The role of oxytocin receptor gene (OXTR) DNA methylation (DNAm) in human social and emotional functioning: a systematic narrative review

Oxytocin (OXT) is a neuropeptide and hormone implicated in prosocial human social and emotional functioning [1]. OXT has been long known for its key role in child birth and postpartum lactation, including changes in infant gut hormones [2]. OXT plays a crucial role in enhancing bonding and moderating the stress response. Mothers’ neuroendocrine response to their baby’s cry, as measured by functional neuroimaging (fMRI), has shown to differ between vaginal and caesarean births, with higher OXT levels linked to enhanced empathy and maternal arousal [3]. Mothers with higher plasma OXT levels during early pregnancy have been shown to be more responsive to their baby after birth, including higher levels of gaze towards their infant’s face, positive affect and affectionate touch [4].

Postpartum, the sucking stimulus by newborns increases OXT release in the mother and thus decreases plasma levels of stress hormones (ACTH and cortisol) [5]. The physiological consequences of disrupted OXT release are thus significant during the perinatal period, however, there is also mounting evidence that disrupted OXT availability could be related to a broad range of social behaviours, including childhood autistic spectrum disorders (ASD) [6], attachment bonds in infancy [7] and mood disorders and response to stress in women [8].

This has led to the development of OXT treatments including using nasal OXT to increase attachment security in healthy males with an insecure attachment pattern [9] and improved social engagement between fathers and their infants [10]. Further, intranasal OXT increased trust and willingness to accept social risks amongst healthy volunteers involved in an experimental game [11], as well as the reduction of the short-term stress response demonstrated by reduced cortisol release [12]. In patients with Autistic spectrum disorder (ASD), intranasal OXT has been shown to provide some improvements in adult social cognition [6] and youth emotional recognition [13].

fMRI studies have shown that during a visual stress task, intranasal OXT administration is associated with reduced activation in the amygdala, which plays a key role in emotion and social behaviour regulation [14]. Furthermore, central OXT has both anxiolytic and antidepressant effects which have been associated with improved emotional behaviours and mental health [15]. This has led to the suggestion that OXT may be an important biomarker of stress exposures [16]. OXT plasma concentrations also correlate with cerebrospinal concentrations and with childhood anxiety [17]. Disrupted peripheral OXT release has been found in non-perinatal women with depression during stress related tasks [8] and with specific types of relationship stress in women [18]. Taken together, these findings suggest an important role for OXT in human social and emotional functioning.

Neurobiologically, OXT is synthesized by the OXT gene in magnocellular neurons in the paraventricular and supraoptic nuclei of the hypothalamus. OXT is then transported along axonal projections to the posterior lobe of the pituitary, where it is stored in secretory vesicles before being released into the peripheral blood. In addition, dendritic release of OXT occurs into the extracellular space with diffusion through the brain. Smaller parvocellular neurons in the paraventricular nucleus also produce OXT and project directly to other regions in the brain involved in human social and emotional functioning including the amygdala, hippocampus, striatum, suprachiasmatic nucleus, bed nucleus of stria terminalis and brainstem - where they act as neuromodulators or neurotransmitters and thereby influence neurotransmission in these areas [19].

Oxytocin receptors are synthesised by the Oxytocin Receptor (OXTR) gene, which expresses both centrally in the brain and within peripheral organs. In this way, OXT has both peripheral and central functions. OXTR is located on human chromosome 3p25.3 [20, 21]. The OXTR gene spans 17 kilobytes (kb) and contains 3 introns and 4 exons. Exons 1 and 2 correspond to non-coding regions while exons 3 and 4 encode the amino acids of the oxytocin receptor (Fig. 2). The oxytocin receptor is a 389-amino acid G-protein coupled transmembrane receptor, enabling the activation of a number of different intracellular secondary messengers facilitating the oxytocin pathway [22].

