Background

The Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort was established in 1991 to study developmental phenotypic variation across the local population in relation to inherited DNA variants and a wide range of dietary, social and environmental exposures. [1] Maternal exposures during and before the study pregnancy was one interest during the first decade, being relevant to the Developmental Origins of Health and Disease (DOHaD) hypothesis [2,3] and the emergence of environmental epigenetics [4,5]. By the second decade, ALSPAC began exploring associations with grandmaternal exposures in the light of experimental evidence of mammalian epigenetic inheritance across several generations (see below). Whilst population-wide traits have been the primary ALSPAC outcome measures, doctor-diagnoses are recorded and, in common multifactorial conditions, the two may be considered together. Two recent examples of this are postnatal refractive eye development focusing on refractive errors underlying myopia (near sightedness) [6] and four childhood behavioural traits predictive of autism spectrum disorders (ASD) and diagnosed ASD [7].

Since the launch of human genome wide association studies (GWAS) in 2007 [8] the mainstream view of the inheritance of complex traits and multifactorial diseases is that the heritability will eventually be explained by the sum total of common and increasingly rare DNA sequence variations inherited by the individual, plus ‘de novo’ mutations arising early in development. As whole genome sequencing takes off, there is some support for this view with respect to the genetics of height [9], no support from studies of Type 2 diabetes [10,11] and a speculative reappraisal of what complex traits actually are [12]. In parallel, the last decade has seen mammalian evidence of non-DNA-sequence-based inheritance induced by parental/ancestral experiences [13]. These effects are called intergenerational if the exposure (or importantly the organism’s response to it, e.g. response to DNA damage) could have reached the germ cells leading to the next generation(s), or transgenerational if this is not the case. The latter implies some molecular ‘memory’ of the ancestral experience, or its response to it, is being passed down via the gametes. We have written inter/trans-generational when both apply.

The molecular basis of intergenerational or transgenerational epigenetic inheritance is poorly understood, particularly in humans, but it does include altered DNA methylation, histone modification and exposure induced non-coding RNAs being transferred to sperm [14,15]. Thus, what has, in the past, been interpreted as ‘likely to be genetic’, is now better considered more broadly ‘as likely to be factors inherited at conception’[16]. In line with this view, ALSPAC has included assessment of associations between parental/ancestral early-life experiences and developmental variation in the study participants. Unlike initial human epidemiological studies by others of associations with parental/ancestral exposure to historic famines [17,18] we selected cigarette smoking in the early 2000’s as a suitable exposure for proof of principle studies in a contemporary population.

As described in detail elsewhere [19] the intention at the start was to see if we could replicate some of the distinctive features of human intergenerational and transgenerational associations found by others, principally the exposure sensitive period(s) during development. We found support for the mid-childhood sensitive period, initially shown by Bygren and colleagues [18], and noted a sex-specific bias in the intergenerational association with paternal onset of regular smoking before 11years. This led to a collaboration with Professor Bygren and colleagues to test for sex-specific effects in the Swedish data; and striking sex-specific (but not sex-limited) associations were found [20]. In view of this, we started a programme on intergenerational association studies within ALSPAC with smoking as the exposure of interest. We began by analysing the child’s growth and anthropomorphic measures in relation to study fathers who started regular smoking before puberty (<11ys) [21] and with either grandmother smoking in pregnancy [22]. We have since extended these intergenerational studies to behavioural and sensory perception measures collected prospectively in the study participants.

The hypersensitivity measures were intolerance of loud sounds and the ability to detect bitter taste using 6-n-propylthiouracil (PROP) [23]. Here we consider intolerance of loud sounds in relation to grandmaternal smoking. Aetiological studies of intolerance to loud sounds per se have rarely been undertaken but may be incorporated into studies of hyperacusis; namely intolerance to everyday or low intensity sounds. Hyperacusis is a feature of certain monogenic disorders, such as Williams syndrome [24] and Smith-Megenis syndrome [25] and occurs in about 20% of children with the multifactorial disorder ASD [26,27]. In adults with hyperacusis who are otherwise neurotypical, the medial olivocochlear (MOC) efferent reflex is stronger than usual, as it is in children with ASD and coexisting hyperacusis [28]. Among the candidate genes for ASD, the contactin genes CNTN5 and CNTN6 code for neural cell adhesion proteins that promote neurite outgrowth and synaptogenesis.[29]. CNTN6 mutations, in particular, are risk factors for abnormal auditory sensory perception amongst those with ASD [30]. Our earlier finding of an association between maternal grandmother smoking in pregnancy and impaired social communication and repetitive behaviour in granddaughters but not grandsons [7] provides an example of intergenerational responses to maternal grandmaternal smoking in pregnancy. However, the present study of intolerance to loud sounds was not intended or designed to focus on ASD.

