Background

Unfavorable environmental conditions during pregnancy have been shown to promote the onset of mental disorders in the offspring [1,2,3] via epigenetic mechanisms [4,5,6]. One epigenetic mechanism that can be changed by adverse intrauterine exposure and influences the development of offspring health is deoxyribonucleic acid (DNA) methylation [5, 7,8,9,10]. DNA methylation is the addition of methyl groups to cytosine-guanine dinucleotides (CpG), with the potential to regulate gene expression [11,12,13,14,15]. For instance, Palma-Gudiel et al. [16] reported increased methylation of the glucocorticoid receptor gene (NR3C1), a gene involved in the regulation of the hypothalamic–pituitary–adrenal (HPA) axis in the offspring, following exposure to prenatal stress. Increased NR3C1 methylation has, in turn, been associated with mental disorders [17,18,19] such as depression [20].

Undernutrition in pregnant women is an unfavorable environmental condition that can affect the intrauterine development and may thus have long-lasting detrimental consequences for the mental health of the offspring later in life [21]. The effect of prenatal undernutrition on mental health can be analyzed through natural experiments (quasi-experimental studies), in which undernutrition (e.g. famine) occurs naturally in a specific population [22, 23]. Meta-analytic results have already demonstrated an increased risk of suffering from psychotic, affective, and personality disorders in adults who were exposed to famine during prenatal development [24].

One important mechanism to explain how unfavorable maternal food consumption leads to an increased susceptibility to mental disorders in the offspring in adulthood may be altered DNA methylation patterns [25,26,27]. Rijlaarsdam et al. [28] reported that an unhealthy high-fat and high-sugar prenatal diet was positively associated with changes in the insulin-like growth factor gene (IGF2) in the offspring, which was in turn related to increased attention deficit hyperactivity disorder (ADHD) symptoms in adolescence [28]. Moreover, hypomethylation of this IGF2 gene has been found in adult offspring who were prenatally exposed to famine [29]. Less is known, however, about whether altered DNA methylation mediates the effects of prenatal famine exposure on mental disorders in the offspring.

In summary, undernutrition during pregnancy appears to increase the susceptibility to mental disorders in the offspring. However, the aforementioned meta-analysis did not include a quality assessment [24]. To date, therefore, no quality assessment has been conducted on the myriad of published studies examining the effects of prenatal famine exposure on offspring mental health. Moreover, it remains to be elucidated whether changes in DNA methylation are the mechanism linking prenatal famine exposure to the development of mental disorders in adult offspring. The purpose of this study is thus to provide the first systematic review of the existing literature on the impact of prenatal famine exposure on offspring mental health and altered DNA methylation, and to integrate the findings by means of a meta-analysis.

Methods

Search strategy

We conducted a literature search of the databases PubMed and PsycINFO to identify relevant records up to September 2022. The search strategies included the words (a) “famine” and related terms, (b) “pregnancy” and related terms, (c) “DNA methylation” and related terms, or (d) “mental disorders” and related terms. The search followed a systematic approach in accordance with the Preferred Reporting Items for Systematic review and Meta-Analysis Protocols (PRISMA-P) guidelines [30]. This systematic review and meta-analysis was registered on the Open Science Framework (OSF): osf.io/3hn5p.

Screening and selection procedure

First, duplicates of the identified records were removed. Titles and abstracts were screened, and records that did not meet the eligibility criteria, such as non-human studies and non-empirical research, were excluded. The articles yielded by the literature search were screened and selected using the following inclusion criteria: (1) offspring whose mothers experienced famine during pregnancy and including either (2) a measure of DNA methylation or (3) a measure of psychopathology. A full-text reading of all remaining articles was performed. Studies were included in the meta-analyses if they (1) used the same questionnaire to measure symptoms of psychopathology, (2) included a categorical outcome (mental disorders) irrespective of which clinical interview was used to establish the diagnosis, and (3) provided adequate data for statistical analysis.

