Background
Emerging evidence indicates modifiable lifestyle factors and time-stable epigenetic determinants influence childhood obesity [
1,
2] and timing of puberty [
3]. Childhood obesity and early puberty are risk factors for several metabolic and reproductive disorders, including: early menarche [
4], adult obesity [
5], diabetes [
6], polycystic ovarian syndrome (PCOS) [
7], and breast cancer [
8]. PCOS and its common childhood antecedent, premature pubarche, are associated with hyperinsulinemia and androgen excess in overweight females [
7,
9]. Similarly, plasma leptin [
10] and local adiposal estrogen [
11] are associated with breast tumor growth in obese postmenopausal cases, and are hypothesized to drive early breast development and menarche in overweight girls [
12‐
15]. Given that the ovaries are largely quiescent pre-menarche and post-menopause, it is conceivable biomarkers that link adiposity to pubertal timing may inform risk of androgen and estrogen-related disease development throughout the life course. In this exploratory pilot study we asked whether promoter methylation marks in
PPARG or
CYP19A1, two genes that connect energy balance to lifetime estrogen exposure, are associated with pubertal development in girls.
The transcription factor PPARγ is a master regulator of adipose differentiation and endocrine function. Human and animal data link
PPARG hypermethylation to reduced PPARγ expression that is observed in cases of diabetes [
16], breast cancer [
17], and hyperandrogenic PCOS [
9]. Because PCOS and early pubarche share hyperinsulemia and androgen excess as common clinical features, it is conceivable
PPARG methylation aberrations may alter timing of pubarche if present in early development [
9].
Aromatase (product of the
CYP19A1 gene) catalyzes all bodily estrogen biosynthesis
via aromatization of androgen precursors, and inhibitors of this enzyme are highly effective therapies for estrogen-sensitive cancers of the breast [
18]. Aromatase expression varies across tissues and individuals owing to differential activation and repression of several tissue-specific gene promoters. Each of these promoters regulates a unique untranslated ‘first exon’ (‘exons I’) that is spliced to the common coding exons II-X and can be mapped back to its promoters for purposes of deducing what factors drive transcription in a particular tissue [
19]. In healthy adipose,
CYP19A1 is expressed in fibroblasts primarily from activation of a distal glucocorticoid-regulated promoter termed ‘pI.4’, and is lost in the path of adipocyte differentiation driven by PPAR
γ[
20]. In malignant breast biopsies,
CYP19A1 is overexpressed 3–4 fold in tumor and proximal adipose tissue from several tissue-specific promoters [
21], but primarily from two largely ‘gonad-specific’ cAMP-responsive promoters termed “
pII/I.3” [
22] that are activated by cancer-associated transcription factors PPARγ suppresses in healthy breast [
17,
23]. Though this phenomenon of increased tissue-specific promoter usage with preference for gonadal promoters (termed ‘promoter switching’) is well-documented in the breast cancer literature [
19,
24], its timing and extent of tissue distribution related to intermediate risk indicators is less understood. Demura and Bulun [
25] recently described hypomethylation of a CpG dinucleotide in the ‘cAMP-response element-like sequence’ (CLS) of
pII/I.3, which they detected in aromatase-overexpressing fibroblasts derived from a skin-punch biopsy of a healthy patient. In light of the aggregate of findings regarding CYP19A1 misexpression from gonadal promoters in breast cancer cases, Demura and Bulun postulated
CYP19A1 pII/I.3 hypomethyaltion may contribute to the phenomenon of ‘promoter switching’ and inter-individual variability in lifetime estrogen exposure.
In the present study we sought to determine whether methylation of this
CYP19A1 pII/I.3 locus [
25] or the average of five CpG dinucleotides in a differentially methylated region of the
PPARG promoter [
9,
16,
26] was associated with timing of pubic hair (PH2) or breast development (B2) in a cohort of New York City, Black and Hispanic girls who were enrolled in a study of pubertal timing between 6–8 years of age.
Discussion
Early breast and pubic hair development have been associated with disordered leptin, insulin, and IGF-1 profiles in overweight girls in numerous studies [
30]. Perturbations in estrogens and androgens, critical drivers of breast and pubic hair development, remain clinically more challenging to detect [
31]. Given national trends, there is great motivation to identify biomarkers that add value to current plasma and anthropometric measures used in predicting puberty onset [
32]. In this exploratory study we aimed to ascertain whether salivary methylation of the
CYP19A1 and
PPARG promoters was related to age at breast or pubic hair development in girls, both independently and in concert with body size. In light of the current literature, we anticipated overweight girls with
CYP19A1 hypomethylation and
PPARG hypermethylation might be predisposed to early breast development [
33‐
35], and those with
PPARG hypermethylation to early pubic hair development [
9,
16].
