Copy number variations in endoglin locus: mapping of large deletions in Spanish families with hereditary hemorrhagic telangiectasia type 1

The hereditary hemorrhagic telangiectasia syndrome (HHT), also known as Rendu–Osler-Weber syndrome [1‐3] is a vascular disorder inherited as an autosomal dominant trait. Careful epidemiological studies have revealed that HHT affects approximately 1 in 5,000 individuals [4, 5] and is therefore considered to be an inherited rare vascular disease.

The clinical symptoms characteristic of HHT are the so-called Curaçao criteria [2], which help in its diagnosis when at least 3 of the 4 criteria are present in a patient. These include: epistaxis (nose bleeds), telangiectasia at mucocutaneous and gastrointestinal sites, arteriovenous malformations (AVMs) most commonly found in pulmonary, hepatic and cerebral circulations, and dominant familial inheritance [3, 6].

HHT is transmitted as an autosomal dominant condition due to a single mutation in Endoglin (ENG; HHT1) [7], Activin Receptor-Like Kinase 1 (ACVRL1/ALK1; HHT2) [8], or MADH4/SMAD4 (JHPT, a combined syndrome of juvenile polyposis and HHT) [9]. The involvement of all these genes in the transforming growth factor (TGF-β) signaling pathway is inherent in HHT pathogenesis [10]. There are at least two further unidentified genes that can cause HHT: HHT3 between 141.9 and 146.4 Mb on chromosome 5q [3, 11] and HHT4 on chromosome 7p between D7S2252 and D7S510.130 [12].

The genes mutated in HHT encode proteins that mediate signaling by the TGF-β superfamily. Members of this superfamily such as TGF-βs, bone morphogenetic proteins (BMPs), activins, nodals, growth/differentiation factors (GDFs) and inhibins regulate diverse cellular functions by binding to a heteromeric complex of type I and type II transmembrane serine/threonine kinase receptors [13]. In the TGF-β signaling cascade, the type II receptor with very high ligand affinity, co-operatively recruits and transphosphorylates the type I receptor by direct contact with the ligand-modified N-terminus of TβRI [14]. In Smad-dependent TGF-β pathways, the type I receptor subsequently phosphorylates and activates receptor-associated (R)-Smads, according to the receptor complex involved. R-Smads bind to Smad4 and translocate to the nucleus where they influence transcriptional activity with co-activators and co-repressors.

In endothelial cells, upon ligand binding, TβRII can associate with two different TGF-β type I receptors ALK-5 or ALK-1 [15]. Endoglin is an auxiliary receptor that modulates both associations in an opposite manner. Thus, while endoglin promotes signaling through ALK-1, it inhibits the ALK-5 pathway [16, 17]. In turn, ALK-1 and ALK-5 activate distinct Smad pathways, resulting in opposing endothelial cell responses in terms of proliferation, migration, and pro- or anti-angiogenic gene expression [15, 18, 19].

So far, more than 600 different mutations have been found in ENG and ACVRL1 in HHT families (HHT mutation database; http://​www.​arup.​utah.​edu/​database/​hht/​). Mutations range from single base-pair changes to major deletions of multiple exons. Recently, pathogenic mutations affecting the 5’ UTR region of ENG leading to new translation initiation sites (TIS) dominant over the normal endoglin TIS have also been described [20, 21].

In the present study we describe for the first time, a series of 4 families, harboring 4 different independent mutations in the ENG gene, not described so far in literature. All of them were large deletions starting in the 5’ upstream region of ENG, and spanning just the promoter region, one or several exons, or even a big 100-Kb deletion encompassing the whole ENG gene and the two downstream genes FGSH and CDK9. These mutations were first detected by MLPA technique, and subsequently the breakpoints were characterized using a customized copy number variation (CNV) microarray, designed to cover ENG and flanking sequences. Interestingly, common breakpoints located within “Alu” sequences were found among these families. To the best of our knowledge, this is the first time a fine mapping of a series of deletion mutations is described in HHT taking advantage of a customized CNV/CGH microarray.

Most of the mutations involved in the pathogenesis of HHT are single base-pair substitutions or small duplications/deletions. In fact, a review of molecular diagnosis of HHT identified for ENG, 17% nonsense, 30% missense, 25% splice, and 28% frameshift mutations, whereas for ACVRL1 17% were nonsense 60% missense, 7% splice, and 15% frameshift mutations [25].