Data extraction

Included articles were examined for information about the first author, year of publication, cohort, sample description, assessment of symptoms of psychopathology, and main results. Articles on DNA methylation were examined for information about chromosome number and location, gene, number of CpGs, method for DNA methylation analysis, and main results. Data extraction was performed by one of the authors (HE) and a research assistant. Risk of bias was assessed using a modified version of the Newcastle–Ottawa scale [31, 32], containing the following seven items: sampling representativeness, sample size, exposure definition, famine severity assessment, confounding adjustment, outcome assessment, and statistical methods. Each item was scored as either good, fair, or poor [31]. The items outcome assessment and sample size were modified for studies on mental disorders, epigenome-wide DNA methylation analyses, and targeted candidate gene analyses (see Additional file 1: Tables S1–S3). Risk of bias assessment was performed by one of the authors (HE) and a senior researcher from our workgroup.

Data analysis

To assess the association between prenatal famine exposure and symptoms of psychopathology or mental disorders in adulthood, we calculated the effect size across studies as the overall pooled log10 odds ratio (logOR) of the number of individuals with and without symptoms or a mental disorder in the prenatal famine group and in the control group. The logOR was used for the depression and schizophrenia studies. The control group consisted of offspring who were exposed to famine during childhood (non-prenatal famine exposure) and/or offspring who were not exposed to famine at all (non-exposure). For two studies that used the Hospital Anxiety and Depression Scale (HADS), we used means and standard deviations to calculate Hedges’ g. One of these studies did not report the specific standard deviations for each of the two subscales of the HADS (anxiety and depression) and instead only provided overall standard deviations, which were therefore used as a reference. Results were considered statistically significant if the p value was < 0.05. Meta-analyses were conducted if at least two studies used the same outcome measurement. Studies with insufficient data were only included in the systematic review, and not in the meta-analyses. Random-effects meta-analyses were conducted using the meta-analysis function integrated in SPSS version 28.0.1.1, which also allowed us to create forest plots. The Q and I2 statistics were calculated to assess the heterogeneity of the included studies. Subgroup analyses were performed to detect whether a more homogenous effect size could be calculated. Following the Cochrane Handbook for Systematic Reviews of Interventions [33], when 10 or more studies were included in our meta-analyses, we used the trim-and-fill procedure and visual inspection of funnel plots to detect publication bias [34].

Results

Search results

The literature search yielded 2697 articles, of which 239 were duplicates and removed. Of the remaining 2458 articles, a further 2382 were excluded due to publication in a language other than English, non-empirical research, or irrelevant title/abstract. Of the final 76 articles assessed for eligibility, 39 were excluded for as they did not assess the outcome, only examined exposure to nutrient deficiency, were exclusively polymorphism analyses, or assessed different exposure periods. Thus, in total, 37 studies were eligible for data extraction and were included in this systematic review. Of these studies, 22 reported effects of prenatal famine exposure on symptoms of psychopathology or mental disorders, and 14 studies reported effects of famine during pregnancy on DNA methylation. The remaining study analyzed the mediating effect of DNA methylation on mental disorders in adults prenatally exposed to famine. Eleven of the 37 studies reported sufficient data to be included in meta-analyses. The study selection is summarized in Fig. 1.

Fig. 1
figure 1

Screening and selection process of studies displayed by a PRISMA flowchart

Full size image

Study characteristics

Characteristics of the included studies are shown in Tables 1, 2, 3 and 4. Articles were published between 1992 and 2022. All participants were adults. The sample size ranged from 13 to 494,684. All studies focused either on the Dutch Famine (1944–1945) or the Chinese Famine (1959–1961), with one exception, the Bangladesh Famine (1974–1975). Individuals without prenatal famine exposure were either born after the famine (non-exposure: had not experienced famine in their life) or before the famine (non-prenatal exposure: experienced famine during infancy, childhood, adolescence, or adulthood). Most DNA methylation studies (67%) used either sibling or time controls. Sibling controls were siblings of prenatally exposed adults and were mostly younger than their exposed siblings. Time controls were adults who were born either before or after the famine. As the respective authors did not specify how many control adults were in each group, it was not possible to assign them to the non-prenatal exposure or non-exposure group. Periconceptional exposure referred to exposure to famine during conception and the 1st trimester.

