Angiomotin like-1 is a novel component of the N-cadherin complex affecting endothelial/pericyte interaction in normal and tumor angiogenesis

Transmission of mechanical force via cell junctions is an important component that molds cells into shapes consistent with proper organ function. Of particular interest are the cadherin transmembrane proteins, which play an essential role in connecting cell junctions to the intra-cellular cytoskeleton. Understanding how these biomechanical complexes orchestrate intrinsic and extrinsic forces is important for our understanding of the underlying mechanisms driving morphogenesis. We have previously identified the Amot protein family, which are scaffold proteins that integrate polarity, junctional, and cytoskeletal cues to modulate cellular shape in endothelial as well as epithelial cells. In this report, we show that AmotL1 is a novel partner of the N-cadherin protein complex. We studied the role of AmotL1 in normal retinal as well as tumor angiogenesis using inducible endothelial-specific knock-out mice. We show that AmotL1 is essential for normal establishment of vascular networks in the post-natal mouse retina as well as in a transgenic breast cancer model. The observed phenotypes were consistent with a non-autonomous pericyte defect. We show that AmotL1 forms a complex with N-cadherin present on both endothelial cells and pericytes. We propose that AmotL1 is an essential effector of the N-cadherin mediated endothelial/pericyte junctional complex.

junctional protein MAGI1b and actin 6,7 . The identification of the Amot binding partners suggests a role of these scaffolds to integrate essential morphogenic cues to control cell shape and proliferation of multicellular tissues.
We have previously shown by gene inactivation in zebrafish and mouse models, that the three members of the Amot protein family exert distinct roles during blood vessel development 8,9 . AmotL2 associates with VE-cadherin in order to connect the adhesion junction complex to the actin filaments thus establishing a mechanical connection between endothelial cells 9 . Using traction force microscopy, we also demonstrated that AmotL2 is essential to generate force in endothelial junctions. This may explain the lack of aortic expansion when genetically interfering with AmotL2 expression during zebrafish or mouse development. The pivotal role of Amot in physiological angiogenesis was demonstrated by the fact that its inactivation results in inhibition of migration of inter-segmental vessels in zebrafish and also affects blood vessel development in causing death in utero 8 . Finally, recent data indicate that Amot, similarly to AmotL2, is also part of a junctional complex as it associates to CDH11 and may be involved in promoting migration in prostate cancer 10 . Less is known however regarding the role of AmotL1 in the formation and maintenance of cell-cell contacts. Morpholino knock-down experiments suggest a perturbed adhesion of stalk cells of inter-segmental vessels and some overlap in function with Amot (at the level of tight junctions) 8,11 . In this report, we have used an endothelial-specific knock-out approach to analyze the role of AmotL1 in normal and pathological angiogenesis in mice. We identify AmotL1 as a novel component of the N-cadherin adhesion protein complexes and provide evidence that AmotL1 is essential for vascular remodeling during angiogenesis.

Results
In order to investigate the role of AmotL1 in the vascular endothelium and other tissues, we used a genetic deletion approach to silence gene expression in mouse. To inactivate AmotL1 in a cell-type specific fashion, we generated mice with loxP sites flanking exons 5 and 6 (Supplementary Figure 1a). The amotL1 flox/flox mice were mated with Cdh5 (PAC)-CreERT2 transgenic mice (hereafter abbreviated as amotL1 ec+/ec+ ) 12 . This model allows the efficient tamoxifen-inducible conditional recombinase expression in the endothelial cell lineage (recombined mice are abbreviated amotL1 ec−/ec− ). Tamoxifen-induced recombination was verified by genomic PCR analysis (Supplementary Figure 1b). The efficiency of tamoxifen injections in inducing recombination events was also analyzed in tumors from MMTV-Cre/amotl flox/flox /MMTV-PyMT mice by western blot (Supplementary Figure 1c).
We analyzed effects of inactivation of AmotL1 protein expression during post-natal vascularization of the mouse retina. The retinas of newborn mice are avascular but become vascularized in a very reproducible manner over the first 10 days after birth 12 . In order to detect AmotL1 protein expression we have previously developed immunoaffinity-purified antibodies that did not cross-react to any of the other family members 11 . Western blot analysis showed expression of both AmotL1 isoforms (p90 and p100) in developing as well as in adult retina (Supplementary Figure 1d). We also used whole mount immunostaining of mouse retinas of post-natal day 6 to analyze AmotL1 expression. AmotL1 showed junctional localization in established vessels of the developing vascular network. No staining above background was detectable in endothelial cells of the front of the expanding vascular network (Fig. 1a).
