 1(7) or the NGFIA binding site between pups separated from their mothers and pups raised normally. Nevertheless, Weaver et al. (Reference Weaver, D'Alessio, Brown, Hellstrom, Dymov, Sharma, Szyf and Meaney2007) demonstrated that high-LG-ABN care was associated with demethylation of the 5′ CpG dinucleotide in the NGFIA response element specifically. Additionally, the authors showed that increased LG-ABN was correlated with greater NGFIA binding, histone acetylation, GR mRNA levels, hippocampal NGFIA expression and increased amounts of GR protein.

Early life adversity and methylation of NR3C1 in humans

One of the first studies in humans to examine the relationship between prenatal adversity and NR3C1 methylation was by Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008). Children of depressed mothers who had received medication (n = 33) were compared with children of untreated depressed mothers (n = 13) and controls (n = 36). An association was seen between prenatal exposure to third trimester maternal depression and increased methylation levels of the NGFIA binding site in exon 1f of the NR3C1 promoter (deemed CpG sites 1 to 3) at birth. CpG site 3, within the NGFIA binding site, was also associated with increased cortisol response at 3 months of age. Antidepressant medication had been associated with increased GR mRNA levels in rodents (Pepin et al. Reference Pepin, Pothier and Barden1992; Pariante & Miller, Reference Pariante and Miller2001; Yau et al. Reference Yau, Hibberd, Noble and Seckl2002) and increased GR density in human peripheral blood cells (Calfa et al. Reference Calfa, Kademian, Ceschin, Vega, Rabinovich and Volosin2003), whilst NR3C1 polymorphisms have been observed to predict antidepressant medication responses (Binder et al. Reference Binder, Salyakina, Lichtner, Wochnik, Ising, Putz, Papiol, Seaman, Lucae, Kohli, Nickel, Kunzel, Fuchs, Majer, Pfennig, Kern, Brunner, Modell, Baghai, Deiml, Zill, Bondy, Rupprecht, Messer, Kohnlein, Dabitz, Bruckl, Muller, Pfister, Lieb, Mueller, Lohmussaar, Strom, Bettecken, Meitinger, Uhr, Rein, Holsboer and Muller-Myhsok2004; Spijker & van Rossum, Reference Spijker and van Rossum2012). However, maternal treatment with selective serotonin reuptake inhibitors in the Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008) study did not have any observable effect on offspring CpG methylation status. Nevertheless, these findings encouraged other studies (e.g. Hompes et al. Reference Hompes, Izzi, Gellens, Morreels, Fieuws, Pexsters, Schops, Dom, Van Bree, Freson, Verhaeghe, Spitz, Demyttenaere, Glover, Van den Bergh, Allegaert and Claes2013) to assess the same portion of the exon 1f promoter and CpG sites. Conradt et al. (Reference Conradt, Lester, Appleton, Armstrong and Marsit2013) reported that newborn offspring exposed to maternal depression in utero had increased methylation at the authors’ CpG site 2 within exon 1f as well as adverse neurobehavioural outcomes.

Radtke et al. (Reference Radtke, Ruf, Gunter, Dohrmann, Schauer, Meyer and Elbert2011) examined DNA methylation using peripheral blood samples taken from children (n = 24) aged up to 19 years old whose mothers had been exposed to violence before, during and after their pregnancy. Increased methylation rates in children were significantly associated with maternal exposure to violence during their pregnancy. Methylation was seen in seven of the 24 children, in five of the 10 CpG sites examined and at rates of up to 10%. Strikingly, there was no association between child NR3C1 methylation and maternal exposure to violence either before or after pregnancy. Maternal NR3C1 methylation was not significantly correlated with methylation levels in their children and was unaffected by exposure to violence. This study was the first in humans to show an apparently sustained dysregulation of the HPA axis associated with previous early life psychological stress. However, the lack of data over such long periods of time, up to 19 years in some instances, and the small sample sizes used meant that innumerable confounders could not be ruled out and the statistical power of the study remained relatively limited.

