Neuroligin 1 Induces Blood Vessel Maturation by Cooperating with the α6 Integrin*

Background: Disregulated adhesion of endothelial cells to extracellular matrix proteins is associated with numerous human diseases. Results: Neuroligin 1 modulates the interaction between α6 integrin and laminin and induces the formation of a normal extracellular matrix. Conclusion: Neuroligin 1 modulates vessel maturation and cell adhesion. Significance: We first report that Neuroligin 1, originally a synaptic protein, modulates extracellular matrix functions in the vascular system. The synaptic protein Neuroligin 1 (NLGN1), a cell adhesion molecule, is critical for the formation and consolidation of synaptic connectivity and is involved in vascular development. The mechanism through which NLGN1 acts, especially in vascular cells, is unknown. Here, we aimed at deepening our knowledge on the cellular activities and molecular pathways exploited by endothelial NLGN1 both in vitro and in vivo. We analyzed the phenotypic consequences of NLGN1 expression modulation in endothelial cells through in vitro angiogenesis assays and the mouse postnatal retinal angiogenesis model. We demonstrate that NLGN1, whereas not affecting endothelial cell proliferation or migration, modulates cell adhesion to the vessel stabilizing protein laminin through cooperation with the α6 integrin, a specific laminin receptor. Finally, we show that in vivo, NLGN1 and α6 integrin preferentially colocalize in the mature retinal vessels, whereas NLGN1 deletion causes an aberrant VE-cadherin, laminin and α6 integrin distribution in vessels, along with significant structural defects in the vascular tree.

The synaptic protein Neuroligin 1 (NLGN1), a cell adhesion molecule, is critical for the formation and consolidation of synaptic connectivity and is involved in vascular development. The mechanism through which NLGN1 acts, especially in vascular cells, is unknown. Here, we aimed at deepening our knowledge on the cellular activities and molecular pathways exploited by endothelial NLGN1 both in vitro and in vivo. We analyzed the phenotypic consequences of NLGN1 expression modulation in endothelial cells through in vitro angiogenesis assays and the mouse postnatal retinal angiogenesis model. We demonstrate that NLGN1, whereas not affecting endothelial cell proliferation or migration, modulates cell adhesion to the vessel stabilizing protein laminin through cooperation with the ␣6 integrin, a specific laminin receptor. Finally, we show that in vivo, NLGN1 and ␣6 integrin preferentially colocalize in the mature retinal vessels, whereas NLGN1 deletion causes an aberrant VE-cadherin, laminin and ␣6 integrin distribution in vessels, along with significant structural defects in the vascular tree.
The extracellular matrix (ECM) 7 provides essential support for the growth and organization of blood vessels (1). During sprouting angiogenesis, endothelial cells (ECs) secrete matrix proteins such as vitronectin, fibronectin, and collagen I leading to the formation of a provisional ECM, which sustains ECs proliferation and migration (2). Moreover, ECM proteins coordinate or at least complement the sprouting process by modulating the differentiation of the so-called tip and stalk cells (3,4). When the angiogenic stimulus drops, ECs arrest the migration and proliferation processes, form a lumen, and re-establish functional VE-cadherin-mediated cell-to-cell adherens junctions (5). ECs are normally quiescent although they are bound to vascular basement membrane (BM), a specialized form of ECM of mature vessels, mainly composed of laminin, collagen IV, heparan sulfate, and proteoglycans. Hence, the primary signals originating from BM inhibit proliferation and promote an environment that facilitates appropriate cell to cell adhesion. The main cellular sensors of cell-ECM adhesion are integrins, a large family of dimeric transmembrane proteins displaying different affinities for the ECM components.
Neuroligins (NLGNs) are neuronal proteins involved in the modulation of synaptic transmission (6). Recently, we have demonstrated that they have a function in vascular biology (7), and that a mature ECM has a pivotal role in vascular development and angiogenesis (7)(8)(9). In particular, the use of a zebrafish model combined with the knockdown of multiple genes (8) suggested that ECM is a critical component of NLGN signaling during vascular development, allowing the organization of ECM-immobilized vascular endothelial growth factor-A (VEGFA) gradients necessary for capillary networking (10) (11). Moreover, in ECs the cellular processes and molecular pathways in which NLGNs are involved remained unknown. In this article we show that NLGN1 promotes cell adhesion selectively on the BM protein laminin, and induces in vitro cord formation of ECs plated on BM by cooperating with ␣6 integrin, an integrin subunit that forms widely distributed endothelial laminin receptors (12). Furthermore, in vivo, NLGN1 deletion causes important organizational/morphological defects in the vascular tree of the developing mouse retina as well as in the structure of single vessels. This is demonstrated by deranged cell to cell junctions and disturbed laminin distribution, especially in the areas where NLGN1 and ␣6 co-localize. To our knowledge, this is the first report of a neuronal synaptic protein that functionally interacts with an integrin to modulate cell to ECM adhesion in the vascular system.

