An epigenetic clock for gestational age at birth based on blood methylation data

Knight et al. Genome Biology (2016) 17:206 DOI 10.1186/s13059-016-1068-z RESEARCH Open Access An epigenetic clock for gestational age at birth based on blood methylation data Anna K. Knight1, Jeffrey M. Craig2, Christiane Theda3, Marie Bækvad-Hansen4, Jonas Bybjerg-Grauholm4, Christine S. Hansen4, Mads V. Hollegaard4,5, David M. Hougaard4,5, Preben B. Mortensen6, Shantel M. Weinsheimer7, Thomas M. Werge7, Patricia A. Brennan8, Joseph F. Cubells1,9,10, D. Jeffrey Newport11, Zachary N. Stowe12, Jeanie L. Y. Cheong2,3, Philippa Dalach2, Lex W. Doyle2,3, Yuk J. Loke2, Andrea A. Baccarelli13, Allan C. Just14, Robert O. Wright14, Mara M. Téllez-Rojo15, Katherine Svensson14, Letizia Trevisi16, Elizabeth M. Kennedy1, Elisabeth B. Binder10,17, Stella Iurato17, Darina Czamara17, Katri Räikkönen18, Jari M. T. Lahti18,19,20, Anu-Katriina Pesonen18, Eero Kajantie21,22,23, Pia M. Villa24, Hannele Laivuori25,26, Esa Hämäläinen27, Hea Jin Park28, Lynn B. Bailey28, Sasha E. Parets10, Varun Kilaru28, Ramkumar Menon29, Steve Horvath30,31, Nicole R. Bush32,33, Kaja Z. LeWinn32, Frances A. Tylavsky34, Karen N. Conneely1,9† and Alicia K. Smith1,10,28*† Abstract Background: Gestational age is often used as a proxy for developmental maturity by clinicians and researchers alike. DNA methylation has previously been shown to be associated with age and has been used to accurately estimate chronological age in children and adults. In the current study, we examine whether DNA methylation in cord blood can be used to estimate gestational age at birth. Results: We find that gestational age can be accurately estimated from DNA methylation of neonatal cord blood and blood spot samples. We calculate a DNA methylation gestational age using 148 CpG sites selected through elastic net regression in six training datasets. We evaluate predictive accuracy in nine testing datasets and find that the accuracy of the DNA methylation gestational age is consistent with that of gestational age estimates based on established methods, such as ultrasound. We also find that an increased DNA methylation gestational age relative to clinical gestational age is associated with birthweight independent of gestational age, sex, and ancestry. Conclusions: DNA methylation can be used to accurately estimate gestational age at or near birth and may provide additional information relevant to developmental stage. Further studies of this predictor are warranted to determine its utility in clinical settings and for research purposes. When clinical estimates are available this measure may increase accuracy in the testing of hypotheses related to developmental age and other early life circumstances. Keywords: Developmental age, Aging, Epigenetic clock, DNA methylation, Preterm birth, Cord blood, Fetus, Blood spot, Biomarker, Medicaid, Socioeconomic status, Birthweight * Correspondence: † Equal contributors 1 Genetics and Molecular Biology Program, Emory University, Atlanta, GA, USA 10 Department of Psychiatry & Behavioral Sciences, Emory University School of Medicine, Atlanta, GA, USA Full list of author information is available at the end of the article © 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Knight et al. Genome Biology (2016) 17:206 Page 2 of 11 Background Differences in gestational age (GA) as small as one week have been shown to have significant impacts on neonatal morbidity and mortality, as well as long-term outcomes [1–6]. In light of this, the American College of Obstetricians and Gynecologists (ACOG) recently recommended revising the categorization of births from term (>37 weeks gestation) and preterm (≤37 weeks gestation) into several subcategories (early preterm, preterm, early term, full term, late term, and post term) that better reflect the developmental differences associated with GA at each of these time points [7, 8]. Accurate classification systems that reflect both developmental time and maturity may improve our ability to predict neonatal risk. Traditionally, GA is estimated using one or more of the following methods: early obstetric ultrasound, last menstrual period (LMP), or neonatal estimation [9]. Ultrasound-based methods are considered to be the gold standard and have proven to be a better predictor of delivery date [10] as LMP estimates may be influenced by uncertainty regarding LMP dates, normal variations in ovulation timing, atypical bleeding, and contraceptive use [9]. Neonatal estimation, which is based on a combination of physical appearance, muscular tone, flexibility, and reflexes, is the only available method for determining GA after birth but is less precise than LMP and ultrasound [9, 11, 12]. In circumstances where LMP date is uncertain and ultrasounds are not available, a more accurate method for estimating GA may be beneficial. Recently, DNA methylation (DNAm) has been used to accurately predict chronological age in children and adults [13–16]. Later work revealed that a methylation-based prediction of age may also associate with physiological consequences in adults when a study reported that an increased methylation age relative to chronological age was associated with an increase in mortality risk [17–22]. However, the predictors optimized in these studies were not designed to estimate GA and did not attempt to differentiate between different GA, as samples taken at birth were either assigned an age of zero or were excluded from the model [13, 14]. Because the accuracy and precision of a prediction model is, in general, weakest at the extremes of the distribution, a predictor developed from primarily adult samples would, by nature, be less accurate in neonates than one that is optimized for that purpose. DNAm differences in specific CpG sites have been associated with GA at birth in multiple studies [23–26]. We hypothesize that a predictor designed specifically for use with umbilical cord blood or blood spots already routinely collected for newborn screening could allow for accurate neonatal estimation of GA that may also be informative of developmental stage. The objective of this study was to develop such a predictor to estimate GA from DNAm data using umbilical cord blood or blood spot samples and to assess its ability to predict other indicators of developmental maturity. Results and Discussion DNAm data from 1434 neonates, representing 15 independent cohorts, were used for this study. For each sample, HumanMethylation27 (...truncated)