Right after giving birth to her daughter in 1962, Marcy Hartman—a newspaper reporter
at the time—watched a nurse remove a syringe full of blood from the placenta. “It
was a small contribution I made to the future,” remembers Hartman, now 75.
Between 1959 and 1967, Hartman and more than 15,000 other pregnant women in the San
Francisco Bay Area enrolled in a long-term cohort called the Child Health and Development
Studies (CHDS).1 The cohort was intended to help scientists better understand what
makes pregnancies, and children, healthy. Researchers are now using the biological
specimens and other data provided decades ago by Hartman and the other cohort members
to tackle one of the most vexing scientific questions about the role of the environment
in health—whether an individual’s exposures can impact the health of his or her descendants.
Unlike males, who make sperm throughout life, females are thought to be born with
all the eggs they will ever produce. That means that each prenatal exposure has the
potential to directly impact three generations, says epidemiologist Barbara Cohn,
principal investigator of the CHDS. They include the mother (known as the
F
0
generation), the fetus (
F
1
generation), and, if the fetus is a girl, all her immature egg cells—any of which
may one day become the
F
2
generation.
When a pregnant woman is exposed to an environmental agent, the exposure can affect
not only herself (
F
0
) and her unborn child (
F
1
), but also the germ cells developing within the fetus, if that fetus is a girl (
F
2
). The woman’s great-grandchild (
F
3
) is the first generation not directly affected by the original exposure. Image: EHP.
Figure illustrating the concept of Fsubscript 0 through F subscript 3 generations
In the past decade, studies in laboratory animals have pointed to the possibility
that exposures to potentially toxic environmental chemicals and other stressors in
utero may affect normal development of not just the growing fetus, but multiple generations
to come.2 In one mouse study, for instance, the
F
2
descendants of mice treated during pregnancy with di-(2-ethylhexyl) phthalate had
lower baseline and stress-induced plasma levels of the stress hormone corticosterone
than untreated mice.3 Another study showed differences in the social interactions
of mice of the
F
2
and
F
4
generations of dams fed bisphenol A compared with those of unexposed dams.4
Unsurprisingly, evidence from human studies has proved elusive.5 People, after all,
reproduce far more slowly than mice, and keeping tabs on thousands of humans for decades
is no small feat. There are very few cohorts in the world that are able to generate
the kind of long-term, multigenerational data that would be needed to address questions
about the potential cross-generational effects of human exposures. Yet the few that
do exist are beginning to yield clues about how a person’s exposures and experiences
could influence the neurodevelopment and behavior of future generations.6
,
7
Findings in Humans Emerge
In 2017, researchers studying a large group of parents and children in the United
Kingdom reported associations between smoking by grandmothers during pregnancy and
autism symptoms and diagnoses in grandchildren.6 Study leader Jean Golding, an epidemiologist
at the University of Bristol, had founded the Avon Longitudinal Study of Parents and
Children (ALSPAC) in 1991 to answer questions about child development.8
Golding did not design the 14,000-family cohort around autism risk, specifically.
“At the time, the incidence of autism was thought to be lower than 1 in 1,000 people,”
remembers Golding. She did not think that 14,000 children would be a large enough
number to address questions about such a rare disorder. In fact, she expected fewer
than 10 cases of autism to be counted among the cohort children.
But Golding became interested in risk factors for autism after studies documented
a fivefold increase over previous years in the incidence of autism in kids born in
the United Kingdom between 1988 and 1995.9 Over the years, the ALSPAC researchers
identified more than 200 study children with possible or diagnosed autism.
Golding and colleagues had reason to think there could be a link between children’s
neurodevelopment and their grandparents’ smoking habits. Previously they had found
preliminary evidence suggesting that among boys whose mothers smoked during pregnancy,
average head circumference was smaller for boys whose fathers, too, were exposed prenatally,
compared with boys whose fathers were not exposed prenatally. (No such evidence was
found for girls, however, and average head circumferences of boys or girls exposed
to maternal smoking in utero did not differ depending on whether their mothers were
also exposed prenatally.)10
Grandparents were not included in the ALSPAC cohort, so researchers asked the study
parents whether their own mothers had smoked during pregnancy. Doctors, for the most
part, did not discourage smoking during pregnancy in the 1950s and 1960s, when many
of the parents were born.
