Going back to the future with Guthrie-powered epigenome-wide association studies

Epigenome-wide association studies (EWAS) can be used to investigate links between early life environment, epigenetics and disease. However, such studies raise the question of which came first: the mark or the malady? A recent study has demonstrated that EWAS can be performed on neonatal 'Guthrie' heel-prick blood spots. As Guthrie cards are collected from all newborn infants and stored indefinitely in many countries, they represent an important timepoint to compare with later disease-associated epigenetic marks.

Th ere is growing evidence that noncommunicable diseases such as cardiovascular disease (CVD), stroke, hypertension and type 2 diabetes may originate in early life, a paradigm known as the Developmental Origins of Health and Disease (DOHaD) [1]. Investigations into mechanisms underpinning DOHaD using animal models and human specimens have suggested the involvement of epigenetics: mitotically heritable changes in gene expres sion controlled by chemical modifi cations to chromo somes without altering the DNA sequence. One striking example implicates a role for the epigenetic mark of DNA methylation in the longterm eff ects of the Dutch famine during the Second World War. By middle age, off spring previously exposed to maternal malnutrition during early gestation had a higher incidence of CVD than their unexposed siblings, together with diff erences in DNA methylation in metabolic and CVDrelated genes [2]. Animal studies indicate that epigenetic modifi cations may be reversible by pharmacological or dietary inter ventions, suggesting approaches for future targeted interven tions in humans.
To accelerate progress towards this goal, EWAS have been advocated. Th ese studies use techniques varying in sensitivity, coverage, sequence bias and amounts of DNA required. A handful of EWAS have already been pub lished and many more are in progress. An important question raised by these studies is that of causality: when diseaseassociated epigenetic diff erences are identifi ed, do they refl ect causal pathological pathways of disease or a subsequent eff ect of disease? Furthermore, false posi tives can be captured in such screens. Th us, ideally, EWAS should be conducted longitudinally where possible. To this end, recent EWAS have found common epigenetic changes in pre and postsymptomatic children with type 1 diabetes (T1D) [3], and diff erences in fresh cord blood at gene loci whose expression was associated with body mass index in late childhood [4].

The utility of Guthrie cards in epigenetic research
Around the world, newborn babies are routinely screened for inborn errors of metabolism and other congenital disorders through testing of neonatal blood spot cards, a technique pioneered in the early 1960s by Robert Guthrie, after whom the cards are named. Collected within a few days of birth from heel pricks, Guthrie cards usually contain four blood spots 6 to 10 mm in diameter. Th e duration of Guthrie card archiving varies between and within countries, ranging from a few months to indefi nitely. Th us, in many countries, Guthries represent a nearperfect national biorepository. In addition to serum analytes, DNA has been extracted from Guthries and has been used for such purposes as carrier screening for cystic fi brosis, detection of HIV and, more recently, genomewide association studies. Modest degradation of DNA occurs during storage and extraction that could aff ect data quality [5], although the quality appears suffi cient for genomic assays. Remarkably, gene expres sion has also been analyzed in RNA from 20yearold Guthries [6]. As for the epigenetic regulators of gene expression, it has been shown that DNA methylation can be analyzed at individual genes using the widely used technique of bisulfite conversion, which converts methy lation differences to sequence differences [7]; this has been applied to Guthriebased methylation screening of the FMR1 (fragile X mental retardation 1) gene to predict cognitive impairment in individuals with fragile X syndrome [5].
A recent paper by Beyan and colleagues [8] has now taken this one stage further, conducting a proofof principle epigenomewide pilot study using Guthrie card methylomics. On average, 200 ng of DNA from each 6 mmdiameter Guthrie spot was extracted and used in two methods of genomescale methylation profiling: one arraybased, the other based on immunoprecipitation of methylated DNA followed by highthroughput sequen cing. For the arraybased method, DNA was extracted from 10yearold Guthries and compared with fresh blood and sperm from unrelated individuals. This approach identified tissuespecific differentially methy la ted regions between sperm and blood. There was an excellent genomewide correlation between archived Guthrie DNA and fresh blood, but a weaker correlation for the subset of regions showing small (<20%) differences in methylation between the two tissues. No comparisons were reported between fresh and aged DNA from the same individual, which would have been an ideal control for the effect of storage of Guthries on measurement of DNA methylation. This represents an important caveat of the present study. For the immunoprecipitationbased method, which usually requires 2 µg of DNA, the method was adapted to work with 200 ng. The team then attempted to define the regions of the genome that differed between individuals but remained constant from birth to 3 years of age. This was an important comparison because it has been proposed that such 'metastable epialleles' are influenced by environmental and stochastic factors in utero, remain constant thereafter, and can act as stable biomarkers for disease risk [9]. For this, the team was careful to exclude genomic regions for which genetic heterogeneity could influence epigenetic varia tion and focused on clusters of variable, stable regions. Unfortunately, due to the low DNA yields, the longi tudinal comparisons were limited to the arraybased technique at birth and the immunoprecipitationbased technique at 3 years of age. Nevertheless, up to a dozen metastable epialleles were identified, two of which had previously been associated with human disease.

