K. J. Whitehead, Molecular Medicine Program, University of Utah, 15 N. 2030 East, Salt Lake City, UT 84112, USA Fax: +1 801 585 0701 Tel: +1 801 585 1694 E-mail: email@example.com
Cerebral cavernous malformations are common vascular lesions of the central nervous system that predispose to seizures, focal neurologic deficits and potentially fatal hemorrhagic stroke. Human genetic studies have identified three genes associated with the disease and biochemical studies of these proteins have identified interaction partners and possible signaling pathways. A variety of animal models of CCM have been described to help translate the cellular and biochemical insights into a better understanding of disease mechanism. In this minireview, we discuss the contributions of animal models to our growing understanding of the biology of cavernous malformations, including the elucidation of the cellular context of CCM protein actions and the in vivo confirmation of abnormal endothelial cell–cell interactions. Challenges and progress towards developing a faithful model of CCM biology are reviewed.
Cerebral cavernous malformation (CCM) is a common vascular disease consisting of clusters of dilated, thin-walled vessels lacking smooth muscle support and prone to hemorrhage. They are found in 1 in 200–250 individuals in the general population [1,2]. Although named for their predilection for the central nervous system (CNS), CCMs are also found in the retina, skin and other organs . CCMs can be sporadic or familial, with the familial form manifesting with earlier onset and a higher number of malformations. The familial form is linked to three genes, KRIT1 (KREV1/RAP1A interaction trapped-1 also known as CCM1) [4,5], CCM2 (also known as OSM or Osmosensing scaffold for MEKK3) [6–8] and PDCD10 (Programmed cell death 10, also known as CCM3) . The genetics of cavernous malformations is reviewed by Riant et al. .
The proteins encoded by these three genes are structurally unrelated and lack catalytic domains. Considerable progress has been made characterizing the interaction partners and the signaling pathways of the CCM proteins. The biochemistry of these pathways is reviewed by Faurobert & Albiges-Rizo . Although such basic mechanistic studies are necessary to come to a more complete understanding of the underlying cellular processes that lead to disease, these studies are difficult to interpret without the context of in vivo correlation. Furthermore, these studies have been performed in a variety of cell types, both primary cultures and established laboratory cell lines. An understanding of the relevant cell types in the formation of vascular malformations in CCM is needed to put these studies into a physiological context. Ultimately, the goal of all research into CCM is to understand the basic processes that have been disrupted, resulting in the vascular malformation. Between observational and genetic studies in humans and biochemical and cellular studies at the bench lies a gap. This minireview explores the contributions of animal models to bridge this gap and add to our growing understanding of CCM pathophysiology.
Conservation of CCM genes
The genes responsible for CCM are very well conserved among different organisms (Fig. 1). These genes are found not only in mammals, fish and other vertebrates, but are also found in much more simple and primitive organisms that lack a closed circulatory system, such as Caenorhabditis elegans. The presence of these genes in genetically tractable organisms has allowed the development of numerous experimental animal models, as discussed below.
Although humans are generally not considered in the category of animal models of disease, one can view the field of human genetics as probing a vast natural mutagenesis screen involving billions of individual organisms. As in any mutagenesis screen, the important information on genotype must be coupled with a detailed characterization of phenotype. All other animal models are relevant to disease to the degree that they help us further understand the human phenotype. Recent investigations have further refined our understanding of this phenotype, and bear reviewing in this manuscript.
In human CCM disease, the lesions exhibit a number of characteristic features; these features will serve as guideposts on the road to developing animal models of CCM disease. Classically, a CCM consists of a cluster of dilated blood vessels [12,13]. Each vessel in the cluster is grossly dilated, earning the name of a cavern; each vessel is lined only with a single layer of endothelium, with the absence of normal vascular support cells, such as smooth muscle cells. To be histologically classified as a CCM, the lesion must contain multiple such vessels adjacent to each other (Fig. 2). Grossly, this cluster gives the lesion an appearance likened to a raspberry. In addition, no brain parenchyma occurs in between the vessels. Single dilated vessels, called capillary telangiectasias, are not CCMs, although it has been hypothesized that the disease progresses from a single capillary telangiectasia that blossoms into a multivessel CCM . Functionally, the lesion vessels are subject to subclinical bleeding, because hemosiderin, a breakdown product of blood, is found in the brain tissues surrounding CCM lesions . Although CCMs have been clinically associated as occurring with developmental venous malformations , it has been shown that these two types of malformations are not linked genetically , and familial cases of CCM are not generally associated with venous malformations. Although these clinical features define CCMs for physicians, little is known about the cellular mechanisms that underlie and result in such characteristics. These mechanisms are what must be discovered, using either animal models or by deeper study of human CCM patients.
