Molecular Bases and Phenotypic Determinants of Aromatase Excess Syndrome

Aromatase excess syndrome (AEXS) is a rare autosomal dominant disorder characterized by gynecomastia. This condition is caused by overexpression of CYP19A1 encoding aromatase, and three types of cryptic genomic rearrangement around CYP19A1, that is, duplications, deletions, and inversions, have been identified in AEXS. Duplications appear to have caused CYP19A1 overexpression because of an increased number of physiological promoters, whereas deletions and inversions would have induced wide CYP19A1 expression due to the formation of chimeric genes consisting of a noncoding exon(s) of a neighboring gene and CYP19A1 coding exons. Genotype-phenotype analysis implies that phenotypic severity of AEXS is primarily determined by the expression pattern of CYP19A1 and the chimeric genes and by the structural property of the fused exons with a promoter function (i.e., the presence or the absence of a natural translation start codon). These results provide novel information about molecular mechanisms of human genetic disorders and biological function of estrogens.


Introduction
Aromatase encoded by CYP19A1 is a cytochrome P450 enzyme that plays a key role in estrogen biosynthesis [1]. It catalyzes the conversion of Δ 4 -androstendione into estrone (E 1 ) and that of testosterone (T) into estradiol (E 2 ) in the placenta and ovary as well as in other tissues such as the fat, skin, bone, and brain [1].
Recently, three types of cryptic genomic rearrangements around CYP19A1 have been identified in 23 male patients with AEXS [2][3][4]. The results provide useful implications not only for the clarification of underlying mechanisms but also for the identification of phenotypic determinants. Here, we review the current knowledge about AEXS.

The Aromatase Gene (CYP19A1 )
CYP19A1 encoding aromatase is located on 15q21.2 adjacent to DMXL2 and GLDN ( Figure 1) [3,9]. It spans ∼123 kb and consists of at least 11 noncoding exons 1 and nine coding exons 2-10 [9][10][11][12]. Each exon 1 is accompanied by a tissuespecific promoter and is spliced alternatively onto a common splice acceptor site at exon 2, although some transcripts are known to contain two of the exons 1 probably due to a splice error [9][10][11]. Transcription of CYP19A1 appears to be tightly regulated by alternative usage of the multiple 2 International Journal of Endocrinology Exons 2-10 (coding region) Figure 1: Simplified schematic representation indicating the genomic structure of CYP19A1. CYP19A1 is located on 15q21.2 adjacent to DMXL2 and GLDN and consists of at least 11 noncoding exons 1 and nine coding exons 2-10 [9,10]. Each exon 1 is accompanied by a tissue-specific promoter and is spliced alternatively onto a common splice acceptor site at exon 2 [9][10][11][12][13].
promoters [9][10][11][12][13]. Actually, CYP19A1 is strongly expressed in the placenta and moderately expressed in the ovary, whereas it is only weakly expressed in a rather limited number of tissues including skin, fat, and hypothalamus [4,13]. Of the 11 noncoding exons 1, exon I.4 seems to play a critical role in the regulation of estrogen biosynthesis in males, because this exon contains the major promoter for extragonadal tissues [9,10].

