Early life adversity (ELA), which means stressful events occurring in the first 1000 days of life (1–3), plays a major role in adult-onset illness (4–6). These stressful events include a series of negative situations such as poor socio-economic status, parental mental disease, abandonment and/or institutionalization. Exposure to ELA has long-lasting effects on both mental and physical health as well as having negative behavioral consequences. It is associated with an increased risk of developing cardiovascular diseases (7), asthma (8), cancer (9) and mental disorders such as depression and anxiety later in life in both humans and in rodent models (10–14). There is now growing clinical and pre-clinical evidence that ELA and the associated negative health risk behaviors act through an altered immune system to induce the later-life disease risk (15–19).

Development of the immune system starts in early gestation. Innate immune cells such as monocytes, neutrophils and NK cells appear in the first trimester of pregnancy. Adaptive immune cells (T and B-lymphocytes), emerge around the start of the second trimester (20–22). During the pre- and perinatal period, this development is highly affected by several maternal factors such as obesity (23), malnutrition (24, 25), anxiety (19, 26–28) and smoking (29, 30), but also environmental factors such as parturition (31, 32), breastfeeding (33, 34) and antibiotic treatment (35). Alterations in specific cellular subsets in the immune system resulting from ELA have been widely documented in clinical studies. Individuals subjected to parental separation and subsequent adoption displayed a higher activation state of the immune system, with decreased levels of circulating central memory T cells and CD8+ T regulatory cells (18). Teenagers who suffered from childhood maltreatment, such as sexual and physical abuse, physical and emotional neglect showed increased circulating levels of NK and NKT cells after ELA (36). In addition, blood levels of the inflammatory marker C - reactive protein (CRP) in teenagers who had a low social-economic status as children was found to be increased when compared to individuals not exposed to early stress (37). Individuals with poor maternal care and harsh discipline during childhood also presented increased levels of CRP in the blood (38). Studies with rhesus monkeys, early isolated from their mothers at early age, also show a significant decrease in the CD4+/CD8+ ratio and increase in the circulating levels of NK cells (39). Even though some animals studies do not show any difference in the cell number after maternal separation (40), others document a decrease of the CD8+ T cells with subsequent increase of the CD4+/CD8+ ratio (13), opposite to what was documented in monkeys. Furthermore, prenatal exposure to alcohol lead to an increase in the CRP serum levels, indicating an inflammatory state of the immune system (41), similarly to what was previously observed in clinical studies (37, 38). Despite these numerous investigations, the literature is still lacking an unbiased overview of the complete cellular immune system.

Although the mechanisms through which these events occur are still not fully understood, increasing evidence shows that the mechanism by which ELA influences the function of CD8+ T cells and, consequently, viral responses, may be through the HPA axis (42). The neuro-endocrine axis plays an important role in the adaptive response to stress. When facing insults, corticotropin-releasing factor (CRF) is released from the hypothalamus, which in turn stimulates the production and release of adrenocorticotropin (ACTH). This hormone’s main target is the adrenal cortex where the production and release of glucocorticoids (GCs) happens. Release of GCs into the bloodstream will trigger the adaptive mechanisms, in a negative feedback manner (43, 44). Stress events in an early period of life are known to have an impact in the HPA axis, programming its effects and responses in adulthood. This leads to decreasing levels of blood corticosterone and cortisol, which consequently affects the response of the peripheral immune system, leading to compromised viral responses (13, 45, 46). Such dysregulation of the HPA axis is thought to occur through the glucocorticoid receptor (GR), by regulation of gene transcription and negative feedback on the HPA axis, which in turn decreases the expression of certain cytokines (47). Clinical studies show an association between increased GR1F promoter methylation and ELA (48, 49). However GR/GC signaling remains undisturbed after ELA despite a slight increase in GR1F promoter methylation (50), raising doubts as to the importance of single-digit changes in promoter methylation levels (51).