Table 1 Effects of prenatal exposure to famine on mental disorders/symptoms in offspring
Full size table
Table 2 Effects of prenatal exposure to famine on (epi)genome-wide DNA methylation of the offspring
Full size table
Table 3 Effects of prenatal exposure to famine on targeted DNA methylation of the offspring
Full size table
Table 4 Effects of prenatal exposure to famine on genome-wide DNA methylation and mental disorders
Full size table

Risk of bias assessment

The risk of bias assessment is presented in Additional file 2: Table S4. Quality ratings ranged from poor to good, with only two studies rated good on all study items [35, 36].

Of the studies examining symptoms of psychopathology and mental disorders, most scored highest on the statistical methods item. Most studies (86%) used proper statistical analyses and conducted sensitivity analyses. The sample size item was generally rated as good for the mental disorders or symptoms studies (77%). Of the 22 studies, 14 studies (64%) defined famine exposure both quantitatively and qualitatively. Half of the studies (50%) used a good outcome assessment by a psychiatrist or clinical psychologist according to International Classification of Diseases (ICD) or Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria. Only 36% of the studies adjusted for confounders and explained why they did so. 32% of the studies had good sampling representativeness. Sampling representativeness was rated as fair if the sample was drawn from only one hospital registry or survey. The lowest ratings were achieved for the item famine severity assessment, with 55% of the studies failing to include excess death rates (EDR), cohort size shrinkage index (CSSI) or global hunger index (GHI) to measure the severity of famine (for more information, see [37]).

Of the DNA methylation studies, most (73%) used proper statistical analyses and conducted sensitivity analyses. Adjustment for confounding factors was good in 53% of these studies. Only 27% defined famine exposure both quantitatively and qualitatively, and only 27% used a good description of the DNA methylation assay. A small proportion of the studies (13%) had good sampling representativeness and sample size. None of the DNA methylation studies were rated as showing a good famine severity assessment (0%).

Effects of prenatal famine exposure on offspring symptoms/mental disorders

Twenty-two studies investigated the effect of prenatal famine exposure on offspring symptoms of psychopathology and/or mental disorders.

As shown in Table 1, one study found higher psychopathology, as measured with the Mental Health Inventory (MHI-5) in individuals who experienced famine during prenatal development compared to individuals who did not [38]. Five studies reported increased depressive symptoms [39,40,41,42,43] in individuals with prenatal famine exposure compared to individuals with non-prenatal exposure and/or non-exposure. One study reported an association between prenatal exposure to famine and increased anxiety and depressive symptoms, as measured with the HADS [44]. In contrast, another study found no significant association between prenatal famine exposure and anxiety and depressive symptoms (HADS) as compared to non-prenatal exposure and non-exposure [45].

With regard to mental disorders, one study found a generally increased risk of mental disorders [46] after prenatal exposure compared to non-exposure. Six studies consistently reported an increased risk of schizophrenia after prenatal exposure compared to non-prenatal and/or non-exposure to famine [35, 36, 47,48,49,50]. In contrast, one study found a higher risk of developing schizophrenia in adults with non-exposure to famine than in adults with prenatal exposure [51]. An increased risk of major affective disorders was found to be linked to in utero exposure to famine as compared to non-exposure in two studies [52, 53]. One study reported an increased risk of antisocial personality disorder [54] and another an increased risk of schizoid personality disorder [55] in men after prenatal exposure compared to non-exposure to famine. Addictive disorders [56] and addictive behaviors [57] in adults were related to prenatal famine exposure but not to non-prenatal famine exposure.

