Cortactin deficiency is associated with reduced neutrophil recruitment but increased vascular permeability in vivo

To investigate the physiological role of cortactin, we used a conditional gene ablation approach by flanking exon 7 with loxP sites. The targeting vector was described previously (Lai et al., 2009), and the gene targeting strategy is outlined in Fig. 1 A. Consecutive crosses with mice systemically expressing Flp (to remove the neo-cassette) and Cre recombinase were performed to produce the null mutation on a mixed genetic background. Mice homozygous for the null mutation were successfully obtained by crossing mice with genotype del/WT. Deletion of exon 7 in the cortactin gene by Cre recombinase expressed in oocytes was expected to stop the reading frame generating a null allele in all cells of the organism. PCR genotyping of tail DNA samples revealed successful deletion of exon 7 (Fig. 1 B), and immunoblotting confirmed the absence of cortactin protein in homozygous del/del animals (KO) in any organ analyzed, as expected (Fig. 1 C). Using fibroblasts derived from double-targeted embryonic stem (ES) cells, in which deletion of exon 7 was induced in a comparable fashion in vitro, RT-PCR with primer pairs corresponding to regions of the 5′ end and the 3′ end of the cortactin message did not reveal any detectable cortactin transcript, as shown by Lai et al. (2009).

In a specific pathogen-free environment, adult Cttn^WT/del mice in a mixed genetic background were backcrossed with C57BL/6 mice. Cttn^del/del mice of the sixth backcross were used in our experiments. Cttn^del/del mice of mixed genetic background and of the sixth backcross showed no obvious phenotype, were healthy by visual inspection, and had a normal life span. Thus, cortactin is not essential for normal development, and its absence does not cause any obvious defects under specific, pathogen-free conditions.

Breeding Cttn^WT/del mice on a mixed 129Sv/C57BL/6 background generated homozygous Cttn^del/del mice with normal Mendelian frequency (Table S1), although we found a significant preference toward female offspring. When heterozygous mice of the sixth backcross (C57BL/6) were mated, Cttn^del/del mice were clearly underrepresented among the offspring (Table S2). The reason for this is currently unknown.

Based on RNA silencing in cultured endothelial cells, it has been suggested that cortactin is required for the enhancement of endothelial cell contact integrity by S1P or ATP; however, basal barrier integrity was unaffected (Dudek et al., 2004; Jacobson et al., 2006). This prompted us to use the Miles assay to test whether cortactin gene deficiency would affect vascular permeability in the skin. To this end, we injected WT and Cttn^del/del mice i.v. with Evans blue followed by intradermal injection of either histamine or PBS 15 min later. Remarkably, in the absence of any permeability-stimulating factor, the permeability in Cttn^del/del mice was more than twofold higher than in WT mice (Fig. 2 A), clearly demonstrating that cortactin controls basal permeability. Stimulation with histamine resulted in an additional increase of permeability in Cttn^del/del mice as compared with WT mice (Fig. 2 A). We conclude that cortactin is required in vivo to maintain basal barrier integrity within blood vessels.

To test whether the composition of membrane proteins at endothelial cell contacts would be disturbed in cortactin-deficient mice, we first analyzed the overall expression levels of vascular endothelial cadherin (VE-cadherin), endothelial cell–selective adhesion molecule (ESAM), JAM-A (junctional adhesion molecule A), claudin-5, and also ICAM-1 in Western blots of the lung. As shown in Fig. 2 B, no difference was found in comparison with WT mice. Likewise, junctional localization of VE-cadherin and ESAM was unaffected by cortactin gene disruption, as shown by whole mount stainings of venules in the cremaster (Fig. 2 C). Relative pixel intensity in venules of the KO tissue was 107.9 ± 12.7% for ESAM and 102.9 ± 8.7% for VE-cadherin with pixel intensity for the WT values set to 100%. Isolation of primary endothelial cells from lungs of Cttn^del/del mice and Cttn^WT/WT mice (murine lung endothelial cells [MLECs]) allowed us to compare constitutive cell surface expression levels of VE-cadherin, ESAM, JAM-A, and PECAM-1 and expression of TNF-induced ICAM-1, VCAM-1 (vascular cell adhesion molecule 1), E-selectin, and P-selectin by flow cytometry. Each of these adhesion molecules was expressed at similar levels on both cell types (Fig. S1).

To elucidate the mechanism whereby cortactin supports cell contact integrity, we established siRNA-mediated down-regulation of the expression of cortactin in cultured human and mouse endothelial cells. Cortactin expression could be inhibited by 89% in human umbilical vein endothelial cells (HUVECs; Fig. 3 A) and by 81% in bend.3 mouse endothelioma cells (Fig. S4). Suppression of cortactin expression was accompanied by a 29.8% (±8.6%) increase of paracellular flux of 250-kD FITC-dextran across the monolayer of HUVECs grown on transwell filters (Fig. 3 C). Similar results were obtained for bend.3 mouse endothelioma cells (Fig. S4). Confluent monolayers of Cttn^del/del MLECs had 68% (±23%) higher permeability than Cttn^WT/WT MLECs for 250 kD FITC-dextran and 19% (±4%) enhanced permeability for 150-kD FITC-dextran (Fig. 3 E). In line with this, we found a 53% (±6.3%) reduction of transendothelial resistance of HUVEC monolayers (Fig. 3 D), confirming our in vivo finding that cortactin is indeed required for maintaining endothelial cell contact integrity.