ELA also accelerates immunosenescence, the natural aging process by which the immune cells begin to deteriorate and lead to weakened immune responses (52). Immunosenescence is accelerated not only after exposure to ELA (53), but also with depression after physical injury (54). T cells are strongly affected. Naïve T cell numbers decrease while memory T cell and terminally differentiated effector T cell (TEMRA) numbers increase, with concurrent telomere shortening (55–58). Furthermore, these cells have decreased expression of the co-stimulatory CD28 molecule and increased expression of the glycopeptide CD57, which leads to increased cytotoxicity and decreased proliferative capacity (53, 55, 59, 60). This increase in T cell senescence after ELA has been reported to be influenced by the exposure to and subsequent reactivation of cytomegalovirus (CMV), as levels of CD57+ cells are increased in patients seropositive for CMV (53, 61–63). Moreover, CMV in ELA individuals was recently reported to be linked to the presence of certain gut bacteria and CD8+CD57+ cells (64), suggesting an impact of ELA through the immune-brain-gut axis.

Although it is not as well documented as for T cells, immunosenescence also occurs in other cell types, such as B (65, 66) and NK cells (67, 68). The expression of CD57 in natural killer cells does not necessarily mean they are senescent but rather that they reached a higher maturation state, which is accompanied by functional changes similar to those observed in senescent T cells: less proliferation and higher cytotoxic capacity (28, 67, 69). Moreover, NK cells are clearly involved in the response to CMV infections (70, 71) and links between NK cells, CMV and immunosenescence are starting to emerge (69, 70, 72).

Using an unbiased screening tool for flow cytometry data visualization, viSNE (73), this study provides a detailed description of the overall immune changes induced in the rat maternal deprivation (MD) model of ELA, identifying unexpected, but clear changes in NK cell properties. Furthermore, we describe the functional profile of NK cells, showing a shift in the maturity and cytotoxic capacities. We validated the NK cell phenotype in samples from our EpiPath ELA cohort (53, 74). This cohort consists of young adults (average age 24) institutionalized or otherwise separated from their biological parents at birth and adopted in early childhood (mean age of adoption 4.5 months) together with control participants in their natal families, all brought up in Luxembourg under similar societal and socioeconomic conditions.

In this study, we demonstrated how ELA, in the form of maternal separation, has a more widespread influence on the immune system than previously thought. To our knowledge, this is the first unbiased viSNE analysis linking ELA with clear changes in the overall immune profile and, more specifically, the first to provide specific mechanisms of maturation, senescence, and changes in cytotoxicity and degranulation profiles of NK cells after ELA. We initially examined the effect of ELA using the rat MD model. As previously reported by ourselves and others (87–89), the separation and deprivation of maternal care during this period has been associated with increased cognitive impairment, HPA-axis dysregulation and anxious-like behavior (13, 90–93).

Stress, in all types of forms, either in early life, adolescence or adulthood activates the HPA axis, leading to the release of cortisol that will bind to GRs and ultimately initiate a cascade of molecular and cellular events (94–96). GRs activation is reported to impact gene transcription (97) and to inhibit immune responses (47). However, we previously reported that this process is not always so clear-cut (50). We saw clear clinical consequences later in life, specifically in the immune system and in the development of chronic and psychological disorders (18, 53), however, these changes were not accompanied by alterations in the expression or response of GRs, although the HPA axis was hypo-responsive (50, 74). Although we did not directly assess the functioning of the GRs, our data confirm our previous report that HPA axis hormones and receptors might not be as intimately involved in the long-term consequences of early life stress as thought. Importantly, we were able to reproduce the immune phenotype previously seen in our EpiPath cohort (18), where CD8+ T cells were found to be more activated (18) and more senescent (53) than the cells from the individuals in the control group. In our paradigm, T cells (CD3+) and their subsets (CD4+ and CD8+) were significantly changed, with CD8+ T cells following the same trend as in our ELA (18), confirming the relevance of the MD model for the biological consequences of ELA. Furthermore, we expanded changes in the immune system to B and NK cells. Our unbiased viSNE analysis did not clearly distinguish NK and NKT-like cells, the latter being T cells that share and express NK cell receptors bridging the innate and adaptive immune responses that are implicated in tumor rejection, cardiovascular and neurological diseases (98–101). Little is known about the long-term effects on NK cells after early-life psychosocial stressors, although NK cell numbers have been reported to be impacted (102). However, our previous report (18), together with the data reported here, suggest that ELA has a minor impact on circulating NK cell numbers, but is accompanied by a higher activation state and a trend towards increased senescence. Our data suggest that the NK cells have a similar phenotype to the CD8 T cells, previously reported. We see a different secretion of CD107a and IFN-γ from the CD56dim NK cell subsets (CD56dimCD16dim and CD56dimCD16+), as well as from the CD56-CD16+. As discussed by Emily Mace (103) and others (104, 105), these subsets are thought to be the most differentiated ones, as loss of CD56 expression and acquisition of CD16 was proposed to be part of the maturation process. These results follow the increased expression of maturation markers observed in the rats. Altogether, this seems to indicate that, although immature cells in both adoptees and stressed animals are still functional, as they become more mature, they lose their functionality, both in terms of cytotoxicity and degranulation. In a similar way to the increased activation (CD25) and senescence (CD57) of CD8+ T cells, the NK cells appear to lose functionality as they mature, although unlike CD8 T cells, this was independent of CMV exposure and titers (18). It would appear that this mechanism is not applicable in NK cells, as there was no correlation between either CMV titers or seropositivity and NK cell activity. Furthermore, as there was no clear HPA axis phenotype, although there was a trend towards an increase in stress-induced gluconeogenesis, we conclude that the HPA axis is unaffected in our MD paradigm and, as such, cannot be responsible for the NK cell phenotype either. We saw clear differences in baseline glucose levels in both our MD groups. We initially hypothesized that this was a function of the altered HPA axis (106). However, it would appear that changes in the external environment did not induce changes in the HPA axis stress response nor in the overall glucocorticoid levels. As such, our recent hypothesis that the disturbed HPA axis fundamentally alters hepatic metabolism such as gluconeogenesis to mechanistically crystalize the early-life adversity associated risk of metabolic diseases may, in fact be erroneous, and that this may be independent of the HPA axis.

