Rho Guanine Nucleotide Exchange Factor ARHGEF17 Is a Risk Gene for Intracranial Aneurysms
Background: Intracranial aneurysm (IA) is usually a late-onset disease, affecting 1% to 3% of the general population and leading to life-threatening subarachnoid hemorrhage. Genetic susceptibility has been implicated in IAs, but the causative genes remain elusive.
Methods: We performed next-generation sequencing in a discovery cohort of 20 Chinese IA patients. Bioinformatics filters were exploited to search for candidate deleterious variants with rare and low allele frequency. We further examined the candidate variants in a multiethnic sample collection of 86 whole exome sequenced unsolved familial IA cases from 3 previously published studies.
Results: We identified that the low-frequency variant c.4394C>A_p.Ala1465Asp (rs2298808) of ARHGEF17 was significantly associated with IA in our Chinese discovery cohort (P=7.3×10−4; odds ratio=7.34). It was subsequently replicated in Japanese familial IA patients (P=0.039; odds ratio=4.00; 95% confidence interval=0.832–14.8) and was associated with IA in the large Chinese sample collection comprising 832 sporadic IA-affected and 599 control individuals (P=0.041; odds ratio=1.51; 95% confidence interval=1.02-Inf). When combining the sequencing data of all familial IA patients from 4 different ethnicities (ie, Chinese, Japanese, European American, and French-Canadian), we identified a significantly increased mutation burden for ARHGEF17 (21/106 versus 11/306; P=8.1×10−7; odds ratio=6.6; 95% confidence interval=2.9–15.8) in cases as compared with controls. In zebrafish, arhgef17 was highly expressed in the brain blood vessel. arhgef17 knockdown caused blood extravasation in the brain region. Endothelial lesions were identified exclusively on cerebral blood vessels in the arhgef17-deficient zebrafish.
Conclusions: Our results provide compelling evidence that ARHGEF17 is a risk gene for IA.
See Editorial by He et al
Intracranial aneurysms (IAs) are weakened areas in the cerebral vascular that lead to abnormal dilation or ballooning.1 Unruptured IAs are usually asymptomatic unless they reach sufficient size to influence surrounding structures. They are often found either through screening high-risk patients or as purely incidental findings during brain imaging for other neurological symptoms. The danger posed by IAs is rupture, leading to life-threatening subarachnoid hemorrhage (SAH).2 About 1% to 3% of the general population has or will develop IAs.3,4 And 0.5% to 1% of people with IAs may experience rupture annually.5 Meta-analysis showed a wide range of SAH fatality varied from 8.3% to 66.7% that decreased 0.4% per year (95% confidence interval=0.5–1.2) coinciding with the introduction of improved management strategies of presymptomatic and symptomatic IAs.6
The formation and rupture of IAs are likely related to multiple risk factors including smoking, hypertension, heavy alcohol consumption,7 and other vascular disease.8 A small percentage of which are related to infections9 and trauma.10 Many studies have shown that there are underlying genetic risk factors for a large fraction of IA and SAH cases. It is known that 7% to 20% of patients have a positive family history.11 First-degree relatives of individuals with aneurysmal SAH have a 4- to 7-fold higher risk of being affected than the general population.12
The genetic studies on IAs have been performed through genome-wide association studies and family studies. To date, several genome-wide association studies have examined sporadic IA cases and identified several candidate loci: 4q31.2313, 2q33.114,15, 8q11.23, and 9p21.314–16; 18q11.2, 13q13.1, 9p21.3, and 10q24.23.16 Unfortunately, follow-up replication studies showed that these loci are unlikely to explain a large fraction of IAs.17 In family studies, linkage analysis of single IA families or sibling pairs have mapped several candidate loci, but no IA-causing mutations has been identified. Recently, 4 studies18–21 applied whole exome sequencing on multiple IA-affected families to analyze the effect of the rare variant on IAs, 3 IA candidate genes (ADAMTS15, RNF213, and THSD1) were identified. However, many IA cases involved in those 4 studies cannot be explained by the mutations of these genes.18–21 Therefore, new genetic causality for IA needs to be elucidated to unravel the cellular and molecular mechanisms underlying IAs pathophysiology.
