Cerebral cavernous malformations (CCM) are vascular lesions which can occur as a sporadic (80% of the cases) or familial autosomal dominant form (20%). Three CCM genes have been identified: CCM1/KRIT1, CCM2/MGC4607 and CCM3/PDCD10. Almost 80% of CCM patients affected with a genetic form of the disease harbor a heterozygous germline mutation in one of these three genes. Recent work has shown that a two-hit mechanism is involved in CCM pathogenesis which is caused by a complete loss of any of the three CCM proteins within endothelial cells lining the cavernous capillary cavities. These data were an important step towards the elucidation of the mechanisms of this condition.
Cerebral cavernous malformations (CCM/OMIM 116860) are vascular lesions histologically characterized by abnormally enlarged capillary cavities without intervening brain parenchyma. From large series based on necropsy and/or magnetic resonance imaging (MRI) studies, their prevalence in the general population has been estimated to be close to 0.1–0.5%. Most CCMs are located within the central nervous system but they sometimes affect either the retina or the skin .
CCM occur both sporadically and in a familial context. The proportion of familial cases has been estimated to be as high as 50% in Hispano-American CCM patients  and close to 10–40% in Caucasian patients . The CCM pattern of inheritance is autosomal dominant with incomplete clinical and neuroradiological penetrance. The presence of multiple lesions on cerebral MRI is one of the main features of familial CCM which is an evolutive condition as assessed by the strong correlation between patient age and the number of lesions (Fig. 1) [1–3]. The average age-of-onset is around 30 years but symptoms can start in early infancy or in old age. The main symptoms include seizures and cerebral hemorrhages. Sporadic cases most often have a single lesion on MRI, are not inherited and do not carry a CCM gene germline mutation. However, some CCM patients who have multiple MRI lesions do not have any known clinically affected relative and therefore present as sporadic cases. Combined use of clinical and MRI screening with molecular testing has helped to clarify what might have first seemed confusing .
Three CCM genes have been mapped and identified in the recent years. These molecular genetics data have provided useful information for clinical care of the patients and were an important step towards the understanding of the mechanisms of this disorder. This minireview summarizes the advances in CCM molecular genetics and the remaining gaps in this field. In addition to the identification of CCM genes, a number of recent biochemical in vitro studies and in vivo CCM animal model studies have helped to unravel the functional roles of these proteins and are be the focus of the two accompanying minireviews by Faurobert and Albiges-Rizo  and Chan et al. .
CCM genes germline mutations
Genetic linkage analyses mapped three CCM loci to chromosome 7q (CCM1), 7p (CCM2) and 3q (CCM3) [6,7]. A strong founder effect has been oberved in Hispano-American CCM patients with most families linked to the CCM1 locus . In Caucasian families, the proportions of families linked to each CCM locus were 40% (CCM1), 20% (CCM2) and 40% (CCM3) . The three genes located at these loci have now been identified (Fig. 2) [9–13].
The CCM1 gene contains 16 coding exons which encode for Krit1, a 736-amino acid protein containing three ankyrin domains and one band 4.1 ezrin radixin moesin (FERM) domain. CCM2, a 10-exon gene, encodes for the MGC4607 protein, also called malcavernin, which contains a phosphotyrosin binding domain. CCM3 includes seven exons which encode for Programmed Cell Death 10 (PDCD10), a protein without any known conserved functional domain, which may be involved in apoptosis. Considerable progress has been made recently in understanding the biochemical pathways in which those proteins might be involved (see Faurobert and Albiges-Rizo ).
More than 150 distinct CCM1/CCM2/CCM3 germline mutations have been published to date [9–23]. Those mutations were highly stereotyped because almost all led to a premature termination codon through different mechanisms including nonsense, splice-site and frameshift mutational events, as well as large genomic rearrangements. These data strongly suggest that a loss of function, through mRNA decay of the mutated allele, is the most likely pathophysiological mechanism involved in CCM patients. Only four ‘missense’ mutations within CCM1 have been reported to date; interestingly, all of them actually activated cryptic splice sites and led to an aberrant splicing of CCM1 mRNA and a frameshift with a premature stop codon [10,24]. The only known missense mutation which did not affect splicing has been located within the C-terminal part of the phosphotyrosin binding domain of CCM2 . This mutation has been shown to abolish the interaction of CCM2 and CCM1, strongly suggesting its causality . Only four inframe deletions have been reported: two affect exons 17 and 18 of CCM1, one deletes exon 2 of CCM2 and one deletes exon 5 of CCM3. The last two deletions have been used to map potential relevant interaction domains of CCM2 and CCM3 [23,26]. However, it is not known if these putative truncated proteins were indeed produced and stable in vivo.