Because it was shown before that activation of the Epac (exchange protein directly activated by cAMP)/Rap-1 pathway leads to stabilization of endothelial cell contacts (Cullere et al., 2005; Kooistra et al., 2005), we tested whether down-regulation of the expression of cortactin affects the levels of GTP-loaded Rap-1. Indeed, siRNA-mediated silencing of cortactin reduced the levels of activated Rap-1 by 41.7% (±9.4%) in HUVECs (Fig. 4 A) and by 74% (±17%) in bend.3 cells (Fig. S4). Likewise, levels of activated Rap1 were reduced by 51% (±6.7%)% in Cttn^del/del MLECs compared with Cttn^WT/WT MLECs (Fig. 4 A). Thus, cortactin is required to maintain the steady-state levels of GTP loading of Rap1.

To test whether the reduction of Rap1 activation levels could explain why the loss of cortactin destabilizes endothelial cell contacts, we investigated whether activation of Rap1 reverses the cortactin knockdown effect on endothelial permeability. To this end, we stimulated HUVECs with the cAMP analogue 8-pCPT-2Me-cAMP (8-(4-Chlorophenylthio)-2’-O-methyl-cAMP; or 007) known to specifically activate the GTPase exchange factor Epac and thereby Rap1. As shown in Fig. 4 C, 8-pCPT-2Me-cAMP treatment of HUVECs transfected with cortactin-specific siRNA completely compensated the increase of FITC-dextran permeability caused by cortactin siRNA. In fact, the cAMP analogue reduced permeability to the same level in cortactin knockdown and control HUVECs. Similar results were obtained for bend.3 mouse endothelioma cells (Fig. S4) and for primary Cttn^del/del MLECs (Fig. 4 D and Fig. S5). This reveals that direct stimulation of Epac/Rap1 can stabilize cell contacts, and this is independent of cortactin function. In agreement with this, direct stimulation of Epac/Rap1 with 007 was not affected by the loss of cortactin in HUVECs or MLECs (Fig. 4 B). Because the loss of cortactin in turn leads to reduced activation levels of Rap1, our results position Epac/Rap1 downstream of cortactin.

The impairment of the vascular barrier function in cortactin-deficient mice prompted us to test whether this would also affect the extravasation of leukocytes. Analyzing TNF-inflamed cremaster tissue by intravital microscopy, we found that despite the loss of barrier integrity, neutrophil extravasation was not enhanced. Instead, it was strongly reduced by 50.8% (±22.8%; Fig. 5 D). Analyzing neutrophil rolling and firm adhesion, we found that the defect in extravasation was caused by impaired interactions of neutrophils with the vessel wall. Neutrophil rolling velocity was increased by 69.5% (±26.7%) in Cttn^del/del mice when compared with WT mice (Fig. 5 A), and the rolling flux fraction was increased by 105% (±27.6%; Fig. 5 B). In line with these results, the number of neutrophils that firmly adhered to the endothelium was reduced by 40.6% (±23.4%) in Cttn^del/del mice compared with littermate controls (Fig. 5 C). Hemodynamic parameters of the analyzed vessels are given in Table I, and representative videos are shown in the supplemental material (Videos 1 and 2). We conclude that the slowing down of rolling neutrophils was less efficient in Cttn^del/del mice leading to faster rolling cells and a defect in neutrophil arrest. Because of this defect, cells continued to roll instead of stopping as firmly adhering cells, leading to increased numbers of rolling cells at the expense of firmly adhering cells. Thus, cortactin is required for optimal neutrophil extravasation efficiency at a step upstream of the actual transmigration process.

To determine whether cortactin acts in endothelial cells or in neutrophils during the extravasation process, we first analyzed whether it is expressed in neutrophils. In agreement with a previous study that reported the absence of cortactin in all bone marrow–derived cells except megakaryocytes (Zhan et al., 1997), we could not detect cortactin in immunoblots of mouse leukocyte lysates (Fig. S2). However, this did not rule out the possibility that low expression levels of cortactin below detection levels might be sufficient to support functions relevant for leukocyte extravasation. To test this, we generated chimeric mice by irradiating WT mice and transplanting them with bone marrow cells from either Cttn^del/del mice or (as controls) from Cttn^WT/WT littermates. Intravital microscopy of the TNF-inflamed cremaster of these mice revealed no differences in rolling velocity, rolling flux fraction, or in the number of adherent or extravasated neutrophils when compared with WT mice transplanted with WT bone marrow (Fig. 5, E–H). These results formally rule out the possibility that cortactin potentially expressed below detection limit in neutrophils participates in their extravasation. In addition, these experiments exclude the possibility that platelets, which do express cortactin (Vidal et al., 2002) and are known to be important for neutrophil recruitment into inflamed tissue (Bixel et al., 2010), would require cortactin for their participation in neutrophil extravasation.

Importantly, cortactin gene disruption did neither reduce peripheral leukocyte counts nor cause any changes in the repertoire of leukocyte surface molecules relevant for leukocyte extravasation. Indeed, we did find a slight tendency toward higher numbers of neutrophils and lymphocytes in the blood of Cttn^del/del animals compared with WT controls; however, the differences were not statistically significant (Fig. S2). Furthermore, neither platelet nor erythrocyte counts were significantly affected by cortactin gene disruption. In addition, surface expression of L-selectin, PSGL-1 (P-selectin glycoprotein ligand 1), Gr-1, CD11a, CD11b, CD44, and CXCR2 was not affected by the systemic loss of cortactin (Fig. S2). Collectively, our results suggest that it is the lack of endothelial cortactin that impairs leukocyte extravasation in cortactin-deficient mice.