Molecular Bases of AEXS
A family with dominantly transmitted gynecomastia of prepubertal onset was first described in 1962 by Wallach and Garcia [14]. After this initial report, several cases have been described [5][6][7][8]15]. Laboratory examinations of the affected males revealed markedly elevated serum estrogen values and estrogen/androgen ratios and significantly increased aromatase activity in fibroblasts and lymphocytes [5][6][7][8]15]. Linkage analyses in two families indicated a close association between CYP19A1-flanking polymorphic markers and the disease phenotype [5,6]. Thus, the condition was assumed to be caused by gain-of-function mutations of CYP19A1, and, therefore, the name of AEXS was coined for this condition [7,8]. However, since direct sequencing and Southern blotting analysis failed to detect mutations or copy number abnormalities in the coding region of CYP19A1 [5,6], the molecular basis of this entity remained elusive until recently. In 2003, Shozu et al. reported a father-son pair and a sporadic case with AEXS in whom they identified heterozygous chromosomal inversions of the chromosome 15 [2]. Subsequently, Demura et al. performed detailed molecular studies for these cases and additional two cases and characterized four types of inversions affecting the 5 region of CYP19A1 [3]. Each inversion has resulted in the formation of a chimeric gene consisting of CYP19A1 coding exons and exon 1 of the widely expressed neighboring genes, that is, CGNL1, TMOD3, MAPK6, and TLN2. These data imply that overexpression of CYP19A1 in the inversionpositive cases are caused by cryptic usage of constitutively active promoters. Consistent with this, in silico analysis revealed the presence of promoter-compatible sequences around exon 1 of CGN1, TMOD3, and MAPK6 in multiple cell types, although such sequences remain to be identified for noncoding exons of TLN2 [4].
We recently studied 18 males from six families with AEXS (families A-F) and identified three types of heterozygous cryptic genomic rearrangements in the upstream region of the CYP19A1 coding exons ( Figure 2) [4]. In families A and B, we identified the same 79,156 bp tandem duplication encompassing seven of the 11 noncoding exons 1 of CYP19A1. Notably, this duplication includes exon I.4 that functions as a major promoter for extragonadal tissues such as fat and skin; therefore, CYP19A1 overexpression in these families would be explained by increasing the number of this promoter. Indeed, RT-PCR analysis detected a splice variant consisting of exon I.4 at the 5 side and exon I.8 at the 3 side in lymphoblastoid cell lines and skin fibroblasts of the patients, indicating that the duplicated exon I.4 at the distal nonphysiological position actually functions as transcription start sites. In family C, we identified a 211,631 bp deletion affecting exons 2-43 of DMXL2 and exons 5-10 of GLDN. This deletion appears to have caused CYP19A1 overexpression because of cryptic usage of DMXL2 exon 1 as an extra transcription start site for CYP19A1. Indeed, RT-PCR revealed the presence of chimeric mRNA clones consisting of DMXL2 exon 1 and CYP19A1 exon 2, supporting the notion that aberrant splicing has occurred between these two exons. Such DMXL2/CYP19A1 chimeric mRNA accounted for 2-5% of CYP19A1-containing transcripts from skin fibroblasts. In families D-F, we identified  Collectively, three types of genomic rearrangements on 15q21 have been identified in AEXS to date, namely, inversion type (four subtypes), duplication type, and deletion type (two subtypes) (Figure 3(a)) [2][3][4]. In this regard, sequence analyses for the breakpoints have indicated that (1) inversion types are formed by a repeat sequence-mediated nonallelic intrachromosomal or interchromosomal recombination or by a replication-based mechanism of fork stalling and template switching (FoSTeS) that occurs in the absence of repeat sequences and is often associated with microhomology [16], (2) duplication type is generated by FoSTeS, and (3) deletions are produced by nonhomologous end joining that takes place between nonhomologous sequences and is frequently accompanied by an insertion of a short segment at the fusion point or by a nonallelic recombination [16].  The DMXL2-CYP19A1 chimeric mRNA has two translation initiation codons and therefore is destined to produce not only CYP19A1 protein but also a 47 amino acid protein which is predicted to undergo nonsense-mediated mRNA decay (NMD). The deletion and the inversion types are associated with heterozygous impairment of neighboring genes (deletion or disconnection between noncoding exon(s) and the following coding exons). The inversion subtype 1 is accompanied by inversion of eight of the 11 CYP19A1 exons 1, and the inversion subtype 2 is associated with inversion of the placenta-specific CYP19A1 exon I.1. (b) Expression patterns of CYP19A1 and the five neighboring genes involved in the chimeric gene formation [4]. Relative mRNA levels against TBP in normal human tissues are shown.
Thus, it appears that genomic sequence around CYP19A1 harbors particular motifs that are vulnerable to replicationand recombination-mediated errors. The results provide novel mechanisms of gain-of-function mutations leading to human diseases.
i These four patients allegedly had gynecomastia that required mastectomy (age unknown).
j The sister has macromastia, large uterus, and irregular menses; the parental phenotype has not been described.
The conversion factor to the SI unit: gynecomastia, small testes with fairly preserved masculinization, obvious or relative tall stature in childhood and grossly normal or apparent short stature in adulthood, and age-appropriate or variably advanced bone ages. Blood endocrine studies revealed markedly elevated E 1 values and E 2 /T ratios in all cases examined and normal or variably elevated E 2 values. In addition, Δ 4 -androstenedione, T, and dihydrotestosterone values were low or normal, and human chorionic gonadotropin (hCG) test indicated normal T responses. Notably, LH values were grossly normal at the baseline and variably responded to GnRH stimulation, whereas FSH values were low at the baseline and poorly responded to GnRH stimulation even after preceding GnRH priming, in all cases examined.
The severity of such clinical phenotypes is primarily dependent on the underlying mechanisms (Table 1). They are obviously mild in the duplication type, moderate in the deletion type, and severe in the inversion type, except for serum FSH values that remain suppressed irrespective of the underlying mechanisms. Likewise, gynecomastia has been reported to be ameliorated with 1 mg/day of aromatase inhibitor (anastrozole) in the duplication and the deletion types and with 2-4 mg/day of anastrozole in the inversion type [4].