Such site-specific findings, again using the portion of the exon 1f adopted by Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008), were seen in a study by Tyrka et al. (Reference Tyrka, Price, Marsit, Walters and Carpenter2012) in which 99 healthy adult subjects showed correlations between NR3C1 methylation at CpG sites 1 and 3 in the exon 1f promoter and previous childhood maltreatment, parental care and parental loss. However, methylation rates at these sites and cortisol response were not correlated and the study did not incorporate gene expression data. Hence, the functional significance of their findings remained debatable.

Depression and methylation of NR3C1

In a study using buccal DNA from healthy individuals (n = 92) Edelman et al. (Reference Edelman, Shalev, Uzefovsky, Israel, Knafo, Kremer, Mankuta, Kaitz and Ebstein2012) were able to show that methylation at a single CpG site within a binding site for NGFIA correlated with cortisol response to stress. Furthermore, as a result of epigenetic modifications being chemically stable yet modifiable in accordance with dynamic environmental factors (Meaney & Szyf, Reference Meaney and Szyf2005; Sweatt, Reference Sweatt2009) NR3C1 methylation has been afforded considerable explanatory potential in trying to understand both HPA axis dysregulation and depression (Turner et al. Reference Turner, Alt, Cao, Vernocchi, Trifonova, Battello and Muller2010). Inconsistent results from NR3C1 SNP studies in depression (Bouma et al. Reference Bouma, Riese, Nolte, Oosterom, Verhulst, Ormel and Oldehinkel2011; Lahti et al. Reference Lahti, Raikkonen, Bruce, Heinonen, Pesonen, Rautanen, Wahlbeck, Kere, Kajantie and Eriksson2011; Lewis et al. Reference Lewis, Collishaw, Harold, Rice and Thapar2012; Engineer et al. Reference Engineer, Darwin, Nishigandh, Ngianga-Bakwin, Smith and Grammatopoulos2013; Galecka et al. Reference Galecka, Szemraj, Bienkiewicz, Majsterek, Przybylowska-Sygut, Galecki and Lewinski2013; Koper et al. Reference Koper, van Rossum and van den Akker2014) have added further impetus to this field of enquiry.

Alt et al. (2010) conducted a study exploring the possible association between methylation of NR3C1 and depression. The authors assessed NR3C1 methylation in post-mortem samples from depressed patients (n = 6) in multiple limbic brain regions compared with controls (n = 6). Hippocampal exon 1f transcripts were reduced in depressed patients and NGFIA was down-regulated within the hippocampus, cingulate gyrus and nucleus accumbens. However, these data demonstrated very low overall levels of methylation in both depressed and control brains, whilst the NR3C1 promoter for exon 1f was completely unmethylated in all of the samples taken. Thus, the mechanism for this down-regulation in depressed brains appeared to be entirely independent of methylation patterns. However, as the authors themselves acknowledged, this study's power was limited by small sample sizes. More recently Na et al. (Reference Na, Chang, Won, Han, Choi, Tae, Yoon, Kim, Joe, Jung, Lee and Ham2014) compared methylation levels of NR3C1's promoter region in depressed patients (n = 45) and controls (n = 72). The authors found hypomethylation, rather than hypermethylation, at two CpG sites in patients. Neither the Alt et al. (Reference Alt, Turner, Klok, Meijer, Lakke, Derijk and Muller2010) nor the Na et al. (Reference Na, Chang, Won, Han, Choi, Tae, Yoon, Kim, Joe, Jung, Lee and Ham2014) studies were able to provide definitive data regarding HPA axis functioning or childhood trauma. This would appear to be crucial, as illustrated by the following studies incorporating early adversity into models of depression, HPA axis dysfunction and methylation of NR3C1 at different life stages.

Can methylation of NR3C1 link early life adversity to depression in later life?

