Epigenetics and genomics in Turner syndrome

1 INTRODUCTION

The genetic background for Turner syndrome (TS) is an absence of all or part of one X chromosome in females. The karyotypic variation is large and encompasses 45,X, 45,X/46,XX, karyotypes with isochromosomes (either consisting of p or q arms), and mosaic compositions with a Y chromosome or fragments thereof (Gravholt et al., 2017). The phenotypic breadth of presentation among patients with TS is very variable with some presenting with textbook features and others presenting with no or almost no traits despite identical karyotypes (El-Mansoury et al., 2007). This hampers early diagnosis (Stochholm, Juul, Juel, Naeraa, & Gravholt, 2006), and suggests that it is difficult or perhaps impossible to make sufficiently meaningful karyotype–phenotype associations. For example, some individuals with 45,X have bicuspid aortic valves and aortic dilation, while others with the same karyotype have normal tricuspid aortic valves without aortic dilation (Hjerrild et al., 2010). Likewise, most individuals with 45,X and no signs of mosaicism, have no signs of ovarian function (Pasquino, Passeri, Pucarelli, Segni, & Municchi, 1997), while others with the same karyotype have normal ovarian function and can even conceive offspring (Mortensen, Rohde, Uldbjerg, & Gravholt, 2010). Furthermore large variability in survival is seen (Stochholm et al., 2012). Many other traits in TS show similar diversity. It thus seems clear that the understanding of genetics behind TS and the phenotypic variability leaves much to be unraveled. Here, we will discuss recent advances in the understanding the role of epigenetic and genomic changes in TS.

2 GENOTYPE–PHENOTYPE ASSOCIATIONS

Turner syndrome is caused by a partial or complete lack of a second X chromosome in a female, resulting in the development of a highly variable genotype and clinical phenotype.

About 40–50% of women with TS have a karyotype of monosomy X (El-Mansoury et al., 2007; Gravholt, Andersen, et al., 2017; Gravholt, Juul, Naeraa, & Hansen, 1996), 15–25% have mosaicism with 45,X/46,XX and about 3% present with 45,X/46,XY. Structural alterations of the X chromosome can be found in 20% of cases with isochromosome Xq being most frequent. Ring X chromosome with deletion of much of the short and long arms of the X chromosome are also present and 10–12% of cases have differing amounts of Y chromosome material (Cameron-Pimblett, La, King, Davies, & Conway, 2017; Gravholt, Andersen, et al., 2017).

Since 99% of 45,X fetuses are thought to be aborted spontaneously during the first trimester the 1% of 45,X cases constitutes a surviving minority (Hassold, 1986). Although not proven, it has been proposed that surviving 45,X individuals have some degree of undiscovered mosaism (Deng, Berletch, Nguyen, & Disteche, 2014; Held et al., 1992; Hook & Warburton, 2014). Since mosaism can occur in all cell lines, not only in peripherial lymphocytes, it challenges the determination of the exact degree of mosaism present in women with TS (El-Mansoury et al., 2007; Sybert & McCauley, 2004). Furthermore, a comparative analysis between karyotype and phenotype is complicated by the variability in the definition of clinical features.