Animals
Mice lacking NLGN1 (Nlgn1 tm1 Bros/J, Jackson Laboratory), previously described in Ref. 13, were maintained on a mixed C57BL/6 ϫ 129 background (designated B6; 129) in the animal facility of the Institute for Cancer Research (IRCC) under standard housing conditions. Mice were sacrificed at postnatal (P) 5, and retinas were isolated for analysis. PCR amplification was used for the genotype characterization using the forward common primer (5Ј-GTGAGCTGAATCTTATGGTTAGAT-GGG-3Ј), the wild type reverse primer (5Ј-CGAGAGTCAGG-TAAATTGAACACCAC-3Ј), and the mutant reverse primer (5Ј-GAGCGCGCGCGGCGGAGTTGTTGAC-3Ј). All animal experiments were performed in compliance with guidelines governing the care of laboratory mice of the University of Torino Committee on Animal Research.
Reagents-The ECMs laminin, collagen I, fibronectin, vitronectin, and collagen IV were from Sigma; Lipofectamine 2000, Oligofectamine, and the Alexa Fluor conjugate were purchased from Invitrogen. Small interference RNAs (siRNAs) on-Target were obtained from Dharmacon Scientific.
DNA Construct-The NLGN1 construct (Mus musculus) with splicing inserts A and B carrying the HA sequence at the N-terminal end in pCAG vector and NLGN2 construct (M. musculus) with splicing inserts A carrying the HA sequence at the N-terminal end in the pNICE vector were from Peter Scheiffele. The retroviral vector overexpressing NLGN1 was generated using as template the KIAA1070 obtained from HUGE (database of Human Unidentified Gene-Encoded Large Proteins Analyzed) cloned in the pBlueScript vector (Invitrogen), and subcloned in the final retroviral expression vector pBABE (pBABE NLGN1). The lentiviral vector pWpXL containing the human ␣6 integrin Myc-tagged sequence was produced as described (15).

Cell Culture
Human umbilical vein ECs and human umbilical artery ECs were extracted from umbilical cords and cultured in EGM2 medium (Clonetics, Cambrex Bio Science, Walkersville, MD). In all experiments, ECs were used between passages 2 and 5. To overcome the intrinsic biological variability of EC, all experiments were performed with a mixture of ECs from at least three different umbilical cords.

Endothelial Cells Infection
The ␣6 integrin-Myc pWpXL lentivirus was produced as described (7). The day before the infection, ECs were seeded at concentrations of 3.5 ϫ 10 4 cells/ml in a 10-cm 2 tissue culture dish. Then the cells were transduced for 24 h with ␣6 integrin-Myc lentiviral particles in the presence of 8 g/ml of Polybrene (Sigma). Before starting every experiment the efficiency of cell infection was determined by FACS analysis (BD Biosciences) using rabbit polyclonal anti-Myc antibody (AbCam).

RNA Extraction, Retrotranscription, and qRT-PCR
The total RNA was extracted using TRIzol (Invitrogen) according to the manufacturer's instructions. The mRNAs obtained were treated with a DNase-free kit (Ambion and Applied Biosystem) and quantified with NanoDrop (ND-1000 and ND-8000 8-Sapectrophotometers). The quality of RNA was checked by the Agilent 2100 bioanalyzer (Agilent Technologies, Waldbronn, Germany). First-strand cDNA was generated from 1 g of total RNA using the High Capacity cDNA Reverse Transcription Kit (Invitrogen). The expression of human NLGNs, mouse NLGN1, and human ␣6 integrin was analyzed by quantitative real-time reverse transcription-PCR (qRT-PCR) using TaqMan Gene Expression Assay (human NLGN1, Hs00208784_m1; human NLGN2, Hs00325309_m1; human NLGN3, Hs00219160_m1; human NLGN4X, Hs00535592_ m1; human NLGN4Y, Hs00208784_m1; M. musculus NLGN1 Mm02344305_m1; human ITGa6, Hs00173952, from Applied Biosystems). The mRNA levels, analyzed in triplicate, were normalized by using as housekeeping genes the human (Hs00427620_m1) or mouse (Mm00446971_m1) TATAbinding box protein. The fold-increase or decrease was determined relatively to a control after normalizing to GAPDH (internal standard) with the formula 2 Ϫ⌬⌬CT .

Adhesion Assay
ECs (10 4 /well) were resuspended in 0.1 ml of EBM-2, 0.2% BSA, plated on 96-well plates (Costar, Acton, MA) coated with laminin (10 g/ml) and left to adhere for 30 min at 37°C. ECs were washed twice in PBS and then fixed 30 min in 8% glutaraldehyde and stained with 20% methanol, 0.1% crystal violet for 30 min. The cells were photographed with a QIcam FAST1394 digital color camera (QImaging) and counted with Image-Pro-Plus 6.2 software (Media Cybernetics). In the case of functional blocking the final concentration was 10 g/ml for anti-␣6 and anti-␣3 antibody treatments or the control isotype.