Leukocyte rolling is initiated by the endothelial selectins and slowed down by the binding of the β₂-integrins LFA-1 (lymphocyte function–associated antigen 1) and Mac-1 (macrophage 1 antigen) to their endothelial ligands (Kunkel et al., 2000; Dunne et al., 2002). Thus, the enhanced leukocyte rolling velocity in cortactin-deficient mice could have either been caused by impaired selectin-mediated rolling or by the inability of endothelial β₂-integrin ligands to slow down leukocyte rolling. To test the second possibility, we determined whether cortactin deficiency would still affect the rolling velocity when LFA-1 and Mac-1 are simultaneously blocked with antibodies. As shown in Fig. 5 I, we found that blocking LFA-1 and Mac-1 resulted in the expected strong increase in rolling velocity in Cttn^WT/WT mice. Importantly, these antibodies had no further effect on the rolling velocity in Cttn^del/del mice compared with Cttn^WT/WT mice (Fig. 5 I). This demonstrates that the lack of cortactin did no longer enhance the rolling velocity if only the endothelial selectins but not the β₂-integrins were available to support rolling. This suggests indirectly that cortactin supports neutrophil rolling via β₂-integrin ligands and not via endothelial selectins.

Our intravital microscopy results (Fig. 5) suggest that the defect in neutrophil extravasation in cortactin-deficient mice is caused by impaired neutrophil rolling and firm adhesion. To investigate how cortactin could affect transmigration, we decided to test whether we can recapitulate this effect in vitro. To exclude the rolling aspect, we performed static transmigration assays with endothelial cell monolayers grown on transwell filters. In agreement with the in vivo results, cortactin siRNA inhibited transmigration of human neutrophils through HUVEC monolayers by 36% (Fig. 6 A) and of mouse neutrophils through bend.5 cell layers by 40% (Fig. S4). Likewise, transmigration of mouse neutrophils through monolayers of primary MLECs isolated from Cttn^del/del mice was reduced by 49% in comparison with MLECs from Cttn^WT/WT mice (Fig. 6 B). Neutrophils adhered to Cttn^del/del MLECs with similar efficiency (51.8 ± 4.1% of applied PMNs) as to Cttn^WT/WT MLECs (52.9 ± 6.8% of applied PMNs), possibly reflecting the fact that assays were performed under static conditions.

Because ICAM-1 is the major endothelial adhesion receptor relevant for firm neutrophil binding, we tested whether cortactin would influence neutrophil transmigration via ICAM-1. If this was the case, knockdown of cortactin expression would have no inhibitory effect on neutrophil transmigration in endothelial cells lacking ICAM-1. To test this hypothesis, we used mouse endothelioma cells deficient for ICAM-1 and, as a control, the same cells retransfected for ICAM-1. Expression levels of cortactin in these endothelioma cells were similar to bend.3 or bend.5 cells (not depicted), and siRNA transfection reduced cortactin expression by 87% in ICAM-1–deficient cells and by 83% in ICAM-1–retransfected cells (Fig. 6 D).

As shown in Fig. 6 E, ICAM-1 deficiency reduced neutrophil transmigration by 70% (±25.8%). Inhibition of cortactin expression in these ICAM-1–deficient endothelial cells did not inhibit neutrophil transmigration any further, whereas in ICAM-1 retransfectants, cortactin siRNA inhibited neutrophil transmigration by 31.2 ± 8.4%. Thus, in the absence of ICAM-1, cortactin does not contribute to neutrophil transendothelial migration, suggesting that cortactin acts by supporting ICAM-1 functions.

Neutrophil transmigration through endothelial cell layers is accompanied by the clustering of ICAM-1 around adhering leukocytes, which strengthens cell adhesion and thereby facilitates the transmigration process (Barreiro et al., 2002; Carman et al., 2003). Therefore, we tested whether ICAM-1 clustering would be dependent on the expression of cortactin.

We observed in control-transfected HUVECs ICAM-1–enriched ring-like structures surrounding adherent neutrophils. Cortactin was also enriched in these structures, partly colocalizing with ICAM-1 (Fig. 7 A). Importantly, after down-regulation of cortactin expression by siRNA, only very few of such structures were still observed, and the vast majority of neutrophils were devoid of these rings (Fig. 7 A). We quantified the effect on the formation of ICAM-1–containing rings at sites of neutrophil contact (Fig. 7 B). More than 79% of PMNs on cortactin-positive cells showed closed ICAM-1–positive ring structures, whereas 21% of adherent cells on cortactin-positive HUVECs were not surrounded by such structures. The situation was completely different for PMNs in contact with cortactin-depleted HUVECs. Here, only 25% showed ICAM-1–positive ring structures, and 75% of these cells showed no ICAM-1 rings. These results demonstrate that the presence of cortactin is required for proper formation of ICAM-1–enriched clusters around adhering PMNs, thus allowing for correctly regulated neutrophil recruitment.

To test whether the lack of cortactin would impair neutrophil transmigration via the reduced level of activated Rap1, we tested whether the specific activator 8-pCPT-2Me-cAMP (007) of Epac/Rap1 would affect transmigration of human PMNs through HUVEC monolayers either treated with control siRNA or with cortactin siRNA. As shown in Fig. 6 F, 007 had no effect on neutrophil transmigration, suggesting that Rap1 is not involved in mechanisms whereby cortactin supports neutrophil extravasation.

RhoG has recently been reported to be required for the formation of leukocyte-surrounding ICAM-1 clusters and leukocyte transmigration (van Buul et al., 2007). The activation of RhoG was found to be triggered by leukocyte engagement of ICAM-1. This prompted us to test whether RhoG activation would be compromised in Cttn^del/del MLECs. Therefore, we incubated primary isolated MLECs of Cttn^WT/WT and Cttn^del/del mice with beads either loaded with antibodies against mouse ICAM-1 or with control antibodies and performed RhoG activation assays using the effector protein ELMO (engulfment and cell motility gene) in pull-down experiments. As shown in Fig. 8, engagement of ICAM-1 triggered activation of RhoG in Cttn^WT/WT but not in Cttn^del/del MLECs. Thus, cortactin is required for ICAM-1–triggered activation of RhoG, a signaling process involved in the transmigration of neutrophils.