The researchers found that after adjusting for known autism risk factors, including
mother’s education and socioeconomic status, children whose maternal grandmothers
smoked were more likely to be diagnosed with autism than those whose maternal grandmothers
did not smoke.6 However, this association was only apparent if the child was also
exposed to maternal smoking in utero.
Then in 2018, an analysis of data from the Nurses’ Health Study II (NHS-II), which
had followed a large group of U.S. women since the late 1980s, also found evidence
of a multigenerational exposure–outcome association. Researchers reported that children
whose grandmothers used diethylstilbestrol (DES) during pregnancy were more likely
to be diagnosed with attention-deficit/hyperactivity disorder (ADHD).7 Specifically,
7.7% of mothers who reported being exposed to DES in utero also reported ADHD in their
children versus 5.2% of mothers who did not report in utero DES exposure.
DES was prescribed from 1938 to 197119 to prevent miscarriage and other pregnancy
complications. Doctors stopped prescribing it to pregnant women after it was shown
to increase cancer risk in the daughters exposed prenatally.20 Use of DES during pregnancy
has also been associated with an increase in hypospadias (a penile birth defect) in
grandsons11 and higher odds of irregular menstrual periods in granddaughters.12 Image:
CC BY-NC-SA 2.0 via https://www.flickr.com/photos/diethylstilbestrol/24559986905.
Photograph of three bottles of diethylstilbestrol tablets from 1963
A few previous human studies had suggested that multigenerational health effects could
be possible with endocrine-disrupting chemicals such as DES.11
,
12 The study of the NHS-II cohort, however, is the first to report an association
between a known endocrine disrupter and potential effects on neurodevelopment, specifically,
across three generations, according to lead study author Marianthi-Anna Kioumourtzoglou,
an epidemiologist at Columbia University.
The research is intriguing, because if replicated, such findings could lead to a paradigm
shift in how we think about the etiology of mental disorders,5 says Joel Nigg, a neuropsychiatrist
at Oregon Health & Science University. Such findings also could help to explain why
some complex disorders, such as ADHD, run in families to what Nigg says is a greater
extent than can be currently explained by known gene effects alone.
Yet it is far from clear, at this point, how such multigenerational changes might
actually occur. How is it possible for people’s environmental exposures to trigger
changes in the brains and bodies of their grandchildren?
Some researchers suspect that epigenetic inheritance may play a role. Others raise
the possibility of a direct hit to either the female or male germ line.2 All agree
there are a lot of puzzle pieces still missing.
Elucidating the Role of the Epigenome
The concept that an organism’s genetic sequence determines its biology—that genes
are the primary regulator for processes like disease and even evolution—has been dominant
in science and medicine since the early 1900s. Now it is becoming increasingly clear
that many diseases do not have a single genetic cause, but are instead due to a combination
of genetic factors modulated by interactions with the organism’s environment.
But how one’s environment influences behaviors and neurodevelopment remains largely
a mystery.5 The new science of environmental epigenetics offers some ideas. “What
we are now seeing is that epigenetics is probably equally important [to genetics]
in every biological system,” says Michael Skinner, an epigeneticist at Washington
State University.
The epigenome contains all the chemical modifications that turn certain genes “on”
in some parts of the body during specific periods of development and “off” in other
parts during other periods. Every cell in the body has the same genetic blueprint,
but not every cell in the body performs the same function. Chemical modifications
to DNA determine how different cells read the same blueprint. It is why brain cells
perform different functions than skin cells, even though they have the same DNA.
Image: EHP.