In Vitro Cord Formation Assay
For in vitro cord formation basal membrane extract (BME) (Matrigel, BD Biosciences) was added to each well at a concentration of 8 mg/ml and incubated at 37°C for 30 min to allow gel formation. ECs (2,3 ϫ 10 4 /well) were plated onto Matrigel and incubated for 4 h at 37°C in 5% CO 2 humidified atmosphere. Cell organization was examined (Leica Microsystem, Heerbrugg, Switzerland) and photographed. The lengths of the capillary-like structures were quantified with the imaging software winRHIZO Pro (Regent Instruments Inc.). In the case of functional blocking the final concentration was 10 g/ml for anti-␣6 and anti-␣3 antibody treatments or control isotype.
Retina images were quantified with Image-Pro Plus 6.2 software (Media Cybernetics, "colocalization highlighter" function) as follow: the immunoreactivity was calculated as the surface area of each antibody staining colocalizing with isolectin B4, normalized on total surface vascular area visualized by isolectin B4; the percentage of colocalization between ␣6 integrin and laminin was calculated as the fraction (%) of colocalized area between ␣6 integrin and laminin normalized on the total surface vascular area visualized by isolectin B4; the colocalization between NLGN1 and ␣6 integrin, in positive regions of interest for both Dll4 and VE-cadherin, was considered as a mean fluorescence of colocalization between two proteins in Dll4 or VE-cadherin positive staining regions of interest, traced with "polygon selection" function of Image-Pro Plus 6.2 software.

Quantification of Angiogenic Parameters in Mouse Retina
Images were analyzed using a customized version of a plugin previously described (27), developed for the ImageJ software and available at the NIH website. Briefly, fluorescence confocal images were first assembled to build full retina pictures using the MosaicJ plugin. These mosaics were segmented and analyzed as follows: low frequencies were removed using the deblurring "unsharp mask" method and noise was reduced by Gaussian convolution and dilation. Pre-treated images were then segmented using the so called "moments" threshold method (17). Holes smaller than vascular meshes were then removed and the gross trees were obtained using the skeletonize function of ImageJ. After that the pruning was performed to remove small artifactual twigs, and junctions, segments, and connecting junctions were detected. Finally, the meshes defined as areas enclosed by segments were revealed. These structures were counted and measured inside a ring of interest, delimited by two circle selections defined by the user, and finally used for statistical analysis. The first ring was located in a range between 500 m (the nearest radius from the optic nerve) and 800 m far from the optic nerve and with an area coverage of 1.22 mm, whereas the second ring was located in a range between 800 and 1100 m far from the optic nerve and including the sprouting vascular front, with an area coverage of 1.79 mm 2 . The vascular radius calculated (Fig. 3C, dashed circle, ImageJ 1.47) as the distance between vascular front and the center of the optic disc was normalized on the retinal radius. For each quantification at least 5 mice/genotype were analyzed and the corresponding values were expressed relatively to the mean retinal radius.
The confocal images were acquired using a confocal laserscanning microscope (TCS SP2 with DM IRE2; Leica). To quantify the degree of colocalization between two proteins we calculated the Mander's coefficient by using the JACoP plugin of ImageJ (28), whereas for the colocalization of three fluorescent proteins we used the "coloc" function on Imaris software (version 6.3.1) in maximum-projected confocal image stacks.

Statistical Analysis
Upon verification of normal distribution, statistical significance of raw data between the groups in each experiment was evaluated using the unpaired Student's t test or ANOVA followed by the Bonferroni post-test. Results are expressed as mean Ϯ S.E. when derived from averaged experiments, or mean Ϯ S.D. when derived from several data points of one experiment. n represents the number of individual experiments. The asterisks (*, **, and ***) in figure panels refer to statistical probabilities (p) of Ͻ0.05, Ͻ0.01, and Ͻ0.001, respectively.