Mosaicism varies with tissue type and age of a given person, and in some women over 50 years of age there can be a loss of one X chromosome in some cells, with resultant 45,X/46,XX mosaism occurring as a variant, which is not to be considered as TS, but may be related to development of diseases, such as autoimmune primary biliary cirrhosis (Denes, Landin-Wilhelmsen, Wettergren, Bryman, & Hanson, 2015; Gravholt, Andersen, et al., 2017; Invernizzi et al., 2004). Furthermore, if the percentage of abnormal cells does not increase above 5% in younger females, it should also not be confused with a diagnosis of TS, unless there are clear phenotypical signs consistent with a diagnosis of TS. Women with TS and 45,X/46,XX mosaicism usually have a milder phenotype and less severe congenital heart disease (Cameron-Pimblett et al., 2017; El-Mansoury et al., 2007) and are more likely to experience spontaneous menarche and even pregnancies (Bernard et al., 2016; Bryman et al., 2011; Negreiros, Bolina, & Guimaraes, 2014). Women with 45,X/47,XXX mosaicism also have a higher rate of spontaneous menarche and a milder phenotype (Sybert, 2002). The presence of Y chromosomal material is sometimes associated with virilization during puberty and infrequent development of gonadoblastoma (Gravholt, Fedder, Naeraa, & Muller, 2000; Oliveira et al., 2009), a benign tumor which can become malignant over time (Gravholt et al., 2017). Further, women with TS and a 45,X/46,XY mosaic karyotype appear less likely to develop autoimmune disorders such as hypothyroidism (Cameron-Pimblett et al., 2017).

Women with TS with isochromosome Xq experience the same comorbidities as women with the 45,X karyotype, but an increased incidence of hearing loss, autoimmunity and congenital heart disease have been reported (Barrenas, Landin-Wilhelmsen, & Hanson, 2000; Prakash et al., 2016). Ring X chromosome can be specifically associated with learning difficulties, and behavioral maladjustment in some cases (Kuntsi, Skuse, Elgar, Morris, & Turner, 2000; Leppig et al., 2004), which is thought to be due to loss of XIST on the ring chromosome, which leads to loss of inactivating capacity on the ring chromosome, and therefore the genes on the ring remains expressed in all cells (Callen et al., 1995). In addition, an increased incidence of metabolic syndrome has been observed (Cameron-Pimblett et al., 2017).

The mechanisms by which monosomy for the X chromosome disrupts development are still not fully understood. The “gene dosage effect” theory proposes that haploinsufficiency of X-chromosomal genes affects the embryonic development into adult life. This explains the cause of TS based on the absence of a limited number of dosage sensitive genes primarily localized on the Xp chromosome. However, the phenotype may not merely be due to genomic imbalance from haploinsufficiency of X-chromosomal genes, since two individuals with the exact same karyotypes, can exhibit very different traits and comorbidities. It is more likely that the haploinsufficiency of X-chromosomal genes in several ways influence the genome via different pathways. These mechanisms might be related to X chromosome inactivation, imprinting of genes located on the X chromosome or epigenetic factors whereby haploinsufficiency affects other autosomal genes within a given network and alters the regulation of gene expression across the genome.

In women with TS the X chromosome have been shown to be maternal in 60–80% of cases, paternal in 35–15% of cases, and due to post-zygotic loss in remaining cases (Abramowitz, Olivier-Van, & Hanover, 2014; Helena & Morris, 2007; Lepage, Hong, Hallmayer, & Reiss, 2012; Mohandas et al., 1992). The parental origin of the X chromosome could affect phenotypic diversity and it has been proposed that some X-linked genes could be imprinted (Skuse et al., 1997). Imprinting is a phenomenon dependent on differential methylation of DNA based on parental origin, which can lead to differential allele expression of a given gene (Abramowitz & Bartolomei, 2012). Cardiovascular disease and cognitive function have been attributed to parental origin of the X chromosome. No specific genes have been proven responsible, though (Abramowitz et al., 2014). Other studies have, however, not been able to validate a differential cognitive profile dependent on parental origin of the X chromosome (Lepage et al., 2012). It is likely that much larger studies will be necessary to settle differences found in current literature.

