Hereditary hemorrhagic telangiectasia: clinical features in ENG and ALK1 mutation carriers

Patients

A total of 135 consecutive adult patients (age 19–74 years), belonging to 65 unrelated families, were referred to our HHT multidisciplinary center over a period of 3 years (April 2001 to March 2004). The patients were clinically diagnosed as having HHT according to Curaçao criteria [18] and provided informed consent for mutational analysis. The instrumental screening was prospectively offered to all patients regardless of the presence of clinical symptoms. The results of the genetic testing were obtained later, as genetic testing requires a lengthy period of time. Once the mutation was identified, the patients with identified mutations were enrolled in the study. The genetic data and the clinical data were recorded independently in two different and separate databases which were set up by both the geneticists and clinicians, respectively. The design of the study was organized to assure that both genetic and clinical analyses were performed in a double-blind condition. After all data were collected in the aforementioned manner, the two databases were cross-matched: thus, the 135 adult patients were divided into three groups according to the genetic findings: HHT1 (ENG-mutated patients), HHT2 (ALK-1 mutated patients) and HHT? (this small group of 13 patients, in whom no mutation was found in either gene, was eliminated from the study).

The 122 adult patients for whom both the gene mutation and clinical data were available were enrolled in the study (53 females and 69 males, overall mean age 46.8 ± 14.7; range 19–74 years). A statistically-significant age difference was noted for the HHT1 and HHT2 groups (P < 0.05), whereas no significant difference was found between the proportion of males and females in both groups (Table 1). All patients provided informed written consent before participating in this study after detailed information was provided. The local ethical committee approved the project.

Molecular analysis

All individuals were subjected to genetic counseling and molecular testing. Mutational analysis was performed by single-strand conformation polymorphism, by denaturing high-performance liquid chromatography, and subsequently by the sequencing of samples showing abnormal electrophoretic or chromatographic patterns, as previously described [17].

Screening for visceral manifestations

Screening for visceral involvement was performed in all consecutive mutation carriers. The clinical diagnostic features, in accordance with the aforementioned criteria, were identified by means of a medical history, physical examination and instrumental diagnostic procedures of the potentially affected organs. Each examination was performed by a single investigator in the various specialties to eliminate inter-observer variability. The intra-observer variability was reduced as a result of the expertise of each operator within the different specialties in identifying the HHT symptoms and signs.

Statistical analysis

All statistical evaluations were performed using the ssps program (version 14.0; SPSS Inc., Chicago, IL, USA) and P-values < 0.05 were considered statistically significant. Comparisons between HHT genotypes/phenotypes were performed using a permutational Fisher’s exact test. Comparisons between ages were performed using the Mann–Whitney U-test.

Epistaxis and mucocutaneous telangiectases

Epistaxis was present in 41/45 (91%) HHT1 and 73/77 (95%) HHT2 patients; the difference between the two groups was not statistically significant (P = 0.4). On the contrary, there was a significant difference in age of onset of epistaxis (P < 0.0005) as this symptom appeared in 57% (26/45) of HHT1 patients within the first decade of life compared with 24% (19/77) of HHT2 patients; however, the number of patients with epistaxis within the second decade (age 11–20 years) did not differ statistically (37/45 HHT1 vs. 55/77 HHT2 patients, P = 0.19).

Telangiectases were present in 42/45 (93%) HHT1 and in 68/75 (91%) HHT2 patients, (P = 0.7, NS).

Visceral AVMs

The proportion of HHT1 and HHT2 patients with visceral arteriovenous malformations are reported in Fig. 1 and Table 1. Symptoms as a result of AVMs are reported in Table 2.

Pulmonary findings

Previous studies on an extensive patient sample for genotype–phenotype correlation by a single center have evidenced a 49–67% prevalence of PAVMs in HHT1 and 5–29% in HHT2 [12,14]. In our study, pulmonary AVMs were significantly more frequent in HHT1 than in HHT2 patients (75% vs. 44%) confirming previous data; however, the absolute incidence of PAVMs in the two groups was markedly higher than that reported thus far. The low rate previously found for PAVMs (49% HHT1 vs. 5% HHT2) can be attributed to the less sensitive screening techniques, as acknowledged by the authors [12]. In a more recent publication [14], PAVMs were observed in 67% HHT1 vs. 40% in HHT2 but contrast echocardiography and CT scanning was used for only a fraction of the patients, and less sensitive techniques in others.

With multislice CT scan, we observed that, among PAVM-positive patients, the percentage of those with large PAVMs was significantly higher in HHT1 (61%) than in HHT2 (17%). In addition, for the first time we now demonstrate that in both HHT1 and HHT2, the prevalence of patients with PAVMs does not augment with increasing age. This observation strongly suggests that PAVMs are usually, if not always, lesions which occur prenatally or in the first two decades of life in both HHT1 and HHT2. However, our study does not affirm that PAVMs do not arise later in life, but it suggests that, if they do, they develop only in those patients who already have PAVMs. Obviously, this finding requires extensive prospective monitoring with highly sensitive screening techniques.