One of the challenges common to genome-wide association studies and familial studies on IAs is the number of patients or families in each study. First, unruptured IAs usually do not manifest any clinical phenotype, and the ruptured IAs have high mortality. Accumulating a large sample size requires a long-term epidemiological screening program involving multiple hospitals or registries. Second, as a late-onset disease, some family members may have deceased or not manifested IA at the time, which poses challenges to ascertain a complete pedigree for segregation analyses. Finally, the incomplete penetrance needs to be taken into considerations in the analysis of IAs.
Because much of the genetic contribution to complex traits cannot be explained by genome-wide association studies that focus on the identification of common variants, we focused our attention on the identification of rare (Minor Allele Frequency <0.5%) and low-frequency (0.5% ≤ Minor Allele Frequency <5%) variants and applied next-generation sequencing, which enables more complete assessments of low-frequency and rare genetic variants. It is plausible that analyses of rare and low-frequency variants could explain additional disease risk or trait variability22 (Figure 1A). There is also recent empirical evidence that low-frequency and rare variants are associated with complex diseases.23–25
It is possible that few genetic alleles with modest effect size and incomplete penetrance can segregate in familial fashion as well as increase the IA risk in the general population (Figure 1B). For example, previous studies on familial late-onset Alzheimer disease showed that a significant excess of rare coding variants in APP, PSEN1, and PSEN2 was noted in probands. These variants did not actually cosegregate with the disease, and some of those rare variants were also identified in sporadic Alzheimer cases,26,27 suggesting that those rare variants may serve to modulate the risk of disease both in a familial format and in a sporadic format. With this possibility in mind, to increase the potential genetic loads of IAs in our study samples, lower the burden of sample collection, and identify these risk alleles underlying both familial and sporadic IAs, we purposefully sampled both familial and sporadic IA cases and compared the alleles to the population controls (Figure 1B). This subject ascertaining strategy was also used in a recent IA association study by Kurki et al,28 in which they sampled 760 cases from both familial and sporadic IAs (40% of familial IAs) from Finnish population. They successfully replicated the 9p21.3 and 13q13.1 IA association signals and identified 4 novel loci: 2q23.3, 5q31.1, 6q24.2, and 7p22.128.
We leveraged recent foundational developments in human genetics, such as the latest in silico predictors, large genetic variant databases (1000 Genomes Project [1000G]29 and Exome Aggregation Consortium30), and Genotype-Tissue Expression information31 to prioritize the IA risk gene. We also incorporated unsolved IA cases (no known genetic risk factors identified) from previous studies18–21 as the replication data sets to replicate the candidate variants and to assess the variant burden of the candidate genes (Figure 1C).
The sequencing data that support the findings of this study are available from the corresponding author on reasonable request.
The article has ethics approval and consent to participate by the medical ethics committee of Tianjin Medical University General Hospital (20170035). Informed consent has been obtained, and this report was processed according to the principles expressed in the Declaration of Helsinki.
Further details of the study participants, data analyses, and functional studies in zebrafish can be found in the Data Supplement.
Genome Sequencing Identifies 30 Candidate Variants
Based on our conceptual design described in the Introduction and Figure 1, we developed a human genetics analysis pipeline that includes 3 major stages: variant discovery, variant replication, and risk gene prioritization (Figure 2). In the discovery stage, we genome-sequenced IA patients and performed variant filtering to identify a list of candidate variants. In the variant replication stage, we performed a replication test on the candidate variants using the unsolved IA cases from the previous studies18,19,21 to examine if any of the candidate variants can explain additional IA cases. Finally, we integrated all the sequencing data of IA samples and tested the variant enrichment in the genes to prioritize the risk gene of IAs (Figure 2). To further use the genetic information from familial cases, we performed the segregation analysis and identity-by-descent analysis to examine the association of candidate variants with IAs in families. We also examined the low-frequency risk variant in additional sporadic IA cases (Figure 2).
