Association of common candidate variants with vascular malformations and intracranial hemorrhage in hereditary hemorrhagic telangiectasia

Abstract Background Hereditary hemorrhagic telangiectasia (HHT) is caused by mutations in TGFβ/BMP9 pathway genes and characterized by vascular malformations (VM) including arteriovenous malformations (AVM) in lung, liver, and brain, which lead to severe complications including intracranial hemorrhage (ICH) from brain VM. The clinical heterogeneity of HHT suggests a role for genetic modifier effects. Common variants in loci that modify phenotype severity in Tgfb knockout mice were previously reported as associated with lung AVM in HHT. Common variants in candidate genes were reported as associated with sporadic brain AVM and/or ICH. We investigated whether these variants are associated with HHT organ VM or with ICH from brain VM in 752 Caucasian HHT patients enrolled by the Brian Vascular Malformation Consortium. Methods We genotyped 11 candidate variants: four variants reported as associated with lung AVM in HHT (PTPN14 rs2936018, USH2A rs700024, ADAM17 rs12474540, rs10495565), and seven variants reported as associated with sporadic BAVM or ICH (APOE ε2, ANGPTL4 rs11672433, EPHB4 rs314308, IL6 rs1800795, IL1B rs1143627, ITGB8 rs10486391, TNFA rs361525). Association of genotype with any VM, lung AVM, liver VM, brain VM or brain VM ICH was evaluated by multivariate logistic regression adjusted for age, gender, and family clustering. Results None of the 11 variants was significantly associated with any phenotype. There was a trend toward association of USH2A rs700024 with ICH (OR = 2.77, 95% CI = 1.13–6.80, p = .026). Conclusion We did not replicate previously reported associations with HHT lung AVM and variants in Tgfb modifier loci. We also did not find significant associations between variants reported in sporadic brain AVM and VM or ICH in HHT.


| INTRODUCTION
Hereditary hemorrhagic telangiectasia (HHT) is caused by mutations in TGFb/BMP9 pathway genes, most commonly ENG (OMIM 131195) or ACVRL1 (ALK1, OMIM 601284). HHT is characterized by vascular malformations (VM) ranging from small skin and mucosal telangiectases to arteriovenous malformations (AVM) in lung, liver, and brain, which can lead to severe complications including intracranial hemorrhage (ICH) from ruptured brain VM. The clinical heterogeneity of HHT suggests a potential role for genetic modifier effects. Identification and validation of such modifiers could inform disease prognosis and guide clinical management of HHT patients and provide potentially useful stratifier biomarkers for clinical trials. A few such genetic associations with HHT disease features have been reported. Common variants in two loci that modify phenotype severity in Tgfb knockout mice (TGFb modifier loci) have been associated with lung AVM in HHT families (Benzinou et al., 2012;Kawasaki et al., 2014). We previously showed that the ACVRL1 c.314-35A>G polymorphism, associated with sporadic brain AVM (Pawlikowska et al., 2005;Simon et al., 2006), is also associated with VM in HHT, but only among patients with ENG mutations (Pawlikowska et al., 2015). Common variants in other candidate genes have previously been reported to be associated with sporadic BAVM Mikhak et al., 2011;Pawlikowska et al., 2005;Su et al., 2010) or with sporadic BAVM ICH (Achrol et al., , 2007Hysi et al., 2007;Pawlikowska et al., 2004Pawlikowska et al., , 2006Weinsheimer et al., 2009). We investigated whether the four TGFb modifier locus variants previously reported in HHT lung AVM and seven variants from sporadic AVM studies are associated with organ VM and with brain VM ICH in a large cohort of Caucasian HHT patients enrolled by the Brain Vascular Malformation Consortium (BVMC) (Akers et al., 2013).

| Ethical compliance
The study protocol was approved by the institutional review board at each recruiting center. All patients provided written informed consent for genetic studies and a blood or saliva sample for DNA extraction.

| Cohort
The study includes 752 Caucasian HHT patients enrolled by the BVMC at multiple recruiting centers in the US, Canada, and the Netherlands between 2010 and 2015.
Cohort recruitment has been previously described (Akers et al., 2013;Pawlikowska et al., 2015). Patients were screened for organ VM and other clinical features according to standard clinical practice and International HHT Guidelines (Faughnan et al., 2011), including: comprehensive history, physical, routine blood tests, and clinical screening for recurrent spontaneous epistaxis and HHTrelated gastrointestinal bleeding. All patients were screened for pulmonary AVM by contrast echocardiography or by chest CT. All patients were screened for brain VM by magnetic resonance imaging. If screening was positive for pulmonary AVM or brain VM (Krings et al., 2015), patients underwent further diagnostic imaging and treatment, where appropriate. For liver VM, all patients underwent clinical screening (chronic right upper quadrant pain, portal hypertension, high-output heart failure, liver bruit on examination, abnormal liver function tests) as well as echocardiography for high-output cardiopathy. If screening was positive, then diagnostic liver imaging was performed (ultrasound Doppler, contrast CT or contrast MRI) and therapy recommended where appropriate. The BVMC HHT cohort targets 25% brain VM-positive patients; other characteristics are similar to other cohorts (Letteboer et al., 2006;Nishida et al., 2012). Brain VM ICH included any ICH during HHT disease course (prior to enrollment or during follow-up, including after brain VM treatment). For lung AVM, we also defined a severe subphenotype as: diffuse lung AVM and/or feeding artery diameter of largest lung AVM ≥5 mm and/or presence of any of the following complications: ischemic stroke, brain abscess, massive hemoptysis, spontaneous hemothorax.