3 X CHROMOSOME INACTIVATION AND INFLUENCE OF THE EXTRA X CHROMOSOME IN KLINEFELTER SYNDROME

The X chromosome contains 155 million basepairs (~900 genes) whereas the Y chromosome contains 55 million basepairs (~200 genes). The X and Y chromosome consists of identical pseudoautosomal regions, PAR1 (24 genes) and PAR2 (4 genes) at the end of Xp and Xq, respectively (Helena & Morris, 2007). These regions allow the X and Y chromosomes to pair and segregate during meiosis in males (Mohandas et al., 1992). X chromosome imbalance requires mechanisms to equalize gene dosage between sexes and relative to autosomes in order to avoid a potentially inexpedient double-dose (Lyon, 1961). This involves two processes; X chromosome inactivation, silencing genes on one of the X chromosomes leading to functional X monosomy (for the majority of X genes), and X chromosome upregulation leading to increased gene expression on the single active X chromosome in males or females. All genes within PAR1 escape X-inactivation and are therefore candidates for the etiology of TS (Blaschke & Rappold, 2006).

The initial stage of X-inactivation is governed by the XIST gene. The gene is transcribed into Xist, a long-non-coding RNA, which coats the X chromosome from which it is transcribed and recruits various complexes to silence the chromosome. The exact mechanisms are still being investigated. In total, about 15% of X-linked genes escape inactivation (escape genes) and 10% shows variable expression (Balaton, Cotton, & Brown, 2015; Carrel & Willard, 2005). Some of the phenotypic features seen in TS, as well as in other sex chromosome aneuplodies such as Klinefelter syndrome and Triple X syndrome may be explained by haploinsufficiency or overexpression of escape genes, which then would be lacking in TS and overexpressed in Klinefelter and Triple X syndrome. Several X-chromosome genes have functionally equivalent Y homologs. Escape genes that lacks a Y paralogue are especially interesting as the cause of sex specific differential gene expression and phenotypes (Berletch, Yang, Xu, Carrel, & Disteche, 2011). One example is KDM6A that regulates reproduction-related genes in females and might be involved in ovarian dysfunction in TS (Berletch, Deng, Nguyen, & Disteche, 2013), and it can also be found mutated in some cases of Kabuki syndrome (Miyake et al., 2013), which is a rare syndrome also characterized by multiple congenital anomalies, variable degrees of mental retardation, short stature, a peculiar facial gestalt, skeletal and visceral abnormalities, cardiac anomalies, and immunological defects, most often caused by mutations in KMT2D or KDM6A (Lintas & Persico, 2018). Furthermore, KDM6A is also thought to be involved in congenital cardiovascular malformations perhaps overlapping with what is seen in congenital malformations of the heart in TS (Trolle et al., 2016). Interestingly, KDM6A is differentially methylated in TS (Trolle et al., 2016) and it is thought to be involved in congenital cardiovascular malformations overlapping with what is seen in TS (Trolle et al., 2016).

The SHOX gene (short stature homebox) is located on Xp22.23 (PAR1), and is the only X-chromosome gene that has been convincingly linked with a phenotypic trait in TS. It belongs to the homebox gene family, which is a transcriptional regulator and associated with short stature and skeletal growth (Rao et al., 1997). The function of the gene is dosage dependent and causes short stature in TS. Extra copies of SHOX leads to increased stature as seen in other sex chromosome aneuploidy conditions such as 47,XXX, 47,XYY, 47,XXY and 48,XXYY syndromes (Ottesen et al., 2010). An increased severity in growth deficits has been reported in Turner women with a ring X chromosome and isochromosome Xq, and even more pronounced than in 45,X (Fiot et al., 2016), suggesting that haploinsufficiency for an unknown Xp gene also decreases growth both pre- and post-natally in individuals with TS.

Transcriptome analyses of the X chromosome have not revealed a direct correlation between the nature of the genomic imbalance and gene expression levels (Sharma et al., 2015; Skakkebaek et al., 2018; Trolle et al., 2016). A novel study investigating sex chromosome dosage revealed that some X-linked genes were expressed in a dosage specific manner, such that having fewer X chromosomes was linked with upregulation (45,X) in comparison with the normal sex chromosome complement (46,XX) or having supernumerary X chromosomes as in Triple X (47,XXX) syndrome (Raznahan et al., 2018). Further, the authors reported that the Y chromosome dosage also influences the X-chromosome gene expression. These findings modify the simple gene-dosage relationship and influence on phenotype. Interestingly, genes encoding for transcription factors are more likely to be compensated to avoid harmful imbalances (Deng & Disteche, 2010; Veitia, Bottani, & Birchler, 2008). Such compensatory mechanisms have been described before and include increased transcription and extended RNA half-life (Prestel, Feller, & Becker, 2010).