Genetic variants, including single nucleotide polymorphisms (SNPs), in the OXTR have been shown to independently influence gene expression and been associated with human emotional responsiveness and social behaviour [23]. Two SNPs rs2254298 and rs53576 have been most investigated regarding the OXTR (Intron 3) (Fig. 2). Rs2254298 (G > A) has also been associated with autistic spectrum disorder (ASD) [24], depressed women and their families [25] and with smaller bilateral amygdala volumes and greater grey matter volumes in a brain imaging study [26], while rs237887 has been associated with human social recognition skills [27]. Rs53576 (A > G) has been associated with reduced positivity [28], reduced empathy [29] and decreased parental sensitivity [30] while rs1042778 has been associated with reduced parenting sensitivity [31] and rs237915 has been associated with brain responsiveness to social cues [32]. A meta-analysis of 11 independent populations of patients with ASD reported associations with rs7632287, rs237887, rs2268491 and rs2254298 [33].

The focus of this review is on the newly emerging field of OXTR epigenetics and socio-emotional functioning in humans. Epigenetic mechanisms are reversible modifications occurring “above” the level of the DNA sequence, which can influence gene expression [34]. The most widely investigated and understood epigenetic mechanism is DNA methylation (DNAm), which commonly involves the addition of methyl groups to the cytosine and guanine dinucleotides termed “CpG” dinucleotides in the DNA sequence. OXTR is of particular interest because, unlike OXT, which has tight regional expression, OXTR is expressed extensively through brain and body and can thus control the final (rate-limiting) step in signal transmission. Differential methylation of the OXTR has been associated with changes in OXTR expression both in animals and in humans.

In mouse models, OXTR DNAm patterns have been correlated with differential OXTR expression in different body tissues [35]. In different brain areas, differing DNAm at specific CpG regions within the OXTR occurred with differential expression of OXTR [36]. The differential OXTR expression is considered to relate to a diversity of animal social behaviours arising from the limbic brain structures. Though animal and human OXTR gene promoter brain regions differ, the inference is that DNAm differences result in OXTR transcription changes with impact on human social and emotional behaviour.

In human samples, a region within the OXTR third intron between regions encoding trans-membrane domains 6 and 7, has been shown to be hyper-methylated within non-expressing tissues, but hypo-methylated in the uterine myometrium at term, when OXT is up-regulated for the birthing process [21]. Further work on OXTR has identified a region between the first exon and the first intron (MT2 segment) that is linked to tissue specific expression OXTR in peripheral blood mononuclear cells (PBMCs), liver, and uterine myometrium (from non-pregnant and term-pregnancies) [37]. Genetic variation may also influence patterns of DNAm within the gene [38–41] and thus can have both combined and independent effects on gene expression [42, 43]. However, lack of replication to date requires cautious interpretation.

The purpose of this review by narrative synthesis is to bring together the emerging evidence on OXTR DNAm and human socio-emotional functioning. The diversity of studies available in this new area of research currently precludes meta-analytic presentation of findings. This review provides a timely summary of a new field of research around OXTR and socio-emotional behaviour, with implications for future promotion of emotional health in both population and clinical settings.

A review of the literature was undertaken to identify all published studies that investigated oxytocin receptor gene (OXTR) DNAm in human social behaviour, with a focus on social and/or emotional functioning. We defined human social and emotional functioning as the experience, expression and management of emotions as well as the ability to establish positive and rewarding relationships with others [44]. This extends to relationships, response to stress and social cognition with a basis in neurobiology, which has become more available to research through functional brain scanning.

Figure 1 presents a PRISMA flow diagram showing initial study identification and stepwise exclusions of ineligible studies. A total 374 articles were initially identified for review. Of these a total of 19 were eligible for inclusion in the final review. Details of each are provided in Table 1. Study populations varied widely in age, heritage and from general population to clinical samples. Sample sizes were variable and predominantly small though size has been increasing in more recent research [45]. Most studies focused on the “MT2 segment” (between Exon 1 and Intron 1), the region identified as functionally significant in the initial gene expression studies (Fig. 2, Area A) [21, 37, 46]. Increased DNAm of this region, including the site − 934, has been linked to decreased OXTR gene transcription in both PBMCs and brain [46]. No study has examined the extent to which DNAm influences gene expression in Exon 3, (Fig. 2, Area B).

