The serotonergic system plays a pivotal role in socio-emotional stress response (10, 11). Serotonin (5-HT) receptors are widely distributed throughout the central nervous system, they appear early during gestation, and the entire serotonergic system rapidly develops during the first months of life (12). The functioning of the serotonergic system is regulated by feedback mechanisms through the serotonin transporter (5-HTT), which is encoded by the SLC6A4 gene (13). The transcriptional activity of the SLC6A4 is regulated both by genetic variations and epigenetic mechanisms. Among the genetic variants, previous studies have mainly investigated the role of a transporter-linked polymorphic region (i.e., the 5-HTTLPR polymorphism) in affecting infants’ stress response (14–17). The 5-HTTLPR has a short (S) and long (L) allelic variants (18), with the former being linked to reduced 5-HTT transcription and heightened risk for adverse developmental outcomes, including socio-emotional dysregulation and stress susceptibility (10).

Notably, the genetically determined variability conveyed by the 5-HTTLPR polymorphism has proved to be insufficient to independently account for differences in socio-emotional stress response and susceptibility (19, 20). As such, during the last decade, epigenetic mechanisms regulating SLC6A4 transcription have been increasingly studied in association with early adversities and behavioral developmental outcomes (21). DNA methylation is the most widely studied epigenetic mechanisms in humans, and it consists in the addition of a methyl group to the specific CpG sites within a gene promoter region, leading to reduced transcriptional activity and, eventually, to gene silencing (22). Converging evidence suggests that increased SLC6A4 methylation might be a marker of early adversities in humans and might play a role in altered developmental trajectories of stress response and susceptibility (23). For example, maternal depression during pregnancy (24), childhood maltreatment (25), and disadvantaged socioeconomic conditions (26) have been found to associate with CpG-specific patterns of altered methylation within the SLC6A4 promoter region.

Socio-emotional stress response refers to the ability to respond adaptively to an interpersonal challenging condition (27). Early infants’ socio-emotional stress response is thought to be predictive of adjustment later in life (28). Better stress response has been shown to contribute to adaptive socio-emotional functioning (29), as well as to the development of secure attachment (30). Past research has suggested that infants’ socio-emotional stress response is particularly apparent in response to a brief relational disruption during the FFSF paradigm (31). During the FFSF, infants first engage in normal face-to-face interaction with their caregiver (Play episode). After that, mothers are asked to exhibit a neutral facial expression, avoid touching, and responding to their infant (Still-Face episode). During the Still-Face episode, infants usually exhibit a typical behavioral pattern of stress response (i.e., still-face effect), characterized by increased negative emotionality (32). Subsequently, mothers and infants resume their face-to-face interaction as in the first episode (Reunion episode). The FFSF paradigm has been extensively used to study infants’ socio-emotional stress response to maternal unresponsiveness (33, 34).

Previous FFSF research documented that infants’ socio-emotional stress response might be affected by both infants [e.g., VPT birth; (35, 36)] and caregiver [e.g., maternal sensitivity; (30, 37)]. A handful of studies have reported poorer regulatory skills in preterm infants compared to their FT counterpart. For example, moderately preterm infants (i.e., gestational age between 31 and 34 weeks) show heightened susceptibility to maternal still-face at 2 months (38), reduced self-soothing attempts at 3 months during the Still-Face episode (39), and more avoiding strategies at 9 months during the Reunion episode (40).

Caregiving behavior is another factor affecting infants’ stress response. The social buffering hypothesis suggests that maternal sensitivity may moderate the effects of early stress exposures on infants’ behavior and physiology (41, 42). Sensitivity is generally defined as a set of maternal caregiving features including the ability to detect infants’ gross and subtle communicative signals, the acceptance of infants’ behavioral signals and requests, the providing of responses contingent in time and intensity to infants’ signals, the mirroring and scaffolding of infants’ socio-emotional experience in tactile and face-to-face interactions. Maternal sensitivity is associated with an increased ability to regulate stress in human infants (43–45), and it is associated with better developmental outcomes later in life (46). Importantly, maternal behavior has been found to mediate the susceptibility effects conveyed by the 5-HTTLPR S allele in 4-month-old infants during the FFFS: high levels of maternal social engagement reduced the negative emotionality exhibited by S-carrier infants to levels comparable to those of L-homozygous subjects (15).

Behavioral epigenetics has been applied to the study of VPT infants’ socio-emotional development (6). Emerging evidence suggests that epigenetic mechanisms might be involved in setting the risk of altered developmental trajectories of stress response in VPT infants. For example, despite no significant difference in SLC6A4 methylation was detected at birth between VPT and FT infants, a CpG-specific methylation increase was observed at discharge within the promoter region of the serotonin transporter gene in VPT infants, as a function of the number of skin-breaking procedures to which they were exposed during the NICU hospitalization (47). When these infants were exposed to the FFSF procedure, they showed heightened stress susceptibility compared to FT controls and the amount of negative emotionality exhibited during both Still-Face#2 and Reunion#2 episodes was predicted by the CpG-specific increase in SLC6A4 methylation observed at NICU discharge (48). In a cross-sectional study, Chau et al. (49) reported a similar association of early pain-related stress exposure in NICU and 7-year-old VPT children, further suggesting that the SLC6A4 methylation at 7 years was significantly correlated with an increased risk of behavioral problems. Taken together, these findings suggest that epigenetic mechanisms might contribute to long-lasting alterations of the SLC6A4 methylation status and to associated detrimental effects on behavioral and socio-emotional development during infancy and childhood.