4 EPIGENETICS AND RNA EXPRESSION IN TS

While X chromosome haploinsufficiency has been the focus of much research so far, studies investigating if epigenetic mechanisms may be part of the genetic mechanism behind sex chromosome aneuplodies are just emerging (Table 1). From these studies it has become evident that epigenetic processes are altered in subjects with sex chromosome aneuploidies (Alvarez-Nava & Lanes, 2018; Rajpathak et al., 2014; Sharma et al., 2015; Skakkebaek et al., 2018; Trolle et al., 2016). Epigenetic modifications are processed by several mechanisms. DNA methylation of CpG sites in the promoter region (Tirado-Magallanes, Rebbani, Lim, Pradhan, & Benoukraf, 2017), regulation of gene expression by noncoding RNAs (Wei, Huang, Yang, & Kang, 2017), remodeling of nucleosomes, or post-translational modifications of histones (Suganuma & Workman, 2011). Since DNA methylation at many positions is a reversible process (there are metastable epialleles, which are genomic regions in which DNA methylation is formed stochastically during fetal life and remain unchanged across tissues throughout life (Dominguez-Salas et al., 2014; Waterland et al., 2010)), it might be a promising avenue for reducing (genetic) imbalances in TS and ultimately increasing patient life expectancy, unlike X monosomy and the resultant loss of expression of many genes from the inactivated X chromosome, which is clearly an irreversible process. The epigenetic sink hypothesis proposes that the silenced X chromosome attracts heterochromatizing factors away from other chromosomes and shifts the epigenetic status and gene expression (Arnold et al., 2016; Lemos, Branco, & Hartl, 2010; Wijchers & Festenstein, 2011).

We recently studied 33 women with 45,X and compared with age-matched normal females and males and found a globally changed methylation pattern extending to all chromosomes (Trolle et al., 2016). We found a pattern of preferential hypomethylation across chromosomes, although we also did see areas of hypermethylation in comparison with both females and male controls (Trolle et al., 2016). Although extending to all chromosomes, we did see preferential enrichment of chromosomes 1, 11, 17 and 22, as well as preferential hypomethylation of proximal promoters. These studies were done with the Illumina 450k array (~480,000 CpG sites in the genome) and in essence validating and extending results published with the Illumina 27k array (~27,000 CpG sites in the genome) (Sharma et al., 2015), however both arrays are still far from covering all 28 million CpG sites in the genome.

When studying not only differentially methylated positions, but also differentially methylated regions (areas including several CpG sites involving genes), we found that the TIMP1 (tissue inhibitor of matrix metalloproteinase 1) gene was affected (Trolle et al., 2016). This is interesting because TIMP1 has previously been implicated in aortic aneurysm formation (Rabkin, 2017), and recently we described that haploinsuffciency of the TIMP1 gene together with certain risk alleles of its paralogue TIMP3 situated on chromosome 22 leads to more than 10 times increased risk of bicuspid aortic valve and aortic dilation in TS (Corbitt et al., 2018). Thus, if these findings are validated in functional studies, it suggests that a two-hit event is necessary, or at least to explain the advent of aortic bicuspid valve development in TS—first loss of a X chromosome resulting in only one copy of the TIMP1 gene and then the presence of a risk TIMP3 allele. It also means that the TIMP1/TIMP3 genes could be involved in bicuspid valve development in other conditions than TS (Pedersen et al., 2019), an area of research that has proven difficult to elucidate (Pedersen et al., 2019).