To determine whether the defect in ICAM-1–triggered RhoG activation in Cttn^del/del MLECs would be responsible for impaired neutrophil transmigration, we tested whether transfection of CA-RhoG would rescue the phenotype. We found that CA-RhoG but not WT RhoG indeed partially restored the formation of ICAM-1–containing rings around neutrophils adhering to Cttn^del/del MLECs (Fig. 8 B). In addition, transmigration efficiency of neutrophils across a monolayer of these cells was largely restored with CA-RhoG (Fig. 8 C). We conclude that the defect in ICAM-1–triggered activation of RhoG is responsible for impaired ICAM-1 clustering and neutrophil transmigration in cortactin-deficient endothelial cells.

Analyzing cortactin-deficient mice, we show for the first time that this actin-remodeling protein is required in endothelial cells, not in neutrophils, for optimal efficiency of neutrophil recruitment to inflamed tissue. Loss of cortactin caused an increase in rolling velocity and a decrease of neutrophil adhesion to the vessel wall. Thus, endothelial cortactin is required for slowing down rolling neutrophils and for firm adhesion to the vessel wall. Interestingly, cortactin was not required for the support of selectin-mediated rolling but for the slowing down of rolling mediated by β₂-integrins and their ligands. In addition, formation of ICAM-1–containing clusters was disturbed, providing a possible explanation for the decreased numbers of neutrophils adhering to the vessel wall. Searching for a mechanism for this defect, we found that cortactin deficiency blocked the activation of RhoG triggered by ICAM-1 engagement, and CA-RhoG partially restored ICAM-1 clustering and neutrophil transmigration across cortactin-deficient endothelial cells. Independent of the role of cortactin in neutrophil extravasation, we found that cortactin is essential to maintain endothelial cell contact integrity. Gene deficiency led to enhanced basal permeability for high molecular weight substances. This effect could be rescued by directly activating Epac/Rap1 with a stable cAMP derivative, suggesting that cortactin acts upstream of Rap1. Thus, cortactin affects leukocyte extravasation and endothelial cell contact integrity by controlling the activity of two different GTPases in two independent mechanisms.

Neutrophil capturing, rolling, and adhesion represent consecutive steps of a complex process that requires the interplay of selectins and neutrophil integrins. It is well documented that P- and E-selectin mediate capturing and rolling of neutrophils (Vestweber and Blanks, 1999). Besides directly mediating rolling via binding to various neutrophil selectin ligands (McEver, 2002; Hidalgo et al., 2007), the endothelial selectins intensify the rolling interactions and thereby slow down rolling velocity by activating the neutrophil integrins LFA-1 and Mac-1, which participate in this process via binding to ICAM-1 and additional endothelial ligands (Kunkel et al., 2000; Dunne et al., 2002; Wang et al., 2007; Zarbock et al., 2007, 2008; Miner et al., 2008). Slowing down neutrophil rolling has so far only been studied as a neutrophil process requiring signaling that triggers integrin activation (Zarbock and Ley, 2009). We show here that also endothelial adhesion molecules require cytosolic components that are necessary to enable them to support neutrophil rolling. With cortactin, we have identified the first intracellular component in endothelial cells that is required in vivo for the slowing down of rolling neutrophils.

In vitro studies have suggested that E-selectin requires cytoskeletal components for its function. Various actin-associated proteins were copurified with E-selectin upon interaction with neutrophils (Yoshida et al., 1996), one of them being cortactin (Tilghman and Hoover, 2002). Although one study suggested that the cytoplasmic tail of E-selectin would be dispensable for the adhesive function of E-selectin (Kansas and Pavalko, 1996), another showed that clustering of E-selectin in either clathrin-coated pits or lipid rafts enhanced neutrophil adhesion under flow (Setiadi and McEver, 2008). Collectively, these results suggest that the cytoskeleton can influence E-selectin function. In this context, it is remarkable that cortactin deficiency, although clearly affecting rolling velocity in vivo, did not affect rolling when the function of β₂-integrins was blocked. Thus, cortactin only contributes to the rolling velocity if the integrins on PMNs are functional. This is indirect evidence suggesting that cortactin is required for the transition of fast to slow neutrophil rolling and this process depends on the adhesive function of β₂-integrin ligands. Of the three β₂-integrin ligands ICAM-1, ICAM-2, and JAM-A, only ICAM-1 has a clearly detectable although only small role in slow rolling (Ley et al., 1998; Steeber et al., 1998), whereas ICAM-2 seems not to serve as a rolling ligand of LFA-1 (Dunne et al., 2003) and JAM-A does not appear to play a role in adhesion under flow (Ostermann et al., 2002). It will be interesting to study in the future how cortactin participates in the incorporation of ICAM-1 and possibly other β₂-integrin ligands into the rolling process. Possibilities range from the support of ligand miniclusters or even selectin/integrin ligand heteroclusters to the organization of a cortical actin network that may provide a platform to stabilize forces that pull on ICAM-1 and/or other β₂-integrin ligands.

In addition to the contribution to neutrophil rolling, our results reveal that cortactin is also needed in vivo for firm adhesion of neutrophils in inflamed venules and for clustering of ICAM-1, which supports firm adhesion of leukocytes to endothelial cells and subsequent transmigration (Barreiro et al., 2002; Carman et al., 2003; Carman and Springer, 2004). Other cytoskeletal adaptors that are important for the formation of such structures are the ERM (ezrin, radixin, and moesin) proteins (Barreiro et al., 2002) and filamin B, which was found to be involved in ICAM-1 recruitment to adhering leukocytes (Kanters et al., 2008). As a mechanism whereby cortactin deficiency impairs the formation of ICAM clusters and neutrophil transmigration, we found that cortactin is essential for the activation of RhoG, triggered by the engagement of ICAM-1 on the surface of endothelial cells. ICAM-1–triggered RhoG activation was recently demonstrated to be critical for the formation of ICAM-1 clusters and also for neutrophil transmigration (van Buul et al., 2007). Interestingly, neither the depletion of RhoG via siRNA (van Buul et al., 2007) nor cortactin deficiency (this study) impaired neutrophil adhesion to endothelial cells in static adhesion assays. In the light of our in vivo results, this suggests that ICAM-1–containing clusters affect neutrophil arrest only under in vivo flow conditions.