Sequencing of all coding exons of the three CCM genes and search for genomic rearrangements using cDNA and/or quantitative multiplex PCR such as multiplex ligation-dependent probe amplification in Caucasian non-Hispano-American CCM multiplex families led to the identification of the causative mutation in 95% of the families [23,27]. Approximately 72% of multiplex families harbored a mutation in CCM1, 18% in CCM2 and 10% in CCM3. The CCM3 proportion was much lower than expected, based on previous linkage data which suggested that 40% of CCM families were linked to the CCM3 locus.
The mutation detection rate was lower in sporadic cases with multiple lesions, ranging from 45% to 67% [22,23,27]. Most of these sporadic cases with multiple lesions had either inherited their mutation from one of their asymptomatic parents because of incomplete penetrance or had a de novo mutation. Sporadic cases with multiple lesions in whom no mutation was detected are nevertheless most likely affected by a genetic form of the disease. Several hypotheses may be raised to explain the absence of any detected mutation, including a somatic mosaicism of a de novo mutation which occured during gestation and is not detectable in DNA extracted from peripheral blood cells. It will be important to solve this in the future because it is of interest for genetic counseling . With regard to sporadic CCM cases with a unique lesion on cerebral MRI, no mutation was detected in reported series [16,17]. Combination of these data with those obtained in familial CCM strongly suggests that sporadic cases with a unique lesion who would harbor a germline mutation are most likely very rare.
Haplotyping data strongly suggested a founder effect in the Hispano-American CCM population; this was confirmed by the detection of a Q455X stop codon mutation in CCM1 in most families with this ethnic background . Recurrent mutations have also been identified in a few additional populations [21,28]. However, in most cases, despite their highly stereotyped consequences, germline CCM mutations are ‘private’ mutations present in only one or very few families.
Biallelic somatic and germline mutations in CCM lesions
Based on the autosomal dominant pattern of inheritance of CCM and the presence of multiple lesions in familial CCM, contrasting with the detection of a single lesion in nonhereditary cavernous angiomas, it has been proposed that a second hit affecting the wild-type allele might be involved in CCM lesions pathophysiology, as reported previously in retinoblastoma or other vascular malformations [29,30]. According to this hypothesis, CCM formation would be caused by a complete loss, within affected cells, of the two alleles of a given CCM gene. Loss of one of the alleles (first hit) would be the result of a germline mutation and loss of the second allele (second hit) will occur somatically.
This hypothesis is not easy to test because of the heterogeneous cellular nature of CCM lesions and the very limited number of endothelial cells lining the capillary cavities. Indeed, direct sequencing of the DNA extracted from a heterogeneous lesion may not detect the mutation depending of the proportion of the cells which harbor this mutation within the lesion. This approach was initially used to screen CCM lesions from both sporadic and a few familial patients and did not detect any somatic mutation except in one sporadic case . In this latter case, two CCM1 missense mutations, F97S and K569E, were detected in the CCM lesion and were shown to be absent in the blood of the patient. However, the data were difficult to interprete because of the nature of the mutations which were not truncating mutations (a possible aberrant splicing effect of these two mutations was not investigated) and the fact that the biallelism of these mutations was not explored.
In 2005, Gault et al. reported the first biallelic CCM1 germline and somatic truncating mutation in a CCM lesion, strongly supporting this ‘two-hit’ mechanism in the formation of lesions, at least in CCM1 patients; they demonstrated recently that this second hit occurred within the endothelial cells [32,33].