In terms of depressive symptoms, two studies [39, 42] provided sufficient data for meta-analysis based on OR, with results varying by exposure period. On the one hand, adults prenatally exposed to famine showed a decreased risk of depressive symptoms compared to adults with no exposure to famine and adults who were exposed to famine after gestation (logOR = 0.96, 95% CI [0.79, 1.14]; Z = 10.75, p < 0.001; Q = 8.56, I2 = 88%). On the other hand, adults prenatally exposed to famine showed an increased risk of depressive symptoms compared to adults with no exposure to famine (logOR = 1.14, 95% CI [0.94, 1.34]; Z = 11.31, p < 0.001; Q = 6.87, I2 = 86%). In terms of anxiety and depressive symptoms as measured by the HADS, meta-analysis confirmed the null-findings (HADS-A: g = 0.08, 95% CI [− 0.05, 0.21]; Z = 1.17, p = 0.241; Q = 0, I2 = 0%; HADS-D: g = 0.06, 95% CI [− 0.08, 0.19]; Z = 0.84, p = 0.403; Q = 0.23, I2 = 0%). Meta-analysis confirmed the increased risk of suffering from schizophrenia in adulthood after prenatal famine exposure compared to non-prenatal exposure and non-exposure together (logOR = 1.13, 95% CI [0.97, 1.29]; Z = 13.97, p < 0.001). Heterogeneity was high (Q = 9.02, I2 = 89%), see Fig. 2. The results remained unchanged when subgroup analyses were conducted for the Dutch and the Chinese famine (two Dutch famine studies: logOR = 1.21, 95% CI [0.85, 1.57]; Z = 6.57, p < 0.001; Q = 1.13, I2 = 11% and five Chinese famine studies: logOR = 1.12, 95% CI [0.92, 1.33]; Z = 10.74, p < 0.001; Q = 18.25, I2 = 95%). Insufficient data were available for meta-analyses on major affective disorders, antisocial and schizoid personality disorder, as well as addictive disorders.

Fig. 2
figure 2

Forest plot of studies comparing adults prenatally exposed to famine with adults non-prenatally and non-exposed to famine regarding risk of developing schizophrenia. Conducting subgroup analyses for the Dutch and the Chinese famine did not alter the results

Full size image

Effects of prenatal famine exposure on offspring DNA methylation (epigenome-wide analysis)

Nine studies, which are listed in Table 2, investigated DNA methylation by conducting (epi)genome-wide analysis in adults prenatally exposed to famine [58,59,60,61,62,63,64,65,66]. All of these used whole blood as tissue.

Four studies determined DNA methylation using the HumanMethylation450 BeadChip microarray, which has a coverage of over 450,000 sites [67, 68]. The first of these four studies did not find significantly differentially methylated regions (DMRs) in adult offspring following prenatal famine exposure as compared to non-prenatal exposure and non-exposure [59]. The second study identified that prenatal exposure to famine during early gestation was significantly associated with 613 DMRs as compared to non-exposure [58]. The authors specifically reported hypomethylated regions in four genes, namely CCDC51, TMA7, ENO2 and ZNF226 [58]. The third study found a variety of hyper- (FAM150B/TMEM18, PPAP2C, SLC38A2) and hypomethylated (OSBPL5/MRGPRG) genes in adult offspring exposed to famine during early gestation as compared to time and sibling controls. In addition, exposure during conception was associated with decreased methylation of TMEM105/SLC38A10, and exposure during any week of gestation was associated with increased methylation of the genes TACC1 and ZNF385A compared to time and sibling controls [60].

Lastly, an association was found between prenatal famine exposure and hypo-methylation of the genes CRELD2, LRRC8D, LOC100132354, OSBPL5/MRGPRG, TXNIP, PFKFB3 as well as hypermethylation of the genes ABCG1, CCDC155, FAM150B, METTL8, PNPO, PPAP2C, SLC38A2, SYNGR1, TACC1 and ZNF385A compared to controls [64].

Two studies used methylation analyses, which cover over 850,000 sites [69]. One study reported evidence of 601 hypermethylated and 360 hypomethylated sites after prenatal famine exposure as compared to time controls [63]. The other study reported no significant differentially methylated sites after controlling for multiple testing [65].

The two studies measuring global DNA methylation via pyrosequencing did not find a link between prenatal famine exposure and altered methylation patterns as compared to sibling controls and time controls [62, 66]. One of these studies also analyzed global DNA methylation via MethyLight and LUminometric Methylation Assay (LUMA), yielding no significant findings [62].