Study	Characteristics	Main Measure	Tissue	OXTR region & measurement	Methylation Association	Adjustment
Gregory et al. (2009) [46]	(i). 119 Autistic Spectrum Disorder (ASD) probands (78% male) & families. (ii). 20 ASD (50% male) & 20 controls. (iii). 16 brain samples in 8 ASD (75% male) & 8 controls (Caucasian US)	ASD (DSM-IV criteria) & ADI-R	PMCS & brain tissue (temporal cortex)	2 regions: (i) exons 1, 2 & 3 incl. MT2 a (ii). 3rd intron No SNPs DNAm Specific sequencing	i) ASD Family study: ↑ DNAm 5/22 CpG sites (− 860, − 901, − 924, − 934 & -959) in region (i) with significant associations at − 860 & - 934 in males & -959 in females ii) ASD vs controls: ↑DNAm 3 CpG sites (− 860, − 934, − 959). iii) Brain samples: ↑ DNAm 4 CpG sites (− 860, − 901, − 924, − 934) iv) ↓ Gene expression: PCR ↓mRNA in ASD male brains	Matched: age & gender.
Jack et al. (2012) [42]	42 healthy adults (55% male) 18–30 yrs. Inclusion: normal vision & no previous psychiatry (67%Caucasian US)	fMRI brain scan of during visual perception animation tasks.	PMCS	- 934 single site No SNPs Pyrosequencing	↑ DNAm site −934 associated with ↑fMRI activity in temporo-parietal junction incl. dACC during tasks.	No racial or gender differences
Unternaehr-er et al. (2012) [48]	76 adults (43% male) 61–67 yrs. (Caucasian Germany).	TSST	3× whole venous blood: before TSST, 10 & 90 mins	2 regions in exon 3: OXTR1 & OXTR2 No SNPs Sequenom EpiTYPER	↑DNAm 7/8 CpG sites OXTR1: pre to 10 mins post-test & ↓DNAm from 10 mins to 90 mins post-test. ↓DNAm trend 2/27 CPG sites OXTR2: 10 to 90 mins. Post-test.	Blood cell counts & batch effects. No sex differences. Multilevel models
Cecil et al. (2014) [53]	84 youth with conduct disorderb (no gender ratio), (UK).	High Callous Unemotional (CU) traits maternal assessed at 13 yrs	PMCS in cord blood, 7 & 9 years	12 probes across gene. SNPs:4 CU focus Illumina array 450 Kc	↑DNAm 3/12 probes on exon 2 in cord blood with ↑ CU in low internalising problems group & with ↑parental risk exposures (prenatal & late-childhood. SNPs rs2301261, rs237915, rs4564970, rs4686302 NSd with DNAm, CU or internalising groups.	Gender stratified analysis (internalising problems)
Dadds et al. (2014) [49]	98 male; 4–16 yrs.: conduct disorder or Oppositional Defiant Disorder (ODD) Exclusion: ASD, neurological /physical illness, ↓IQ. (Australia).	i).ODD /Conduct Disorder assessmente ii). CU traitsf	Blood	11 probes: exon 1 - intron 1. No SNPs Sequenom MassARRAY	Children 9-16 yrs., ↑ DNAm associated with ↑ CU traits. Examining the 6 individual CpG sites, these findings were predominantly with CpG 5 Subset 37 youth ↑DNAm associated with ↓ circulating OXT protein levels, in older children only.	Covariates tested.
Kim et al. (2014) [39]	51 females, 18 + yrs. (15 Anorexia Nervosa (AN); 36 healthy controls (South Korea) Exclusion: ASD, psychosis & substance use disorder.	SCID (DSM-IV) & EDE-Q, BMI, AQ (ASD traits), DBI & STAI.	Buccal	48 probes: exon 1-intron 1. No SNPs DNAm specific sequencing	↑DNAm 5/48 CpG sites & averaged across region associated with AN. ↓BMI correlated ↑DNAm at 5/48 CpG sites ASD traits did not associate with DNAm	Age, BMI & clinical variables (ASD, anxiety & depression)