We then analyzed effects of inactivation of AmotL1 protein expression during post-natal vascularization of the mouse retina. In amotL1 inactivation experiments, pups were injected with tamoxifen on P1-3 and retinas were harvested on P6. The developing vasculature was visualized by whole mount staining using isolectin B4 12 . Overall, the density of the vessel network of the amotL1 deficient retinas was reduced and covered a smaller area of the retina (Fig. 1b,e). The vascular density and the number of branching points, evaluated using the Angiotools program 13 , were significantly lower in the periphery and center of the amotL1 ec−/ec− retinas (Fig. 1c,d,f-i). In contrast inactivation of amotL1 in adult mice did not have any effect on vessel branching or density (Fig. 1j,k).
The reduced expansion of the vessel network in amotL1 ec−/ec− animals suggested a defect in cell migration. Luminal endothelial cells convert into tip cells that direct vessel migration and contribute to the development of a multicellular stalk 14 . Our analysis of tip cells did not reveal any obvious defect in tip cell morphology (Supplementary Figure 2a). Furthermore, the number of filopodial extensions per tip cell was not significantly different (Supplementary Figure 2b). However, the total number of tip cells per length unit was lower in amot-L1 ec−/ec− mice, which perhaps could be ascribed to the lower vascular density (Supplementary Figure 2c).
Endothelial cells depend on other cell types such as microglia, astrocytes and pericytes, to be able to form a functional circulatory network 14 . Each cell type is engaged in distinct functions such as controlling anastomoses (microglia), guiding network branching (astrocytes) and vessel maturation (pericytes). We went on to analyze whether inactivation of amotL1 affected endothelia-interacting cells in a non-autonomous manner. Whole mount immunostaining of P6 amotL1 ec−/ec− retinas, performed to visualize microglia and astrocytes did not reveal any striking difference (Supplementary Figure 2d,e). Immunostaining using the NG2 pericyte marker revealed that mutant blood vessels of amotL1 ec−/ec− retinas were associated with pericytes (Fig. 2a). However, the pericytes showed a decreased area of coverage as the mural cells displayed a changed morphology in that they "bulged" out from the vessels (Fig. 2a,b).
We then assessed the role of AmotL1 in two tumor model systems, the transgenic model of breast cancer MMTV-PyMT and the Lewis Lung Carcinoma (LLC) transplanted tumor model. In the first model, the MMTV promoter drives the expression of the Polyoma Middle-T oncogene in the mouse mammary epithelium 15 allowing studies of progressive breast tumor development over a period of 6-14 weeks. The amotL1 ec+/ec+ mice were crossed into the MMTV-PyMT background. Genetic deletion was induced by intraperitoneal tamoxifen injections at week 4. Analysis of the infiltrating blood vessels in the tumor area revealed a marked difference in morphology between the control and AmotL1 deficient mice. amotL1 ec−/ec− vessels were quite enlarged and dilated as shown by CD31 and Collagen IV immunofluorescent stainings ( Fig. 3a and quantification   A lower functionality of the tumor vessels was indicated by the longer lag time until tumors were detected in the amotL1 ec−/ec− mice, whereas tamoxifen administration did not affect tumor progression per se (Fig. 3d).