NLGN1 Modulates in Vitro EC Cord Formation and
Adhesion to Laminin-By virtue of the functional differences between members of the NLGN family at synapse, we began by screening the expression of the different NLGN isoforms in vascular tissues and cells. Fig. 1, A and B, shows that the most expressed isoforms of NLGN in ECs were NLGN1 and -2. These two isoforms were chosen as targets and their role in endothelial functions related to angiogenesis was investigated by gain and loss of function approaches (validated in Fig. 1, C  and D).
Although down-regulation of NLGN1 or -2 did not affect cell migration, haptotaxis, and cell proliferation (data not shown), the silencing of NLGN1 modified the spontaneous cord formation of ECs seeded on BME. As shown in Fig. 2A (upper panel) NLGN1-silenced ECs formed a significantly less structured network compared with the control and NLGN2-silenced ECs. To further verify the specific involvement of NLGN1 in vascular organization, we analyzed ECs overexpressing NLGN1 and demonstrated that these cells formed a more complex network related to control cells ( Fig. 2A, lower panel).
Because the process of cell adhesion to different ECM proteins is an important step in endothelial morphogenesis, we tested the adhesion capability of NLGN1 or -2 down-regulated ECs. The endothelial cells were plated on several components of provisional and mature ECM, i.e. fibronectin, collagen I, collagen IV, vitronectin, or laminin (Fig. 2B). The screening indicated that NLGN1 (but not NLGN2)-silenced cells displayed a reduced adhesion on laminin but not on any of the other matrix proteins, when compared with control cells. We then verified that NLGN1 overexpression in EC enhanced cell adhesion to laminin (Fig. 2C). To confirm the specificity of the NLGN1-induced phenotype in siRNA-treated cells, we performed a "rescue" experiment. The overexpression of the mouse HA-NLGN1, which is not targeted by the human 3Ј-UTR targeting siRNA (siRNA17, Fig. 2D), fully rescued the EC adhesion to laminin (Fig. 2E). These results demonstrated that NLGN1 is specifically involved in the adhesion of ECs to the ECM protein laminin.
NLGN1 Modulates in Vivo Angiogenesis in the Mouse Retina-Given the role of NLGN1 in zebrafish vascular development (8) and in endothelial cord formation in vitro, we investigated post-NLGN1 and ␣6 Integrin Cooperatively Induce Vessel Maturation JULY 11, 2014 • VOLUME 289 • NUMBER 28 natal retinal angiogenesis in P5 WT and NLGN1 null mice. In WT retinas, NLGN1 (Fig. 3A, upper panel, red) was expressed in vessels in most of the region marked by isolectin B4 (Fig. 3A, upper panel, green), whereas it was not detectable in NLGN1 null mice (Fig. 3A, lower panel, red).
Although the GFAP-positive astrocytic network, which is known to provide guidance to the growing vessels (18), displayed no morphological difference between WT and NLGN1 null mice retinas (Fig. 3B), significant anomalies were identified in the retinal vascular tree of NLGN1 null mice by staining with isolectin B4 (Fig. 3, C-I). The defects included a shorter vascular radius (Fig. 3C), a reduced total vascularized area (Fig. 3D), and a decrease in the complexity of the vascular network (Fig. 3,  E and F). This last feature consisted in a reduced number of junctions (Fig. 3G) and meshes (Fig. 3H) as well as in an increase in the size of the meshes (Fig. 3I). Interestingly, the reduction in the complexity of the vascular network was especially evident in the mature region of the retinal plexus, and measurements were made in a specific circular area of the plexus as illustrated by the schematization in Fig. 3F. The involvement of NLGN1 in the maturation/stabilization of the vessels was in agreement with the fact that NLGN1 did not modulate migration/haptotaxis or proliferation of ECs. Moreover, NLGN1 was involved in EC adhesion to laminin, which is a component of mature ECM, but not to "angiogenesis promoting" matrix proteins such as collagen I, fibronectin, and vitronectin (2). Indeed, NLGN1 deletion caused a significant loss of the presence of VE-cadherin at cellto-cell junctions (Fig. 3J), whereas Dll4 expression and distribution, an indicator of sprouting vasculature (19), remained unchanged between WT and NLGN1 null mice (Fig. 3K). Moreover, prompted by our in vitro data we analyzed laminin and collagen IV localization in WT and NLGN1 null mice. Intriguingly, we found that the distribution of laminin (Fig. 3L), but not that of collagen IV (Fig. 3M), was affected in the retinal vascular plexus of NLGN1 null mice. In particular, laminin appeared delocalized from vessels in many sites throughout the network. These data indicated a role for NLGN1 in the formation of EC junctions and in the correct patterning of laminin in the BM. Ultimately, this can lead to an alteration in the morphology of the vascular tree and a reduction of the vascularized area.