Our results are in agreement with studies by Yang et al. (2006a,b) who showed that cortactin supports the clustering of ICAM-1 into ring-like structures around neutrophils attached to HUVECs. However, although they found that silencing of cortactin inhibited neutrophil transmigration through endothelial cell layers under flow conditions, they found no effect on neutrophil adhesion to the apical surface of endothelial cells. Rolling velocity was not reported in these studies, and the fact that no inhibitory effects on neutrophil attachment were detectable might be caused by differences between in vitro and in vivo conditions, or not least to the fact that siRNA cannot block expression as completely as gene deficiency. Interestingly, however, the study by Yang et al. (2006b) and also our in vitro results may hint toward an additional function of cortactin downstream of neutrophil rolling and adhesion, which could be a role in the diapedesis process itself or in the initiation of this process. In the light of these results, it is important to note that ICAM-1 cross-linking triggers tyrosine phosphorylation of cortactin. This suggests that cortactin is involved in ICAM-1 signaling (Durieu-Trautmann et al., 1994), which is in good agreement with our finding that cortactin is required for ICAM-1–triggered activation of RhoG. ICAM-1 clustering triggers various signals that may be involved in the destabilization of endothelial junctions (Adamson et al., 1999; Etienne-Manneville et al., 2000; Allingham et al., 2007; Turowski et al., 2008). Whether cortactin participates in neutrophil diapedesis in vivo can currently not be decided because of its prominent role in neutrophil rolling and firm adhesion.

Our in vivo analysis of vascular permeability and also our in vitro results showed that cortactin is required to maintain the barrier function of the vasculature. Whereas our in vitro results clearly show that it is endothelial cortactin that is required for endothelial cell contact integrity, it is also possible that in vivo cortactin in perivascular cells may additionally contribute to the integrity of the vessel wall. Previous in vitro studies suggested that cortactin is required for signaling mechanisms that enhance the stability of endothelial cell contacts. Treatment of pulmonary endothelial cells with either S1P or ATP led to barrier enhancement via junctional translocation of cortactin accompanied by translocation and activation of Rac1 and MLCK (myosin light chain kinase; Dudek et al., 2004; Jacobson et al., 2006). However, silencing of cortactin expression in these studies blocked S1P or ATP-stimulated enhancement of the barrier function but had no effect on the baseline permeability of cultured pulmonary endothelial cells. Thus, a requirement of cortactin for the maintenance of baseline barrier integrity as shown in our cortactin-deficient mouse model has not yet been described. Our findings that cortactin expression supports Rap1 activation and that stimulation of Epac/Rap1 with 8-pCPT-2Me-cAMP can rescue endothelial cell contact integrity in cells lacking cortactin suggests that Rap1 acts downstream of cortactin in the control of endothelial cell contact integrity. In contrast, it has been reported that activation of Epac/Rap1 by 8-pCPT-2Me-cAMP results in enhanced barrier integrity through improved VE-cadherin–dependent cell adhesion and increased cortical actin formation, which was accompanied by translocation of cortactin to cell junctions (Fukuhara et al., 2005; Kooistra et al., 2005). In addition, Epac/Rap1 activation was suggested to lead to Rac1 activation and translocation to cell junctions, which was accompanied by cortactin recruitment to endothelial junctions (Baumer et al., 2008). Thus, a positive feedback loop may exist where cortactin supports Rap1 activation and at the same time activated Rap1 stimulates the recruitment of more cortactin to cell contacts.

In numerous in vitro studies, cortactin has been implicated to be involved in many important cellular processes that require actin remodeling (e.g., Selbach and Backert, 2005; Ammer and Weed, 2008; Ren et al., 2009). Through physical interactions with actin filaments and the Arp2/3 complex, cortactin is thought to promote Arp2/3-dependent actin filament nucleation. However, whether the vascular phenotype of cortactin-deficient mice we describe here is indeed caused by a lack of a direct impact of cortactin on the actin cytoskeleton remains to be determined.

Recent studies investigating ES cell–derived fibroblasts harboring a null mutation analogous to the one used here or primary cells derived from independently generated cortactin-deficient mice consistently demonstrated a nonessential role for cortactin in actin remodeling (Lai et al., 2009; Tanaka et al., 2009). Although we found a slight induction of actin fibers that cross through the body of cortactin-deficient cells (Fig. S3), the nature of these fibers and potential functions are unclear. Instead of major effects on actin remodeling, cortactin was found to be involved in platelet-derived growth factor–induced signaling to Rho-GTPases (namely Rac-1 and Cdc42) and in cell migration (Lai et al., 2009). The data suggest that cortactin may operate in signaling to actin polymerization more indirectly than previously anticipated, possibly through contributing to the temporal and spatial regulation of activation of selected Rho GTPases. This is in good agreement with our results on the role of cortactin for the regulation of Rap1 and RhoG activity in endothelial cells.