Genes situated in the pseudoautosomal regions on the X (and Y) chromosomes have for long attracted special attention, since they escape X-inactivation and are thought to be under- or over-expressed in sex chromosomal disorders (Gravholt et al., 2018). The SHOX gene is prototypical of this, and is implicated in the clinical phenotype in TS explaining part of the short stature. We found that several pseudoautosomal genes were differentially methylated in TS compared with controls, including IL3RA and CSF2RA. We speculate that IL3RA and other genes related to immune functioning could be linked to the grossly increased risk of autoimmune diseases among females with TS (Berglund et al., 2018; Lleo, Moroni, Caliari, & Invernizzi, 2012) and CSF2RA has previously been implicated in early intrauterine lethality (Urbach & Benvenisty, 2009).

Four escape genes (KDM6A, UBA1, STS, USP9X) were differentially methylated and KDM6A was also differentially expressed, has a known Y homolog, is a histone demethylase, has importance for reestablishment of pluripotency and germ cell development, and is involved in Kabuki syndrome (Miyake et al., 2013), and probably also involved in development of congenital cardiovascular malformations. We also hypothesize that KDM6A could be involved in gonadal dysgenesis, and we recently showed that it is differentially expressed in Klinefelter syndrome as well (Skakkebaek et al., 2018), a condition also accompanied by hypergonadotropic hypogonadism and apoptosis of the gonads. A sex chromosome effect on expression of KDM6A has recently been shown in a study of 45,X, 46,XY, 46,XX, 47,XXX, 47,XXY and 47,XYY (Raznahan et al., 2018). In this latter study, the researchers found an apparent upregulation of some X-linked genes in comparison with normal 46,XX females, which runs counter to previous beliefs of a reduced expression of X-linked genes escaping X-inactivation (Raznahan et al., 2018).

On the X chromosome we found three genes (LANCL3, RPS4X, JPX) normally known to escape X-inactivation (Bellott et al., 2014), were differentially expressed. RPS4X has been implicated in TS before (Fisher et al., 1990; Omoe & Endo, 1996; Rajpathak et al., 2014; Zhang et al., 2013). RPS4X encodes the ribosomal protein S4 and transcription is dosage sensitive, therefore may play a role in TS (Fisher et al., 1990; Just, Geerkens, Held, & Vogel, 1992). RPS4X has a homolog gene on the Y chromosome RPS4Y which encodes for a isoform of the ribosomal protein S4 (Watanabe, Zinn, Page, & Nishimoto, 1993). Bakalov et al. speculated that RPS4X was involved in the pathogenesis of diabetes, since overexpression was seen in 46,X,i(Xq) along with a higher frequency of diabetes (Bakalov et al., 2009). However, these findings need further validation. Nevertheless RPS4X continues to surface in RNA-seq TS studies (Rajpathak et al., 2014; Zhang et al., 2013). We found other X-chromosomal genes, normally X-inactivated, that were differentially expressed, like CD40LG and KDM5C. KDM5C participates in transcriptional repression of neuronal genes, and thus could play a role in the distinct neurocognitive profile of TS, and we recently also found KDM5C to be differentially methylated in Klinefelter syndrome (Skakkebaek et al., 2018), which seems to represent a mirror image of TS on a genomic scale with preferential global hypermethylation.

5 CONCLUSIONS

It is clear that the epigenetic profile of white blood cells among primarily adults with TS is distinctly changed in comparison with appropriate controls. It is also clear that the expression profile of primarily white blood cells is distinct from controls. However, a number of issues remain to be elucidated—for instance, how is the temporal development during the life of a female with TS, what happens in other tissues than blood, and will specific changes in epigenetics or RNA expression explain different phenotypic traits in TS. Such data sets would take the understanding of TS development and associated morbidities to a new level.

DISCLOSURE OF INTERESTS

The authors have nothing to disclose.