Our main observations were that relative hypomethylation of a CpG in the gonadal
CYP19A1 promoter termed
“pII “ was associated with earlier age at B2 among overweight girls only (Table
3), and with earlier age at PH2 independent of body size (Table
2). While only correlative and based on a relatively small number of samples, our B2 findings are supported by a case report authored by Demura and Bulun [
25], which describes hypomethylation of
pI.3/II in CYP19A1-overexpressing fibroblasts relative to CYP19A1-quiescent fibroblasts derived from punch biopsies of four healthy subjects. In their report, CYP19A1 activity was robustly induced in the former upon cAMP stimulation, while fibroblasts from the other three subjects were cAMP-refractory. Further investigation revealed CpG dinuleotides within and proximal to the CLS (CRE-like sequence) of gonadal
pI.3/II were relatively hypomethylated in cAMP-responsive CYP19A1-overexpressing fibroblasts, and were relatively hypermethylated in non-responsive fibroblasts. These results support the hypothesis that
CYP19A1 hypomethylation may be an early ‘permissive’ event, which renders one susceptible to subsequent intrinsic/extrinsic transcriptional activators of
CYP19A1, and concomitant local or systemic estrogen excess. Such a ‘two-hit’ mechanism of derepression (hypomethylation) and activation (e.g., obesity-related cytokines) may also explain why
CYP19A1 hypomethylation was associated with early B2 in overweight, but not normal weight girls in the present study.
Aromatase catalyzes estrogen biosynthesis from androgen precursors. Elevated androgen, insulin, and IGF-1 signaling are widely accepted co-determinants of early pubarche in overweight girls [
7,
30,
36]. Thus, our finding that
CYP19A1 hypomethylation (theoretical increase in expression) was related to earlier age at PH2, independent of BMI (Table
2), was unanticipated. While intriguing, the statistical significance of this association was attenuated after adjustment for covariates, and we can only speculate as to its implications without further study.
We assessed methylation of the −383 to −281 bp region of the
PPARG promoter as PPARγ suppresses CYP19A1 expression in breast tissues in culture [
20,
33], and relative hypermethylation of this region has been associated with reduced PPARγ expression in hyperandrogenic PCOS [
9] and diabetes models [
16]. Though we detected no statistically significant effects related to
PPARG methylation in the present study, puberty-associated methylation patterns may exist in genes for PPARγ co-factors, effectors, or downstream targets in salivary or other surrogate tissue DNA. Indeed, methylation biomarkers of childhood adiposity and maternal BMI have been described in
RXRA and
PPARGC1A when assayed in umbilical tissue [
2,
37].
This exploratory investigation has several limitations regarding generalizability, including but not limited to: small sample size, lack of perceived stress assessments, use of candidate genes, and DNA derived from whole saliva samples collected only from Black and Hispanic girls. We describe salivary
CYP19A1 hypomethylation not as a ‘causal’ event, but merely as a ‘surrogate biomarker’ that with further study may have utility in predicting risk of premature breast development in overweight girls. Specifically, the CpG we describe is contained in a critical transcription factor binding site (CLS), located in a strong
CYP19A1 gonadal promoter termed ‘
pII’, which is activated by the ubiquitous pleiotropic second messenger cAMP in the follicular phase of the menstrual cycle [
38]. DNA methylation is highly tissue-specific, and CYP19A1 is not likely expressed in buccal epithelial cells from gonadal
pII to any significant degree. However, weighing these considerations together supports the notion that hypomethylation of
CYP19A1 pII in whole saliva- a tissue in which expression from gonadal
pII is likely silenced, may in fact represent a methylation aberration, possibly established early in life. Such a mark could be deemed a ‘surrogate’ aberration if it portends risk generalizable to more functional tissues with niche transcriptional machinery requisite to affect CYP19A1 expression changes that promote disease, as has been extensively reported for
pII and estrogen-related disorders [
22,
24].
Our findings are only suggestive and only extend to saliva samples we collected from Black and Hispanic girls. Procurement of effector/target tissues (e.g. adipose/breast) to investigate the validity of salivary
pII methylation as a risk surrogate in a statistically robust manner in more diverse pediatric populations is precluded by ethical, logistical, and economic considerations. However, animal and cell line co-culture models designed to capture adipose and developing breast tissue interactions are emerging [
39,
40], and it will be interesting to follow developments that functionally characterize the complex biological and environmental interactions that orchestrate epigenetic factors related to thelarche and pubarche onset.
Competing interests
All authors declare that they have no competing interests.
Authors’ contributions
MSW and SLT are the co-PIs of the parent study. TRS and JC contributed to the design of this sub-study. TRS performed pyrosequencing analysis, all related laboratory procedures, and assumes the primary role of composing the manuscript with the supervision of JC. SLT and AP performed statistical analysis. All authors were responsible for critical revision of the manuscript and approved the final copy.