Gene imprinting

Genomic imprinting is the mechanism by which monoallelic expression is achieved in a parent‐specific fashion. Several human genes are known to be imprinted (Nazor et al. 2012). These comprise epigenetic changes, when the imprints are established in the germline, or somatic changes, when they arise during early embryonic development. Genomic imprinting defects are associated with developmental disorders, including Silver‐Russell, Beckwith‐Wiedemann, and Prader‐Willi syndromes. Genomic imprints are affected by environmental factors, and also associate with several human cancers. Gene imprints were then analyzed for paternal or maternal patterns and gene expression regulation (www.geneimprint.org).

DNA hypermethylation modules

Association with DNA hypermethylation modules was evaluated as described (Easwaran et al. 2012). Briefly, at a genome‐wide level, hypermethylated genes are marked by Polycomb complex signatures. Most of these genes comprise developmental regulators, which contain DNA hypermethylation modules, within bivalent, that is divergently regulated, chromatin.

DNA methylation dynamics

The status of DNA methylation of regulatory regions of height‐associated genes was assessed to provide information on DNA methylation dynamics (www.methdb.de/;202.97.205.78/diseasemeth/) (Lv et al. 2011).

Regulation of height‐associated genes by DNA methylation

Mechanistic relevance of DNA methylation was assessed versus a regulatory role of promoter methylation on the expression levels of height‐associated genes.

Signaling networks of height‐associated genes

To detect potential associations to cancer growth control pathways, network‐based analyses were designed:

SNOW (studying networks in the omics world; babelomics.bioinfo.cipf.es/) builds protein–protein contact networks and maps lists of genes or proteins over a reference interactome, where nodes are proteins and edges are interaction events. A SNOW interactome (minimal connected networks, MCN) of height‐associated genes was built using the HPRD (human protein reference database), IntAct, BIND (biomolecular interaction network database), DIP (database of interacting proteins) and MINT (molecular interaction database). Topological parameters included the distributions of node connections degrees (i.e., of the number of edges for each node); the betweenness (i.e., a measure of centrality of all nodes and of their distribution); the distribution of clustering coefficients (i.e., of the connectivity of the neighborhood of each node) and the number of components of the network (i.e., the different groups of nodes that are generated in a network analysis). MCNs were identified adopting the option of only one bridging protein between any pair of proteins analyzed. Statistical significance was computed by comparing the obtained networks with the entire reference interactome or with networks of random composition. All networks presented below have significantly higher betweenness and lower connections degrees as compared with irrelevant networks (P < 0.0001) (Fig. ​(Fig.11).

Open in a separate window

Figure 1.

Pathway analysis. Graphical representation of the height‐associated proteins relationships retrieved through SNOW network analysis. Proteins are represented as nodes (hubs), the biological relationships between the nodes (edges) are represented as lines. Height‐associated proteins are in red; linker proteins are in white; miRNA are in gray. Major hubs are in magenta; SMAD isoforms are in blue.

MetaCore^™ network analysis (http://lsresearch.thomsonreuters.com/pages/solutions/1/metacore) builds on a curated database. MetaCore^™ is based on a proprietary manually curated database of human protein‐protein, protein‐DNA and protein‐com pound interactions, metabolic and signaling pathways for human, mouse and rat, supported by proprietary ontologies and controlled vocabulary. Over 2000 multi step canonical pathway maps of human protein‐protein and protein‐DNA interactions are utilized as reference maps. The analyze network TF feature was exploited to identify transcriptionally regulated pathways. The shortest pathway algorithm was used to link the height‐associated genes with additional database objects along a directed path, enforcing the stringent option of one‐intermediate‐only.

Ingenuity pathways analysis (Ingenuity Systems, www.ingenuity.com) was utilized to map height‐associated genes onto the Ingenuity knowledge base (focus points). Networks of focus molecules were generated by maximizing specific connectivities. Networks were ranked by score (negative log P‐value by right‐tailed Fisher's exact test). Each score takes into account the number of molecules in the network, the final network size, as well as the dataset size and the total number of network‐forming molecules in the Ingenuity Knowledge Base.