There is a substantial experimental animal literature on the ways in which specific environmental exposures can influence the phenotype of subsequent (unexposed) generations through non-DNA sequence-based inheritance [13,31,32]. In relevant animal experiments exposures are often delivered to a pregnant dam thereby exposing her offspring throughout fetal life. The dam is traditionally labelled F0, so we have adopted the same notation for the study grandmothers, F1 for their offspring–the study parents, and F2 for the study participants. The results of animal experiments support the human epidemiological observations and provides numerous examples of sex-specificity or sex-bias in both the route of transmission down the generations and the phenotypic outcomes in descendants. As mentioned above, sex-specificity was a feature of initial human studies and this is true of early rodent experiments [33,34,35].

The transmissions (progressing from intergenerational to transgenerational) can be complicated as illustrated by the study of maternal high-fat diet exposure in mice [36]. This resulted in an increase in body size and reduced insulin sensitivity in both sexes that persisted across two generations (F1 & F2) via both maternal and paternal lineages. However, at the next generation (F3) only the females had increased body size and this effect was only passed on via the paternal lineage. Another example is a study that induced chronic social instability in adolescent mice (F0) and then followed their unexposed descendants (F1, F2 & F3) [37]. Only female descendants exhibited anxiety-like behaviour and social deficits, but after F1 this phenotype is only transmitted through males who themselves do not manifest anxiety-like behaviour and social deficits. This latter study was the starting point for a commentary that explored why sex matters in such intergenerational and transgenerational responses.[38] It could not provide any conclusive explanations for the sex-specificity but recognised that the phenomenon was common. It is important to emphasise that intergenerational phenotypic responses can be widespread, for example combine metabolic and behavioural features [39] and include beneficial outcomes [40].

There is not necessarily transmission of the initial phenotype, although inheritance of an acquired characteristics can occur. Sometimes the F1 or F2 phenotypic can change to be in the opposite direction, as if the response is a corrective adaptation [41]. Increasingly animal experiments permit organ/cell specific gene expression and epigenomic studies. These can include germline, sperm, ovum or early embryo analyses to elucidate the molecular mechanisms that mediate intergenerational transmissions [31,32,42]. Such opportunities are limited in humans, although there have been ingenious studies to elucidate the nature of the epigenome in the conceptus and its trajectory into the human pre-implantation embryo [43].

Overall human results from epidemiological, observational studies over the last two decades have been in line with rodent experiments (reviewed in [44]). The inter/trans-generational responses can be through either the paternal or maternal line; and like rodent experiments, human observations often demonstrated a sex-specificity (but not necessarily sex limitation) in either the parental route of transmission and/or the offspring phenotypic outcomes. This is true of most of the human studies referenced below since 2005 and why we have hypothesised that there will be sex-specific differences in any transgenerational associations detected in this study of intolerance to loud sounds. Human studies have defined ‘exposure-sensitive periods’ during early development; namely fetal life and mid-childhood just before puberty. These exposure times are broadly supported by responses to swings in F0 food supply [17,19,33,45,46] or F0 smoking [21,47,48, 49]. However, some F0 pre-conceptional exposures in (early) adulthood can also be associated with phenotypic changes in the next generation [21].

Exposure to famine is likely to be confounded by psychological stress, and F0 exposure to stress per se has also been reported to be associated with altered psychological state or early development in the next generation [50,51]. As described above, smoking was selected in the early 2000’s as an exposure (with available data) in a contemporary population with which to valid the exposure sensitive periods of fetal life and mid-childhood as indicated by studies of historic famines. Smoking can clearly initiate intergenerational responses, which raises the question as to how it can induce the relevant molecular change(s) in the gametes or the resulting conceptus. It is important to note that tobacco smoking causes DNA damage [52]. This damage or the body’s response to it [53] may well be the key ‘signal’ that is transmitted to the next generation(s); and with such a generic genomic stress, the impact on development may be widespread.

Material and methods

Results

Maternal grandmother smoking and child intolerance to loud sound

The variation in unadjusted odds ratios for intolerance to loud sounds for all children showed little sign of any association with maternal grandmother smoking (Table 1). However, there was an increase in odds among the boys at age 6, especially when the mother did not smoke (P<0.10). Although the odds ratios for the girls were no different than would be expected by chance, the difference in odds between boys and girls showed a significant interaction (P < 0.05). There were no associations with intolerance to loud sounds at age 13.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Unadjusted risk of the child hating loud sounds at ages 6 and 13 if the maternal grandmother had smoked in pregnancy.