One study used reduced representation bisulfite sequencing (RRBS) to assess DMRs and found hypermethylation in 60.8% out of 181 identified sites and hypomethylation in 39.2% following periconceptional exposure to famine compared to sibling controls [61]. In the present analysis, we solely reported on genes for which there was a significant association between DNA methylation and prenatal famine exposure. Using the data published in the included papers, we verified whether genes that were significant in some studies were also significant in others, and mostly found no concordance. For instance, only six genes identified by Tobi et al. [60] were replicated in another study by Tobi et al. [64], even though methylation analysis was performed on the same sample. Meta-analysis was not suitable due to different DNA methylation microarrays/data processing approaches and partially unavailable data.

Effects of prenatal famine exposure on offspring DNA methylation (candidate gene analysis)

As can be seen in Table 3, candidate gene DNA methylation analyses revealed significant associations between prenatal famine exposure and a variety of hyper- and hypomethylated genes as compared to the different control groups.

Compared to sibling controls, periconceptional famine exposure was associated with hypomethylation of KLF13 [61], IGF2 [29, 66], and INSIGF [66, 70]. Besides periconceptional exposure, prenatal exposure during late gestation was associated with hypomethylation of the GNASAS gene [70]. Compared to sibling and time controls, prenatal exposure to famine was related to hypermethylation in several genes (CDH23, CPT1A, INSR, SMAD7 [61]; ABCA1, IL-10, LEP, GNASAS and MEG [70]). Compared to time controls only, prenatal famine exposure was related to hypomethylation of the AGTR1 and PRKCA genes [63] and hypermethylation of the IGF2 and INSR genes [72].

As compared to non-prenatal exposure and non-exposure, adults prenatally exposed to famine showed decreased methylation of the ZFP57 and PRDM9 genes and increased methylation of the PAX8 gene [59]. Moreover, prenatal exposure to famine was related to hypomethylation of VTRNA2-1 and EXD3 compared to non-prenatal exposure only [59]. One study reported no association of GR 1-C, LPL, PI3kinase, and PPARy with in utero exposure to famine compared to non-prenatal exposure and non-exposure [73].

In sum, the candidate genes most affected by prenatal famine exposure are IGF2 and INSR. In addition, prenatal famine exposure was not associated with several other candidate genes, which are reported in Table 3 [59, 61, 70, 73, 74].

Although a few significant candidate genes were replicated in other studies, it is possible that methylation analyses were performed on the same sample. Candidate-gene studies were not eligible for meta-analysis due to the heterogeneity of affected genes and partially unavailable data.

DNA methylation as a mediator between famine exposure during pregnancy and mental disorders

Table 4 presents a more recent study by Boks et al. [75], who analyzed changes in DNA methylation in individuals exposed to famine during the first 3 months of prenatal development and their susceptibility to schizophrenia in adulthood. The authors reported that prenatally exposed adults with schizophrenia showed hypermethylation of the DUSP22 gene compared to non-exposed patients and healthy controls [75].

Discussion

In the present systematic review and meta-analysis, we investigated the association between prenatal famine exposure, DNA methylation and mental disorders in adult offspring. We report three main findings: First, meta-analysis confirmed that exposure to famine during prenatal development increases the offspring’s risk of suffering from schizophrenia. With regard to depression, meta-analyses yielded contradictory findings, showing either increased or decreased risk of depressive symptoms depending on exposure periods. Anxiety and depressive symptoms, as measured with the HADS, were not associated with prenatal famine exposure. Prenatal famine exposure was further associated with addictive disorders and behaviors as well as antisocial and schizoid personality disorder. Second, we found that prenatal famine exposure is associated with hypo- and hypermethylation of a variety of genes. The largest number of studies reported differences in DNA methylation of the IGF2 gene. Third, only one mediation study has been conducted to date, which described altered DNA methylation of the DUSP22 gene as a potential mechanism underlying the association between prenatal famine exposure and schizophrenia in adult offspring.