Data meta‐analysis (Tables S1–S3) validated identity and relationships between network members (www.signaling-gateway.org/; www.wikipathways.org/index.php/Wiki-Pathways and www.genome.jp/kegg/pathway.html). Yes/no relevance for specific signaling pathways (SMAD; c‐Myc; p53; ERα) or processes (development, cell growth; apoptosis) was verified by wet‐lab approaches (www.ncbi.nlm.nih.gov/sites/entrez; and manuscript in prep.).

Gene imprinting

Gene imprinting was analyzed for paternal or maternal patterns and regulation of gene expression (www.geneimprint.org). Listed in Table S3 are the height‐associated genes/gene clusters that were found associated with imprinting (Nazor et al. 2012). Twenty five height‐associated genes were found to be imprinted. Additionally, imprinting of GNAS was shown to lead to either gynogenetic or androgenetic patterns. Growth factor receptor bound protein 10 (GRB10) expression was shown an imprinted gene (Li et al. 2008). PLAGL1 lies within a DNA methylation imprinting center on human chromosome region 6q24. Imprinting of PLAGL1 was also shown by expression and phenotypic profiling of parthenogenetic fetuses (Bischoff et al. 2009). Hypomethylation patterns were demonstrated for both PLAGL1 and GRB10. Using Affymetrix GeneChip microarrays (Carletti et al. 2006) and/or semiquantitative PCR, organs expression patterns were profiled, and PLAGL1 was found expressed in a tissue‐specific manner; imprinting was confirmed by pyrosequencing. A paternally imprinted regulatory role was also shown, as paternal noncoding RNAs at this locus can interact with regulators of active chromatin (Iglesias‐Platas et al. 2012).

DNA hypermethylation modules

The functional relevance of DNA methylation of height‐associated genes was first assessed by association with DNA hypermethylation modules (Easwaran et al. 2012) (Table S3), as marked by Polycomb‐complex signatures within bivalent chromatin. Most of Polycomb‐signed genes comprise developmental regulators, and were shown to contain DNA hypermethylation modules (Easwaran et al. 2012). As compared to control H3K4Me3 genes, which are hypermethylated in 1.5% of cases (Easwaran et al. 2012), DNA hypermethylation modules were detected in 42/87 height‐associated genes (P = 8.02e⁻¹⁷; Fisher exact test, two tailed).

DNA methylation dynamics

The extent of DNA methylation of regulatory regions of height‐associated genes was experimentally assessed as described by Lv et al. (2011) (www.methdb.de/; 202.97.205.78/diseasemeth/) (Table S3). All genes where DNA methylation plays a regulatory role are expected to undergo shifts in methylation states. We challenged this prediction in our gene‐set. Only five genes (ACAN, ANKS1, FBP2, NACA2, ZBTB38) were found to have no evidence of DNA methylation. The remaining genes (94.3%) were shown to undergo broad changes of DNA methylation levels across experimental conditions (Table S3), consistent with shifting between distinct, DNA methylation‐associated regulatory states.

Regulation of height‐associated genes by DNA methylation

A mechanistic relevance of these findings was investigated.

All assessed genes were shown to possess a CpG island in the promoter region that overlaps the TSS (Tables S1, S2), consistent with functional relevance on gene transcription (Esteller 2008).

Loss of ADAMTS proteases in nonsmall‐cell lung carcinomas is caused by hypermethylation in the ADAMTS genes promoters (Dunn et al. 2004).

BMPs are important regulators of cell growth, differentiation, and apoptosis. CpG island methylation in the BMP2 promoter causes loss of BMP‐2 protein expression in transformed cells (Wen et al. 2006). Shut‐down of the BMP6 gene by promoter methylation was observed in malignant lymphomas (Taniguchi et al. 2008).

Cdk5 and Abl enzyme substrate 1 (CABLES1) is a cyclin‐dependent kinase binding protein. Loss of nuclear CABLES1 expression is due to epigenetic modifications of the CABLES1 locus. Full ablation is caused by LOH of the transcriptionally‐competent allele (Sakamoto et al. 2008).

The expression of the deleted in colorectal cancer (DCC) gene is frequently lost in intestinal cancers. In up to three quarters of the cases the loss of expression of DCC is due to DNA methylation (Park et al. 2008).

Loss of the deleted in lymphocytic leukemia 7 (DLEU7) gene is frequently observed in chronic lymphocytic leukemia, due to hypermethylation of the DLEU7 promoter (Hammarsund et al. 2004).