Recent research also highlights that variations in the quality of caregiver behavior has an impact on the epigenetic status of stress-related genes (9). Rodent offspring raised by mothers characterized by low levels of sensitive caregiving exhibited high methylation and reduced expression of the gene encoding for glucocorticoid receptors (i.e., NR3C1) and had lower stress response skills during adulthood (50). Nonetheless, when these rodents were cross-fostered to mothers characterized by high levels of care quality, these effects were reversed (9). Although preliminary, similar evidence has recently emerged for humans. Conradt et al. (7) documented that mothers who were more responsive and who engaged in more appropriate touch during face-to-face interactions had infants with less CpG-specific methylation of a stress-related gene (i.e., NR3C1) compared to mothers characterized by low sensitivity. This evidence gives new support to the hypothesis that quality of early maternal sensitivity might moderate the effects of early epigenetic alterations due to stressful conditions even in human infants. The potential buffering effect of maternal caregiving on epigenetic changes observed in VPT infants has been recently suggested (51, 52), but there is no evidence to date.

Previous research documented that NICU-related stress might contribute to altered SLC6A4 methylation in VPT infants (47) and that these epigenetic variations might lead to increased socio-emotional stress susceptibility at 3 months in response to the FFSF procedure (48). Using the FFSF paradigm, the present study was aimed at exploring the potential buffering role of maternal sensitivity in moderating the relationship between CpG-specific SLC6A4 methylation and 3-month socio-emotional stress response in VPT and FT infants (corrected age for prematurity, CA). Consistent with the abovementioned literature on maternal buffering effect, we hypothesized that high levels of maternal sensitivity would be associated with better socio-emotional stress response in both groups.

The present study is part of a longitudinal research project on the epigenetic and behavioral effects of early adverse experiences in NICU in VPT infants. In previous works, we have reported partial data about SLC6A4 methylation and infants’ behavioral development during the first months of life (47, 48, 53). Here, we focus on the potential role of maternal sensitivity in moderating the association between altered SLC6A4 methylation and stress response at 3 months CA. A cohort of 88 infants and their mothers was consecutively recruited between October 2011 and April 2014. Of the initial sample, 53 infants had complete infant and maternal data at 3 months CA and were included in the present study. Included and excluded infants did not differ for any of sociodemographic and perinatal variables. VPT infants (N = 27, 14 females) were enrolled at the NICU of the Department of Clinical Sciences and Community Health, Fondazione IRCCS Ca’ Granda Ospedale Maggiore Policlinico of Milan (MI, Italy). Inclusion criteria for VPT infants were: gestational age less than 32 weeks and/or birth weight ≤1,500 g, no major brain lesions on cerebral ultrasound (intraventricular hemorrhage >2 according to Papile, periventricular leukomalacia >1), no neuro-sensorial deficits (retinopathy of prematurity >2), no genetic syndromes, and/or major malformations. FT infants (N = 26, 14 females) were recruited at the Pediatric Unit of the Sacra Famiglia Hospital, Erba (CO, Italy). FT infants were all healthy, with gestational age ≥37 weeks, no postnatal complications, or prenatal/perinatal risk factors. For both groups, mothers’ inclusion criteria were as follows: Italian nationality, age over 18 years and no single parent. Medical charts were used to screen and exclude mothers documenting cognitive impairments, manifest psychiatric disorders, prenatal depression or anxiety, use of psychotropic medication during pregnancy or drug/alcohol addiction. Mothers and fathers of all participating infants provided written informed consent. According to previous studies on this topic (47), the present sample size appears to be adequate to detect significant variations in CpG-specific methylation of the SLC6A4 gene with alpha error set at 5% and beta error set at 10% (i.e., statistical power = 0.90). The research project has been conducted in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki, seventh revision, 2013) and the study has been approved by the Ethics Committees of the Scientific Institute IRCCS Eugenio Medea, Bosisio Parini (LC, Italy) and participating hospitals.

An overview of the research study plan and the timing of epigenetic and behavioral procedures is reported in Figure 1. The NICU was a traditional open space level III unit, where parents had 24-h access (54). Developmental care was not clustered in protocols (e.g., Newborn Individualized Developmental Care and Assessment program, NIDCAP). The NICU had a maximum of 23 incubators, without single-room facilities. The parents had reclining chairs available throughout the stay. Consistently to previous research (24, 47) and to control for post-conceptional age at the epigenetic assessment, methylation was assessed from cord blood at NICU discharge in VPT infants and from cord blood at birth in FT controls. Blood samples were obtained by trained nurses to avoid hemolysis and immediately stored at −20°C and sent to the lab. Infants’ clinical variables were obtained from medical records. At 3 months CA, the FFSF procedure occurred during a home visit and mothers completed a sociodemographic survey including the collection of neonatal variables and maternal sociodemographic characteristics as well as maternal depressive and anxious symptoms.