Expression Pattern of CYP19A1 and the Chimeric Genes as One Phenotypic Determinant
Phenotypic severity is much milder in the duplication type than in the deletion and the inversion types. This would be explained by the tissue expression pattern of CYP19A1 and the chimeric genes. Indeed, RT-PCR analysis using normal human tissue samples revealed that CYP19A1 is expressed only in a limited number of tissues such as placenta, ovary, skin, and fat, while the five genes involved in the formation of chimeric genes are widely expressed with some degree of variation (Figure 3(b)). Therefore, it is likely that the duplication types would simply increase CYP19A1 transcription in native CYP19A1-expressing tissues, whereas the deletion and the inversion types lead to CYP19A1 overexpression in a range of tissues, because expression patterns of chimeric genes are predicted to follow those of the original genes. Furthermore, it is also likely that the native CYP19A1 promoter is subject to negative feedback by elevated estrogens [17], whereas such negative feedback effect by estrogen is weak or even absent for the chimeric genes in the deletion and the inversion types.

Structural Property of the Fused Exons as Another Phenotypic Determinant
Phenotypic severity is also milder in the deletion type than in the inversion types, despite a similar wide expression pattern of genes involved in the chimeric gene formation (Table 1, Figure 3(b)). In this context, it is noteworthy that a translation start codon and a following coding region are present on exon 1 of DMXL2 of the deletion type but not on exons 1 of the chimeric genes of the inversion types (Figure 3(a)). Thus, it is likely that DMXL2/CYP19A1 chimeric mRNAs transcribed by the DMXL2 promoter preferentially recognize the natural start codon on DMXL2 exon 1 and undergo nonsense-mediated mRNA decay and that rather exceptional chimeric mRNAs, which recognize the start codon on CYP19A1 exon 2, are transcribed into CYP19A1 protein. By contrast, such a phenomenon would not be postulated for the inversion-mediated chimeric mRNAs. Consistent with this, it has been shown that the DMXL2/CYP19A1 chimeric mRNA is present only in 2-5% of CYP19A1-containing transcripts from skin fibroblasts, whereas the CGNL1/CYP19A1 chimeric mRNA and the TMOD3/CYP19A1 chimeric mRNA account for 89-100% and 80% of transcripts from skin fibroblasts, respectively [2,4]. In addition, the genomic structure caused by the rearrangements would affect efficiency of splicing between noncoding exon(s) of neighboring genes and CYP19A1 exon 2. For example, in the inversion subtype 1, the physical distance between CGNL1 exon 1 and CYP19A1 exon 2 is short, and, while a splice competition may be possible between exon 1 of neighboring genes and original CYP19A1 exons 1, eight of 11 CYP19A1 exons 1 including exon I.4 have been disconnected from CYP19A1 coding exons by inversion (Figure 3(a)). This may also enhance the splicing efficiency between CGNL1 exon 1 and CYP19A1 exon 2 and thereby lead to relatively severe overexpression of the CGNL1-CYP19A1 chimeric gene, although this hypothesis would not be applicable for other chimeric genes.

Implication for the Hypothalamus-Pituitary-Gonadal Axis Function
It is notable that a similar degree of FSH-dominant hypogonadotropic hypogonadism is observed in the three types, although E 1 and E 2 values and E 2 /T ratios are much higher in the inversion type than in the duplication and deletion types (Table 1). In particular, FSH was severely suppressed even after GnRH priming in the duplication type [4]. This implies that a relatively mild excess of circulatory estrogens can exert a strong negative feedback effect on FSH secretion primarily at the pituitary. This would be consistent with the results of animal studies that show strong inhibitory effect of E 2 on transcription of FSH beta-subunit gene in the pituitary cells and almost negligible effect on synthesis of LH beta-subunit and secretion of LH [18,19]. In this regard, while T responses to hCG stimulation are normal in the duplication and the deletion types and somewhat low in the inversion type, this would be consistent with fairly preserved LH secretion in the three types and markedly increased estrogen values in the inversion type. In addition, whereas fertility and spermatogenesis are normally preserved in the three types, this would be explained by the FSH-dominant hypogonadotropic hypogonadism, because FSH plays only a minor role in male fertility (spermatogenesis) [20].

Conclusions
Current studies argue that AEXS is caused by overexpression of CYP19A1 due to three different types of cryptic genomic rearrangements including duplications, deletions, and inversions. It seems that transcriptional activity and structural property of the fused promoter constitutes the underlying factor for the clinical variability in most features of AEXS except for FSH-dominant hypogonadotropic hypogonadism. Thus, AEXS represents a novel model for gain-of-function mutation leading to human genetic disorders.