. Screening of NLGN isoforms expression and validation of NLGN silencing and overexpression efficiency in ECs.
A, NLGN1, -2, -3, -4X, and -4Y expression is presented for primary cells (human umbilical vein and human umbilical artery ECs) and organs (brain, heart, and carotid). NLGN mRNA levels were assessed by qRT-PCR and the graph shows the single threshold cycle (C t ), which is inversely correlated to the expression level. The data revealed NLGN1 and NLGN2 as the most expressed isoforms in the ECs. Values are mean Ϯ S.E. of three independent experiments. B, immunoprecipitation assay was performed in ECs using an anti-pan-NLGNs (L067) antibody and immunoblotting was conducted with a monoclonal antibody (4F9) able to recognize NLGN1 and NLGN2 isoforms (top panel). The membrane was then stripped and incubated with a polyclonal antibody recognizing only NLGN2 (bottom panel). The 120-and 100-kDa bands detectable in ECs correspond to NLGN1 and NLGN2, respectively. The images shown are representative of 1 of 3 reproducible experiments. dimers, our first approach was to use anti-␣6 or anti-␣3 integrin-blocking antibody to treat NLGN1 overexpressing ECs in the adhesion and cord formation assays. We found that functional blocking of integrin ␣6 inhibited the NLGN1induced cord formation on BME (Fig. 4A). The anti-␣3 integrin antibody slightly inhibited cord formation of ECs carrying the empty vector, but it was fully ineffective on ECs overexpressing NLGN1. Similarly, the anti-␣6 antibody (GOH3) totally abolished the cell adhesion induced by NLGN1 on laminin, whereas the anti-␣3 antibody was ineffective (Fig. 4B).
NLGN1 Modulates the ␣6 Integrin-laminin Distribution in Retinal Vessels-We then investigated NLGN1 and ␣6 integrin cooperation and if their role in controlling adhesion and cord formation in vitro was matrix dependent.
We performed an immunofluorescence assay in ECs plated on laminin or collagen I. The HA-tagged NLGN1-transfected ECs were probed with the anti-HA and GOH3 antibodies. As shown in Fig. 5A, HA-NLGN1 and ␣6 integrin co-localized at the cell membrane of ECs plated on laminin significantly more than in ECs plated on collagen I.
Although it is known that the ␤4 integrin forms a laminin receptor with ␣6 in a small subset of ECs (12), we wanted to clarify which ␤ subunit (␤1 or ␤4) of the integrin ␣6 heterodimer was functionally involved with NLGN1 in ECs. Confocal imaging analyses performed in ECs plated on laminin showed a significant colocalization between NLGN1 and ␣6␤1 dimer rather than with ␣6␤4 dimer (Fig. 5B). Accordingly to the functional results cited above, we did not observe any significant colocalization of NLGN1 with ␣3 integrin (Fig. 5C). Hence, the ␣6␤1 dimer appeared as the main laminin receptor that cooperates with NLGN1 in ECs.
As an indication that NLGN1 and ␣6 could be part of the same molecular complex we transduced ECs with lentivirus FIGURE 2. NLGN1 is specifically involved in EC morphogenesis and adhesion to laminin. Cord formation (A) and adhesion (B, C, and E) assays were performed on ECs transfected with control siRNAs (SCRL) or siRNAs targeting either NLGN1 (A5, A6, and 17) or NLGN2 (B9), or ECs overexpressing either NLGN1 (HA-tagged mouse NLGN1, mNLGN1-HA, or retroviral vector pBABE NLGN1, NLGN1) or control pBABE vectors (EMPTY or GFP). A, ECs were seeded on 48-well plates coated with BME (8 mg/ml) and photographed after 4 h to evaluate cord formation. Images are representative of 1 of 3 reproducible experiments. The graph displays the mean EC tube length, quantified through WinRHIZO Pro, of three representative experiments Ϯ S.E. One-way ANOVA with Bonferroni test: *, p Ͻ 0.05; ***, p Ͻ 0.001. B, ECs were allowed to adhere on 96-well plates coated with fibronectin (1 g/ml), collagen I (10 g/ml), collagen IV (2 g/ml), vitronectin (1 g/ml), or laminin (10 g/ml) for 30 min at 37°C. After crystal violet staining, cells were counted employing Image ProPlus, to evaluate the adhesion rate. Results are expressed as the percentage of adherent cells relative to non-targeting siRNA. Values are expressed as mean Ϯ S.D. of one representative experiment (of two) performed in triplicate. One-way ANOVA with Bonferroni test: **, p Ͻ 0.01. C, adhesion assay was performed on ECs transfected as indicated and plated on laminin (10 g/ml), as in panel E. Values are mean Ϯ S.E. of three independent experiments. Unpaired Student's t test, two tailed: **, p Ͻ 0.01. D, NLGN1 mRNA levels, evaluated by qRT-PCR, in ECs transfected with siRNA NLGN1_17 (which targets the 3Ј-UTR sequence of human NLGN1) 24 h before transfection of mouse HA-tagged NLGN1 cDNA. The graph highlights that siRNA NLGN1_17 does not affect the overexpression of the mouse HA-tagged NLGN1 mRNA. Results are expressed as fold-change relative to control siRNA (SCRL). Values are mean Ϯ S.E. of three independent experiments, ***, p Ͻ 0.001. E, ECs were transfected as indicated and the adhesion assay (as in panel B) was performed on laminin. Values are expressed as mean Ϯ S.D. of a representative experiment of three, performed in triplicate. One-way ANOVA with Bonferroni test: ***, p Ͻ 0.001. carrying a Myc-tagged integrin ␣6 protein. Then, we plated these cells on BME to perform a co-immunoprecipitation experiment. Each of the 2 proteins was immunoprecipitated and the resulting precipitate was immunoblotted with the antibody against its respective partner. In these conditions endogenous endothelial NLGN1 and exogenous Myc-␣6 integrin coprecipitated (Fig. 5D), whereas VEGFR2 (KDR), which is well expressed in ECs, did not.
To characterize the respective localization of NLGN1 and ␣6 integrin in vivo, we analyzed the colocalization in the retinal vascular plexus. We found that in P5 WT retinas, NLGN1 and ␣6 integrin colocalized (arrowheads in Fig. 6A) in retinal vessels. Interestingly, we found that NLGN1/␣6 integrin colocalization was higher in areas of mature vessels marked by the presence of VE-cadherin at cell-to-cell junctions, rather than at the growing front/tip cells, as marked by Dll4 (Fig. 6B). Impor-FIGURE 3. NLGN1 null mice display an abnormal retinal angiogenesis. A, P5 mouse retina of WT (Nlgn1 ϩ/ϩ ) and NLGN1 null (Nlgn1 Ϫ/Ϫ ) mice were immunostained with isolectin B4 to visualize the retinal vascular vessels (green) and with the anti-NLGN1 antibody (clone H-45, red), which revealed no specific signal in NLGN1 null mice (bottom panel) compared with WT mice (top panel). Scale bar: 100 m, n ϭ 6 mice for each genotype. B, P5 retinas of both WT (Nlgn1 ϩ/ϩ ) and NLGN1 null (Nlgn1 Ϫ/Ϫ ) mice were immunostained with the anti-GFAP antibody to visualize the astrocytes. Images highlight the correct organization of astrocytes in NLGN1 null mice. n ϭ 12 retinas for each genotype. Scale bar: 200 m. C-I, evaluation of angiogenic parameters on P5 retinas of both WT and NLGN1 null mice (n ϭ 15 retinas for each group) immunostained with the vascular marker isolectin B4. C, the confocal image highlights the differences in terms of vascular radius (C, left panel), quantified through the dashed circle tool of ImageJ 1.47, and the vascular area (D), evaluated through the Angiogenesis Analyzer plug-in for ImageJ. E and F, the images illustrate the altered vascular pattern of NLGN1 null mice (E) and the snapshot exhibiting the obtained model (F) used for the analysis of the number of junctions (G), the number of meshes (H), and the size of meshes (I). These parameters were calculated with the Angiogenesis Analyzer plug-in for ImageJ, in a circular area between 500 and 800 m from the optic nerve (covered area of 1. tantly, the morphological defects in NLGN1 null mice were more evident in the mature vessels (Fig. 3F).
Based on these data, we evaluated if the NLGN1 loss would induce a redistribution of ␣6 integrin localization, by quantifying the percentage of ␣6 integrin localized in vessels in WT and NLGN1 null retinas. In accordance with the abnormal distribution of laminin in the vascular BM of NLGN1 null mice (Fig. 3L,  arrowheads), ␣6 integrin showed a decreased percentage of colocalization with the endothelial marker isolectin B4 in NLGN1 null mice compared with WT (Fig. 6C). To evaluate if, in the absence of NLGN1, ␣6 integrin was also delocalized from its ligand, we analyzed the co-distribution of vascular laminin and ␣6 integrin. The analysis revealed that the laminin/␣6 integrin colocalization in NLGN1 null mice was about 50% lower than in normal mice (Fig. 6D). These data imply that NLGN1 modulates the interaction between ␣6 integrin and laminin and contributes to the formation of a normally structured basal lamina.