In our discovery stage, we ascertained 20 Chinese IA samples (Figure IA and IB in the Data Supplement) and interrogated their genetic variants using whole exome sequencing or whole genome sequencing technologies. We focused on the genetic variants in the coding regions by subsetting the WGS to the exome regions that captured by whole exome sequencing (ie, Agilent SureSelect Human All Exon 50 MB). Using our ensemble variant calling pipeline (Figures and Methods in the Data Supplement), we identified 115 912 single nucleotide variants and 6033 indels in total. On average, we found 42 115 single nucleotide variants and 3267 indels per exome and 97.51% of those single nucleotide variants and 83.92% of indels were in dbSNP141. The variant set had a mean transition-to-transversion (Ti/Tv) ratio of 2.56 for all single nucleotide variants per exome (Table VI in the Data Supplement).
We first kept variants in our cases with allele frequency ≤5% in the East Asian population of both the 1000G and Exome Aggregation Consortium databases (Figure 2). Then we performed a case-control analysis to further ensure the putative risk allele was significantly enriched in the cases and was potentially associated with IA. We aggregated the 20 familial and sporadic cases and chose 208 Han Chinese subjects in the 1000G as controls after verifying the congruence of population structure by principal component analysis (Figure IV in the Data Supplement). We applied the 2-sided Fisher exact test on each variant in cases and controls and kept variants (6685 variants left) that were significantly enriched in cases with a P value <8×10−3 after multiple hypotheses correction using Benjamini-Hochberg procedure at False Discovery Rate=5% (Figure V and Methods in the Data Supplement).
Then we kept the variants annotated as loss-of-function variants. For missense variants, we adapted widely used bioinformatics tools including GERP++32, CADD,33 SIFT,34 PolyPhen-235, LRT,36 and MutationTaster37 to predict their functional effects38 (Figure V and Methods in the Data Supplement). We found 117 variants passed these bioinformatics filters. Finally, we required that at least 2 IA-affected families share a candidate variant, and 30 variants remained as candidates after this filter (Figure V in the Data Supplement). Three of them are loss-of-function variants that occurred at essential splicing sites, and 27 of them are damaging missense variants predicted by multiple in silico algorithms (Methods and Table VII in the Data Supplement).
Replicate Candidate Variants in Additional Familial Cases
We next examined the 30 candidate variants in additional unsolved IA cases from previously published studies.18,19,21 There are 86 unsolved familial IAs (24 families) with whole exome sequencing data available. We named the Japanese cohort as replication 1, the European American cohort as replication 2, and the French-Canadian cohort as replication 3 (Methods and Table I in the Data Supplement).
For replication 1, we included 37 IA cases from 11 families (Methods and Figure IIA in the Data Supplement). The control samples for replication 1 were from 2 databases that were used in the original publication by Yan et al19: 104 1000G Japanese samples and 1208 Japanese samples from Japanese Genetic Variation Consortium Database39 (Methods and Table I in the Data Supplement). For replication 2, we included exome sequencing data of 34 European Americans IA cases from 7 families18 (Figure IIB in the Data Supplement). The control samples were 99 Northern Europeans from Utah (CEU) subjects from 1000G. The ethnicities of cases and controls were examined by principal component analysis in Figure VI in the Data Supplement. For replication 3, we included 15 French-Canadians IA cases (Figure IIC in the Data Supplement). We also applied the 99 CEU subjects from 1000G as controls for replication 3 because the ethnicities of the patients were confirmed in the original publication by Zhou et al.21
We applied our variant calling and prioritization methods to these replication datasets (Methods in the Data Supplement). Among the 30 candidate variants identified in the initial Chinese patients, 6 of them were observed in replication cohorts as well: 5 (rs115753757 of ADAM15, rs2298808 of ARHGEF17, rs35217482 of AOX1, rs114777682 of AKAP13, and rs150645471 of ACSM5) were found in replication 1, and one (rs2397084 of IL17F) were found both in the replication 2 and replication 3 (Figure 3A; Table VIII in the Data Supplement). After comparing the allele count of the variants to the population controls using 2-sided Fisher exact test, we ranked these 6 variants based on their P values by ascending order and plotted it against the P values of 30 variants in the discovery cohort (Figure 3A). Two out of the 6 variants (rs115753757 of ADAM15 and rs2298808 of ARHGEF17) were top ranking with a P value <0.05 (Fisher exact test; Figure 3A; Table VIII in the Data Supplement), suggesting a suggestive association with IAs.