Bell et al. (2015) [34]	269 female postnatal depression (131 antenatal depression); 276 matched controls (135 antenatal depression). 2/3 s: 25–34 yrs. (Caucasian UK)	EDPS: antenatal & postnatal	Whole antenatal blood	- 934 single site SNPs rs53576 & rs2254298 Pyrosequencing	↑DNAm for females with postnatal depression only in rs53576 GG	Matched: age, parity, depress-ion. Adjusted Models: psychosocial covariates.
Chagnon et al. (2015) [33]	43 females > 65 yrs. (18 prev. Anxiety & MDD, 1 with MDD; 24 controls. Excluded: psychosis, schizophrenia or MMSE score < 22. (Canada).	Anxiety & MDD (DSM-IVR)g	Saliva	9 probes in 2 regions: exon 3. SNPs: rs53576 Pyrosequencing	No difference between groups overall. ↑DNAm (× 2) for females with anxiety/depression who were rs53576 AA when averaged across region 1 (4 probes) and for 1/5 CpG sites of region 2.	Matched: no details
Puglia et al. (2015) [47]	98 healthy adults, (43% male) 18–30 yrs.: normal vision. (Caucasian, US)	fMRI brain scan during emotional face matching task	PMCS	- 934 single site No SNPs Pyrosequencing	↑ DNAm associated with fMRI responses: ↑ face & emotion processing areas (amygdala, fusiform & insula) & ↓ connectivity between social perception systems & ↓ coupling amygdala & emotional regulation regions.	Gender stratified analysis. Age: NS
Reiner et al. (2015) [35]	85 premenopausal female adults (43 depressed & 42 controls), aged 19–52 yrs. Excluded: psychosis, medical, personality, eating & substance use disorders. (Caucasian Germany).	MDD (SCID-DSMIV)	PMCS	Mean DNAm: exon 1(36 sites); exon 2(6 sites) SNP: rs53576 DNAm Specific sequencing (ESME software)	↓DNAm in exon I (not exon 2) in depressed females. ↓DNAm in depressed females rs53576 GG > AG/AA but in healthy controls ↑DNAm in GG > AG/AA. ↓DNAm in exon 2 associated with rs53576 GG & not depression	Matched: age and education. Mixed models: age, alcohol, smoking, BMI antidepressant use.
Unternaehr-er et al. (2015) [52]	85 university students with maternal care: low (45) and high (40) (79% female) aged 19–66 (median 24). (Caucasian, Switzerland)	PBI (maternal care) when < 16 years: retrospective assessment	Whole blood	2 regions in exon 3 (OXTR TS1 & OXTR TS2 ). (35 CpG units) No SNPs Sequenom EpiTYPER	↑ DNAm in low versus high maternal care in OXTR TS2 ↓DNAm in males compared to females independent of maternal care in OXTR TS2 .	Adjusted: age, gender, batch effects. Mixed model analysis: cell count, BMI & depression considered.
Ziegler et al. (2015) [51]	220 adults: 110 un-medicated social anxiety (76 female & 34 male); 110 controls (77 female & 33 male) Excluded: medical, psychosis & substance use disorders. (Caucasian Germany)	Psychological: SIAS, SPS; cortisol response to TSST, fMRI of amygdala during social anxiety word task.	Whole blood	Region in exon 3 with 12 CpG sites. SNP: rs53576 DNAm Specific sequencing (ESME software)	↓DNAm at 1/12 CpG site with psychological measures, cortisol response & fMRI. ↓average DNAm across all 12 CpG sites in social anxiety patients compared to controls. ↓average DNAm & at 7/12 CPG sites for rs53576 AA/AG > GG.	Matched: age and sex