Given the high risk of a paradoxical embolism and neurological complications of PAVMs which can be prevented by embolotherapy, screening for these lesions is recommended, even when asymptomatic [22]. The fact that small PAVMs are more numerous in HHT2 patients and that, in some instances, they can also increase in size thereby reaching the embolization threshold, indicates that both HHT1 and HHT2 patients should be targeted for periodical pulmonary survey for PAVMs.

Of the 21 HHT1 patients with large PAVMs, 12 had a history of brain complications as a result of their large PAVMs. None of the HHT2 patients with large PAVMs had experienced neurological sequelae, but neurological complications secondary to PAVMs have been reported in HHT2 patients in literature, though to a lesser extent than that observed in HHT1 [19,24]. As neurological complications mainly occur in PAVMs with a feeding artery > 3 mm [22,25], few HHT2 patients are expected to suffer from brain involvement secondary to PAVMs. It is noteworthy that 6/12 HHT1 patients with PAVM-related neurological complications were younger than 45 years; moreover, 11 HHT1 and only two HHT2 patients with large PAVMs were observed at an age < 45 years, suggesting that not only large PAVMs are more frequent in HHT1, but they are also rare in young HHT2 patients. In addition, 7/21 HHT1 and only 1/6 HHT2 large PAVM patients had respiratory symptoms. All these findings demonstrate that HHT1 implies a more frequent brain and lung involvement than HHT2, regarding not only the incidence of PAVMs, but also the age of incidence and/or onset of respiratory and cerebral symptoms. Consequently, HHT1 patients seem to refer to our multidisciplinary HHT center at a younger age than those with HHT2. Furthermore, HHT1 patients also seem to be more convinced of the need for cerebral screening than HHT2 (see below), most likely as a result of the appearance of more severe cerebral events (secondary to either large PAVMs or CAVMs) in either themselves or family members. Therefore, the combined clinical and molecular screening performed show that useful criteria for prediction of a disease-causing mutation in the ALK1 or ENG gene could be based on the prevalence of ALK1 and ENG mutations in the specific population to which the patient belongs and the presence and size of PAVMs in the individual patient. As the prevalence of large PAVMs is uncommon in HHT2 patients < 45 years of age, whenever a large PAVM in a < 45-year-old patient is documented, the ENG gene should be first targeted for mutational screening.

Cerebral findings

Previous studies on a genotype–phenotype correlation performed in a single center evidenced a 15–16% prevalence of CAVMs in HHT1 and 1–2% in HHT2 [12,14]. In our sample, CAVMs were only found in HHT1 patients; however, a possible recruitment bias might have occurred as a significantly higher number of HHT2 patients refused brain screening. Thus, the likelihood of finding a CAVM-positive patient in the HHT2 cohort was further reduced, as the prevalence of CAVMs in HHT2 is already low. Nevertheless, it is important to emphasize that genetic and clinical analyses were performed in a double-blind condition (see Patients and methods section). Our explanation for the difference in the two patient groups who refused cerebral screening is that there was a greater incidence of neurological symptoms in the HHT1 group. In fact, 15 HHT1 patients in our cohort experienced various HHT-related neurologic symptoms whereas these symptoms were absent in the HHT2 cohort. These neurological symptoms can be considered to be secondary to either CAVMs or large PAVMs (see above and Table 2). This was a powerful motive which convinced the HHT1 patients as well as their relatives included in our study of the usefulness of cerebral screening.

Although there is still no consensus on the best therapy for CAVMs (cerebral embolization, surgical resection, etc.), CAVMs are considered an important cause of morbidity [26,27]. Based on our findings, we suggest that routine screening for CAVMs should be performed for the HHT1 cohort. However, as previous papers do report CAVMs in HHT2 [12–16,24], although at a very low rate, further studies on larger samples are required to definitely assess whether screening for CAVMs should be recommended for HHT2 as well as HHT1 patients.

Liver involvement

Previous studies investigating a genotype–phenotype correlation have evidenced a 2–8% prevalence of HAVMs in HHT1 and 28–41% in HHT2 [12,14]. We found a significantly higher prevalence of HAVMs in HHT2 with respect to HHT1 patients (84% vs. 60%). HAVMs were systematically investigated in all patients, even those asymptomatic, with highly sensitive imaging methods (MDCT scan). As already stated in our previous report [20], a MDCT scanner can increase the diagnostic accuracy of these imaging methods [28] and allow the detection of small HAVMs, often overlooked by other conventional techniques.