Gene-Level Variant Enrichment Analysis Identified a Significantly Higher Variant Burden of ARHGEF17 in IA Cases Than in Controls
To assess the variant burden of the 6 genes with variants rediscovered in replication, particularly ADAM15 and ARHGEF17 with nominal significance in additional replications, we performed a gene-level variant enrichment test, a simplified gene-based association test: if a gene is associated with IA, we expect to identify more distinct rare deleterious variants at gene level in cases than in controls, which can help us evaluate association for multiple variants in a gene22 (Figure 2; Figure VII in the Data Supplement).
We combined all 4 data sets of IA patients (106 in total) and considered the entire gene as a testing unit. We used 103 Han Chinese in Beijing subjects, 104 JPT subjects, and 99 CEU subjects from 1000G as control individuals (in total 306 control samples). We applied the same variant prioritization methods that we developed (Figure VIII in the Data Supplement). For each gene, we categorized the damaging variants into 3 categories: exclusively identified in cases, exclusively identified in controls, and identified in both cases and controls (Figure 3B). We compared the number of distinct damaging variants that exclusively observed in cases to those exclusively observed in controls, and we found a higher number of distinct damaging variants of ARHGEF17 in cases than in controls. Four distinct ARHGEF17 rare deleterious variants were exclusively observed in IA-affected individuals, whereas only one was exclusively observed in the 1000G control individuals (Figure 3B; Table). For the other 5 genes, we did not observe higher number of distinct damaging variants in cases than in controls (Figure 3B; Table IX in the Data Supplement).
We next examined the frequency differences of these ARHGEF17 deleterious variants in cases and controls. We found that 21 of the 106 cases (21%) and 11 of the 306 1000G controls (3.6%) have the ARHGEF17 deleterious variants (P=8.1×10−7 by 2-sided Fisher exact test; odds ratio=6.6; 95% confidence interval=2.9–15.8; Table). When we consider the family as a unit by pruning the related subjects, 9 out of 36 families (25%) carried the deleterious mutations of ARHGEF17 (P=4.3×10−5 by 2-sided Fisher exact test; odds ratio=8.8; 95% confidence interval=3.5-Inf; Table and Table X in the Data Supplement).
We also examined the expression profile of the 6 genes across different human tissues using the Genotype-Tissue Expression data.31 We found that ARHGEF17 is the only gene among the 6 that is highly expressed in blood vessels (Figure 3C), which provided additional evidence that ARHGEF17 could be a candidate gene of IAs and plays a role in human blood vessels. Based on the Exome Aggregation Consortium database, ARHGEF17 is extremely intolerant to loss-of-function variants (probability of being loss-of-function intolerant=1.00), and ARHGEF17 is also intolerant to missense variants with Z score equals 3.18.
Pedigree and Identity-by-Descent Analyses on ARHGEF17 Variants
We went back to the pedigrees for segregation analysis. We found that 3 out of the 5 ARHGEF17 variants (c.4394C>A_p.Ala1465Asp, c.5168G>A_p.Arg1723Gln, and c.5327G>A_p.Cys1776Tyr) were cosegregated with IAs in at least one of the pedigrees (Figure 3D; Table). The variant c.3511G>A_p.Ala1171Thr was a novel variant found in a Chinese sporadic IA patient. The variant c.5581C>T_p.Arg1861Cys was found in 2 patients from a French-Canadian family (Figure 3D; Table). We performed identity-by-descent analysis on the families with ARHGEF17 mutations. Among the 4 families (discovery P1, discovery P4, replication 1 P6, and replication 2: family D) with ARHGEF17 mutations cosegregated with IA, 3 (discovery P1, replication 1 P6, and replication 2: family D) had shared identity-by-descent haplotype that cover ARHGEF17 gene region (Table XI in the Data Supplement).