Very preterm and FT infants were compared for neonatal and sociodemographic variables by means of t-test and chi² test for continuous and categorical variables, respectively. Socio-emotional stress response was compared between VPT and FT infants through mixed model analysis of variance, with negative emotionality as the output variable and FFSF episodes as the within-subject factor (5 levels: Play, Stil-Face#1, Reunion#1, Still-Face#2, Reunion#2) and Group as the between-subject factor (2 levels: VPT vs. FT). Maternal sensitivity during the Play episode was compared between the two groups. To assess the moderating role of maternal sensitivity, two sets of stepwise multiple linear regressions were performed separately for VPT and FT infants.

Separate regression models are generally used when comparing preterm vs. FT infants’ outcomes [e.g., Ref. (49)] because VPT and FT infants’ life experience is very different and different confounders need to be controlled for. To select predictors, preliminary bivariate correlations were run to test the association between (a) maternal sensitivity and infants’ negative emotionality across the FFSF episodes; (b) SLC6A4 CpG2 and CpG5 methylation and infants’ negative emotionality across the FFSF episodes. The final model was tested on infants’ negative emotionality considering only the FFSF episode(s) for which significant preliminary had emerged, and it included the following predictors: methylation of CpG sites significantly correlated with infants’ negative emotionality during the specific episode, maternal sensitivity during Play, and the interaction between CpG-specific SLC6A4 methylation and maternal sensitivity. Finally, the regression model was controlled for clinical confounders. For both VPT and FT infants, confounders were: gestational age (weeks), maternal age (years), maternal educational level (years of study), family SES, maternal depression, maternal anxiety. Moreover, for VPT infants only, pain-related stress index, days on ventilation, and length of stay in the NICU was added as confounders. All the analyses were carried out using IBM SPSS Statistics 21, at α < 5%.

Descriptive statistics for the sample are reported in Table 1. Gender distribution did not differ significantly between VPT and FT groups, X² = 0.11, p = 0.73.

Significant main effects of FFSF episodes, F(4, 48) = 20.89, η²_p = 0.29, p < 0.001, and Group, F(1, 51) = 12.20, η²_p = 0.19, p = 0.001, were better specified by the interaction effect FFSF episodes X Group, F(4, 48) = 4.34, η²_p = 0.27, p = 0.004 (see Figure 2). Negative emotionality during Play did not differ between groups, t(51) = 1.90, p > 0.05. Compared to FT counterpart, VPT infants showed higher levels of negative emotionality during Still-Face#1, t(51) = 2.45, p = 0.018, Reunion#1, t(51) = 2.49, p = 0.018, Still-Face#2, t(51) = 3.51, p = 0.001, Reunion#2, t(51) = 3.65, p = 0.001.

Preliminary correlations are reported in Table 2. No significant correlations emerged for both SLC6A4 methylation and maternal sensitivity with infants’ negative emotionality during Still-Face#1 and Still-Face#2 as well as during Reunion#1. As such, only negative emotionality during Reunion#2 was tested in the regression model assessing the moderating role of maternal sensitivity on infants’ stress response. Moreover, as only SLC6A4 methylation at CpG2 was found to be significantly correlated with infants’ stress negative emotionality, methylation at CpG5 was not included in the regression models.

The regression model for VPT infants was significant, R² = 0.12, p = 0.04. The only significant predictor of VPT infants’ negative emotionality during Reunion#2 was CpG2 methylation at VPTs’ discharge, B = 0.27, β = 0.40, 95% CI (0.01; 0.52), t = 2.15, p = 0.04. Higher CpG2 methylation at VPT infants’ discharge from the NICU was predictive of heightened negative emotionality during Reunion #2. When the model for VPT infants was controlled for birth weight as a confounder, the significant effect of CpG2 methylation on infants’ socio-emotional stress response was maintained.

The regression model for FT infants was significant, R² = 0.31, p = 0.002. No significant effect of CpG2 methylation at FTs’ birth emerged, whereas maternal sensitivity significantly associates with FT infants’ negative emotionality during Reunion #2, B = −0.29, β = −0.49, 95% CI (−0.46; −0.12), t = −3.54, p = 0.002. Higher maternal sensitivity during the Play episode was predictive of less negative emotionality during Reunion #2 in FT infants. Moreover, a significant interaction effect between CpG2 methylation and maternal sensitivity emerged for the FT group, B = −0.43, β = −0.61, 95% CI (−0.65; −0.24), t = −4.43, p < 0.001. FT infants from mothers characterized by low maternal sensitivity showed a significant positive association between SLC6A4 methylation of CpG2 and negative emotionality during Reunion #2, whereas no significant association emerged for FT infants from mothers with high maternal sensitivity (Figure 3). When the model was controlled for confounders, both the independent and interaction terms remained significant.