One aspect of disease discovered in humans is that CCMs are associated with an inflammatory response. CCM lesions harbor a variety of immune cells , and oligoclonal banding of IgG has been observed in the CCM tissue . What is still not known, however, is whether this inflammatory response is a secondary reaction to antigens exposed by the defective blood–brain barrier of a CCM  or if inflammatory action is part of the mechanism of pathogenesis leading to formation of aberrant blood vessels. It is intriguing that the mitogen-activated protein kinase kinase kinase MEKK3 plays a key role in immune signaling , and CCM2 protein has been shown to function as a scaffold for MEKK3 in response to stress . The finding of immune involvement in CCM illustrates the complexity of the disease; CCM is a vascular disease localized mainly to neural tissues with an additional immune component. The involvement of multiple cell and tissue types raise the question of where the CCM genes primarily function, and in which cell type their loss leads to pathogenesis of disease.
Aside from the question of tissue specificity of CCM gene function, another important question of disease pathogenesis is that of a triggering event – what events on a molecular, cellular or physiological level lead to the formation of these isolated malformations? A clue comes from studying patients with sporadic CCM and those with familial CCM. People with an inherited form of CCM have a larger number of lesions and more frequent sequellae, such as seizure and hemorrhage. These features are reminiscent of the cancer retinoblastoma, which led to the Knudson ‘two-hit’ hypothesis. Similarly, a two-hit hypothesis has been proposed for the pathogenesis of CCM, in which an inherited mutant allele is a silent, but predisposing hit, and a second mutation acquired during life leads to a disease phenotype. The data supporting this hypothesis have been reviewed by Riant et al.  in an accompanying minireview. In addition to genetic and epigenetic events leading to CCM, these studies do not explore physiologic stressors as potential disease triggers in the heterozygous patient. For example, serum levels of the angiogenic vascular endothelial growth factor have been correlated with disease progression in case reports [21,22].
Recent cell biology observations, supported by data from mice, call to mind an important observational study . Using detailed ultrastructural examination of surgically excised CCM specimens, the investigators observed abnormal endothelial cell junctions from the cavernous malformation. An important component of the normal blood–brain barrier, tight junctions form between endothelial cells and can be observed by electron microscopy. Although the cavernous malformation was found in the CNS where such tight junctions are the rule, the investigators observed numerous regions with impaired or deficient tight junctions between adjacent endothelial cells. These areas of junctional breakdown were associated with hemosiderin pigment as functional evidence that junction breakdown was associated with pathologic vascular leak, one of the defining features of CCMs.
Hailed for its transparency and genetic tractability, a significant body of work has been carried out in zebrafish to determine the functions of the CCM genes. Initial results were described for santa (san, the zebrafish orthologue of KRIT1) and valentine (vtn, the zebrafish orthologue of CCM2). Zebrafish with loss-of-function mutations in san or vtn share a common phenotype with fish lacking heart of glass (heg). Although mutations in the human orthologue of heart of glass (HEG1) have not been identified in patients with CCM, this gene has been shown to be functionally and genetically related to santa and valentine.
Heg is a single-pass transmembrane protein. Zebrafish with a nonsense heg mutation exhibit a dilated heart phenotype. The myocardium proliferates to a normal number of cells, but instead of building into concentric layers to form the walls of the heart, the myocardial cells form into a single layer, resulting in a dilated, thin-walled heart whose structure is reminiscent of a CCM vessel. Heg has two soluble splice variants in addition to the transmembrane isoform, but it is the transmembrane isoform that is essential in cardiac patterning. Although the defect is one of myocardial patterning, heg is expressed in the endocardial cells, indicating that this cell layer signals to the myocardium via Heg .