McGowan et al. (Reference McGowan, Sasaki, D'Alessio, Dymov, Labonte, Szyf, Turecki and Meaney2009) compared methylation rates in hippocampi of post-mortem samples from suicide victims, with and without histories of childhood abuse, and controls (n = 12 for each of these three groups). Two-thirds of the suicide victims, whether abused or not, were retrospectively diagnosed with mood disorders via psychological autopsies using Structured Clinical Interviews for DSM-III-R (SCID). Levels of childhood abuse in suicide victims correlated with higher levels of methylation of the NR3C1 promoter as well as lower NR3C1 mRNA levels, both overall and for the exon 1f splice variant alone. There was no significant difference in methylation rates of the exon 1f NR3C1 promoter or NR3C1 mRNA levels between non-abused suicide victims and controls. In abused suicide victims CpG site-specific increases in methylation were associated with reduced NGFIA binding and NGFIA-induced transcription. However, this study was limited statistically by the small sample sizes used for each group and the removal of outliers in their final analysis. Also of note was that McGowan et al. (Reference McGowan, Sasaki, D'Alessio, Dymov, Labonte, Szyf, Turecki and Meaney2009) reported exon 1f levels accounting for up to 60% of the total amount of expressed NR3C1 promoters. Alt et al. (Reference Alt, Turner, Klok, Meijer, Lakke, Derijk and Muller2010), meanwhile, gave a figure in accordance with previous studies of less than 1%. Such relatively low levels of exon 1f could detract from its apparent functional importance in comparison with other alternative first exons, whilst dramatically different expression rates between studies may also reduce the confidence with which results can be generalized.

Perroud et al. (Reference Perroud, Paoloni-Giacobino, Prada, Olie, Salzmann, Nicastro, Guillaume, Mouthon, Stouder, Dieben, Huguelet, Courtet and Malafosse2011) investigated the correlation of the severity of childhood maltreatment with methylation rates in NR3C1 for patients diagnosed with borderline personality disorder (n = 101), depression (n = 99) or depression with co-morbid post-traumatic stress disorder (n = 15). A portion of the exon 1f promoter was analysed with reference to the sequence used by Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008). They were able to show highly significant associations between methylation of NR3C1 and the severity and number of sexual abuse episodes. In a study by Melas et al. (Reference Melas, Wei, Wong, Sjoholm, Aberg, Mill, Schalling, Forsell and Lavebratt2013) salivary DNA from depressed adults (n = 92) was compared with controls (n = 82). They examined whether differential methylation rates of NR3C1 were associated with various adversities experienced in childhood. Of the 47 CpG sites spanning the exon 1f promoter that were analysed, increased methylation was seen at a single CpG site, near to the NGFIA binding region. This was significantly associated with early parental death. However, this study did not assess childhood abuse, the numbers involved in the adversity subgroups were not large and only females were included.

Very few studies have identified candidates for human maternal behaviours equivalent to rat LG-ABN. However, in rats the effects of LG have been shown to be mimicked by stroking pups with a brush (Mulligan et al. Reference Mulligan, D'Errico, Stees and Hughes2012). A study by Sharp et al. (Reference Sharp, Pickles, Meaney, Marshall, Tibu and Hill2012) demonstrated moderation of the effects of prenatal maternal depression upon emotional and physiological outcomes in human infants through mothers stroking their babies in their first weeks of life. A very recent follow-up study by Murgatroyd et al. (Reference Murgatroyd, Quinn, Sharp, Pickles and Hill2015) has showed reduced NR3C1 methylation associated with maternal stroking in these children, hence bolstering the possible role of epigenetic mechanisms in the long-term effects of early life stress and maternal care. Interestingly, the same study also found interactive effects between prenatal and postnatal maternal depression on methylation of NR3C1's exon 1f. Infants of mothers with low prenatal depression showed increased methylation when exposed to increased postnatal depression – consistent with an interplay between prenatal and postnatal environments. In general terms this is supportive of the fetal origins hypothesis of human disease according to which environmental exposures in utero lead to adaptive modifications in fetal development that act to increase fitness in similar postnatal environments (Table 1).

Table 1.

Studies investigating interactions between methylation of NR3C1, depression and early adversity^a

CpG, Cytosine–guanine; GR, glucocorticoid receptor; PTSD, post-traumatic stress disorder; BPAD, bipolar affective disorder. ^aTo aid clarity, CpG sites are numbered according to Daskalakis & Yehuda's (Reference Daskalakis and Yehuda2014) exon 1f promoter start point (see Fig. 1).

