4. Discussion

This study conducted a detailed analysis of the clinical phenotypes and genetic characteristics of four families with hereditary hemorrhagic telangiectasia type 1 (HHT1) carrying ENG gene variants.

In Families 1–4, patients presented primarily with PAVM as the initial clinical manifestation, with some patients (II‐1 of Family 2 and II‐3 of Family 4) exhibiting severe clinical manifestations and rapid disease progression. PAVM can be classified into primary and secondary types based on the presence or absence of underlying causes, with primary PAVM accounting for approximately 80% of all PAVM cases [12]. About 80%–90% of PAVM patients are eventually diagnosed with HHT [13], whereas at least 50% of HHT patients present with PAVM [14]. Primary PAVM patients with HHT exhibit more pronounced symptoms, faster disease progression, a higher likelihood of multiple PAVMs, and an increased incidence of complications. Patients with PAVMs exhibit diverse clinical manifestations, primarily including cyanosis, fatigue, dyspnea, dizziness, and decreased oxygen saturation. In severe cases, the condition can be life‐threatening. Thrombosis is prone to form within the vascular malformations of PAVMs, and emboli may enter the systemic circulation, leading to conditions such as cerebral ischemia and cerebral infarction (e.g., II‐1 of Family 2 and II‐3 of Family 4) [15]. Anderson et al. [16] suggested in a retrospective study that PAVM might be the only clinical criterion for genetically confirmed HHT. Clinicians and patients are generally aware that PAVMs often occur in a familial setting, most commonly secondary to HHT. For a patient with PAVM and confirmed HHT, each first‐degree relative carries a 50% risk of HHT, and at least 50% of such affected relatives are predicted to have PAVMs. Genetic testing serves as an effective method for diagnosing HHT, regardless of the presence of clinical manifestations, and has become an essential element in the management of patients presenting with one or more PAVMs [1]. Therefore, when encountering patients with one or more PAVMs, clinicians should enhance clinical suspicion and actively pursue HHT genetic testing.

According to the Curaçao Criteria, patients from four families met three diagnostic criteria for HHT, including recurrent spontaneous nosebleeds, mucocutaneous telangiectasia, and a positive family history [11]. Consequently, genetic testing was performed on all four families, revealing ENG gene variants in each. Based on the clinical phenotypes and genetic testing results, we confirmed that four families were diagnosed with HHT1. HHT is a rare autosomal dominant vascular disorder with an estimated prevalence of 1 in 5000 to 1 in 8000 [17]. HHT patients exhibit significant phenotypic heterogeneity, with different families displaying varying clinical features, and even within the same family, patients carrying the same pathogenic mutation may present with diverse clinical manifestations. Several studies have indicated that patients with HHT1 are at a greater risk of developing PAVMs and cerebral arteriovenous malformations (CAVMs) compared with those with HHT2 [18, 19]. As demonstrated in the four families reported in this study, clinical phenotypes varied among families, and the severity and progression of the same clinical features also differed (Table 3).

The ENG gene is located on human chromosome 9q34.11 and consists of 14 exons [8]. This gene encodes ENG, a homodimeric integral membrane glycoprotein primarily expressed on the surface of vascular endothelial cells. ENG is a crucial component of the TGF‐β receptor complex. By associating with the signaling receptors RI and RII, ENG modulates the activity of the TGF‐β signaling pathway. ENG mutations may prevent endothelial cells from properly responding to TGF‐β signals during angiogenesis, leading to abnormal vascular development, manifested as telangiectasia and arteriovenous malformations [10, 20].

ENG is a Type I transmembrane protein characterized by a large extracellular (EC) domain comprising two functionally distinct regions: an NH2‐terminal orphan domain and a membrane‐proximal zona pellucida (ZP) domain [21]. The orphan domain specifically binds TGF‐β superfamily ligands (including activins, TGF‐β, and BMPs), consistent with ENG′s function [22]. Meanwhile, the ZP domain facilitates receptor oligomerization, mediates the formation of ENG homodimers/heterodimers, and maintains the stability of the TGF‐β receptor complex [23]. The ZP domain also encodes an Arg‐Gly‐Asp (RGD) tripeptide that is a prototypic member of a family of motifs involved in integrin‐based interactions with EC matrix and certain cell surface proteins [24].

