The angiogenic sprout has been compared to the growing axon, and indeed, many proteins direct pathfinding by both structures 1 . The Roundabout (Robo) proteins are guidance receptors with well-established functions in the nervous system2 , 3; however, their role in the mammalian

Cerebral cavernous malformation (CCM), a disease associated with defective endothelial junctions, result from autosomal dominant CCM1 mutations that cause loss of KRIT-1 protein function, though how the loss of KRIT-1 leads to CCM is obscure. KRIT-1 binds to Rap1, a guanosine triphosphatase that maintains the integrity of endothelial junctions. Here, we report that KRIT-1 protein is expressed in cultured arterial and venous endothelial cells and is present in cell–cell junctions. KRIT-1 colocalized and was physically associated with junctional proteins via its band 4.1/ezrin/radixin/moesin (FERM) domain. Rap1 activity regulated the junctional localization of KRIT-1 and its physical association with junction proteins. However, the association of the isolated KRIT-1 FERM domain was independent of Rap1. Small interfering RNA–mediated depletion of KRIT-1 blocked the ability of Rap1 to stabilize endothelial junctions associated with increased actin stress fibers. Thus, Rap1 increases KRIT-1 targeting to endothelial cell–cell junctions where it suppresses stress fibers and stabilizes junctional integrity.

Endothelial cell–cell junctions regulate vascular permeability, vasculogenesis, and angiogenesis. Familial cerebral cavernous malformations (CCMs) in humans result from mutations of CCM2 (malcavernin, OSM, MGC4607), PDCD10 (CCM3), or KRIT1 (CCM1), a Rap1 effector which stabilizes endothelial cell–cell junctions. Homozygous loss of KRIT1 or CCM2 produces lethal vascular phenotypes in mice and zebrafish. We report that the physical interaction of KRIT1 and CCM2 proteins is required for endothelial cell–cell junctional localization, and lack of either protein destabilizes barrier function by sustaining activity of RhoA and its effector Rho kinase (ROCK). Protein haploinsufficient Krit1+/− or Ccm2+/− mouse endothelial cells manifested increased monolayer permeability in vitro, and both Krit1+/− and Ccm2+/− mice exhibited increased vascular leak in vivo, reversible by fasudil, a ROCK inhibitor. Furthermore, we show that ROCK hyperactivity occurs in sporadic and familial human CCM endothelium as judged by increased phosphorylation of myosin light chain. These data establish that KRIT1–CCM2 interaction regulates vascular barrier function by suppressing Rho/ROCK signaling and that this pathway is dysregulated in human CCM endothelium, and they suggest that fasudil could ameliorate both CCM disease and vascular leak.

Cerebral cavernous malformations (CCMs) are vascular lesions of the central nervous system appearing as multicavernous, blood-filled capillaries, leading to headache, seizure and hemorrhagic stroke. CCM occurs either sporadically or as an autosomal dominant disorder caused by germline mutation of one of the three genes: CCM1/KRIT1, CCM2/MGC4607 and CCM3/PDCD10. Surgically resected human CCM lesions have provided molecular and immunohistochemical evidence for a two-hit (germline plus somatic) mutation mechanism. In contrast to the equivalent human genotype, mice heterozygous for a Ccm1- or Ccm2-null allele do not develop CCM lesions. Based on the two-hit hypothesis, we attempted to improve the penetrance of the model by crossing Ccm1 and Ccm2 heterozygotes into a mismatch repair-deficient Msh2(-/-) background. Ccm1(+/-)Msh2(-/-) mice exhibit CCM lesions with high penetrance as shown by magnetic resonance imaging and histology. Significantly, the CCM lesions range in size from early-stage, isolated caverns to large, multicavernous lesions. A subset of endothelial cells within the CCM lesions revealed somatic loss of CCM protein staining, supporting the two-hit mutation mechanism. The late-stage CCM lesions displayed many of the characteristics of human CCM lesions, including hemosiderin deposits, immune cell infiltration, increased endothelial cell proliferation and increased Rho-kinase activity. Some of these characteristics were also seen, but to a lesser extent, in early-stage lesions. Tight junctions were maintained between CCM lesion endothelial cells, but gaps were evident between endothelial cells and basement membrane was defective. In contrast, the Ccm2(+/-)Msh2(-/-) mice lacked cerebrovascular lesions. The CCM1 mouse model provides an in vivo tool to investigate CCM pathogenesis and new therapies.

Endothelial cells lining the vasculature have close cellcell associations that maintain separation of the blood fluid compartment from surrounding tissues. Permeability is regulated by a variety of growth factors and cytokines and plays a role in numerous physiological and pathological processes. We examined a potential role for the p21-activated kinase (PAK) in the regulation of vascular permeability. In both bovine aortic and human umbilical vein endothelial cells, PAK is phosphorylated on Ser 141 during the activation downstream of Rac, and the phosphorylated subfraction translocates to endothelial cell-cell junctions in response to serum, VEGF, bFGF, TNF␣, histamine, and thrombin. Blocking PAK activation or translocation prevents the increase in permeability across the cell monolayer in response to these factors. Permeability correlates with myosin phosphorylation, formation of actin stress fibers, and the appearance of paracellular pores. Inhibition of myosin phosphorylation blocks the increase in permeability. These data suggest that PAK is a central regulator of endothelial permeability induced by multiple growth factors and cytokines via an effect on cell contractility. PAK may therefore be a suitable drug target for the treatment of pathological conditions where vascular leak is a contributing factor, such as ischemia and inflammation.