Overall, in our Chinese cases, we found 2 ARHGEF17 mutations (c.4394C>A_p.Ala1465Asp and c.3511G>A_p.Ala1171Thr). We screened the 2 mutations in the IA-unaffected Chinese family members that were available to our study. The variant c.3511G>A_p.Ala1171Thr was only found in IA-affected family members. We found that 4 unaffected individuals had the mutation c.4394C>A_p.Ala1465Asp and 3 of them are younger than the IA-affected family members (Figure 3D; Figure IA in the Data Supplement). Although not manifesting IA, the clinical history records of those 4 individuals showed that 3 of them had cerebrovascular abnormalities, including cerebral infarctions, cerebral artery stenosis, and bilateral fetal posterior cerebral artery. Two of them also had hypertension (Table XII in the Data Supplement).
Association Analysis in Sporadic IAs Provided Additional Evidence Supporting That ARHGEF17 Is a Risk Gene of IA
Because c.4394C>A_p.Ala1465Asp (rs2298808) is a low-frequency variant in East Asians and is carried by IA-unaffected individuals as well, we further examined whether it was enriched in sporadic IAs cases. We independently screened this variant in additional 832 unrelated Chinese SAH patients caused by IA-rupture and 599 unrelated Chinese individuals with no evidence of IAs or other malformations (Table II in the Data Supplement).
We found the alternative allele of rs2298808 showed nominal enrichment in IA-affected individuals compared with control individuals (AFCases=3.73%; AFControls=2.50%; P=0.041 by 1-sided Fisher exact test, odds ratio=1.51; 95% confidence interval=1.02-Inf; Table XII in the Data Supplement). In addition, the homozygous rs2298808 were only observed in 5 patients with ruptured IAs (Table XIII in the Data Supplement). They provided additional lines of evidence that ARHGEF17 is a risk gene of IAs.
Functional Studies of ARHGEF17 Function and Its Mutations
We performed additional functional analyses on ARHGEF17 using in silico methods. We first performed a literature search to examine whether ARHGEF17 is biologically related to IAs. In addition, we also performed computational modeling of ARHGEF17 domains using structural homologies to examine effects of mutations on the domain structures.
In our literature search, we found that ARHGEF17 encodes a Dbl family Rho GEF (guanine nucleotide exchange factor), which plays an important role in activating the Rho family GTPases.40–42 The proteins of Rho subfamily (RhoA, B, and C) promote the formation of actin stress fibers that are essential for cellular processes, including cell shape, polarity, migration, cell-cell, and cell-matrix interactions.40,43,44 We also found that the depletion of ARHGEF17 leads to defective human umbilical vein endothelial cell junctions,45 suggesting that it could play a role in maintaining the integrity of blood vessels.
The predicted domains of ARHGEF17 include the catalytic Dbl-homology (DH) domain for Rho GTPase activation, PH (pleckstrin homology) domain for lipid interaction, and the β-propeller fold that may mediate protein-protein interaction.46,47 The variants c.3511G>A_p.Ala1171Thr and c.4394C>A_p.Ala1465Asp map to the DH domain and the PH domain, respectively. Two variants (c.5327G>A_p.Cys1776Tyr and c.5581C>T_p.Arg1861Cys) map to the β-propeller domain (Figure 4A). All 5 mutations affect evolutionarily conserved residues (Figure 4B). The variant c.5168G>A_p.Arg1723Gln is the first amino acid of the predicted β-propeller domain that connects to the linker sequence between 2 domains (Figure XIA in the Data Supplement).
Computational modeling of ARHGEF17 important domains using protein homology showed that 4 variants (c.4394C>A_p.Ala1465Asp, c.5168G>A_p.Arg1723Gln, c.5327G>A_p.Cys1776Tyr, and c.C5581T_p.Arg1861Cys) are exposed on the surface of the protein domain, suggesting that these variants might alter the enzyme in its interactions with substrate or cofactors or in its structural stability. Thr at position 1171 can introduce a bulky side chain compared with the wild-type Ala and might result in clashes in the domain structure (Figure XIA and XIB in the Data Supplement).
arhgef17-Deficient Zebrafishes Display Intracranial Hemorrhage
We found that zebrafish Arhgef17 is 67% identical to the human ARHGEF17 with a high degree of conservation in the DH domain, PH domain, and β-propeller48 (Figure XII in the Data Supplement). Whole-mount in situ hybridization displayed abundant expression of arhgef17 in the zebrafish head region, including the forebrain, the midbrain, and the hindbrain during development (Figure 5A through 5C). Using ARHGEF17 antibody, we detected that Arhgef17 is expressed in the brain blood vessels at 72-hour postfertilization (Figure 5D, 5DI, 5DI-I, and 5DI-II).
To validate the functional impact of ARHGEF17 in vivo, we developed a zebrafish model with arhgef17 deficiency (Methods). We designed 2 splice-blocking morpholinos, I1E2-MO, and I9E10-MO to knockdown arhgef17 by skipping exon 2 (DH domain) and exon 10 (PH domain), respectively (Figure XIIIA in the Data Supplement). RT-PCR analyses of embryos injected with I1E2-MO and I9E10-MO determined the knockdown efficiency (Figure XIIIB through XIIID in the Data Supplement). arhgef17 transcripts skipping exon 2 or partial exon 2 were detected in embryos with microinjection of I1E2-MO (Figure XIIIB and XIIIE in the Data Supplement), confirmed by DNA sequencing (Figure XIIIF in the Data Supplement). Microinjection of I9E10-MO caused a frameshift, resulting in nonsense-mediated mRNA decay of arhgef17, as the new electrophoretic band was absent, and the PCR product of unedited sequence was reduced (Figure XIIIC and XIIID in the Data Supplement).
arhgef17-deficient embryos appeared to be morphologically indistinguishable from their wild-type siblings up to 42-hour postfertilization (Figure 5E and 5F). From 42-hour postfertilization to 80-hour postfertilization, embryos injected with I1E2-MO or I9E10-MO displayed intracranial hemorrhage in multiple areas in the zebrafish head region, including the forebrain, the midbrain, and the hindbrain, in comparison to control embryos (Figure 5E and 5F; Figure XIVA and XIVB in the Data Supplement). Whole-mount o-Dianisidine analyses for hemoglobin staining49 confirmed erythrocyte extravasation in arhgef17-deficient embryos (Figure 5G; Figure XIVC and XIVD in the Data Supplement), suggesting the presence of defects in vessel integrity or development. To identify endothelial lesions that cause intracranial hemorrhage, we used double transgenic zebrafish Tg (kdrl:EGFP; gata1:DsRed), in which the endothelium and the erythroid are marked by EGFP (enhanced green fluorescent protein) and DsRed(Discosoma sp. red fluorescent protein), respectively (Figure 5H). We observed erythrocyte extravasation marked by DsRed in the midbrain and hindbrain region (Figure 5I; Figure 6A and 6B), as well as the forebrain (Figures 5J and 6B) in arhgef17-deficient embryos. These intracranial bleeding sites are adjacent to the prosencephalic artery (Figure 5J) and the hindbrain vascular network, comprising the basilar artery, primordial hindbrain channels, and central arteries50 (Figure 5H). However, overall blood vessel morphology and development appeared to be normal (Figure 5A, 5B, and 5G through 5I). Furthermore, confocal imaging analyses identified multiple endothelial lesions, localizing specifically to a central artery (Figure 6AI and 6AI-I), ACeV (Figure 6BI and 6BI-I), mesencephalic vein (Figure 5BII and 5BII-I) and posterior cerebral vein (Figure 6BIII and 6BIII-I). When we coinjected wild-type human ARHGEF17 mRNA into zebrafish zygotes with arhgef17 morpholinos, we observed a statistically significant reduction of intracranial hemorrhages (Figure 6C). However, coinjection with rs2298808 variant human mRNA can only partially reduce hemorrhage phenotypes in arhgef17-deficiency embryos (Figure 6C). These results reveal that endothelial leakage in multiple cranial vessels causes erythrocyte extravasation in arhgef17-deficient embryos and ARHGEF17 is a risk gene of IAs.