Methodological issues and possible future studies

Whilst these human studies have added overall support to a role for epigenetic modification in the link between early adversity, HPA axis dysregulation and depression vulnerability, attention is increasingly being drawn to inconsistencies in study design that may have prevented causal inferences being made (Daskalakis & Yehuda, Reference Daskalakis and Yehuda2014; Turecki & Meaney, Reference Turecki and Meaney2014). In this final section an attempt will be made to give an overview of such inconsistencies in order to offer a potential direction for future studies.

Unlike the majority of animal studies examining NR3C1 methylation, many types of potential stressors, sometimes at different developmental stages, have been used to represent early human adversity. For example, Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008) used prenatal exposure to maternal depression whilst Radtke et al. (Reference Radtke, Ruf, Gunter, Dohrmann, Schauer, Meyer and Elbert2011) examined the impact of pre- and perinatal exposure to violence against the mother. Both McGowan et al. (Reference McGowan, Sasaki, D'Alessio, Dymov, Labonte, Szyf, Turecki and Meaney2009) and Perroud et al. (Reference Perroud, Paoloni-Giacobino, Prada, Olie, Salzmann, Nicastro, Guillaume, Mouthon, Stouder, Dieben, Huguelet, Courtet and Malafosse2011) used histories of childhood abuse. Studies examining the impact of early adversity in humans is clearly more limited in design by ethical considerations when compared with animal studies. Given the relatively complex nature of human interactions and stress diathesis there is a need to minimize confounders by standardizing the assessment of stressors whenever possible. With regards to using stressors at different developmental stages, the natural history of site-specific methylation, such as the exon 1f promoter of NR3C1, in individual subjects has not been explored. Prenatal stress exposure, as used by Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008) and Radtke et al. (Reference Radtke, Ruf, Gunter, Dohrmann, Schauer, Meyer and Elbert2011), has helped to establish the temporal boundaries of what appears to be a developmentally sensitive period for a possible causal chain of adversity, epigenetic modification, HPA dysregulation and subsequent depression. However, substantial differences can be expected in the nature of stresses prenatally compared with postnatally, as well as their developmental consequences, and this again risks introducing many confounders when attempting to interpret data. Future studies need to comprehensively detail adverse events, as is common practice on psychiatric in-patient wards and to a lesser degree in the community setting, over extensive periods of time and to combine this with regular assessments of epigenetic modifications.

Human studies have also undertaken analysis in different types of tissue. Peripheral blood has often been used due to its relative acceptability from the patient's perspective and its clinical practicality. Radtke et al. (Reference Radtke, Ruf, Gunter, Dohrmann, Schauer, Meyer and Elbert2011) and Perroud et al. (Reference Perroud, Paoloni-Giacobino, Prada, Olie, Salzmann, Nicastro, Guillaume, Mouthon, Stouder, Dieben, Huguelet, Courtet and Malafosse2011) used DNA extracted from peripheral whole blood whilst Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008) investigated mononuclear cells from the cord blood of newborns. Other studies have looked at post-mortem tissue: McGowan et al. (Reference McGowan, Sasaki, D'Alessio, Dymov, Labonte, Szyf, Turecki and Meaney2009) analysed hippocampal specimens whilst Alt et al. (Reference Alt, Turner, Klok, Meijer, Lakke, Derijk and Muller2010) looked at several different limbic regions. Melas et al. (Reference Melas, Wei, Wong, Sjoholm, Aberg, Mill, Schalling, Forsell and Lavebratt2013), meanwhile, used DNA from saliva. It has already been highlighted that methylation levels may differ between cell types (Glossop et al. Reference Glossop, Nixon, Emes, Haworth, Packham, Dawes, Fryer, Mattey and Farrell2013; Simar et al. Reference Simar, Versteyhe, Donkin, Liu, Hesson, Nylander, Fossum and Barres2014), meaning that comparisons between studies using entirely different tissues could be very challenging. However, evidence has emerged that peripheral blood may be an appropriate tissue to identify biomarkers for depression in the context of genetic studies (Rollins et al. Reference Rollins, Martin, Morgan and Vawter2010; Hepgul et al. Reference Hepgul, Cattaneo, Zunszain and Pariante2013), whilst in a review by Tylee et al. (Reference Tylee, Kawaguchi and Glatt2013) the methylome was shown to be more highly correlated between blood and brain samples than the transcriptome. Provençal et al. (Reference Provençal, Suderman, Guillemin, Massart, Ruggiero, Wang, Bennett, Pierre, Friedman, Côté, Hallett, Tremblay, Suomi and Szyf2012) showed that in rhesus macaques variations in mothering (surrogate v. mother reared) led to differential methylation rates including the A2D681 gene which is the homologue of NR3C1 in humans. A weak but significant correlation was seen in differential methylation between prefrontal cortex samples and T lymphocyte cells. The use of blood samples does have an advantage over the use of post-mortem and placental tissue given that these samples are taken at markedly different physiological states. Saliva samples involve cells from different developmental lineages to both brain and blood tissue and hence the validity of its use is unclear. These various confounders mean that many more studies will be needed before effects directly attributable to early life trauma can be separated from those relating to tissue type. Future investigations also need to involve repeated peripheral samples taken from individuals who have nominated themselves for future brain donation. The measurement of NR3C1 methylation levels across various brain regions in these individuals will allow the consolidation of findings from different tissue types and could lead to effective and clinically acceptable therapeutic interventions. Additionally, efforts must be made to isolate specific cell types, primarily those cells found in peripheral blood, in order to establish cell-specific methylation profiles (El-Sayed et al. Reference El-Sayed, Haloossim, Galea and Koenen2012).