[P values <0.10 are in bold].

https://doi.org/10.1371/journal.pone.0229323.t001

Examination of the variables that were associated with intolerance to loud noises at 6 (Table 2 and S2 Table) identified the year of birth of each maternal grandparent (F0), the age of her father (F0) at the time of birth of the study mother (F1), and the ethnic background of each of her parents (F0) as potential confounders.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Proportion (n) of children who hated loud noises at age 6 according to features of their grandparents.

[P values <0.10 are in bold].

https://doi.org/10.1371/journal.pone.0229323.t002

Once these were taken into account (Table 3), the associations for boys at age 6 became significant at P<0.05, and there were still significant interactions between the sexes with the girls being less likely to be intolerant to loud noises and the boys more so. This was true for all children as well as for those whose mothers did not smoke. In contrast there were no such associations for the children who were intolerant to loud noises at age 13.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Adjusted risk of the child hating loud sounds at ages 6 and 13 if the maternal grandmother had smoked in pregnancy.

[P values <0.10 are in bold]. Adjusted for year of birth of each grandparent, age of grandfather at birth of the mother, ethnic origins of both grandparents.

https://doi.org/10.1371/journal.pone.0229323.t003

Paternal grandmother smoking and child intolerance to loud sounds

The patterns were quite different if the paternal grandmother had smoked in pregnancy (Table 4). The unadjusted data for intolerance to loud noises at age 6 showed a reduced risk for all children and for girls, as well as for the children whose mothers did not smoke. Again, there were significant differences between the sexes, with the girls being less likely than the boys to be intolerant of loud noises (P<0.05). At age 13, all children were significantly more likely to have a reduced risk, with boys and girls having similar odds ratios. This was true of those whose mothers did not smoke, but when the mother did smoke, there was a significant difference between the sexes–the girls being more likely than the boys to be intolerant to loud noises.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Unadjusted risk of the child hating loud sounds at ages 6 and 13 if the paternal grandmother had smoked in pregnancy.

[P values <0.10 are in bold].

https://doi.org/10.1371/journal.pone.0229323.t004

Examination of the variables related to paternal grandparents that were associated with intolerance to loud noises at 6 and 13 (Table 2 and S2 Table) identified the year of birth of each of the grandparents (F0), the age of the grandparents (F0) at the time of birth of the study father (F1), and the ethnic background of each of his parents (F0) as potential confounders. The adjusted odds for children intolerant to loud noises were still reduced for all children but there were no longer interactions between the sexes with the exception of the group where the study mother (F1) had smoked: as in the unadjusted data at age 13, the girls were more likely than the boys to be intolerant to loud sounds (Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Adjusted risk of the child hating loud sounds at ages 6 and 13 if the paternal grandmother had smoked in pregnancy.

[P values <0.10 are in bold]. Adjusted for year of birth of each paternal grandparent, age of grandfather at birth of the father, ethnic origins of both grandparents.

https://doi.org/10.1371/journal.pone.0229323.t005

Grandchild’s setting of music listening level.

The test of music listening level was repeated on two occasions (named run 1 and run 2). In order to determine whether choice of run made any difference to the results we compared the two sets of unadjusted mean results. They showed the same associations–we therefore decided to use Run 1 for further analyses since there were marginally more results for this than for Run 2 (S3A and S3B Table). The comparisons showed that the grandchildren whose grandmothers had smoked when expecting either parent tended to set their volume level higher than those whose grandmothers had not smoked in pregnancy, and that the granddaughters set their level higher than the grandsons.

In order to compare our results with those who were intolerant loud noise, we have analysed the volume listening results assuming that those who were intolerant of loud sounds would be less likely to choose the highest 10% volume levels. The unadjusted results for maternal grandmothers smoking in pregnancy (Table 6) indicate that the grandchildren, especially the granddaughters, were more likely to choose a louder volume. As before, we assessed the backgrounds of the grandparents to identify what variables to allow for (S4 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Odds ratio showing risk of stereo run 1 volume level being in the top 10% where maternal grandmother smoked.

Adjustment A concerns allowing for maternal grandmother’s year of birth, her ethnicity, her education level; Adjustment B additionally takes the maternal grandfather’s social class into account.

https://doi.org/10.1371/journal.pone.0229323.t006

The factors that were relevant for the maternal line were the maternal grandmother’s year of birth, her ethnicity, her education level and the maternal grandfather’s social class. The relationships without and then including social class are shown in Table 6. Whether adjusted or not, there were significant interactions between the sexes–granddaughters whose mother had been exposed as a fetus to their grandmother smoking were significantly more likely to choose loud sounds than the grandsons, who tended to be less likely to make this choice. This was true of the population as a whole as well as of those grandchildren whose mothers did not smoke. If there was any association among grandchildren whose mothers smoked in pregnancy, the numbers of children involved were too small to demonstrate this.