Decrease or loss of EGF‐containing fibulin‐like extracellular matrix protein 1 (EFEMP1) has been shown in human breast cancer. EFEMP1 promoter methylation is a major cause of down‐regulation (Sadr‐Nabavi et al. 2009).

Hedgehog (Hh) signaling activation is frequently mediated by the epigenetic repression of the Hh‐interacting protein (HHIP) gene, a negative regulator of Hh signaling. Consistently, HHIP transcription is rescued by demethylation (Tada et al. 2008).

Hypermethylation of MOS is associated with the development of acute lymphoblastic leukemia (Scholz et al. 2005).

CpG methylation inhibits proenkephalin (PENK) gene expression by interfering with the binding of the stimulatory transcription factor AP‐2 (Comb and Goodman 1990).

Differential methylation of the zinc finger protein gene pleomorphic adenoma gene 1 (PLAGL1) was found in diabetes and cancer and was shown to modulate the expression levels of this gene (Arima and Wake 2006).

Patched homolog 1 (PTCH1) is the Hh receptor. Methylation of the PTCH1 promoter inhibits PTCH1 expression in breast cancers (Wolf et al. 2007).

The suppressor of cytokine signaling (SOCS) 2 CpG islands are hypermethylated in ovarian cancers. Aberrant methylation of this gene correlates with transcriptional silencing also in hepatocellular carcinomas (Martinez‐Chantar et al. 2008).

DNA methylation suppresses the expression of thyroid transcription factor‐1 (TTF‐1) in 60% of undifferentiated thyroid carcinomas (Kondo et al. 2009).

c‐Myc is a widely expressed TF, that can bind 10–15% of all genes (Orian et al. 2003). c‐Myc regulates at least seven height‐associated genes (CDK6, COIL, HMGA1, LIN28B, RBBP8, RPS20, TRIM25/EFP), and its binding to genomic loci is dependent on chromatin structure and CpG methylation.

Epigenetic defects and hereditary growth anomalies

Distinct epigenetic defects have been linked to separate hereditary growth anomalies, providing evidence for a broad regulatory role of DNA methylation on body growth. The Beckwith‐Wiedemann syndrome (130650) is caused by deregulation of imprinted genes within the 11p15 chromosomal region, i.e., KIP2, H19 and LIT1, whether alone or as interacting regulatory units (Niemitz et al. 2004). Hypermethylation at the 11p15 telomeric imprinting control region (ICR1), are observed in about 5 to 10% of affected patients (see below for opposite epigenetic changes in Silver‐Russel patients). Both H19 and LIT1, which encode untranslated RNAs, and IGF2 are either maternally imprinted genes with growth enhancing activity or paternally imprinted genes with growth suppressing activity. Key features of the syndrome are exomphalos, macroglossia, and gigantism in the neonate, supporting a direct involvement of epigenetic regulatory mechanisms in body size and body development. Consistently, alterations include visceromegaly, adrenocortical cytomegaly and overgrowth of the external genitalia in both males and females. Affected children reach an average height of 2.5 SD above the mean at or after puberty, and their growth velocity is above the ninetieth percentile until 4–6 years of age.

Prader–Willi syndrome (176270) is a complex childhood disorder that affects the nervous and hormonal systems and leads to excessive weight gain. It arises from the loss of activity of several imprinted genes on chromosome 15, that are usually expressed from the paternally inherited copy. Affected children suckle poorly for the first months of their lives; their voracious appetite and consequent obesity develop from weaning onwards.

Angelman's syndrome (AS) (105830) arises from opposite imprinting defects to those described in Prader–Willi syndrome. The Angelman syndrome is characterized by mental retardation, movement disorders, altered behavior, and speech abnormalities. Most cases are caused by absence of a maternal contribution to the imprinted region on chromosome 15q11‐q13, whereas the Prader‐Willi syndrome results from deletion of the same region in the paternal chromosome. Other patients with an Angelman syndrome carry mutations in the gene encoding methyl‐CpG‐binding protein‐2 (MECP2).