DISCUSSION
In this article we have deepened our knowledge on the cellular activities and molecular pathways of endothelial NLGN1. In particular we have demonstrated that NLGN1 (a) modulates vascular morphogenesis in vitro and in the mouse retina, (b) modulates cell adhesion on the extracellular matrix protein laminin but not on collagen I/IV, fibronectin, or vitronectin, and (c) cooperates with ␣6 integrin in modulating cell adhesion to laminin. Moreover, in vivo, NLGN1 and ␣6 integrin preferentially colocalize in the stable/mature retinal vessel, whereas the loss of NLGN1 causes significant disturbances to both VEcadherin and ␣6 integrin/laminin localization. A, ECs overexpressing retroviral vector pBABE, EMPTY, or containing NLGN1 cDNA (pBABE NLGN1) were seeded on 48-well plates coated with BME (8 mg/ml) and treated 4 h with blocking antibody for ␣6 integrin (GOH3), ␣3 integrin, or non-related rat or mouse IgGs. Tube length was quantified through WinRHIZO Pro and is represented in the histograms. Images and graphs shown are representative of 1 of 3 reproducible experiments, each in triplicate. Data are expressed as mean Ϯ S.D. One-way ANOVA with Bonferroni test: ***, p Ͻ 0.001 for pBABE EMPTY versus pBABE-NLGN1 and for pBABE-NLGN1 versus pBABE-NLGN1 ϩ anti-␣6 (GOH3). B, the adhesion assay was performed by plating cells on 96-well plates coated with laminin (10 g/ml). Cells infected with the retroviral vector pBABE, EMPTY, or containing NLGN1 cDNA (NLGN1) were treated with the ␣6 integrin blocking antibody (GOH3), the ␣3 integrin blocking antibody or non-related rat or mouse IgGs and allowed to adhere at 37°C for 30 min. Data show that the GOH3 antibody completely abolished the NLGN1-induced adhesion on laminin. Data are expressed as mean Ϯ S.D. of a representative experiment (of three) performed in triplicate. One-way ANOVA with Bonferroni test: ***, p Ͻ 0.001 for pBABE-EMPTY versus pBABE-NLGN1 and pBABE-NLGN1 versus pBABE-NLGN1 ϩ anti-␣6 (GOH3). NS, not significant. JULY 11, 2014 • VOLUME 289 • NUMBER 28