| Statistical analysis
Genotypes were collapsed for analysis into risk genotype carriers versus noncarriers according to published associations with sporadic BAVM (Achrol et al., , 2007Hysi et al., 2007;Kim et al., 2009;Mikhak et al., 2011;Pawlikowska et al., 2004Pawlikowska et al., , 2005Pawlikowska et al., , 2006Su et al., 2010;Weinsheimer et al., 2009). For the four TGFb modifier variants, the published analysis (Benzinou et al., 2012;Kawasaki et al., 2014) was for familial transmission, which we cannot perform in our nonfamilial cohort, so we used an additive model for the minor (risk) allele, as this is the most general genetic model for a case-control analysis with the fewest additional assumptions. Association of genotype with presence of Any VM (pulmonary, liver or brain), pulmonary AVM, liver VM, brain VM, and ICH from brain VM was evaluated by multivariable logistic regression adjusted for age at last follow-up, gender and family clustering. The same direction of effect (risk genotype) as previously published was required. The statistical significance threshold was adjusted for 11 variants tested and set at p = .0045. We also performed secondary analyses stratified by HHT mutation (ENG or ACVRL1) for all phenotypes except ICH (where the N of BAVM patients with known HHT mutations was too small to accommodate multivariate analyses for nine variants), and of the severe lung AVM phenotype (to evaluate the effect of a more stringent phenotype definition). One variant-phenotype analysis could not be evaluated with multivariable logistic regression due to perfect prediction; we instead provide results from univariable exact logistic regression. Table 1 shows demographic and clinical characteristics of the study cohort. Among 752 Caucasian HHT patients, 52% had pulmonary AVM, 20% had liver VM and 21% had brain VM. 20% (30/153) of brain VM cases had ICH (prior to enrollment or during follow-up).

| RESULTS
None of the 11 variants was significantly associated with VM in any organ or with ICH (Table 2). We did not replicate the previously reported associations of four TGFB modifier variants (Benzinou et al., 2012;Kawasaki et al., 2014) with HHT lung AVM. There was a trend toward association of one of these variants, USH2A rs700024, with ICH from brain VM (OR = 2.77, 95% CI = 1.13-6.80, p = .026).

| DISCUSSION
We did not replicate the previously reported associations between four TGFB modifier loci variants with HHT lung AVM (Benzinou et al., 2012;Kawasaki et al., 2014)  overall BVMC cohort. In analysis stratified by HHT gene, among ENG mutation carriers only, there was a significant association of one of the four variants, ADAM17 rs12474540, and a trend toward association of two other TGFB modifier SNPs with liver (but not lung) VM. There was a trend in the same direction in the overall cohort. The lung AVM associations originally reported were also stronger among ENG mutation carriers (Kawasaki et al., 2014). There was also a trend toward association of another of these variants, USH2A rs700024, with ICH from brain VM. These findings may indicate that the TGFB modifier loci variants are associated with multiple HHT phenotypes, but a larger cohort will be required to confirm this. It is not clear why we do not detect a lung AVM association in our cohort, and do not replicate the original finding (Kawasaki et al., 2014). To check for a possible effect of different lung AVM ascertainment on the genetic association results, we defined a severe lung AVM phenotype closer to the lung AVM phenotype definition used in the cohort studied in the original report (Letteboer et al., 2015;Pawlikowska et al., 2015), but did not detect statistically significant associations with any of the 4 TGFB modifier variants. The disparity in genetic association results may also be influenced by methodological differences: the published analysis of lung AVM was performed in families, testing for transmission of risk alleles; we performed a case-control association analysis in a large cohort of mostly unrelated subjects. Another difference between our cohorts was that the BVMC targeted enrollment of 25% brain VM positive patients, resulting in cohort with 21% brain VM positive patients. This proportion is within the upper end of the range of prevalence of brain VM positive patients reported by other cohorts (5%-23% (Akers et al., 2013)), so we do not expect it to significantly influence the results of our association analysis, however it does represent a divergence from completely unbiased enrollment. Finally, our cohort was North American Caucasians, the original findings were in Dutch and French Caucasians (Benzinou et al., 2012;Kawasaki et al., 2014). It is possible for population stratification even among/within Caucasian cohorts to influence genetic association results. We did not detect significant associations between any of the 7 variants previously reported in sporadic BAVM or ICH with HHT brain VM or ICH. Possible explanations include different genetic influences on sporadic and HHT brain AVM, that the original findings (from small cohorts and unreplicated) were false positives, and a false negative, due to lack of power to detect small effects in our cohort. In the case of ACVRL1 c.314-35A>G, the sporadic brain AVMassociated variant that we found to also be associated with VM in HHT (Pawlikowska et al., 2015), the original finding had been replicated in multiple sporadic brain AVM cohorts, and the variant was in an HHT gene. The BVMC cohort of 752 subjects is substantial for a rare disease such as HHT, but may not be sufficient to detect associations of small effect size. We did observe several associations with the same direction of effect as in previous studies; a larger cohort is required to validate these associations. The BVMC is recruiting a second cohort of 800 HHT patients.
Our findings suggest the effects of genetic modifier variants are complex, but given the functional data from animal and in vitro models for the involvement of ADAM17 and PTPN14 in HHT biology (Benzinou et al., 2012;Kawasaki et al., 2014), it is of interest to revisit these in larger cohorts. Identification and validation of genetic modifiers could inform prognosis and guide clinical management of HHT patients and provide potentially useful stratifier biomarkers for clinical trials.