Interestingly, fish with nonsense mutations in san and vtn were later shown to exhibit the same phenotype as the heg mutant fish – that of the dilated heart covered by a single layer of myocardium. The similarity of the phenotype in these nonsense alleles suggested that these three proteins share a common developmental function. In addition, co-morpholino experiments demonstrated synergy among the three genes, putting them into a common genetic pathway . Another group refined the characterization of the santa and valentine phenotypes using different mutant alleles. Focusing on the vasculature instead of the heart, they found that these fish developed dilated, thin-walled vessels that failed to form lumens. The dilated, thin-walled, closed vessels, like the dilated, thin-walled heart of these fish, are very reminiscent of human CCM vessels and the closed vessels seen in CCM knockout mice (see below). This dilation was attributed to abnormal endothelial cell spreading, a potential mechanistic insight into CCM pathogenesis. Of note, these abnormal vessels were able to be rescued by the transplantation of endothelial cells from wild-type fish, again hinting that the endothelial cell is the cell type that most needs the function of the CCM proteins . Later work also showed that loss of heg or vtn via morpholino knockdown resulted in non-patent vessels that patterned normally, similar to the phenotypes seen in the Krit1 and Ccm2 knockout mice (see below) . Most recently, it has been shown that a deletion mutation of pdcd10 (ccm3), which is duplicated in the zebrafish genome, results in the same developmental defects as mutations in santa and valentine , making the zebrafish the first non-human model organism to link all three CCM genes phenotypically. Specifically, these defects are caused by the loss of Ccm3 interaction with the kinases serine/threonine kinase 25 (STK25) and mammalian sterile twenty-like 4 (MST4), giving a hint to the signaling pathway in which Ccm3 belongs, as both STK25 and MST4 belong to a family of kinases that are thought to act upstream of the mitogen-activated protein kinases (see the accompanying minireview  on the biochemical interactions of the CCM proteins).
Furthering the pursuit of genetic interactions, co-morpholino experiments were performed to examine the interactions between the CCM genes and rap1b, a Ras family small GTPase known to regulate cell junctions  and notable as being closely related to RAP1A, the binding partner bait originally used to identify KRIT1 . Knockdown of rap1b via morpholino resulted in defective endothelial cell junctions and intracerebral hemorrhage in the fish, reminiscent of both the slow, unpredictable blood leak and the frank hemorrhage associated with CCMs . The dose of rap1b morpholino was then titrated down so that the hemorrhage phenotype was seen in only a small percentage of fish. Combining this low dose of rap1b morpholino with a similarly low dose of san morpholino resulted in a synergistic increase in both the intracerebral hemorrhage phenotype of rap1b and the cardiac developmental phenotype of san.
The zebrafish experiments demonstrate the role of the CCM genes in cardiac and vascular development; the genetic tractability of the fish also provided a powerful way to discover genetic interactions between the CCM proteins and other proteins such as rap1b and the previously unknown heg. A mystery remains as to why HEG1 mutations are not found in humans with CCMs. The synergistic effects of low-dose knockdown of the CCM genes and their partners imply that a similar mechanism may be responsible for pathogenesis in humans; however, as previously stated, such polygenic effects have yet to be identified in human tissue samples.
Mice have long been favored as a model organism for laboratory studies and are the closest relative to humans commonly used in genetic studies. Knockout mice lacking Krit1  and Ccm2 [26,32–34] have been generated and described. Although an experimental model of CCM lesions in the CNS was desired, neither mice with heterozygous knockout of Krit1 nor Ccm2 develop CNS vascular lesions with any useful frequency [26,31–34]. Although disappointing, this lack of faithful disease modeling has generally been the case for most mouse genotype equivalents of human disease [35–38].
An important role for mouse models of genetic disease is to identify essential roles for protein function in vivo, especially in development where the proof of essential function is often embryonic lethality in complete knockouts. Indeed, mice lacking either Krit1 or Ccm2 die in mid-gestation with vascular defects at the same developmental stage, and with a similar phenotype [26,31,33,34]. The complete loss of Krit1 or Ccm2 results in vascular defects with a failure to connect the developing heart to the developing aorta with a functioning, patent first branchial arch artery. The associated rostral portions of the aorta are similarly narrowed (Fig. 3). As a result, circulation is not established as expected at E8.5 , and developmental arrest and death ensue. Prior to developmental arrest, cardiac and neural development proceeds normally.
Cavernous malformations are vascular lesions that form predominantly in the CNS. The basis for this anatomic predisposition is uncertain, but one possibility suggested by the abundant neuronal expression of the CCM genes [32,39–42] is a mechanism by which there is impaired signaling from neuronal cells to the endothelium, with a primary defect in the neuronal cell. Alternatively, the defect may lie primarily in endothelial cells, and the CNS selectivity of the disease could be a result of a unique sensitivity of the CNS vasculature to CCM gene function. To address these possibilities, mice with tissue-specific deletions of Ccm2 using the Cre–Lox inducible recombination system have been generated and described. Two separate floxed Ccm2 alleles were generated by different research groups [33,34]. Using Cre recombinase driven by the Tie2 promoter to direct recombination in endothelial cells, both groups found an absolute requirement for Ccm2 in the endothelium during development. The neuronal expression of Ccm2 was not required for development (as shown by deletion using the Nestin promoter-driven Cre recombinase).