With regard to the first finding, additional studies confirm the increased risk for the development of schizophrenia in offspring prenatally exposed to a (natural) disaster such as an earthquake [76, 77], a terrorist attack [78], infections, and lead exposure [79]. There are several potential reasons for this effect of unfavorable environmental circumstances on an increased susceptibility to schizophrenia. According to the neurodevelopmental hypothesis proposed by Weinberger [80] and Murray and Lewis [81], such conditions impair the neurodevelopment of the fetus by adversely altering gene expression [81,82,83,84,85,86,87]. In particular, shortly after fertilization, a complete demethylation of the genome occurs, which is then re-established during embryogenesis [88]. Adverse environmental circumstances during this periconceptional period can thus permanently alter the DNA methylation of genes involved in neural pathways, impair brain development, and predispose the offspring to an increased risk of schizophrenia [84]. Moreover, researchers have found that schizophrenia shares common features with other mental disorders such as schizoaffective disorders and depression [89, 90], suggesting that the same epigenetic mechanisms are involved in its pathogenesis. However, the inconclusive findings of the meta-analyses on depressive symptoms may also be explained by the fact that environmental conditions influence DNA methylation at other life stages, in addition to early prenatal development [91]. Indeed, offspring exposed to famine in infancy or childhood exhibit more depressive symptoms than offspring exposed to famine prenatally. Nevertheless, prenatal exposure to famine increases the risk of depressive symptoms in adult offspring compared to offspring who have never been exposed to famine. Furthermore, the inconclusive findings regarding depressive symptoms and the null findings regarding anxiety may be attributable to the fact that only two studies could be included in the meta-analyses due to the heterogeneity of the examined exposure periods and different methods of statistical analysis.

With respect to the finding that IGF2 appears to be the gene that is most affected by prenatal famine exposure, the studies in this review revealed both hyper- and hypomethylation of the IGF2 gene in offspring. The reason for this finding of both increased and decreased methylation, despite the fact that all offspring were prenatally exposed to famine, might lie in a dose–response relationship in terms of duration and severity of prenatal famine exposure and IGF2 DNA methylation. Specifically, the Chinese famine was more severe and lasted for longer (3 years) compared to the Dutch famine, which was less severe and lasted for only 6 months [92]. More severe and longer exposure may have led to increased DNA methylation [72], whereas shorter and less severe exposure may have resulted mainly in decreased methylation of the IGF2 gene [29, 66]. This assumption is in line with the study by Shen et al. [92], who reported increased methylation of the IGF2 gene in offspring exposed to severe famine compared to offspring exposed to moderate famine. Moreover, different genomic positions annotated to the IGF2 gene were examined [29, 66], which could be another reason for differences in the direction of DNA methylation.

As for the third finding, there is evidence that DUSP family genes are involved in neural functions and play a role in the pathophysiology of mental disorders such as depression, bipolar disorder, and schizophrenia [93]. This supports the involvement of the DUSP22 gene in the etiology of schizophrenia in adults prenatally exposed to famine [75]. In addition, we suggest that altered DNA methylation of the aforementioned IGF2 gene may contribute to an increased risk of mental disorders, as this gene is also involved in neuronal functions. Specifically, it is an important contributor to fetal growth and development of the central nervous system [94,95,96], with increased methylation of the IGF2 gene in the placenta, for example, showing an association with higher birth weight [94]. However, another study found that increased methylation of this gene (in maternal blood) was associated with lower birth weight [97], and others found no significant association [98]. In terms of the central nervous system, dysregulations of this gene are associated with various mental disorders such as depression and schizophrenia [99].