In Family 1, the proband carried the c.613del: p.Arg205Glyfs∗17 mutation; in Family 3, the proband carried the c.1498dup: p.Glu500Glyfs∗28 mutation; and in Family 4, the proband carried the c.322del: p.His108Ilefs∗55 mutation. Among pathogenic mutations in the ENG gene, small insertions and deletions account for approximately 50%, representing a common type of pathogenic mutation in this gene. In Family 1, the frameshift mutation (c.613del: p.Arg205Glyfs∗17) disrupts the OR1 domain and causes premature protein termination, significantly impairing TGF‐β signaling. In Family 4, the frameshift mutation (c.322del: p.His108Ilefs∗55) generates a severely truncated protein (163 aa), with premature protein termination (Figure 9). Phenotypic analysis revealed that Family 4 patients exhibited early‐onset severe symptoms with rapid disease progression, whereas Family 1 patients showed relatively milder clinical manifestations (Table 3). In Family 3, the mutation (c.1498dup: p.Glu500Glyfs∗28) primarily affects the ZP domain. Although ligand‐binding capacity of OR1 and OR3 domains is preserved, impaired oligomerization results in reduced signal transduction efficiency (Figure 9). In Family 3, the clinical phenotype is relatively mild due to retained partial protein function (Table 3). These mutations primarily cause disease by leading to insufficient expression of the ENG gene [25, 26]. In this study, the mutations in these three families were all frameshift mutations (c.613del, c.1498dup, and c.322del), resulting in altered protein sequences and premature termination of translation, leading to the likelihood of generating a functional protein is negligible which is the primary mechanism of disease development [26].

The proband of Family 2 carried the c.1428 + 2 T > C variant, located at the 5 ′ splice donor site of intron 11, where the second base (thymine) was replaced by cytosine. The nucleotides directly adjacent to the intron–exon junctions within the splice region are highly conserved. According to the canonical GT‐AG rule, the 5 ′ (donor site) and 3 ′ (acceptor site) bases of introns are almost always GT and AG, respectively. In mammalian genomes, the probability of splice sites conforming to the canonical GT‐AG combination is as high as 98.71% [27]. Therefore, splice acceptor site variants occurring in this conserved region are highly likely to be pathogenic. Minigene assay and Western blot results provided strong evidence for the pathogenicity of c.1428 + 2 T > C, confirming that this novel variant disrupts splice acceptor function, leading to complete skipping of exon 11. Abnormal splicing resulted in the deletion of 39 amino acids (p.Lys438_Gln476del), disrupting the EC domain of the ENG protein, which is critical for ligand binding and receptor complex formation (Figure 9). The loss of glycosylation at Asn444 and Asn465 may affect protein stability and trafficking, as evidenced by the accumulation of nonglycosylated forms in MUT cells. Thus, we conclude that this novel splice‐site variant is definitively pathogenic.

In conclusion, through whole‐exome sequencing, we identified variants in the ENG gene (c.613del, c.1428 + 2 T > C, c.1498dup, and c.322del) in four families. The pathogenicity of the novel heterozygous intronic variant c.1428 + 2 T > C in the ENG gene of Family 2 was verified using an in vitro minigene assay and Western blot analysis. Abnormalities in intronic splicing sites are crucial for predicting splicing variants, which in this case led to aberrant ENG mRNA splicing, resulting in the skipping of exon 11 (c.1312_1428del) and the production of a shorter protein with deletion of the predicted glycosylation motifs (p.Lys438_Gln476del). The mutations in Families 1, 3, and 4 (c.613del, c.1498dup, and c.322del) are all frameshift mutations, causing premature termination of translation. Furthermore, our study expands the spectrum of ENG variants and highlights the significant phenotypic heterogeneity and genetic complexity associated with ENG gene variants in HHT1. These findings not only enhance our understanding of the molecular pathogenesis of HHT but also remind clinicians to consider the possibility of concurrent HHT in patients with PAVMs and emphasize the importance of early genetic testing and personalized therapeutic interventions.