NK cells are known to be affected by current acute and chronic stress. The early work by Schedlowski et al. showed that acute stress in adulthood, in this case novice parachute jumpers, had significant changes in the circulating lymphocyte subsets as well as functional differences in NK cells immediately post-stress. We expanded on this to demonstrate the kinetics of NK cell redistribution throughout the day, coupled to the circadian HPA axis rhythm, though in the work of Schedlowski it appeared to be associated with noradrenaline levels (107, 108). Both studies suggested that this rapid mobilization of NK cells was a natural physiological reaction to an external stressor, in agreement with both their natural role as an immediate initiator of the immune response before adaptation starts, and as an evolutionary mechanism, preparing the body to fight injury or infection after encountering an acute stressor. Sympathetic nervous system control of NK cell action via noradrenaline has been suggested to be an advantage because of the speed with which the immune system can be primed to act after a stressful encounter, as well as the speed in which the priming can be terminated and homeostasis re-established (109) through the inflammatory reflex (110). There is a similar dearth of literature on the effects of chronic stress on NK cell functioning. In a similar manner to our observation of decreased NK cell functionality, chronic low-dose glucocorticoid administration reduced histone acetylation levels around promoters for two essential NK cell produced effectors: perforin and granzyme B. This was associated with lower mRNA transcript levels, lower protein levels, and NK cells were functionally impaired in a manner similar to what we report in both our rat model and in the EpiPath cohort. The lower perforin and granzyme B levels decreased their cytolytic activity. Inversely, the same administration regime increased histone acetylation of the IFN-γ and IL-6 promoters, up-regulating transcription and functional protein levels (111). The situation is, however, far from clear-cut. Children with current chronic stress from maternal mental health had higher levels of psychiatric symptoms as well as an increase in the number of illness episodes that was associated with increased NK cell cytotoxicity (102). However, none of the data available so far addresses the long-term effect of early life psychosocial stress and adversity, and the differences in NK cell functionality when the stressor is no longer present. In our previous report from the EpiPath cohort, multiple correction testing during our survey of the complete immune system meant that NK cells only narrowly missed significance (18), and the only other comparable study did not investigate NK cells (64, 112). Both of these studies reported that ELA induced a long-term immunosenescence and reduced T-cell functionality that was most probably due to continued re-activation of viruses such as CMV. It would seem logical that the exposure to a period of chronic stress in both models presented here has had a similar effect on NK cells. The two experimental systems show that once the period of ELA has resolved, NK cells are programmed with a long-term hypo-reactivity. As for acute stress preparing the NK cells to deal with an immediate infectious threat or potential wound, we suggest that this long-term hypo-reactivity is a similar evolution. The sensitive early life period would appear to have prepared the NK cells for an environment in which they can expect to be more regularly activated, and as such, to avoid any negative effects associated with NK cell secreted effector molecules or cytokines.