Cappi et al. (2016) [40]	73 adults: 42 OCD (9 CBT or fluoxetine /33 treatment naïve); 31 controls: 18–65 yrs. Excluded: psychosis, suicide risk, prev. Head injury, medical & substance use disorders (Brazil)	OCD(DSMIV): Y-BOCS score ≥ 16 (obsessions & compulsions) or ≥ 10 (obsessions or compulsions), BDI, BAI, YGTTS.	Peripheral blood leucocyte ≤2 weeks treatment	2 regions exon 3:OXTR 1 (4 CpG sites) & OXTR 2 (5 CpG sites). No SNPs Pyrosequencing	↑DNAm at 2/9 CpG sites in OCD > controls. ↑DNAm correlated with ↑OCD severity ↑DNAm at 8/9 CpG sites with ↓BDI score. No difference DNAm between 33 treatment naïve or 9 fluoxetine/CBT. No DNAm correlation between OCD & tics.	Examined age, sex, education, BDI score, BAI score, tics, YBOCS score
Elagoz Yuksel et al. (2016) [41]	66 children: 27 ASD (23 male & 4 female); 39 controls (6 female & 33 male), 22–94 months. Excluded: medication, ↓development, neuro-degenerative, medical & psychiatric disorders(Turkey)	ASD (DSMIV-TR) CARS (≥30 for ASD).	Peripheral blood	4 regions: exon 1 – exon 3 (MT 1–4). No SNPs DNAm Specific sequencing	↓DNAm in 2/4 regions: MT1 & MT3 in ASD > controls. MT1 site (exon 1) DNAm %: 44.4% ASD & 71.8% healthy controls MT3 site (intron 1, exon 2, intron 2) DNAm %: 29.6% ASD & 61.5% healthy controls.	Matched: age & gender.
Kimmel et al. (2016) [43]	51 females postnatal depression: average 30.6 yrs. (70% Caucasian). Previous mood disorder (66% MDD, 33% Bipolar Disorder) (US)	MDD (DSMIV) ≤ 1st month >birth.	Whole blood T3h	18 CpGs sites SNP: rs53576 Illumina array 450 K (Brain Cloud tool)	↓DNAm 1/18 CpG site in female postnatal depression. rs53576 NS	Adjusted cell proportions.
Rubin et al. (2016) [50]	242 adults: 167 with psychosis (affective/bipolar disorder & non-affective/schizophrenia (92 female & 75 male); 75 healthy controls (38 female & 37 male). Excluded: previous head injury, ↓ reading, medical & substance use disorders (US)	Psychosis (DSMIV); Penn Emotional Recognition Task; fMRI during facial recognition task (79% sample)	Whole blood	A single CpG site (−934) No SNPs Pyrosequencing	↑DNAm schizophrenia > bipolar disorder ↑DNAm psychotic females >males ↑DNAm with ↓emotional recognition in females & controls > males. ↑DNAm with ↓ brain volumes: temporal-limbic and prefrontal regions (social cognition) in healthy controls & schizophrenia females. Plasma OXT levels and DNAm NS in psychotic/ controls Gender differences in plasma OXT & DNAm: (−) males & (+) for females.	Age, race, intracranial brain volume Sex stratified.
Smearman et al. (2016) [36]	393 adults (70.7% female) 18–77 yrs.: childhood abuse & current anxiety/depression (African American, US).	CTQ, HAMA, BDI Traumatic Events Inventory	Whole blood	18 CpG sites (before exon 1-intron 3) SNPs 44 Illumina array 450 K	↑ DNAm associated childhood abuse at 2/18 CpG sites but NS. DNAm not a mediator of psychiatric symptoms. 44 proximal SNPs: 68% associated DNAm of nearby CPG sites.	Age, sex & cell type
Unternaehr-er et al. (2016) [44]	39 infants with mothers mean age 31.9 yrs. (no sex ratio)(Switzerland)	Pregnant mother: EDPS, TICS-K, ILE, saliva cortisol in T2i, T3h.	Cord blood	13 CpG sites (exon 3) No SNPs Sequenom EpiTYPER	↓ DNAm associated with ILE, EDPS, T2 maternal cortisol (AUCg).	Batch effects, demographics, pregnancy & births.
Rijlaarsdam et al. (2017) [45]	743 children (51% male) ≤6 years yrs. (Netherlands)	ASD 6 yrs. with parental ratings: SRS. Maternal prenatal stress	Cord blood	3 CpG sites across gene. SNP: rs53576 Illumina array 450 K	DNAm & prenatal maternal stress or ASD: NS ↑DNAm associated ↓social communication: rs53576 GG	Child sex, age, cell type, maternal smoking & technical array.