Biallelic somatic and germline mutations in each of the three CCM genes were recently reported by Akers et al. . These authors amplified and sequenced a large number of clones from 10 CCM lesions resected from patients harboring a heterozygous germline mutation in either CCM1 (two patients), CCM2 (five patients) and CCM3 (two patients). One CCM lesion was analyzed for each patient. They were able to convincingly establish the presence of a biallelic somatic and germline deleterious mutation in four of these lesions from two CCM1 patients, one CCM2 patient and one CCM3 patient. The proportion of amplicons carrying the somatic mutation ranged from 4% to 16%. None of these mutations was detected through direct sequencing of lesion DNA, emphasizing the lack of sensitivity of direct sequencing of lesion DNA. These data established the existence of biallelic somatic and germline mutations, whatever the nature of the CCM gene involved, at least in some lesions. No mutation was detected in the six remaining lesions. Several hypotheses may be raised to explain this absence of mutation including the incomplete sensitivity of this type of approach which would miss a second hit consisting in either large genomic deletions and/or epigenetic silencing mechanisms.
Interestingly, the authors showed, using laser capture, that the somatic mutation occurred in endothelial cells and not in the intervening neural tissue. The proportion of endothelial cells which harbor the somatic mutation was estimated in one lesion and shown to be close to 30%, suggesting the mosaicism of this somatic mutation. These data are in agreement with those obtained very recently with an immunohistochemistry-based approach which showed a mosaic loss of expression of CCM proteins in endothelial cells lining CCM caverns . This question would, however, require additional investigations. It would also be important to analyze several lesions from a given patient to test for the presence of the same mutation in multiple lesions. A unique somatic mutation has indeed been detected in multifocal lesions in another hereditary vascular condition suggesting a common origin for abnormal endothelial cells lying in distant sites .
Altogether these data strongly suggest that CCM, as several other hereditary vascular conditions, show a paradominant inheritance. It remains to determine when do occur the somatic, second hit, events.
Are there additional CCM genes?
Previous linkage data obtained on 20 large North American families suggested that the three CCM loci on 7p, 7q and 3q would most likely account for all CCM families . However, despite extensive screening of exonic sequences for point mutations and deletions, no mutation was detected in 5% of familial CCM cases and a larger proportion of sporadic cases wth multiple lesions . In addition, the proportion of families showing a mutation within PDCD10 (10%) at the CCM3 locus on chromosome 3q25, was much lower than expected based on linkage data (40%).
Several mutually nonexclusive hypotheses may explain these data such as: (a) the existence of mutations affecting cis-regulatory elements located at long distances from known CCM transcription units; (b) epigenetic silencing of these three genes; and (c) the existence of additional nonidentified CCM genes, one of which is possibly located close to PDCD10.
Recently, an additional gene, Zona Pellucida-like Domain containing 1 (ZPLD1), has been reported to be disrupted in a CCM patient harboring a balanced translocation between chromosome X and chromosome 3q . ZPLD1 is located on chromosome three centromeric to PDCD10. The expression of the mRNA in lymphoblastoid cell lines of the patient was shown to be significantly decreased suggesting that the interruption of this gene may be causal. However, the same authors screened this gene in 20 additional CCM patients without any mutation in CCM1/CCM2/CCM3 and did not detect any mutation. These data suggest that either this gene is involved in very rare CCM patients or its interruption does not cause CCM but that the translocation present in this patient deregulated the expression of a gene unidentified yet.
Additional work currently conducted in several teams should help in the next future to identify the molecular anomalies of CCM patients ‘without’ mutations.
Conclusions and future
The recent identification of the three CCM genes is an important step towards the elucidation of the mechanisms of this condition. It helped to clarify several features of this condition including its incomplete clinical and MRI penetrance as well as the molecular basis of sporadic cases with multiple lesions. Additional large series studies are needed to evaluate genotype–phenotype correlations (particularly the prognosis) depending of the nature of the mutated gene. Several additional questions, however, have to be adressed. What is the nature of the molecular anomaly in familial CCM cases in whom no mutation has been detected? Are sporadic cases CCM patients with multiple lesions showing a mosaicism for a germline mutation? Are there modifying genes that may explain the intrafamilial clinical variability? In addition to these questions, one main challenge is to understand the mechanisms of this condition. The recent identification of several of the biochemical pathways involving CCM proteins as well as the analysis of several fish and mouse CCM animal models has already provided a number of clues to this goal (see Faurobert and Albiges-Rizo  and Chan et al. ).