Illustration of various epigenetic mechanisms. Diagram showing processes involved
in epigenetic changes to the mother’s and father’s genomes, which may persist through
fertilization and change the expression of genes in the embryo. Illustration of a
chromosome: Sperm development is a time of unique susceptibility to epigenetic changes;
as the immature sperm cell’s DNA unwinds during meiosis, the nucleotides within are
briefly exposed. That is when epigenetics tags latch on. Illustration continues with
unwinding chromatin leading genes wound around histones: When DNA is inaccessible,
the gene is inactive; DNA winds around histones for compaction and gene regulation.
Illustration shows histone modification: Enzymes and protein complexes that bind to
histone “tails” cause modification of histone proteins. These modifications can alter
the accessibility of DNA. Illustration shows unwound DNA exposing a gene: When DNA
is accessible, the gene is active. Illustration shows histones with tails with enzymes
and protein complexes at the end of the tails. Illustration ends with DNA unwinding
to show major details of double-stranded DNA with methyl groups attached to a gene
to show DNA methylation: Methylation, or the attachment of a methyl group to DNA at
regions where cytosine and guanine are paired, has tended to receive the most attention
in epigenetic research so far. Methyl groups can tag DNA and activate or repress genes.
This process affects whether factors that would normally cause the gene to be expressed
will do so.
For many years, scientists thought that all these modifications, or epigenetic “marks,”
were reset with each subsequent generation.2 Early on in the formation of a new embryo,
the epigenetic marks from the parent are erased. This normal biological process ensures
the pluripotency of embryonic stem cells—their unique potential to grow into any cell
in the body.
But sometimes the process does not work that way. Scientists now know that some of
these epigenetic marks are not erased or reset. Instead, they get imprinted on the
genome and passed on. It is not yet clear how and why this happens normally.2 However,
Skinner and colleagues have shown in animal models that exposures to environmental
toxicants may lead to new regions of the genome where epigenetic marks are not erased.13
“Essentially, you are taking the normal sites and creating a shift,” says Skinner.
If these shifts occur at just the right time during development, they may become permanently
programmed and passed on, he explains.
That’s the idea behind epigenetic inheritance. Animal studies suggest that environmental
exposures can cause changes in an organism’s epigenome, and these changes can be transmitted
from one generation to the next long after the initial exposure.14
Skinner and colleagues were among the first to demonstrate this phenomenon. In 2005,
they reported that reproductive abnormalities, such as decreased sperm motility, carried
out to the fourth generation after dosing pregnant mice with the fungicide vinclozolin.15
Skinner hypothesized that these fourth-generation offspring had inherited the molecular
effects of a previous generation’s exposure without ever being exposed to the fungicide
themselves. With subsequent animal studies, he reported evidence that several other
environmental toxicants may increase the susceptibility of future generations to reproductive
and kidney problems, obesity, and cancer.16
,
17
“Toxicological studies—rodent studies—are extremely helpful, but I think to have more
confidence we need both toxicological and human studies,” says Columbia’s Kioumourtzoglou.
“At the end of the day, rodents are not humans.” Yet epidemiological studies on their
own are not ideal, because they are limited in what they can prove or show. “If we
have corroboration from both, the evidence becomes much more powerful,” says Kioumourtzoglou.
It is impossible to prove whether epigenetic inheritance is responsible for the associations
of tobacco smoking and DES exposures with autism and ADHD symptoms in grandchildren
observed in the ALSPAC and NHS-II cohorts. Despite the tantalizing clues they offer,
the two studies have a lot of limitations. Both, for instance, relied on self-reports
from one generation’s recall of a previous generation’s exposures and health habits.6
,
7
The researchers cannot rule out the possibility that the associations they saw were
due to other exposures, information errors, or chance. These studies also cannot account
for the influence of other environmental factors that may have been shared across
generations. And it is difficult—if not impossible—to directly assess epigenetic status
in the living human brain. Epigenetic markings in DNA are tissue specific, and there
is no good way, at present, to obtain and analyze brain tissue from healthy people.5
Skinner points out another potential caveat if umbilical cord blood samples from past
human cohorts were to be used in epigenetic studies of human generations. Studies
with cord blood or blood require purification of cell types for optimal epigenetic
analysis, he says. But it can be difficult to purify frozen cell populations.