For the paternal line, the factors adjusted for were the paternal grandfather’s year of birth, his highest educational qualification, and his social class. Although the results showed few unadjusted associations, again there were significant interactions, with granddaughters more likely to choose loud sounds and boys less so (Table 7).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Odds ratio showing risk of stereo run 1 volume level being in the top 10% where paternal grandmother smoked.

Adjustment A concerns allowing for paternal grandmother’s year of birth, her ethnicity, and her education level; Adjustment B additionally takes the paternal grandfather’s social class into account.

https://doi.org/10.1371/journal.pone.0229323.t007

Discussion

Our aim was to assess whether the grandchild’s intolerance to loud sounds had an intergenerational association with the grandmother’s smoking during the pregnancy that had resulted in the birth of the grandchild’s parent. We compared the grandchildren whose grandmothers smoked during pregnancy with the grandchildren whose grandmothers did not, analysing separately the maternal and paternal grandmothers. As in our previous studies [6, 21, 22], we considered the population as a whole as well as subdividing by sex of the grandchild and by whether the grandchild’s mother had smoked or not during the pregnancy.

We showed that about 11% of the 6-year-old grandchildren were said to be intolerant to loud sound. If the maternal grandmother had smoked in pregnancy the boys were more likely than expected, and the girls were less likely than expected to be so (P for interaction <0.05). This pattern was found for all grandchildren considered together, as well as for those whose mothers did not themselves smoke in pregnancy. There were no such relationships for 13-year-old grandchildren (Tables 1 and 3).

For paternal grandmothers’ smoking, the grandchildren tended to be less likely than expected to be intolerant of loud noises at both age 6 and 13, but the only interaction between the sexes was in the relatively small group comprising both mother and paternal grandmother smoking prenatally (the 13-year-old grandsons were less likely and the granddaughters more likely to hate loud sounds (Tables 4 and 5)). However, with much larger numbers, the granddaughters of grandmothers who had and mothers who had not smoked prenatally were significantly less likely to have such a reaction to loud noises.

Since the report of intolerance to loud sounds was based on subjective reports from parents, we used the preferred volume listening level set by the child themselves as a more objective measure. We assumed that children who were intolerant to loud sounds would be less likely to choose the loudest part of the register. As predicted, we found that, if the grandmother had smoked in pregnancy, not only were the grandsons more likely than the granddaughters to be said to be intolerant to loud sounds; confirmatory evidence was shown on testing–the grandsons were less likely than the granddaughters to choose to listen at higher volume levels (Tables 6 and 7).

There was no association of peripheral auditory function with grandmother’s smoking, indicating that the observed effect is not likely to be explained by differences in hearing thresholds. The question about sound intolerance asked in this study specifically relates to loud sounds, rather than intolerance to everyday or low intensity sounds (known as hyperacusis). Intolerance to loud sounds may relate to the phenomenon of loudness recruitment, a characteristic of cochlear hearing loss [60]. However, the lack of an association with hearing threshold suggests this is not a peripheral, cochlear phenomenon. The findings are consistent with models of hyperacusis which propose a central rather than peripheral cause, resulting from increased central auditory gain [61]. However as noted above, the measure used in this paper is not synonymous with hyperacusis [62].

At present we are largely ignorant of the causal pathways underpinning this and similar intergenerational responses to grandmaternal smoking in pregnancy. As indicated in the Introduction, the exposure of either (F1) parent to tobacco as a fetus may have resulted in a direct xenobiotic exposure to both their developing somatic tissues and their emerging germline, ultimately destined for any (F2) grandchildren. Alternatively, it may be the generalised DNA damage caused by (FO) grandmaternal smoking and/or the consequent DNA damage response of the fetus that modifies the emerging germline and alters F2 embryonic brain development. Clearly very complex, it is premature to speculate further.

Explaining the sex differences observed here, and in those reported in many earlier studies of inter/trans-generational responses, might seem more tractable. However, a recent lengthy review of the role of sex in the genomics of human complex traits emphasises that this has been a neglected topic [63]. These authors note that nearly all human complex traits and disease phenotypes exhibit some degree of sex differences, but attempts to attribute this solely to classic genetic x environmental effects has proved unfruitful. Despite devoting some space to epigenome analysis and genomic imprinting, the authors only reviewed single generation studies and assumed throughout that biological inheritance is purely down to transmission of DNA sequence differences. A meta-analysis of 2,335,920 twin pairs and 2,608 phenotypes reports that only 3% demonstrated significant sex differences in heritability [64]. Furthermore, classic twin studies are likely to be confounded by epigenetic super-similarity in monozygotic twins [43]. One conclusion from the failure of these huge studies to explain little of the sex biases in complex diseases and traits is that non-DNA sequence-based inheritance may be an important source of sex differences in developmental traits.

Supporting information

Acknowledgments

We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses.