Genotype‐phenotype correlations in multiple lesions of familial cerebral cavernous malformations concerning phosphatidylinositol 3‐kinase catalytic subunit alpha mutations

Dear Editor, Our study suggested that in familial multiple cerebral cavernous malformations (CCMs), symptomatic intracerebral haemorrhage (ICH) lesions and dot-sized lesions harbour distinct clinical features and genotypes. Bevacizumab, a vascular endothelial growth factor (VEGF) inhibitor, might be a potential therapeutic agent for familial CCMs. CCMs affect approximately 0.16% to 0.8% of the general population, which is a major cause of ICH in young adults. Sporadic cases often occur as a single lesion without a family history, whereas familial cases are characterized by multiple lesions with a positive family history.1 Based on magnetic resonance imaging (MRI), two subtypes of lesions of familial CCMs were found: ICH lesions exhibit subacute haematoma, while dot-sized lesions show haemosiderin staining.2 The genotype-phenotype correlations of sporadic CCMs are well studied. However, the phenotype, genotype and relationship of these multiple lesions in familial CCMs remain unclear. To clarify the clinical features of the two subtype lesions, 10 familial CCM patients were enrolled (Table S1). Among 10 patients, there were 12 ICH lesions and 505 dot-sized lesions (Figure 1A,B), and the pattern of these lesions both distributed throughout the brain parenchyma.3, 4 After a follow-up time of 3230.7 lesion-years, haemorrhage events occurred in 11 lesions and Kaplan-Meier analysis revealed that the annual incidence rate for haemorrhage events of ICH lesions (22.0 per 100 lesion-years) was significantly higher than that of dot-sized lesions (0.03 per 100 lesion-years) (p < 0.001) (Figure 1C–E). To investigate the genotypes of ICH and dot-sized lesions, DNA sequencing was performed on the surgical samples from familial CCMs. First, we investigated the genotypes between pairings of symptomatic ICH lesions and their adjacent independent dot-sized lesions (Figure S1). Combining whole exome sequencing (WES) and droplet digital polymerase chain reaction (ddPCR), 3 ICH lesions harboured CCM germline mutations and phosphatidylinositol 3-kinase catalytic subunit alpha (PIK3CA) somatic mutations, whereas three dot-sized lesions only had CCM germline mutations (Figure 2A and Table S2). Furthermore, through WES analysis of 12 ICH formalin-fixed paraffin-embedded (FFPE) samples from 10 patients enrolled in imaging follow-up (Figure S1), CCM germline mutations were detected (Tables S1 and S2). Because the total DNA extracted from FFPE samples was limited, ddPCR was applied to detect the prevalent PIK3CA mutations.5 PIK3CA mutations were found in 10 samples (Figure 2A,B, Figures S2–S6 and Table S3). Overall, 86.7% (13/15) ICH lesions harboured somatic PIK3CA plus CCM germline mutations, whereas 100% (3/3) dot-sized lesions had only CCM germline mutations (p = 0.012) (Figure 2C, Figure S2–S6 and Table S3). Additionally, three (16.7%) specimens were detected with somatic CCM mutations by WES (Table S4). In familiar cases, CCM germline mutations occur earlier than somatic PIK3CA mutations, nevertheless, CCM mutations may be a secondary event following the mutation of PI3K pathway genes in sporadic cases.3-5 Two subgroups' lesions harboured different genotypes and related phenotypes, which implies an underlying molecular mechanism within them. Recent studies indicated that PIK3CA mutation was associated with haemorrhagic events in CCMs.6-8 To confirm the mechanism of PIK3CA mutation in familial CCMs, we activated PI3K signalling in human umbilical vein endothelial cells (HUVECs) by transfecting with short-interfering PTEN (siPTEN). Compared to the control group, the levels of phosphorylated AKT (p-AKT), VEGF A (VEGFA), phosphorylated rho-associated kinase 2 (p-ROCK2) and thrombomodulin (TM) in the PTEN-knockdown HUVECs were elevated. The TM expression was reversed by MK-2206 (PI3K inhibitor) and fasudil (ROCK inhibitor) (Figure 3A). Furthermore, we investigated the effect of PI3K pathway activation in Krit1-knockdown HUVECs (Figure S7). We found that in Krit1-knockdown HUVECs, PI3K pathway activation also increased the expression of p-AKT, VEGFA, p-ROCK2 and TM (Figure 3A). TM expression was also reversed by MK-2206 and fasudil. We then validated the genotype-phenotype correlations of the two lesion subgroups in vivo. C57BL/6 mice with endothelial cell-specific deletion of Krit1 (Krit1iECKO) were induced by 4-hydroxytamoxifen