a. MT2 segment is a genomic region on OXTR identified by Kusui, 2001; b. Diagnostic Interview Schedule for Children, Adolescents & Parents; c. 450 K: Illumina Human Methylation 450 Bead Chip array (Illumina, USA); d. Not Significant; e. Antisocial Process Screening Device/ASPD & pro-social subscale of Strengths & Difficulties/ PSSSDQ; f. Youth with early onset & persistent conduct problems assessed by Strengths & Difficulties Questionnaire ‘Conduct problem’ subscale; g. Diagnostic Interview Schedule and Composite International Diagnostic Interview for anxiety and depressive disorders; h. pregnancy second Trimester; i. pregnancy third Trimester

Only one of the 19 studies directly examined the relationship between OXTR DNAm and gene expression [46]. Two studies investigated a possible association between DNAm changes at the OXTR and subsequent oxytocin neuropeptide synthesis [47, 48]. As DNAm marks can be cell and tissue specific, it is important to note that most studies have used DNA derived from blood samples. One study each used buccal [49] and saliva samples [38], the latter of which may contain buccal epithelium and blood cells, including peripheral blood mononuclear cells (PBMCs). A third study investigated a small subset of brain tissue collected post-mortem [46].

The role of OXTR DNAm in human social and emotional functioning is a relatively new area of investigation. There are currently few studies in the field and methods and findings are highly variable, precluding firm conclusions about the role of OXTR DNAm. However, within this context, several non-replicated associations between OXTR DNAm and outcomes in various domains are emerging in the literature. These include positive associations between OXTR DNAm and callous-unemotional traits in youth [47, 58], social cognitive deficits in ASD [45, 46], rigid thinking in anorexia nervosa [49], affect regulation problems [41, 57] and mood deficits [38–40] as well as limbic regions linked with facial and emotional recognition [48, 50, 51]. In contrast, reduced OXTR DNAm has been associated within indicators of perinatal stress [55], postnatal depression [56], social anxiety [53] and autism in children [60].

Taken together, findings suggest a role for increased OXTR DNAm (which may indicate reduced receptor expression) in general impairments in social, cognitive and emotional functioning, and decreased OXTR DNAm (which may indicate increased receptor expression) in specific patterns of impairment related to mood and anxiety disorders (but not all). However, patterns of association both within and across outcomes are difficult to clearly demarcate, and most likely reflect the broad range of differences that exist between studies at this early point in the development of the field. For example, reported associations for mood disorders are in both direction, which may reflect striking differences between studies, including type of mood disorder under investigation, epigenome regions studied, underlying variation in genotypic and hormonal factors [39, 56] as well as the effect of aging [38].

Furthermore, exceedingly little is known about the functional implications of OXTR DNAm in humans. Only one study has reported a relationship between increased OXTR DNAm and reduced OXTR expression [46]. Another has reported a relationship between mean OXTR DNAm across eleven probes (spanning exon 1 to intron 1) and circulating oxytocin protein levels; however, this was not replicated in another study examining a single CpG site (− 934) in intron 1 [48]. This is an important area for future research. Genetic association findings are equally non-conclusive with results differing considerably across the studies. Lack of reproducibility is common in genetic association studies and likely reflects small sample sizes of unknown representation with underpowering. Future progress on genetic influences will depend on access to larger samples and benefit by a shift from its current focus on candidate SNPs to genome-wide analysis (GWAS). Furthermore, replication opportunities are essential for establishing the credibility of reported genetic associations. Earliest canvassing of these opportunities would permit optimised alignment of research samples and instrumentation.

