There is a growing discussion about the intergenerational determinants of health risks.  Conditions in parental and grandparental generations have been shown to influence both health and social outcomes in the child generation. A major research question is to identify and disentangle various causal mechanisms behind observed associations. The Swedish psychiatrist Gustaf Jonsson coined the concept of “social inheritance” more than half a century ago.  This was based on the observation that children growing up in poor family circumstances often develop similar problems to those experienced by their parents.  Genetic inheritance is an alternative causal mechanism which has figured prominently in the discussion. There has previously been a great deal of antagonism between these two views. Today, however, there is a will to understand how genes interact with the social, cultural and physical environment.
We have seen the birth of a new field of research, epigenetics, which is concerned with factors which influence gene expression. In this research field, developmental plasticity (taking shape in utero, infancy and during childhood), is of interest. The concept of “biological programming” is a way of describing mechanisms behind development. For instance, suboptimal nutrition in utero influences foetal growth and regulates gene expression, with life-long consequences for the individual with regard to blood pressure, cardiovascular disease, cognitive ability and more. Starvation and trauma during the foetal period have been observed to have epigenetic consequences for the two following generations. This has been referred to as epigenetic inheritance, to be distinguished from the social or genetic inheritance. A more neutral term is transgenerational response: a change in development in response to environmental conditions in previous generations, without assumptions about which causal mechanisms are operating.  

Our project aimed to test a specific hypothesis, the so-called Pembrey-Bygren hypothesis from their oft-quoted 2006 paper. In this, the authors claim to ”have proof of principle that a sex-specific male-line transgenerational effect exists in humans”. They are very careful in their wording, but they do state that this effect may be epigenetic.  They suggest that one period of childhood, the prepubertal period (“the slow growth period”) may be open for germ-line epigenetic reprogramming.

This latter specification made it possible for us to formulate an empirical test. Could severe childhood trauma (such as parental death) trigger a cascade of events influencing development (physical, emotional and/or cognitive) across generations, detectable both in the exposed children themselves and in their children? If so, is there a specific period of childhood which constitutes a “critical window” of extra susceptibility for such a trans-generational response to evolve? Finally, is any such critical period the same for boys and girls and the response similar along paternal and maternal lines of transmission?
We addressed those questions using a database which spans three generations; the last generation (G3) includes all births in Sweden 1972-2002. Thus, for almost 765,000 G3 individuals we traced both parents (G2) and all four grandparents (G1).
We were therefore able to compare the offspring (G3) of boys and girls (G2) who had suffered parental (G1) death during their childhood with the offspring of other children.  We studied G3 birth weight and prematurity by G2 experience of parental loss at specific ages. According to the pembrey-Bygren hypothesis, the pre-pubertal period (defined as ages 8-12) should be the most sensitive period (a “susceptibility window”).  In addition, we analysed how parental loss influenced some aspects of the life course for G2 individuals. For a sub-cohort (of G2 men, n=250.427) we were able to study how blood pressure, BMI and psychological resilience were linked to parental death. However, our main research question throughout has been outcomes in the subsequent generation, G3.

Results
/Study 1 (Vågerö & Rajaleid 2017a). We concluded that exposure to childhood trauma, such as the death of a parent, triggers a cascade of events, both psychological, behavioural, metabolic and social. While this was true for both boys and girls suffering parental death, the transmission of this response to their offspring differed markedly. Along the female line it was most of all a socially-mediated (partly through education) transgenerational response which gave rise to reduced birth weight and increased prematurity risk in offspring. Along the male line results were compatible with the hypothesis formulated by Pembrey and Bygren. Our tentative conclusion was that the period just before puberty may be extra sensitive to trauma, resulting in germ-line epigenetic change in men which is transmittable to offspring; in our case influencing offsprings’ foetal growth. The latter finding was explored further in Study 3 (see below).

/Study 2 (Vågerö & Rajaleid 2017b). Peter Hõrak, an evolutionary biologist, wrote a response to study 1. We took great care to further analyse our data to address his concerns.  Genetics, according to Hõrak, determine whether people have a “fast” or a “slow life history strategy”. The relevance of the concept of “fast life history strategy” is that it both involves a strategy of taking risks and a “reduction of the amount of maternal somatic investment in fetal development, resulting in shortened gestation and low birth weight”. Men and women with a “fast life history” tend to marry each other according to Hõrak, so called assortative mating. Genetic correlations in life history traits between spouses arise because of this assortative mating. This could therefore be an explanation of our results, according to Hõrak.  
Assortative mating is a relevant concept, well-known also in the social sciences, where it is usually applied to education. In our population we were able to calculate the correlations in educational achievement between G2 spouses. This is 0.31. The same correlation between G1 grandparents is 0.29. Clearly, assortative mating based on education exists in both G1 and G2 in our study.
However, we find no evidence that the particular outcomes that we focus on, birth weight and gestational length in G3, are influenced by assortative mating in G2. For a segment of generation 2 we have data on birth weight (n=43,308 pairs) and gestational age (n=42,971 pairs) for both G2 spouses.  Birth weight correlations between spouses/parents in G2 are 0.02. Correlation in gestational length (measured in days) between spouses is even smaller: 0.008. It is obvious that there is no assortative mating whatsoever based on these traits.