IA is usually a late-onset complex disease with both genetic and environmental risk factors.17 Similar to other common complex diseases such as Alzheimer disease, breast cancer, and diabetes mellitus, it is also genetically heterogeneous, so a large proportion of the patients does not follow a canonical Mendelian inheritance pattern.26,51,52 Many affected families may include both inherited and sporadic components of the disease. Meanwhile, the sporadic cases can include inherited components as well, likely because of the incomplete penetrance of genetic risk factors.
In this study, we pooled familial and sporadic IA cases together and performed a case-control analysis in candidate variants filtering. We think this method may help (1) increase the genetic loads in cases; (2) lose the burden of collecting large families with multiple family affected; and (3) identify the risk factors underlying the general population.
Aggregation of the unsolved IA cases from previous studies provided us the opportunity to study the genetic risk factors in multiethnic familial IA samples. We found ARHGEF17 had a significantly high mutation burden in the IA cases compared with control individuals (21/106 versus 11/306; P=7.4×10−12), suggesting that ARHGEF17 mutations may be IA risk factors in both Asian and European populations.
We found 4 IA-unaffected individuals in familial samples had the ARHGEF17 damaging mutation (rs2298808, Minor Allele Frequency=3.24% in EAS from Exome Aggregation Consortium, Table XII in the Data Supplement), which is likely because of the incomplete penetrance of this variant. We also found that the 4 heterozygous carriers of rs2298808 from 2 families are IA negative and relatively younger. One family had 2 heterozygous IA patients and the mean diagnose age is 50, whereas the mean age of heterozygous carriers is 43 (Figure IA in the Data Supplement; family P4). The other family contained one 58-year-old heterozygous IA patient, whereas the mean age of the 2 heterozygous carriers is 53.5 (Figure IA in the Data Supplement; family P6). Therefore, at this point, we cannot entirely exclude the possibility that some of the unaffected heterozygotes will develop IA when they grow older. We further examined rs2298808 in sporadic Chinese IA cases. We found that the alternative allele frequency of rs2298808 is 3.73% in 832 unrelated cases, whereas 2.50% in 599 controls. This difference achieved nominal statistical significance (P=0.041 1-sided Fisher exact test; odds ratio=1.51; Table XII in the Data Supplement), suggesting that it is a genetic risk factor of IAs. More interestingly, we only observed the homozygous rs2298808 in IA-affected individuals but not in the control individuals, and all of them have ruptured IAs (Table XIII in the Data Supplement). Our rescue experiments using rs2298808 variant mRNA and found that rs2298808 mRNA can only partially rescue the hemorrhage phenotype in Arhgef17-deficiency embryos, suggesting that s2298808 mutation cannot disrupt the ARHGEF17 function completely in zebrafish. Nevertheless, this result does not contrast our conclusion that ARHGEF17 variant (s2298808) is associated with human IA with incomplete penetrance.