Whereas Krit1, Ccm2 and Pdcd10 have similar widespread expression patterns in the mouse [32,39–42], the expression of the mouse orthologue of heart of glass (Heg1) is restricted to the endothelium and endocardium. Unlike zebrafish, Heg1 knockout mice do not phenocopy Krit1 or Ccm2 knockouts . Rather, Heg1 knockout mice die later in gestation or in early postnatal stages with a variety of cardiac, vascular and lymphatic defects. Although pulmonary hemorrhage, cardiac rupture or chylous effusions may variably be the mechanism of death, a common theme throughout was disruption of the cell–cell junctions within the endothelial or endocardial cells. Heg1 and Ccm2 were also shown to genetically interact in the mouse as previously seen in fish . Mice with both homozygous knockout of Heg1 and heterozygous for Ccm2 were found to have a much more severe phenotype than either mutant in isolation. These dual knockouts phenocopy mice with homozygous knockout of Ccm2 or Krit1.
Multiple lines of investigation implicate a role for impaired cell-to-cell communication and endothelial cell junction integrity in states of CCM protein deficiency. Endothelial cell tight junctions are required to retain cells and macromolecules within the vasculature and to prevent vascular leak. Although mice heterozygous for Ccm2 do not frequently develop CCM lesions like their human counterparts, these mice were shown to have abnormal vascular leak in the dermis when stressed with vascular endothelial growth factor . Tight junctions are significantly regulated by the Rho family of GTPases. Endothelial cell culture experiments had implicated abnormally increased activity of RhoA in vitro. A role for increased RhoA activity in vivo was suggested by the ability of statins – known inhibitors of Rho GTPases  – to rescue the abnormal vascular leak of Ccm2 heterozygous knockout mice .
It is not clear what is responsible for the difference in susceptibility between mice and humans for the cerebral vascular lesions. Although differences in lifespan and brain mass may contribute to the lack of vascular lesions in Krit1 or Ccm2 heterozygous knockout mice, modifying factors are being sought which increase the risk of CCM lesion formation. As discussed above, observational studies in CCM patients suggest that a ‘two-hit hypothesis’ may underlie some lesions. Taking advantage of the high rate of spontaneous mutations in mice lacking the tumor suppressor p53, Krit1 heterozygous knockouts have been mated onto a p53 knockout background . It was hoped that this model would reproduce both the human genotype (two genetic hits) and phenotype (cavernous malformations). Cerebral vascular lesions were observed in a high proportion of animals on this background with characteristics varying from capillary telangiectasias to more complex cavernous malformations, but the potential second hit mutation was not found. Unfortunately, mice lacking p53 have a shortened lifespan because of a high frequency of spontaneous tumors, including occasional brain tumors [13,45], It is unfeasible to study the natural history of CCM disease in these mice as they die from tumor burden shortly after developing CCM lesions. Great caution must be taken when interpreting the genetic contribution to vascular lesions on this background with potential for cancer-related vascular dysregulation and other physiologic stressors that may contribute to CCM lesion development. These results suggest, however, that a two-hit model may produce malformations useful for study; this second hit could come from recombination mediated by an inducible promoter driving Cre recombinase, thereby eliminating the confounding effects of the p53 null background.
Vascular malformations result in considerable morbidity and mortality, especially with respect to lesions of the central nervous system. The ability to prevent or treat such lesions requires a greater understanding of the underlying biology of lesion formation. In this regard, cerebral cavernous malformation as a genetic disorder offers unique opportunities to understand the biology of vascular malformations. Initial insights regarding the biochemistry of the CCM genes left a considerable gap in understanding between protein function and lesion biology. By exploring the function of the CCM genes in animal models this gap is being bridged. Animal models have demonstrated the central importance of endothelial cell–cell interactions in the pathogenesis of CCM vascular disease. Endothelial cells need to be coordinated to organize into proper sized lumens and to maintain vascular barrier function. As a result of research into CCM, it becomes apparent that vascular malformations may result from the loss of genes crucial to vascular stability.
In addition to providing the important in vivo context for insights gained from biochemistry, animal models can allow an acceleration of translational research to ultimately impact the patients and families with CCM. Recent work in mice shows the promise of this approach, in that testable phenotypes can be identified and potential therapies can be evaluated in mice genetically similar to CCM patients. Manipulations of the current animal models to more closely mimic human disease also appear promising. Ultimately, we hope that a complete model of CCM lesion biology can be developed to act as a vital link between bench and bedside.
This work was funded by the US National Institutes of Health (K.J.W. and D.Y.L.), including training grant T32-GM007464 (A.C.C.), the American Heart Association (K.J.W. and D.Y.L.), the H.A. and Edna Benning Foundation, the Juvenile Diabetes Research Foundation, the Burroughs Wellcome Fund and the Flight Attendants Medical Research Institute (D.Y.L.).