Thus, if foetal growth and risk taking behavior are both indicators of a genetically driven “fast life history strategy” and men and women choose partners based on such strategies, we would expect birth weights and gestational age to be correlated between spouses. Since they were not we rejected the “fast life history” explanation.

/Study 3 (Rajaleid & Vågerö 2018). The purpose of study 3 was twofold. Firstly to disentangle those causal mechanisms for which accumulation of stress load over time is most important, from mechanisms based on the existence of susceptibility periods in early life.  Here we focused on pathways along the paternal line which we discovered in study 1. Secondly, we aimed at identifying potential mediators between parental death and offspring birth outcomes.

We, firstly, applied the structured life course approach (Mishra 2009) and explored the effect of G2 men’s age at parental death on their own stress resilience, BMI, diastolic and systolic blood pressure measured at military conscription, as well as on their offspring’s birth weight and length of gestation. The method defines a number of regression models corresponding to the different life course hypotheses, and uses a model selection procedure to identify the model that explains most of the variance in the outcome.

Best fit to data was a susceptibility model, where parental loss at ages 8-17 influences offspring birth weight and gestational length. This does not exclude the possibility that a process of long-term accumulation of stress load also plays some role, although we found this to be less evident in our study.

Secondly, we used the four-way decomposition analysis developed by Vanderweele [2014] to identify mediators and interaction. We decomposed the effect of age of G2 men at parental death on G3 birth outcomes, assuming mediation by G2 stress resilience, BMI, diastolic or systolic blood pressure. The method allows us to study the overall effect of an exposure on an outcome when a mediator is present, and decomposes this effect into four components: solely due to mediation, solely due to interaction between the exposure and mediator, due to both mediation and interaction, and due to neither mediation nor interaction.

We concluded that psychological resilience in men was itself influenced by parental loss and, in addition, interacts with it. Further, it mediates a considerable part of the influence of parental loss on offspring birth weight.  


/Summary conclusion.
Childhood trauma, such as loss of a parent, brings consequences for the next generation. These are at least in part sex-specific. Parental loss will influence development, and later the intra-uterine environment, of girls, more or less regardless of their age when this trauma occurs. We refer to this as a social rather than (genetic) pathway. Boys´development will also be affected, for instance in the form of reduced resilience, a somewhat higher BMI or a marginal increase in diastolic blood pressure at age 18. However, looking at the consequences in the next generation, there appears to be only a limited period ofone childhood that is of importance, namely the period before or during puberty. This observation is compatible with epigenetic changes in male germ cells, which in turn will influence offspring foetal development.

Psychological resilience seems to play an important part in this pathway. Numerous hormones, neurotransmitters and neuropeptides are involved in the immediate response to stress. Could gametes in some way capture the response to trauma from the hypothalamic-pituitary-adrenal (HPA) axis, and thus influence the male germ line epigenome? We are only at the beginning of understanding this phenomenon.
 


   
Presentations of the project
1.    International Conference on ”Epigenetics as a meeting point between nature and nurture”,  Uppsala, spring 2015. A background to the project was presented: ”Transgenerational responses to social experience of humans. A challenge to epidemiology and the social sciences”.  
2.    Seminar at CHESS, Centre for Health Equity Studies, Stockholm University, 2016
3.    Participating in Swiss Epidemiology Winter School:  “Epigenetic Epidemiology” with Caroline Relton and George Davey Smith; Wengen, 18-20 januari 2016
4.    Seminar on ”Epigenetic epidemiology” at the Institute for Stress Research, 2016
5.    Presentation of the project and its background at Institute of Family Medicine and Public Health, Tartu University, 2016
6.    Lecture on Epigenetic epidemiology and the project at Estonian Society of Learning (Estniska Lärdomssällskapet i Sverige) Stockholm, 2016
7.    Interview at the web page of Riksbankens Jubileumssfond, 2017
8.    “Stress Research Seminar and Networking Event”, organized by KI and SfoEpi, December 11th 2017.


Publications and submitted work

1.    Vågerö D, Rajaleid K (2017a). Does childhood trauma influence offsprings’ birth characteristics? International Journal of Epidemiology Vol. 46, no 1, p. 219-229. On-line May 4th, 2016 doi: 10.1093/ije/dyw048
2.    Vågerö D, Rajaleid K (2017b). Transgenerational response and life history theory: a response to Peeter Hõrak. Vol. 46, no 1, p. 233-234
3.    Rajaleid K, Vågerö D (2018). Stress resilience as a mediator of transgenerational response in men. Ms. Submitted.