Although NK cells are often associated with a positive regulation of the immune response, they are also associated with the development of immunopathologies. In chronic hepatitis B virus infection, NK cells contribute to both liver inflammation and injury (113, 114), and aggravate and increase the lethality of bacterial infections in murine models (115, 116). NK cell activity was found to be impaired in patients that suffered from multiple sclerosis (117, 118), type-1 diabetes (119) and cardiovascular diseases (120–122), and found to sustain joint inflammation in rheumatoid arthritis patients (123); the risk of the latter three are all increased by exposure to ELA (16, 124–126). This raises the possibility of the long-term alteration of the NK cell phenotype underlying the pathophysiological effects of ELA. It may seem counterintuitive that we see an accumulation of NK cells of a more mature phenotype, but of lower functionality. This does, however, very clearly mirror our prior data on CD8+ T cells. Although it is not as well documented as for T cells, immunosenescence also occurs in other cell types, such as B (65, 66) and NK cells (67, 68). Our data suggests that, as suggested by Judge et al. (67), this corresponds to the accumulation of senescent NK-cells. Although NK cells have received far less attention, anergy, senescence and exhaustion are distinct dysfunctional entities that parallel the same processes in T cells and are all characterized by a significant reduction in both proliferation and NK-effector functions. There is the caveat that the expression of CD57 in natural killer cells does not necessarily mean they are senescent but rather that they reached a higher maturation state, which is accompanied by functional changes similar to those observed in senescent T cells: less proliferation and higher cytotoxic capacity (28, 67, 69). Furthermore, an alternative interpretation may be that at the single cell level ELA reduces NK cell effector functions, while at the population level the NK cells were more mature. This is somewhat confirmed by our previous report of ELA reducing activation CD69 levels on both total NK cells and CD56dimCD16+ NK cells. Unfortunately, in our previous report, this result was highly significant prior to, but not after, post-hoc multiple testing correction (18)

The limitations of our pre-clinical study include potential litter effects and the absence of a clear HPA axis phenotype. Similarly, the number of EpiPath participants analyzed was limited, however, based on the rat MD data, our power calculation suggested that to see the same phenotype in the cohort only 14 participants were required to have 80% power at alpha 0.05, which we largely exceeded. However, these are clearly outweighed by the reproduction of the functional NK cell phenotype in the EpiPath cohort. Furthermore, the identical phenotype in the MD model and the cohort allowed us to exclude the two most prominent mechanistic hypotheses from the literature – HPA axis control, and continual CMV reactivation. Nevertheless, ELA has a direct impact in the maturation state and later exhaustion of NK cells that may impair their activity and lead to uncontrolled reactions in adulthood.

It is clear that all cells of the immune system are not equally affected by ELA. Here, we have expanded our prior observation of T cell immunosenescence to a novel, unbiased examination of the immune system, identifying NK cells as functionally affected by ELA in both the rat MD model and in our human institutionalization – adoption cohort and NKT-like cells to be differently expressed in the rats. The immature NK cells appear to retain their functionality, however, as they mature towards CD56dim NK cell subsets and CD56- phenotype, their cytotoxic and degranulation potential are reduced. It is now evident that alterations in the HPA axis, either as stress-induced cortisol/corticosterone production or gluconeogenesis, are not responsible for the immune phenotype. The challenge is now to understand how ELA is inducing such changes and the role of both T cell and NK/NKT-like cells functional and expression loss in the long-term ELA-induced disease risk.

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by Luxembourg National Research Ethics Committee (CNER, No 201303/10 v1.4) and the Ethics Review Panel (ERP, University of Luxembourg, No 13-002). The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by local Animal Welfare Structure (DII-2017-18).

Conceptualization: SF and JT. Literature review: SF and JT. Data collection: SF, NP, SM, ME, FL, and MT. Data analysis: SF, NP, JZ, and JT. Manuscript writing and editing: SF, NP, JZ, and JT. All authors contributed to the article and approved the submitted version.

This study was funded by the Fonds National de Recherche Luxembourg grants FNR-CORE (C16/BM/11342695 “MetCOEPs”), (C12/BM/3985792 “Epipath”) and the Ministry of Higher Education and Research of Luxembourg.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.