injection at postnatal day 1 (P1) as described previously (Figure S8).9 Then, adeno-associated virus (AAV)-control or AAV-PIK3CAH1047R was administered to Krit1iECKO or Krit1fl/fl mice at P60 via retro-orbital sinus injection. Brain MRI and histologic examination were performed at P90 (Figure 3B). On MRI, typical subacute ICH lesions were observed in the AAV-PIK3CAH1047R+Krit1iECKO group, while only chronic small lesions were observed in the AAV-control+Krit1iECKO group (Figure 3C,D). The mean numbers of lesions in the AAV-PIK3CAH1047R+Krit1iECKO and AAV-control+Krit1iECKO group were 98.2 and 44.5, respectively (Figure 3D and Figure S9A), and a higher number of larger lesions (> 10 000 µm2) were observed in the PIK3CAH1047R+Krit1iECKO group than in the AAV-control+Krit1iECKO group (45.0 vs. 11.0, Figure 3E and Figure S9A). Furthermore, compared with Krit1fl/fl+AAV-PIK3CAH1047R or AAV-control+Krit1iECKO mice, the brain endothelial cells of AAV-PIK3CAH1047R+Krit1iECKO mice showed elevated expression of VEGFA and TM (Figure 3F and Figure S9B). Vitro and vivo experiments both indicated that PI3K-VEGF pathway activation plays a vital role in the genotype-phenotype correlations. We hypothesized that VEGF inhibitors might restrain bleeding. Bevacizumab, a VEGF inhibitor, is an effective therapy for cancers. To investigate the effect of bevacizumab, P60 Krit1iECKO mice were injected with AAV-PIK3CAH1047R via retro-orbital sinus and bevacizumab through the tail vein, and the control mice were injected with AAV-PIK3CAH1047R and vehicle alone. Then, the mice were subjected to MRI and histological examination at P90 (Figure 4A). Surprisingly, bevacizumab treatment inhibited ICH formation (Figure 4B,C). Furthermore, haematoxylin and eosin staining revealed that bevacizumab dramatically decreased the number of CCM lesions and prevented the formation of larger CCMs (Figure 4D–F). Our study has several limitations. First, for ethical considerations, we cannot obtain more surgical specimens of dot-sized lesions in the familial CCMs; second, some paired blood samples were unavailable in FFPE samples and the mutation type of surgical samples (germline or somatic) was identified by minor allele frequency (MAF); finally, for the suboptimal quality of sample DNA, the MAF of CCM germline mutation were below 50%, particularly in the FFPE samples. Here, we identified that in familiar CCMs, the two subgroup lesions have distinct genotype-phenotype correlations concerning PIK3CA mutations, which implies differentiated management strategy might be taken. Bevacizumab, as a promising therapeutic agent for familial CCMs, requires further prospective clinical study to evaluate its safety and efficacy due to the risk of complications such as hypertension and cardiac ischemia.10 Jiguang Wang and Yong Cao conceived and designed the study; Jie Wang contributed to clinical data collection, samples collection, vivo and vitro experiment, data analysis and wrote the manuscript; Jihong Tang and Yingxi Yang contributed to bioinformatics analysis; Ran Huo and Hongyuan Xu contributed to experimental design and implementation; Yingfan Sun, Yuming Jiao, Qiheng He, Shaozhi Zhao and Qifeng Yu helped with collecting the samples. Shuo Wang, Jizong Zhao and Yong Cao performed the operation and provided CCMs samples and control specimens; Jiguang Wang supervised bioinformatics studies. Jiguang Wang and Yong Cao provided overall oversight of the research. We thank Professor Xiangjian Zheng (Department of Pharmacology, Tianjin Medical University, China) and Xi Yang (Department of Pharmacology, Tianjin Medical University, China) for kindly providing transgenic mice and experimental assistance in our study. The authors declare no conflict of interest. This study is funded by Genomics Platform Construction for Chinese Major Brain Disease-AVM (PXM2019_026280_000002-AVM); Beijing Advanced Innovation Center for Big Data-based Precision Medicine (PXM2020_014226_000066). Research in Wang lab is supported by RGC grants (16102522, C6021-19EF), ITC grant (ITCPD/17-9), a project of Hetao Shenzhen-Hong Kong Science and Technology Innovation Cooperation Zone (HZQB-KCZYB-2020083), and the Padma Harilela Professorship. The study was approved by the Institutional Review Board and the Ethics Committee of Beijing Tiantan Hospital. (KY2017-035-02). The Animal Welfare and Ethics Committee of Beijing Neurosurgical Institute Laboratory approved all animal ethics and protocols. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.