Other cortactin-deficient mouse models have recently been presented with various phenotypes. Tanaka et al. (2009) generated mice homozygous for a cortactin allele devoid of exon 5 and succeeded in isolating embryonic fibroblasts on embryonic day (E) 15 of development. Although it was not mentioned whether these mice are viable beyond E15, considering the results described here, such a scenario seems likely. In contrast, Yu et al. (2010) generated a cortactin-deficient mouse model using a gene-trap vector. Although they reported heterozygous animals being viable and fertile, they failed to obtain homozygous offspring. As an explanation for this defect, the authors concluded that cortactin plays a role in asymmetric division of oocytes. Although a lack of embryonic lethality in the mouse model described by Tanaka et al. (2009) remains to be experimentally demonstrated, it is tempting to speculate that the strong phenotype observed by Yu et al. (2010) may be explained, at least in part, by dominant-negative effects caused by expression of N-terminal cortactin residues fused to β-galactosidase in this gene-trap mutant. A potential role of cortactin in spermiogenesis was previously proposed by Kai et al. (2004) based on the analysis of mice (C57BL/6 × C3H mixed genotype) carrying a large genomic deletion including the cortactin locus. At least for mice of the genotype analyzed here (129SV × C57BL/6 mixed background and C57BL/6 sixth backcross), we can exclude this function for cortactin because our cortactin-deficient male mice are fertile.

In conclusion, our results establish cortactin as an essential physiological component of the molecular machinery that controls basal vascular permeability by supporting barrier integrity of the endothelium in vivo. In addition, cortactin is required for proper functioning of two types of ICAM-1–mediated endothelial–neutrophil interactions in vivo: the slow rolling of neutrophils, which depends on the contribution of ICAM-1, and the firm adhesion that is mediated by clustered ICAM-1. Thus, cortactin is crucial for stabilizing the endothelial barrier at endothelial cell contacts and supporting neutrophil accumulation at the luminal surface of the endothelium leading to neutrophil extravasation.

The following commercial primary antibodies were used: monoclonal anticortactin (clone 4F-11; Millipore), polyclonal anticortactin (clone GK18; Sigma-Aldrich), anti–VE-cadherin (clone C19; Santa Cruz Biotechnology, Inc.), anti–ICAM-1 (clone R6-5-D6 [BioXCell]; clone YN1/1.7 [Takei, 1985]), anti–α-tubulin (clone B-5-1-2; Sigma-Aldrich), monoclonal anti-CD11a and anti-CD11b (clones M17/4 and M1/70; BD), and anti–PECAM-1 (clone MEC13.3; BioLegend). Antibodies against mouse P-selectin (clone RB40/34; Bosse and Vestweber, 1994), mouse E-selectin (clone UZ4; Hammel et al., 2001), mouse VE-cadherin (polyclonal serum C5 and clone 11D4), and ESAM (VE19) have been described previously (Gotsch et al., 1997; Nasdala et al., 2002). pAbs against JAM-A (Khandoga et al., 2009) and mAb against mouse VCAM-1 (Hahne et al., 1993) are as previously described. Fluorescently labeled secondary antibodies were purchased from Invitrogen. Peroxidase-labeled secondary antibodies were purchased from Dianova. F-actin was stained using Alexa Fluor 568–conjugated phalloidin (Invitrogen). Epac was activated using 8-pCPT-2Me-cAMP (Biolog). The mAb against RhoG and the expression construct for ELMO-GST were gifts of M.A. Schwartz (University of Virginia, Charlottesville, VA). ELMO-GST was purified from BL-21 bacteria using glutathione–Sepharose 4B (GE Healthcare) as described previously (Meller et al., 2008).

Two independent targeting vectors harboring neomycin and puromycin cassettes, respectively, were previously used to allow consecutive targeting of both cortactin alleles in ES cells that were then injected into blastocysts giving rise to chimeric embryos that were used to isolate puromycin- and neomycin-resistant fibroblast cultures (Lai et al., 2009). To generate cortactin-deficient mice, ES cells (line IGD3.2) were modified by homologous recombination of only one allele with the neomycin-containing construct shown in Fig. 1 A. In brief, the targeting vector comprised exons 6–9 of the Cttn gene, in which a neomycin cassette (flanked by Flp recombinase recognition sites) was inserted after exon 7 and two loxP sites were inserted flanking this exon. Targeted Cttn mut/WT ES (F1) cells were injected into C57BL/6J host blastocysts, which were then transferred into pseudopregnant females. Germline transmission of the targeted ES cells was confirmed by genotyping following standard procedures (DeChiara, 2001). The neomycin cassette was removed by crossing with mice expressing Flp recombinase (Rodríguez et al., 2000). Systemic Cre recombinase–mediated excision of exon 7 was achieved by crossing males carrying the floxed Cttn allele with transgenic females expressing Cre under the control of the human K14 promoter, allowing deletion in the oocyte (Hafner et al., 2004).

Genomic DNA was prepared from mouse tail samples as described previously (DeChiara, 2001). Genotyping for the WT, Cttn-targeted, and neomycin-containing deleted (del^neo) alleles was described previously (Lai et al., 2009). Genotyping for the floxed (neo deleted) versus WT alleles was performed in a standard PCR setup using the following set of PCR primers: forward, fS, 5′-CTCTGCTCTGTGCTTTGACC-3′; and reverse, fA, 5′-GTGCTGTTCATCCACAATGC-3′, producing PCR fragments with a size of ∼150 bp (WT allele) and ∼370 bp (floxed allele). For the genotyping in Fig. 1 B, PCR primer sets forward, dS, (5′-AGGGTCTGACCATCATGTCC-3′) and reverse, fA (see above), were used, and PCR products generated were ∼400 bp for the deleted (KO) and ∼900 bp for the WT allele.