Of particular note in the studies published to date is that there has been considerable heterogeneity in exactly where, and to what extent, within the NR3C1 gene and its promoter regions, methylation has been assessed (Labonte et al. Reference Labonte, Yerko, Gross, Mechawar, Meaney, Szyf and Turecki2012; Daskalakis & Yehuda, Reference Daskalakis and Yehuda2014). Much of the work that has been done has involved a specific location of the NR3C1 gene itself, namely exon 1f and its promoter region incorporating the binding site for NGFIA. In the first of such studies Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008) looked at 13 CpGs in an area comprising exon 1f, its promoter and a further section downstream of this, using the Weaver et al. (Reference Weaver, Cervoni, Champagne, D'Alessio, Sharma, Seckl, Dymov, Szyf and Meaney2004) study as their main reference. McGowan et al. (Reference McGowan, Sasaki, D'Alessio, Dymov, Labonte, Szyf, Turecki and Meaney2009), meanwhile, looked at 39 CpGs across exon 1f and its promoter. Perroud et al. (Reference Perroud, Paoloni-Giacobino, Prada, Olie, Salzmann, Nicastro, Guillaume, Mouthon, Stouder, Dieben, Huguelet, Courtet and Malafosse2011) looked at eight CpG sites across the exon 1f promoter and a further downstream region. Finally, Melas et al. (Reference Melas, Wei, Wong, Sjoholm, Aberg, Mill, Schalling, Forsell and Lavebratt2013) analysed what are thought to be all of the 47 CpG sites throughout the main body and promoter region of NR3C1's exon 1f. Although some effort has been made to correlate individual CpG sites across different studies this has not always been possible, and despite occasional agreement in which sites are differentially methylated no conclusive patterns have yet emerged. Hence, there would be benefit in researchers adopting a unified and comprehensive approach to the nucleotide sequence being assessed as well as the individual CpG sites within it. Given the extensive adoption so far of the nucleotide sequence from the Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008) study it is recommended that this continues to be used as the minimum and necessary sequence coverage for future NR3C1 methylation studies in depressed patients with a history of early maltreatment.

Investigators have examined different functional correlates for the changes observed in NR3C1 methylation. For example, Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008) used salivary cortisol measurements in the morning, following a stressor and in the evening, whereas McGowan et al. (Reference McGowan, Sasaki, D'Alessio, Dymov, Labonte, Szyf, Turecki and Meaney2009) assessed NR3C1 mRNA levels as well as NGFIA binding and NGFIA-induced transcription. Once again this heterogeneity of approaches has potentially impaired efforts to establish causal relationships. Although investigators have amassed a considerable amount of evidence for an association between differential methylation and HPA axis function in humans, a causal relationship still needs to be fully established.