ARHGEF17 encodes one of the Dbl family GEFs and many Dbl family GEFs have been shown as oncogenes. Similarly, ARHGEF17 was identified originally as a gene whose expression was upregulated in endothelial cells during tumor cell-induced angiogenesis.53 This is for the first time that ARHGEF17 is implicated in cerebrovascular disease. Recent studies on ARHGEF17 showed that (1) ARHGEF17 is localized to the actin cytoskeleton via sequences in the N-terminus and activates RhoA at the intercellular junctions46; (2) the depletion of ARHGEF17 leads to defective human umbilical vein endothelial cell junctions.45 Importantly, it has been shown that the specific activation of RhoA at intercellular junctions is critical for proper endothelial junction formation and integrity.54 Other known risk loci in IAs are also implicated in the endothelial cell function and integrity. The recent study identified THSD1, which plays an important role in the endothelial cell focal adhesion and attachment20; the other 2 IA candidate genes ADAMTS15 and RNF213 have antiangiogenic activity and might also be involved in endothelial cell-cell junction stabilization.19,21 The loss of endothelial-specific expression of Sox17 promoted IAs in an elastase-treated mouse model55; the destruction of tight junctions may facilitate macrophage migration and cerebral aneurysm formation in rats.56 Therefore, it is plausible that ARHGEF17 mutations disrupt the RhoA activation that consequently leads to the endothelial junction destabilization in blood vessels, conferring risks for an IA formation and rupture.
We performed functional studies in the zebrafish model and showed that arhgef17 is abundantly expressed in the head region of zebrafish, including brain blood vessels. We have followed the technical guidelines for morpholino use in zebrafish,57,58 and showed an injection of 2 splicing-blocking MOs at low dosages (≈3 ng) causes intracranial hemorrhage without overall embryonic abnormalities. The morpholinos-based knockdown leads to the perturbation of DH domain or results in nonsense-mediated mRNA decay. Importantly, the intracranial hemorrhage could be rescued by coinjection of human wild-type Arhgef17 mRNA. Thus, the prominent phenotypes of intracranial hemorrhage are caused by arhgef17 deficiency rather than morpholino toxicities and off-target effects. Human mRNA with rs2298808 variant can partially rescue the hemorrhage phenotype in Arhgef17-deficient embryos, suggesting that rs2298808 variant cannot completely disrupt the ARHGEF17 function in zebrafish and could be a risk factor for IAs.
We thank Sau Wai Cheung, Graeme Mardon, Chad Shaw, and Christian Schaaf for constructive suggestions, and Yufeng Shen, Richard A. Gibbs, and Lee-Jun Wong for critical reading of the article. We are grateful to all of the patients and their families for participating in this study.
Sources of Funding
This study was supported by grants from the US National Institutes of Health and National Human Genome Research Institute (R01HG008115 to Dr Yu), the National Natural Science Foundation of China (31530044 and 31471357 to Dr Zhong; 81571144 to Dr Yang, and Dr Zhang; 81330029 to Dr Zhang), the Tianjin Science and Technology Commission (16JCZDJC35700 to Dr Yang and Dr Zhang, 15ZXLCSY00060 to Dr Zhang), the US National Institutes of Health-National Heart, Lung, and Blood Institute (5R01HL125957 to Drs Yu and Dong), and the National Institute of Neurological Disorders and Stroke (R01NS094535 to Dr Kent).
Dr Kent is a member and shareholder in Acelerox, LLC to commercialize the medical use of carbon nanomaterial. The other authors report no conflicts.
The Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGEN.117.002099/-/DC1.
- Received December 15, 2017.
- Accepted May 22, 2018.
- © 2018 American Heart Association, Inc.
- Rinkel GJ,
- et al
- Teunissen LL,
- et al
- Benoit BG,
- et al
- Foroud T,
- et al
- Yan J,
- et al
- Santiago-Sim T,
- et al
- Zhou S,
- et al
- Rivas MA,
- et al
- Melé M,
- et al
- Chun S,
- et al
- Richards S,
- et al
- Yamaguchi-Kabata Y,
- et al
- Hall A,
- et al
- Ridley AJ,
- et al
- Ngok SP,
- et al
- Rümenapp U,
- et al
- King MC
- Lee S,
- et al
- Moulton JD