When no mutations are found in either ENG or ACVRL1 by PCR and sequencing of coding exons and intron boundaries, then the multiplex quantitative PCR (qPCR)-based assay method MLPA is the standard routine technique applied to samples. MLPA allows the identification of dosage for proximal promoter and exons in the HHT-related genes, including deletions or duplications of whole exons. However, this method does not yield a systematic mapping of the breakpoints corresponding to these chromosomal aberrations. So far, mutations affecting the copy number variation of ENG or ACVRL1 causing HHT, as evidenced by MLPA, represent around 7% of all the pathogenic mutations detected in these genes [25, 26].

More recently, CNV custom CGH arrays have emerged as an alternative technique to MLPA to detect large deletions or insertions, with the advantage of allowing the mapping and detailed analysis of breakpoints. Thus, this technique can reveal the presence of “hot spots” or points prone to breaking, which may lead to common mutation events in independent families, and coincident with repetitive sequences. In the present work we have identified and mapped four large deletions affecting ENG that have not been previously reported in publications and which were first detected by MLPA. To the best of our knowledge, this is the first time that the breakpoints of large deletions in ENG have been mapped in independent HHT families.

In all cases the deletions started in the region of the ENG promoter and were the first reported deletions affecting the promoter and implicated in HHT. Interestingly, the affected members of the family NMEx were heterozygous for a 9-Kb deletion of the promoter expanding up to 28 bp upstream of the transcription initiation start site. In this case, the loss of the promoter region leading to a hemizygous allele is the likely cause of HHT. Indeed, the proximal promoter region of ENG plays a critical role in the basal transcription, mainly through binding of the transcription factor Sp1 to consensus GC-rich motifs [24, 27]. Additionally, the promoter activity of ENG can be regulated by several physiological stimuli. In this regard, Sp1 can directly bind, in protein-DNA complexes, to the transcription factors KLF6, Smad3/Smad4 or HIF-1α, activated by vascular injury [28], treatment with TGF-β1 [27] or hypoxia [29], respectively. In turn, the multicomplex formed by all these transcription factors leads to synergistic transcriptional cooperation with the promoter activity of ENG. Therefore, basal and stimuli-dependent transcription of ENG would be abrogated in those HHT patients harboring a promoter deletion, contributing to the haploinsufficiency of ENG.

HHT patients from family GUM are hemizygous not only for ENG, but also for the genes FPGS, CDK9 immediately downstream, and at least part of SH2D3C. However, the additional heterozygous deletion of these genes does not seem to affect the severity or type of clinical symptoms. This would explain the lack of a reported pathology associated with heterozygous mutations of these genes in relevant scientific publications.

The detailed mapping of the deletions has revealed interesting hot spots in ENG introns and their upstream promoter region where common breakpoints have been found. Thus, we have identified two different breakpoints placed in non-coding regions corresponding to allocation sites for Alu mobile elements, and both derived from independent recombination events in different families. One of the breakpoints is within a cluster of Alu sequences, 9-Kb upstream of the ENG transcription start site (Figure 4B), and the other is on an Alu element, placed 900 bp upstream of the ENG transcription start site (Figure 4A). In agreement with this finding, Wooderchak et al. [30] described the breakpoints for two deletions affecting ENG in a single HHT family, one of them encompassing exon 3 and the other involving exons 4 to 7 [30]. Interestingly, both deletions share a common breakpoint location in intron 3. Furthermore, a large 117-bp repetitive DNA sequence was identified near the breakpoints in introns 2, 3, and 7 of ENG. These repetitive sequences had a sequence identity of approximately 85%, had similar orientation, and were each found to contain Alu elements. These results suggest that this type of mobile element is a common target of recombination in HHT genes. Supporting this view, Alu sequences have been involved in the generation of genomic deletions in different human genetic disorders [31, 32]. Further studies have yet to be carried out to better understand the mechanisms of recombination in HHT. In this regard, it would be interesting to screen the published deletions and duplications found in ENG and ACVRL1, taking advantage of the CNV array technique to obtain a map of deletions or duplications from as little as 15-bp to many kilobases.

The systematic hybridization of DNA from HHT families, with deletions or duplications, to custom designed microarrays, could allow the mapping of breakpoints, coincident with repetitive Alu sequences that might act as “hot spots” in the development of chromosomal anomalies. New data, generated this way, would then enrich the HHT mutation databank by including the specific sequences of the chromosomal rearrangements, in addition to the current mutations.