Where do EWAS go from here?
The study of Beyan et al. highlights the utility of Guthries for longitudinal EWAS in which retrospective case control studies can produce data more quickly and cheaply than birth cohort studies. However, the latter are designed to collect data on maternal exposures with minimal recall bias, which is not possible in retrospective studies. Furthermore, it is likely that a portion of the epigenome is still susceptible to environmental and stochastic influences in early postnatal life, making a case for repeat sampling. One disadvantage of using whole blood is that methylation levels represent an average of the levels in each of its component cell types, the proportion of which may change over time. However, if the early environment leaves an epigenetic legacy in multiple tissues, this will be a minor issue.
Future longitudinal EWAS (Figure 1) will need to be sufficiently powered to detect diseaseassociated 'epialleles' in contrast to the pilot study of Beyan and colleagues, which was designed to detect large methylation differ ences (>20%) in a small number of comparisons (n = 3 individuals). The group previously compared methylation profiles of 15 monozygotic twin (MZ) pairs discordant for T1D, identifying 132 T1Dassociated methylation variants with withinpair methylation discordance of 0.1 to 6.6%. Crosscohort validation was performed with four additional T1Ddiscordant MZ twins and the temporal origin of T1Dassociated methylation differences were assessed with blood sampled from seven children with T1D before and after presentation, the latter using profiles from the same individuals positive for diabetes associated autoantibodies but negative for typical symp toms of T1D [3]. Another recent study looked for methy lation events that covaried with body mass index at two time points in 74 individuals [9]. Although different techniques were used, importantly, both of these studies used the same array platform at each time point.
It is also worth noting that there may be ethical barriers to longitudinal EWAS, as the use of Guthries without consent has been a major issue in some locations [10]. Currently, cards can be used for limited forensic purposes and deidentified in research. But should consent always be sought for use of Guthries in epigenetic research? Research studies using small numbers of samples are not generally problematic because it is easy to get consent from the individuals or parents. However, studies requir ing large numbers of samples, such as wellpowered EWAS, are a problem because it may not be practical to obtain consent from all the individuals involved. Newborn infant screening programs have recognized these ethical issues and parents are presently better informed about potential uses of stored Guthries, with some programs having introduced a consent process for future deidentified research.

Conclusions
Beyan and colleagues have shown that it is possible to perform longitudinal EWAS starting with blood samples from cases and controls after disease presentation and adding in blood samples obtained from Guthries at birth (Figure 1), a feat only previously achievable (in reverse order) through large birth cohort studies. Limitations that still need to be overcome include optimization of the amount and quality of DNA extracted from Guthries, identification of any technical artifacts associated with long term storage, an increase in study power and over coming ethical barriers. In addition, longitudinal birth cohort studies should aim to sample at multiple time points to determine which diseaserelated epigenetic changes are present at birth and which develop after birth in response to postnatal environmental exposures. Nevertheless, the central message of the paper by Beyan and colleagues is that we now have another arrow in our quiver with which to reach the ultimate target of EWAS: to discover early, reversible biomarkers for human disease. We should move forward in an ethically responsible manner.

Competing interests
MC and JMC are currently performing EWAS with DNA from Guthrie cards. Figure 1. Strategy for future, longitudinal EWAS using Guthrie cards. Epigenome-wide comparisons comparing blood from cases and controls at birth (Guthries) and after disease presentation (blood in any form). After data filtering, differentially methylated regions will be compared to provide stable, disease-associated methylated events. Cross-cohort validation will be an essential step in biomarker discovery.