A further driver of heterogeneity of DNAm findings is choice of tissue for analysis. Studies in humans are limited to peripheral tissue collections i.e. buccal, blood or saliva. The common embryological origin of brain and buccal cells suggests that these two cell types may have more in common than blood cells, and that using buccal cells as a peripheral biomarker might provide a closer approximation to central brain processes, relative to other cell types [61]. At minimum, the tissue specific nature of many epigenetic processes is likely to be an important contributing factor. The different methods of analysis of OXTR DNAm from individual probe methylation to EWAS analysis reflect the challenges of cost, in a new area of investigation which further complicate result comparisons.

From a statistical methods perspective, the majority of studies have investigated methylation at numerous individual CpG sites [38, 40, 41, 45, 46, 49, 52–55, 57, 58], but only a few studies have considered adjustment for multiple comparisons in their analysis, [41, 47–49, 53, 55, 56], to reduce the risk of false positive findings (type 1 errors). This is particularly problematic for studies which have investigated a large number of gene regions and CpG sites. It thus remains possible that at least some of the findings shown here are chance associations, and replication in an independent sample is crucial. On the other hand, some of the studies have used factor analysis to reduce the number of CpG sites and thus tests [40, 45, 47, 58]. The downside of these studies is that factor analysis is a data driven process and does not necessarily make biological sense. Further, conclusions cannot be drawn about the impact of DNA methylation at individual CpGs.

Longitudinal cohort studies beginning in the perinatal period provide the optimal opportunity to study humans at crucial time points during life. This informs knowledge as to how epigenetics contributes to the changing phenotype. Ng et al. identified that in thirty-four life course studies related to the epigenetic mechanism of DNAm, only four studies included information taken from more than 1-time point. Of these four, one study involved epigenetic changes over a relatively short time period (28–180 days) [62]. The Avon Longitudinal study of Parents and Children (ALSPAC) has collected serial samples across the life-span from individual participants providing the gold standard in longitudinal cohort studies for epigenetic analysis [63]. Longitudinal studies which collect biomarkers, at multiple time-points, starting in the perinatal period for epigenetic investigation are unique as an assessment of the changing life-course phenotype. Cross-sectional analyses of epigenetic data lack utility in attempting to understand developmental change or causation direction in epigenetic epidemiology [64].

This narrative literature review examines findings from nineteen papers, which have investigated OXTR DNAm associated with social and emotional functioning in humans. The broad and heterogeneous nature of these studies preclude definitive conclusions. This is typical of a new field of research where the study diversity in design, methodology, phenotype and outcome make meaningful conclusions challenging.

Despite these considerable limitations, emerging evidence points to increased OXTR DNAm in general impairments of social, cognitive and emotional functioning, and decreased OXTR DNAm in specific patterns of impairment related to mood and anxiety disorders (but not in all). These higher-level patterns are speculative at best in this stage of the field’s early development. The essential “next step” is to establish with greater certainty the functional relationship between OXTR DNAm and actual OXTR gene expression. Related to this is greater clarity about potential differences in OXTR DNAm across peripheral tissues currently collected in human studies in relation to central brain tissue OXTR processes. If these fundamentals can be established, progress would be facilitated by increased homogeneity in study design, including phenotypic definition and measurement. This includes greater investment in high quality measurement, such as micro-coding of social interaction using gold-standard protocols, for example, human attachment behaviour, within observational designs [65]. Future progress would also be enhanced by greater investment in embedded DNAm studies within longitudinal designs that enable temporal ordering of relationships with biological collections at multiple time-points. The Developmental Origins of Health and Disease (DOHaD) paradigm emphasises the centrality of exposures within the first 1000 days from conception, however sensitive periods may occur both before and after birth [66]. Longitudinal cohort studies, particularly those with mature cohorts that cross more than one generation, will provide important future opportunities for investigating change over time. Of particular interest are longstanding cohorts prospectively tracking offspring development, given their unique opportunity to explore effects beyond a single generation, to those that transmit to subsequent generations.