Lastly, a recurrent issue in studies examining childhood adversity is that of confounding factors relating to recall bias and the participant's current mental state. Ideally, long-term follow-up of children up to and beyond the period of maximum risk of the development of psychiatric illnesses, with objective and detailed documentation of the reported maltreatment, will minimize this complication and simultaneously allow more effective exploration of the consequences of particular maltreatment categories.

Summary

Studies have continued to emerge that implicate hypermethylation of NR3C1's exon 1f promoter occurring following early trauma with this being associated with HPA axis dysfunction and depression. Results have, however, been inconsistent at times and not without occasional controversy (Dyer, Reference Dyer2014). How much of the observed variation of findings is due to inter-subject variability in underlying pathophysiology, as opposed to experimental design, remains to be seen.

There is a need for researchers to adopt more consistent approaches to document the natural history of methylation patterns at individual CpG sites within NR3C1 after early life adversity. This natural history will need to include other environmental factors such as age and diet (Mathers et al. Reference Mathers, Strathdee and Relton2010; Murgatroyd & Spengler, Reference Murgatroyd and Spengler2011; Suderman et al. Reference Suderman, McGowan, Sasaki, Huang, Hallett, Meaney, Turecki and Szyf2012; Tyrka et al. Reference Tyrka, Price, Marsit, Walters and Carpenter2012; Bakulski & Fallin, Reference Bakulski and Fallin2014). A diet low in folate and high in methionine, for example, has already been associated with increased NR3C1 methylation in mice (Sulistyoningrum et al. Reference Sulistyoningrum, Singh and Devlin2012). The use of newer, more powerful technologies such as epigenome-wide association studies and single cell analysis are exciting but will bring challenges in terms of consistency, defining cell type, cost and data analysis (Plessy et al. Reference Plessy, Desbois, Fujii and Carninci2013; Callaway, Reference Callaway2014; Robinson et al. Reference Robinson, Kahraman, Law, Lindsay, Nowicka, Weber and Zhou2014).

A significant challenge in the epigenetics of mental illness also continues to be the complex nature of these disorders and their aetiology (Caspi et al. Reference Caspi, Sugden, Moffitt, Taylor, Craig, Harrington, McClay, Mill, Martin, Braithwaite and Poulton2003; Eaves et al. Reference Eaves, Silberg and Erkanli2003; Bowes & Jaffee, Reference Bowes and Jaffee2013; Ehlert, Reference Ehlert2013; Brown et al. Reference Brown, Craig, Harris, Herbert, Hodgson, Tansey and Uher2014; Castillo-Fernandez et al. Reference Castillo-Fernandez, Spector and Bell2014; Kanherkar et al. Reference Kanherkar, Bhatia-Dey and Csoka2014). As is the case for most research in psychiatry there is the ever-present issue of difficulties with psychiatric nosology. Most authorities believe that the diagnostic category of ‘major depressive disorder’ contains a heterogeneous collection of disorders with differing underlying pathophysiologies (Schmidt et al. Reference Schmidt, Shelton and Duman2011). This will clearly hamper the interpretation of epigenetic data. This may be addressed by exploring the relationship between epigenetic status and endophenotypes such as those defined using the Research Domain Criteria initiative (Insel et al. Reference Insel, Cuthbert, Garvey, Heinssen, Pine, Quinn, Sanislow and Wang2010; Casey et al. Reference Casey, Craddock, Cuthbert, Hyman, Lee and Ressler2013).

Studies on depression and NR3C1 methylation are exciting in two main ways. First, they suggest a hitherto untapped approach that could help to synthesize a comprehensive and necessarily eclectic theory of depression, aiding our understanding of causal factors and identifying disease biomarkers. Second, they have the potential to point towards novel therapeutic targets since epigenetic changes are potentially reversible and therefore amenable to intervention, as has been seen in cancer, cardiovascular disease and neurological disorders (Heerboth et al. Reference Heerboth, Lapinska, Snyder, Leary, Rollinson and Sarkar2014; Sandhu et al. Reference Sandhu, Roll, Rivenbark and Coleman2015). Epigenetics offers a way to optimize the diagnosis, prognosis and treatment of depression. In doing so it could reduce depression's global burden whilst simultaneously providing a truly personalized medicine.