Our results also demonstrate that hepatic involvement is more frequent than previously reported. Data published thus far, referring to liver screening with the use of less sensitive imaging methods only for symptomatic patients with abdominal bruit or those who had been targeted for pulmonary embolization, resulted in an underestimation of the real incidence, as acknowledged by the same authors [12,14]. In fact, for many years there was no consensus concerning the need for highly-sensitive routine screening for hepatic AVMs, as effective treatment for HAVMs was limited, and orthotopic liver transplantation is the only definitive therapeutic option for the most severely affected patients [29]. In addition, according to literature, these patients become symptomatic in a small number of cases [20,30].

In our study, despite the presence of portal hypertension detected by a CT scan in a few HHT1 patients, cases with liver involvement and evident clinical manifestations were found only in the HHT2 subgroup (see Table 2). This raises the question whether HHT1 patients should be subjected to liver screening. Further studies are needed to verify whether liver AVMs can become symptomatic in HHT1 patients, so that HAVM screening would be justified.

The controversial issue of liver screening in HHT has been discussed in a recent paper [30] reporting general consensus guidelines, in which extensive highly sensitive liver screening, independent of the presence of symptoms and/or clinical signs, is indicated for all HHT patients as a research tool to clarify different aspects of liver involvement and/or in patients with uncertain HHT diagnosis and without easily available genetic analysis. Our results demonstrate that, if clinical diagnosis is required, CT-based liver screening is an appropriate diagnostic tool given its ability to detect the high incidence of HAVMs in both HHT1 and HHT2.

This is the first report which demonstrates an increase of HAVMs with age in both HHT1 and HHT2 patients. As HAVMs are much more frequent in HHT2, these observations indicate that the HHT2 genotype is more vulnerable than HHT1 to those malformations which tend to occur later in life, such as HAVMs. Our results support the hypothesis regarding the existence of separate pathways in the pathogenesis of HHT lesions, with HHT2 mainly involving vessels in tissues in which active vascular remodeling occurs significantly later in life (liver), and HHT1 involving PAVMs or CAVMs which are more likely to be early developing or congenital [14]. However, as HAVMs also occur in a significant proportion of HHT1 patients at a later age, we propose that defective vascular remodeling with increasing age can also contribute to the onset of HAVMs in HHT1 patients, but to a lesser extent than in HHT2.

Different mechanisms might be evoked to explain the tendency of this organ towards AVM formation in HHT2: the liver may produce specific angiogenic factors, mainly ALK1-specific, acting in a paracrine manner on its vasculature. The importance of paracrine signals in TGF-β-mediated pathways in vascular development has recently been demonstrated in ENG, ALK5, and TGFβRII null mice [31]. In agreement, a recent report demonstrated that BMP-9, a member of TGF-β/BMP superfamily, which is predominantly expressed in the liver, is capable of binding to and activating ALK1 in a specific manner [32].

GI involvement

Screening has led to the detection of GI telangiectases in both HHT1 and HHT2 (60% vs. 51%) in agreement with a previous report (72% vs. 65%) [12], even if our prevalence is lower as GI screening was consecutively performed in all patients including those without GI bleeding. There was no significant difference in the prevalence of GI telangiectases between the two groups; however, patients with a history of GI bleeding pertained only to the HHT2 group while in the literature, GI hemorrhaging has been reported in both HHT1 and HHT2 patients. Given that the number of patients with GI bleeding was scarce in our study, we cannot formulate any hypothesis to explain the lack of hemorrhaging in the HHT1 cohort.

Correlation of phenotype with mutation type

The accepted model for HHT1 is haploinsufficiency. Based on molecular studies, several reports have demonstrated that both truncating and non-truncating endoglin mutations behave as null alleles. This implies that all mutations give rise to the same degree of clinical severity, as the variable expressivity is a result of other genetic or environmental modifiers [2]. In fact, no mutation was observed to be linked to a particularly severe HHT1 phenotype, but a milder phenotype has been reported in HHT1 carriers of missense mutations [14]. Haploinsufficiency has also been proposed for HHT2. Truncating mutations of ALK1 have been shown to behave as null alleles, in that they trigger nonsense-mediated-decay; many missense mutations have also been found to be non-functional, because they give rise to unstable misfolded proteins [2]. In contrast, some mutations are expressed as stable proteins and a dominant-negative action has been evidenced in vitro [33]. Although dominant-negative alleles would be expected to determine a more severe phenotype, this is not the case according to the observations by Fernandez-Lopez et al. [34]. To date, no correlation has been established between a series of patients with different ALK1-mutational types. In our HHT2 subset, the patients were classified into three different classes: truncating, missense and c.626-3C > G mutation carriers. While no significant difference was found between truncating and missense mutations, the carriers of c.626-3C > G mutations show a significantly higher prevalence of PAVMs. Nine out of 10 carriers of this mutation show PAVMs, and all had HAVMs. However, to assess definitely whether the c.626-3 G > C mutation causes a more severe phenotype than other ALK-1 mutations, further studies are required to evaluate this mutation at the protein level.