JOURNAL OF BIOLOGICAL CHEMISTRY 19473
The link between NLNG1 function and ECM that we describe here is in accordance with our previous observations in zebrafish. In that case, we showed that NLGN1 loss causes developmental defects in the caudal plexus and subintestinal vessels of the zebrafish embryo, through a synergistic activity with both the ECM binding isoforms of VEGFA and the ECMmaturation promoting enzyme heparan sulfate 6-O-sulfotransferase 2 (8) (9).
Our experiments show that the ablation of NLGN1 induces retinal vascular defects characterized by a reduction of the total vascularized area as well as a shorter vascular radius, suggesting a possible role of NLGN1 in sprouting angiogenesis. However, further analysis revealed that these defects may be a consequence of impaired vessel maturation. In particular we detect an altered pattern of the network meshes that becomes more evident close to the optic nerve, combined with a significant loss of the presence of VE-cadherin at the cell-to-cell junctions and a loss of laminin-blood vessel contacts. Moreover, NLGN1 does not affect EC proliferation or migration in vitro or the number of Dll4-positive tip cells in vivo. Finally, NLGN1 cooperates with ␣6 integrin, which is involved in vessel stabilization (1). The role of ␣6 integrin in angiogenesis appears to be context dependent. Functional blockade of ␣6 integrin inhibits the early phases of angiogenesis (15). However, complete genetic FIGURE 5. NLGN1 and ␣6 integrin: colocalization and co-immunoprecipitation in ECs. A-C, confocal microscopy analysis performed on ECs transfected with mouse NLGN1-HA (mNLGN1-HA) or NLGN1-mRFP and plated 2 h on laminin (10 g/ml) or collagen I (10 g/ml). A, NLGN1 and ␣6 integrin colocalize on laminin. ECs were immunostained with anti-HA (green) and anti-␣6 integrin (red) antibodies. The images are representative of 1 of 3 reproducible experiments. Scale bar: 10 m. The graph shows the colocalization index Ϯ S.E. (n ϭ 3, total cells analyzed n ϭ 10), unpaired Student's t test, two-tailed (**, p ϭ 0.0011) for laminin versus collagen I. B, NLGN1 and ␣6␤1 integrin co-localize in ECs plated on laminin. ECs were immunostained with anti-HA (red), anti-␣6 integrin (blue) in combination with anti-␤1 integrin (green, upper panel) or anti-␤4 integrin (green, lower panel) antibodies. The images are representative of 1 of 5 reproducible experiments. Scale bar: 10 m. The graph shows the percentage of colocalization between NLGN1/␣6/␤1 integrin and NLGN1/␣6/␤4 integrin Ϯ S.E. (n ϭ 3, total cells analyzed n ϭ 10), unpaired Student's t test, two-tailed (*, p ϭ 0.023). C, NLGN1 and ␣3 integrin do not colocalize on laminin. ECs were immunostained with anti-HA (red), anti-␣6 integrin (blue), and anti-␣3 integrin (green) antibodies. The images are representative of 1 of 3 reproducible experiments. Scale bar: 10 m. The graph shows the percentage of co-localization between NLGN1/␣6 integrin and NLGN1/␣3 integrin Ϯ S.E. (n ϭ 3, total cells analyzed n ϭ 10), unpaired Student's t test, two-tailed (**, p ϭ 0.008). D, co-immunoprecipitation assays were performed on ECs infected with a lentiviral vector containing the Myc-tagged ␣6 integrin cDNA and plated on BME (10 g/ml) for 2 h. Proteins were immunoprecipitated either with the rabbit anti-pan-NLGNs (L067) or the mouse anti-Myc antibodies. Immunoblottings (IB) were carried out using the monoclonal antibody (4C12) specifically recognizing NLGN1 (band detected at 120 kDa), the rabbit anti-Myc antibody recognizing ␣6 integrin (band detected at 100 kDa) or, as negative control, the rabbit anti-VEGF receptor 2 antibodies (band detected at 220 kDa). The images are representative of 1 of 3 reproducible experiments.
ablation of ␣6 integrin leads to perinatal death of mice without evidence of developmental angiogenesis defects (21), whereas endothelial-specific elimination of this integrin increases tumor angiogenesis and growth in a xenograft model (22). Here, we show that in cooperation with NLGN1 ␣6 integrin induces vessel stabilization in one of the most useful models for analyzing the molecular and cellular mechanisms regulating angiogenesis. The in vivo influence of NLGN1 on vessel stabilization is further supported by the in vitro data, which demonstrate that ECs adhesion to purified laminin is impaired upon NLGN1 loss.
Furthermore, laminin found in the vascular basal lamina is a mixture of several protein isoforms (23). The laminin protein that we used for the adhesion assays (Fig. 1) is not a specific FIGURE 6. NLGN1, ␣6 integrin and laminin colocalize in vivo in the retinal vascular tree. A, in vivo co-localization of NLGN1 and ␣6 integrin. P5 retinas of NLGN1 WT mice were immunostained with anti-NLGN1 (red), anti-␣6 integrin (magenta) antibodies, and isolectin B4 (green). The arrowheads highlight the colocalization areas between NLGN1/␣6 integrin/retinal vessels, resulting in white spots (n ϭ 6 retinas for each group). B, NLGN1/␣6 integrin colocalization in the mature region of retinal vessels. Immunofluorescence assays were performed on NLGN1 WT P5 retinas using either the anti-Dll4 antibody to stain "tip cells" (top panels magenta, pseudocolored) or the anti-VE-cadherin antibody to stain "mature vessels" (bottom panels, magenta, pseudocolored), in combination with the anti-NLGN1 (red, pseudocolored) and the anti-␣6 integrin antibodies (green, pseudocolored). Data are expressed as the percentage of colocalized NLGN1 and ␣6 integrin signals in Dll4 or VE-cadherin positive areas Ϯ S.E. (n ϭ 3 retinas for each genotype), unpaired Student's t test, two-tailed (**, p ϭ 0.0015). The graph shows that NLGN1 and ␣6 integrin colocalize preferentially in the VE-cadherin positive vessels. C, immunofluorescence stainings were performed on both WT and NLGN1 null mice P5 retinas using the isolectin B4 (green) and the anti-␣6 integrin (red) antibodies. The graph shows the percentage of colocalization calculated as the surface area of ␣6 integrin staining colocalized with isolectin B4, normalized on the total surface vascular area. Values are expressed as mean Ϯ S.E. (n ϭ 3 retinas for each genotype), unpaired Student's t test, two-tailed (*, p ϭ 0.035). Data revealed a higher expression of ␣6 integrin in blood vessels of WT compared with NLGN1 null mice. D, immunofluorescence stainings were performed on both WT and NLGN1 null mice P5 retinas using isolectin B4 (magenta, pseudocolored), anti-␣6 integrin (green, pseudocolored), and anti-laminin (red) antibodies. The graph shows the percentage of co-localization of ␣6 integrin and laminin, normalized on total surface vascular area. Values are expressed as mean Ϯ S.E. (n ϭ 3 retinas for each genotype), unpaired Student's t test, two-tailed (*, p ϭ 0.039). Data revealed that, in the presence of NLGN1, ␣6 integrin is significantly more co-localized with laminin. Scale bars in A-D, 50 m. JULY 11, 2014 • VOLUME 289 • NUMBER 28 isoform but is extracted from BM; similarly the antibody that we used (Fig. 2) is raised against a mixture of laminins extracted from BM. Hence, we cannot argue about which laminin isoform is involved in the NLGN1/␣6 integrin activity. In this context, ␣6 integrin has been shown to induce Dll4 expression upon laminin-111 (24) and laminin 511-411 (4) engagement. Given the absence of an effect on Dll4 expression in vivo upon NLGN1 loss, we suggest that by cooperating with NLGN1, ␣6 integrin activity shifts toward vessel stabilization.

NLGN1 and ␣6 Integrin Cooperatively Induce Vessel Maturation
Some final observations can be made about the implications and parallels that our data may have in neurobiology. First, we show that NLGN2 does not mediate the same cellular activities of NLGN1 in ECs. This is a very interesting similarity with the selective and differential NLGN1 and NLGN2 activities at excitatory and inhibitory synapses, respectively, and perhaps depends on the recruitment of alternative downstream partners in ECs as in neurons (6). Ongoing experiments in our laboratory are dedicated to understanding the role of NLGN2 in ECs. Second, it is possible that the role of NLGN1 in cell adhesion to laminin extends to the central nervous system, where laminin is also present. On one hand, this is conceptually supported by the fact that Neurexin, the natural pre-synaptic partner of NLGN1 contains LNS domains (25), which are present in laminin (26).
In conclusion, our article brings novel data into the vascular role of NLGN1 and introduces some novel working hypotheses on its activity in the nervous system. In particular it would be interesting to analyze if the ECM, as in vascular development, also plays a role in neural plasticity.