MECP2 (300005) is the gene mutated in the Rett syndrome, a neurological (regression of acquired skills, loss of speech, stereotypical movements, seizures, and mental retardation) and developmental disorder that occurs in females. Notably, affected girls are characterized by microcephaly and arrested development between 6 and 18 months of age.

Imprinting defects are associated with other developmental disorders, for example Silver‐Russell dwarfism. Up to 60% of cases of Silver‐Russell syndrome are caused by hypomethylation at the ICR1 on chromosome 11p15, involving the H19 (103280) and IGF2 (147470) genes (Penaherrera et al. 2012). Affected patients demonstrate severe intrauterine growth retardation, poor postnatal growth, craniofacial alterations, and a variety of minor malformations.

Epigenetic regulation of body size

The findings on Beckwith‐Wiedemann syndrome, Prader–Willi syndrome, Angelman's syndrome, Rett syndrome and Silver‐Russell syndrome suggest an epigenetic regulation of body size. Consistently, several other imprinting disorders have been shown to affect placental and fetal size (Constancia et al. 2004). The connection between imprinting and placentation is supported by the finding that erasing all genomic imprints results in the outgrowth of extra‐embryonic tissues and placenta in animal models. Correspondingly, imprinting defects in the female human germline cause the appearance of hydatidiform moles, that is uncontrolled growth of placental cells.

Height‐associated regulators of epigenetic heredity

Height‐associated genes include several key epigenetic regulators of gene expression (Esteller 2008). The activity of transcription factors is modulated by associations with co‐repressors, including histone deacetylase 7. BMP2 regulates the transcriptional activity of HDAC7 by inducing its export from the nucleus. The zinc finger and BTB domain containing (ZBTB38) is a methyl‐DNA‐binding transcriptional repressor gene that binds single methylated CpGs. Chromatin immunoprecipitation indicates that ZBTB38 recognizes the methylated alleles of the H19 and insulin‐like growth factor II (IGF‐2) genes and represses their transcription (Filion et al. 2006). Methylation‐linked inactivation of IGF‐2/H19 can be inherited in a parental‐specific manner (Riccio et al. 2009).

The DOT1‐like and NSD1 histone methyltransferases, the two high‐mobility group A (HMGA) genes, HMGA1 and HMGA2, and the histone clusters 1 and 2 are all involved in the assembly of chromatin structure. Notably, haploinsufficiency of the histone methyltransferase NSD1 causes the Sotos syndrome, that is characterized by very high stature (Kurotaki et al. 2002) and the gene can be inactivated epigenetically in tumors (Berdasco et al. 2009). PRMT5 contributes to coupling DNA methylation and chromatin organization via the methylation of histone H4R3, which then recruits DNMT3A. DNMT3A mediates additional regulatory loops in chromatin organization, as the HMGA1 gene itself is regulated by DNA methylation.

Heterozygous 17q11 microdeletions that encompass NF1 and RNF135 cause a subset of neurofibromatosis type 1 (Douglas et al. 2007). Notably, individuals with microdeletions are typically taller than individuals with NF1 mutations, consistently with the removal of a gene that negatively regulates human growth (RNF135 haploinsufficiency) (Douglas et al. 2007).

Prenatal environmental factors and persistence of DNA methylation changes in the adult

Epigenetic patterns in germ cells can persist into adulthood, and many imprinted genes continue to be expressed from a single parental copy. Environmental factors during pregnancy affect epigenetic marks, and in utero conditions have consequences on adult health and disease (Gluckman et al. 2008). This has been proven due to DNA methylation differences after exposure to prenatal famine (Tobi et al. 2009). Animal studies and human data on the imprinted IGF2 locus indicated a link between prenatal nutritional and DNA methylation. Additionally, methylation of the INSIGF, IL10, LEP, ABCA1, GNASAS and MEG3 genes was persistently modified by exposure to prenatal famine (Tobi et al. 2009).

In animal models manipulation of the maternal diet during pregnancy leads to a persistent shift in average DNA methylation levels of specific genes in offspring resulting in permanent changes in coat color or tail shape. Moreover, widespread dynamics of DNA methylation were shown in response to biotic stress (Dowen et al. 2012). Methyl donor supplementation prevents transgenerational amplification of obesity (Waterland et al. 2008) and maternal methyl supplements were shown to increase DNA methylation in offspring (Waterland et al. 2006).