3 h before cremaster muscle exteriorization, Cttn^del/del and sex- and age-matched Cttn^WT/WT mice received 500 ng TNF (PeproTech) in 0.3 ml saline intrascrotally. Mice were anesthetized with an intraperitoneal injection of 125 mg/kg ketamine hydrochloride (Sanofi) and 12.5 mg/kg xylazine (TranquiVed; Phoenix Scientific), and the cremaster muscle was prepared for intravital microscopy as described previously (Zarbock et al., 2007, 2008). Postcapillary venules with a diameter between 20 and 40 µm were recorded using an intravital upright microscope (Axioskop; Carl Zeiss) with a 40× 0.75 saline immersion objective. Leukocyte rolling velocity, leukocyte rolling flux fraction, and leukocyte firm adhesion were determined by transillumination intravital microscopy. Leukocyte extravasation was investigated by reflected light oblique transillumination microscopy (Mempel et al., 2003; Mueller et al., 2010). Recorded images were analyzed using ImageJ (National Institutes of Health) and AxioVision (Carl Zeiss) software. The leukocyte rolling flux fraction was calculated as the percentage of total leukocyte flux. Emigrated cells were determined in an area reaching out 75 µm to each side of a vessel over a distance of 100 µm vessel length (representing 1.5 × 10⁴ µm² tissue area). The microcirculation was recorded using a digital camera (Sensicam QE; PCO Inc.). Blood flow centerline velocity was measured using a dual photodiode sensor system (Circusoft Instrumentation). All animal experiments were approved by the Animal Care and Use Committees of the University of Münster and the local government. Animals were kept in a barrier facility under special pathogen-free conditions.

Mixed chimeric mice were generated by performing bone marrow transplantation. In brief, bone marrow cells isolated from cortactin-deficient and WT littermate mice were separately injected i.v. into lethally irradiated WT mice (9.5 Gy). Experiments were performed 8 wk after bone marrow transplantation.

Modified Miles assays were performed as described previously (Oura et al., 2003; Mamluk et al., 2005; Wegmann et al., 2006). In brief, three 8–12-wk-old Cttn^del/del and three sex- and age-matched Cttn^WT/WT mice were used per assay. Evans blue dye was injected into the tail vein, and 10 min later, 50 µl PBS or 100 ng histamine in 50 µl PBS was intradermally injected into the back skin that was shaved 24 h before the assay. 30 min later, animals were sacrificed, and skin areas surrounding the sites of injections were excised. Evans blue dye was extracted from the skin by incubation in formamide for 5 d at room temperature. Subsequently, the extracted dye was measured at 620 nm using a spectrophotometer. To exclude the possibility that the increase of Evans blue simply reflected the increase in blood flow, we measured the quantity of hemoglobin in the tissue by noninvasive white light spectroscopy using an O2C spectroscopy device (LEA Medizintechnik). The quantity of hemoglobin in the histamine-injected skin only increased by ∼20%. Thus, the increase of Evans blue detected in the histamine-stimulated skin was mainly caused by leakage and not by increased quantity of blood in the microvessel system.

HUVECs were isolated as described previously (Baumeister et al., 2005) and propagated in EGM endothelial cell growth medium (Lonza). ICAM-1^−/− endothelioma bendI1.1 and ICAM-1–retransfected bendI1.1 cells (provided by B. Engelhardt, University of Bern, Bern, Switzerland) were propagated as described previously (Lyck et al., 2003). The mouse endothelioma cell lines bend.3 and bend.5 cells were cultivated as described previously (Nottebaum et al., 2008). Human PMNs were isolated from the blood of healthy volunteers by Ficoll density gradient centrifugation as described previously (Moll et al., 1998). Murine bone marrow–derived leukocytes were isolated as described previously (Bixel et al., 2007). Primary MLECs were isolated from 8-wk-old Cttn^del/del and littermate Cttn^WT/WT mice as described previously (Schniedermann et al., 2010).

siRNAs directed against the messenger RNAs of cortactin (QIAGEN) were transfected into HUVECs or mouse endothelioma cell lines by nucleofection (Lonza) as described previously (Nottebaum et al., 2008). Target sequences of oligonucleotides were as follows: human cortactin-5 (5′-CACCAGGAGCATATCAACATA-3′), human cortactin-6 (5′-ATGCAACTTATTGTATCTGAA-3′), murine cortactin-5 (5′-CAGAGATGTGCTAGTGGCTTA-3′), and murine cortactin-7 (5′-CACCAACATAGAAATGATTGA-3′). The AllStars Negative Control siRNA (no sequence provided; QIAGEN) served as nonsilencing negative control. Efficiency of knockdown was evaluated by Western blot or immunofluorescence staining of cortactin. Experiments were started 48 h after transfection. The two mouse and two human oligonucleotides each inhibited cortactin expression, permeability, and transmigration with similar efficiency. Experiments are shown with human cortactin-5 and mouse cortactin-5. The cDNA construct for CA and WT RhoG (Meller et al., 2008) was transfected into MLECs by nucleofection using the same conditions as for siRNA.

Equal protein amounts of cell lysates were separated by SDS-PAGE and transferred electrophoretically to polyvinylidenefluoride membranes (Millipore). After incubation in PBS containing 5% skim milk for 1 h to block unspecific binding, the membranes were probed in primary antibodies either for 4°C over night or for 1 h at 37°C, washed three times in PBS containing 0.5% skim milk and 0.05% Tween 20 for 10 min each, and incubated with species-specific peroxidase-conjugated secondary antibody (Dianova) for 1 h at 37°C. Chemiluminescence signals were recorded on Hyperfilm x-ray films (GE Healthcare). Tubulin blots were performed as loading controls.