The phenotype of adults prenatally exposed to famine may additionally be caused by altered DNA methylation of candidate genes in the neuroendocrine and immune systems [17, 100, 101]. Specifically, the LEP gene affects the HPA axis activity by inhibiting the release of corticotropin-releasing hormone (CRH), thereby suppressing its activity and reducing glucocorticoid production [102,103,104]. Hypermethylation of the LEP gene can lead to decreased gene expression [105] and possibly inhibits its role in suppressing HPA axis activity. In addition, hypermethylation of this gene has been associated with schizophrenia [106], and hyperactivity of the HPA axis is an underlying biological mechanism of depression [107, 108]. The findings of our review demonstrate that prenatal famine exposure is associated with hypermethylation of the LEP gene in adult offspring [70]. Furthermore, the function of the neuroendocrine system is closely linked to the function of the immune system, and the HPA axis acts as a mediator between the two systems [109,110,111,112]. The IL-10 gene, an anti-inflammatory cytokine of the immune system, influences the HPA axis activity [112,113,114] by increasing the production of CRH and adrenocorticotropic hormone (ACTH) in the pituitary [109, 110]. Differences in its gene expression have been found in adults suffering from a major affective disorder or schizophrenia [115,116,117]. Evidence indicates that prenatal exposure to famine is related to increased methylation of the IL-10 gene in adult offspring [70].

The present review is the first to systematically and quantitatively present the effects of prenatal famine exposure on both mental disorders or symptoms of psychopathology and DNA methylation. Its strengths include the comprehensive literature search and rigorous quality assessment (risk of bias). However, the results of the meta-analyses, particularly the omission of a meta-analysis for the whole-genome DNA methylation results, should be interpreted with caution because the authors did not to obtain all affected genes from all whole-genome DNA methylation analysis studies. In addition, we are unable to rule out publication bias due to the very small number of studies suitable for meta-analyses. All methylation studies presented in this review used whole blood as a tissue. One might consider whether DNA methylation in peripheral specimens serves as a marker for DNA methylation in brain tissue as there is evidence that epigenetic differences in peripheral specimens do not always correlate with differences in brain tissue [118, 119]. For example, Walton et al. [120] found that only 7.9% of CpGs were broadly correlated between blood and living brain tissue from the same individuals. However, they were able to identify CpG markers from blood tissue that significantly correlated with brain tissue and were involved in biological pathways affected in individuals with schizophrenia [120]. As a further limitation, the heterogeneity of genes affected by prenatal famine exposure might result from the lack of power of small sample sizes and different DNA methylation techniques across the included studies. However, it is noteworthy that most of the associations found were statistically significant at the p < 0.001 level (Tables 2, 3 and 4), even after Bonferroni correction [65, 70, 72, 74] and Benjamin-Hochberg adjustment [60, 66] for multiple testing. Candidate gene analyses have the distinct advantage of enabling a more thorough investigation of specific regions of interest by assessing the overall methylation of a target region and allowing researchers to identify specific CpG sites involved in disease pathogenesis [121]. Epigenome-wide DNA methylation analyses enable the analysis of the entire genome, as generally speaking, more than one gene is involved in the pathogenesis of diseases [122], but cover only small numbers of CpG sites per gene [123, 124]. Moreover, as the examined famine cohorts were geographically diverse, the different methylated genes may be attributable to ethnicity. For instance, Elliott et al. [125] found large differences in DNA methylation between European and South Asian individuals due to ethnically different cell composition. Additionally, the cause of the famines also differed, with the Dutch famine being the result of a food embargo during World War II [23] and the Chinese famine being due to political and economic mismanagement combined with drought [126]. This may further have exposed the two cohorts to distinct psychosocial stressors, which might have influenced their DNA methylation differently.

Conclusion

Prenatal famine exposure has been associated with altered DNA methylation of genes involved in neuronal, neuroendocrine, and immune processes, which may causally promote the development of mental disorders, specifically schizophrenia and depression in adult offspring. Further genome-wide and hypothesis-driven candidate gene mediation analyses, preferably with a longitudinal design and large sample sizes, are warranted to obtain a complete picture of the role of DNA methylation in the association between prenatal exposure to famine and the development of mental disorders. A better understanding may improve the diagnosis and treatment of schizophrenia and depression, as DNA methylation can be reversed by pharmacological drugs [127,128,129], and may inform the development of nutritional intervention programs for pregnant women affected by famine.