Multigenerational studies are challenging because investigators cannot rule out the
possibility that the associations they observe are due to other exposures in the initial
generation, exposures shared across generations, or chance. Image: © iStockphoto/kali9.
Photograph of an older woman, a young woman, a young girl, and an infant who belong
to three different generations of the same family.
That said, from a public health perspective, epigenetics offers the promise not only
for a better understanding of how the environment contributes to disease risk, but
potentially an inkling of how to intervene, according to Brandon Pearson, a neuroscientist
at Columbia University who studies epigenetics and environmental health. Currently,
however, the idea of epigenetic interventions for complex disorders remains a stretch.
“We do not yet understand what makes the epigenetic risk for something like ADHD,”
says Pearson. “There’s unlikely any single epigenetic mark or profiles in genetic
locations that could lead to ADHD or other neuropsychiatric disorder.”
A New Research Effort
It is impossible to say how many of the chemicals to which people are exposed today
may have neurodevelopmental and endocrine effects. “You have a rolling set of environmental
circumstances and new exposures created in every generation,” says Cohn. How these
layers of exposures may compound or interact with each other—and with our genes—remains
largely unknown. These associations are extremely difficult to quantify. “There are
no human studies that are perfect,” says Cohn. She knows that hers will not be, either.
In a second-floor walkup perched over a French deli in the heart of Berkeley, California’s
“Gourmet Ghetto,” Cohn and her colleagues at the CHDS are culling data from more than
20,000 births between the mid-1980s and 2013. These births represent a portion of
the cohort’s third generation—the grandchildren of the expectant mothers and fathers,
including Marcy Hartman and her husband, who volunteered for the study more than five
decades ago.
Cohn and colleagues have linked the birth records of the study grandchildren to an
existing database of autism cases maintained by the California Department of Health.
The researchers are now working to identify and analyze any links between grandparents’
exposures and behaviors—things like smoking, alcohol, and medication use—and autism
diagnoses in the third generation.18
Whereas the ALSPAC and NHS-II studies relied on the second generation to report their
parents’ exposures, CHDS researchers now have direct evidence. They have data from
medical records for pharmaceutical exposures such as DES. Both mothers and fathers
provided blood and urine shortly before and after their children were born.
Those samples were prepared and have been stored in a biorepository for the last several
decades. Now Cohn and colleagues can use the urine and blood to measure how much of
certain chemicals the grandparent generation would have had in their bodies at the
time. They can confirm, for instance, whether and how much a mother may have smoked
during pregnancy by measuring urine levels of cotinine—a metabolite of nicotine. Having
these measurements, according to Cohn, will help to increase confidence that any associations
they find are real.
Another unique thing about the CHDS dataset is that the researchers have information
on environmental exposures in both mothers and fathers of the first generation. Down
the paternal line, three generations is far enough to rule out the possibility of
direct exposure, notes Cohn.
Each generation has its own set of potential environmental circumstances and exposures.
Some of these exposures remain relevant to the day-to-day lives of subsequent generations.
But even for those that become obsolete, potential health effects may persist. Image:
CC BY 2.0 via https://www.flickr.com/photos/simpleinsomnia/25128717624/.
Archival photograph of a woman painting her fingernails, circa 1940s
And initial counts of the California autism database show just over 100 cohort children
with autism. That’s a large enough number, say the researchers, to find major associations,
but may not allow them to pick up on minor factors that would lead to weaker associations.
Each new study, though imperfect, adds a little bit more knowledge. New methods, such
Mendelian randomization—a technique that may strengthen causal inferences from epidemiological
data in some contexts—are providing researchers with more powerful tools to assess
potential effects of environmental pollutants. Nigg and colleagues recently used this
approach to further support evidence that lead exposure may contribute to ADHD.5
As the scientific landscape expands and researchers are able to address increasingly
complex questions about interactions between health and environment, the small contributions
made years ago by Hartman and other members of prospective longitudinal cohorts offer
a wealth of data that will keep on giving for years to come.