Cells were cultured as described in Cell culture, rinsed with PBS, and fixed in 100% ethanol at −20°C for 20 min. After washing, cells were incubated for 1 h in PBS containing 3% BSA to block unspecific binding. Coimmunolabeling with primary antibodies was performed for 1 h followed by washing and incubation with species-specific fluorescently labeled secondary antibodies (Invitrogen). Preparations were mounted in fluorescent mounting medium (Dako) and examined in a confocal laser-scanning microscope (Axiovert 200 M LSM510; Carl Zeiss).

For the determination of ICAM-1 clusters, adhesion of leukocytes to TNF-activated HUVECs was analyzed as described previously (Carman et al., 2003), except for that platelet-activating factor and MnCl₂ treatments were omitted. In brief, 500,000 PMNs were directly added to confluent HUVEC layers on a round 12-mm coverslip. After 20 min, cells were fixed in 4% paraformaldehyde and stained for microscopy.

HUVECs or bend.3 endothelioma were nucleofected, seeded on 6.5-mm-diameter transwell filters (Corning) with 0.4-µm pore size, coated with 0.01% fibronectin, and cultured for 48 h before starting the assay. 4 × 10⁴ MLECs were seeded per 6.5-mm-diameter transwell filters with 3-µm pore size, coated with 0.01% fibronectin, and cultured for 1 wk before starting the assay. 0.25 mg/ml FITC-dextran (250 or 150 kD; Sigma Aldrich) was added to the top chamber, and diffusion was allowed for 30 min in the presence or absence of 100 µM 8-pCPT-2Me-cAMP at 37°C. Fluorescence in the lower chamber was measured using a spectrofluorimeter (Synergy-2; BioTek). Confluence of cells was determined by immunofluorescence staining for VE-cadherin for each assay. Transendothelial electrical resistance of HUVEC monolayers transfected and cultivated as described in Cell culture and Transfections was measured using a CellZScope device (nanoAnalytics) according to the manufacturer’s instructions. The resistance of empty filters of 8.7 ± 0.6 Ω*cm² was subtracted from the values measured for the HUVEC monolayers.

PMN transmigration experiments with HUVEC and bend.5 monolayers, which were activated for 16 h with 5 nM TNF, were performed on 6.5-mm-diameter transwell filters with 5-µm pore size 48 h after transfection with siRNAs. 4 × 10⁴ MLECs were seeded per 6.5-mm-diameter transwell filters with 3-µm pore size, coated with 0.01% fibronectin, and cultured for 1 wk before starting the assay. In brief, cells were washed with 199 medium containing 20% FCS and 25 mM Hepes. The upper reservoir was then filled with 100 µl of supplemented 199 medium containing 0.5 × 10⁶ PMNs. After 1 h, the numbers of transmigrated PMNs in the lower reservoirs containing 500 µl of supplemented 199 medium were quantified using a Casy Cell Counter (Innovatis). PMN migration into the lower reservoirs is expressed as the percentage of total applied PMNs. Endothelial monolayers were controlled by immunofluorescence staining for VE-cadherin for each assay.

To test for the activation levels of the small GTPase Rap1 in the absence of cortactin, GST-RalGDS-RBD pull-down experiments of Rap1-GTP were performed using the Rap1 Activation kit (Millipore) according to the manufacturer’s instructions. Levels of active Rap1 were detected by Western blot using the antibody provided with the kit (Millipore). Western blots were scanned and quantified using ImageJ. To activate RhoG by ICAM-1 cross-linking, a 100-µl slurry of Dynabeads (Invitrogen) was loaded with 30 µg anti–ICAM-1 mAb YN1/1.7 (overnight) and added to a confluent monolayer of TNF-activated MLECs (7 × 10⁶ cells per 100-mm dish) for 30 min at 37°C. Unbound beads were washed off with ice-cold PBS before cell lysis. Aliquots of the cell lysates were snap frozen to determine later the total RhoG content. Active RhoG was precipitated from the remaining cell lysate using 50 µg ELMO-GST per sample (100-mm dish) as previously described (Meller et al., 2008). Levels of total and active RhoG were determined by Western blot using an anti-RhoG mAb provided by M.A. Schwartz.

Data are presented as mean with SD of the mean. Statistical significance was assessed using the Student’s t test. χ² tests were performed for the mating statistics to test the probability that deviation between the observed and the expected values is caused by chance.

Fig. S1 shows that the expression of various endothelial surface receptors is not altered in the absence of cortactin. Fig. S2 shows that cortactin is undetectable in murine leukocytes and that systemic cortactin deficiency does not cause alterations in leukocyte counts or in the expression level of various leukocyte surface receptors. Fig. S3 shows actin and VE-cadherin stainings of WT and KO MLECs. Fig. S4 shows cortactin knockdown effects in mouse endothelioma cells on Rap-1 activation, permeability, and neutrophil transmigration. Fig. S5 shows that cortactin deficiency enhances permeability of primary endothelial lung cell layers over time, which can be compensated by treatment with 007. Videos 1 and 2 show intravital microscopy of an inflamed cremaster muscle venule of a cortactin WT or cortactin KO animal, respectively. Tables S1 and S2 show mating statistics of heterozygous animals of mixed genetic background and from the sixth backcross to C57BL/6, respectively.

We are grateful to Dr. Britta Engelhardt for providing the ICAM-1–deficient and ICAM-1–retransfected endothelioma cell lines and to Dr. Martin A. Schwartz for providing anti-RhoG and constructs for ELMO-GST, WT RhoG, and CA-RhoG. We also thank Tanja Möller, Brigitte Denker, and Kirsten Marlen Kleemann for excellent technical assistance and advice and Werner Müller for infrastructure and discussion.

This work was supported in part by grants from the Max Planck Society (to D. Vestweber) and the Deutsche Forschungsgemeinschaft (RO2414/1-2 to K. Rottner and ZA428/3-1 to A. Zarbock).

The authors declare that they have no conflict of interest or financial interests.