The LLC tumor transplantation model has been extensively used to study tumor vasculature and the efficacy of angiogenesis inhibitors 16 . The LLC tumors grow rapidly and form highly vascularized lesions within 10 days after injection. It has been previously shown that most of the LLC tumor vessels are immature and may regress upon treatment with inhibitors of the VEGF signaling pathway 17    are closely connected in mature blood vessels where they jointly produce the blood vessel basal membrane. The two distinct cell types communicate by paracrine signaling factors as well as direct physical contact in cell-cell adhesions mediated by N-cadherin. In our previous studies we have shown that AmotL2 is part of the VE-cadherin junctional complex and affects endothelial radial actin filaments and cell morphology in the aorta 9 . Co-immunoprecipitation analysis in mouse endothelial cells showed that VE-cadherin was primarily associated with AmotL2 and weakly with Amot or AmotL1 (Fig. 4a). Endothelial cells also express N-cadherin that is, in contrast to VE-cadherin, not endothelial specific as it is expressed in a number of mesenchymal cells such as pericytes 18 . N-cadherin is thought to participate in vessel stabilization by interacting with peri-endothelial cells during vessel formation and therefore a possible binding partner for AmotL1. However, no association could be detected between AmotL1 and N-cadherin in the co-immunoprecipitation analysis (Fig. 4a). Previous findings have shown that in vitro and in vivo N-cadherin junctional localization is controlled by VE-cadherin expression level [19][20][21] . In wild type (wt) endothelial cells in vitro, N-cadherin displays a diffuse localization on the cellular membrane whereas in VE-cadherin depleted cells N-cadherin is localized at adherent junctions 22 . It was possible that a homotypic interaction between N-cadherin in different cell types was required for the recruitment of AmotL1 to N-cadherin associated adhesion complexes. The co-immunoprecipitation analysis was repeated in VE-cadherin −/− and VE-cadherin +/+ endothelial cells with the result that AmotL1 was being pulled down in complex with N-cadherin only in the VE-cadherin −/− cell line (Fig. 4b,c, Supplementary Figure 5). Similar results were obtained in VE-cadherin siRNA depleted endothelial cells (Fig. 4d). Co-localization of AmotL1 and N-cadherin in cellular junctions was also detected in VE-cadherin −/− cells (Fig. 4e).
We also analyzed AmotL1 and N-cadherin expression in pericytes. In the developing retina, AmotL1 could be detected in PDGFR-beta positive pericytes associated to IB4-postive retinal blood vessels (Fig. 5a). Immunofluorescent analysis showed overlap between AmotL1 and N-cadherin (Fig. 5b). Furthermore, AmotL1 could be pulled down associated to N-cadherin in human pericytes (Fig. 5c). In conclusion, we show that AmotL1 is a specific partner of N-cadherin in both endothelial cells and pericytes and our data provide evidence that this interaction is important for vascular network organization and proper endothelial/pericytes interaction (Fig. 5d,e).

Discussion
In this paper we provide insight regarding the role of the scaffold protein AmotL1 in normal and pathological angiogenesis. We show for the first time that AmotL1 is associated to the cellular adhesion molecule N-cadherin. This highlights the importance of Amot proteins as specific interactors of cellular adhesion proteins with distinct roles in controlling the cellular architecture of blood vessels during angiogenesis.
We provide evidence that AmotL1 is essential for the development of a proper vascular network during normal vascularization of the mouse retina. Our previous findings have shown that AmotL2 links the VE-cadherin and confers mechanical force via contractile actin filaments via cell-cell junctions. In contrast, AmotL1 does not associate to VE-cadherin and it is apparently not involved in the homotypic interactions of endothelial cells. In support of this notion, we could show that AmotL1 does not bind VE-cadherin and that knock-down of AmotL1 does not visibly affect actin filaments associated with VE-cadherin. In vitro, VE-cadherin mediates adhesion between adjacent endothelial cells and is located at adherens junctions whereas N-cadherin has a diffusely dispersed distribution all over the outer membrane. It has previously been shown that VE-cadherin displaces N-cadherin from endothelial cell junctions and that reduction of VE-cadherin allows N-cadherin to form homotypic cell-cell interactions 19 . In this report, we could only find evidence of AmotL1 binding to N-cadherin in the absence of VE-cadherin. This suggests that the recruitment of AmotL1 is dependent on homotypic N-cadherin interactions in cell junctions. The VE-cadherin-dependent displacement of N-cadherin from endothelial cell contacts allows binding to perivascular N-cadherin expressing cells in vivo. Indeed, multiple reports have shown that N-cadherin is required for endothelial-pericyte cross-talk. N-cadherin is typically expressed in cells of mesenchymal origin including vascular mural cells such as pericytes and smooth muscle cells. A shared basal membrane separates mural cells and endothelial cells but these cells make direct contact through cavities in the basal membrane. Several pieces of evidence suggest that N-cadherin is essential for this interaction between endothelial cells and pericytes. Ji-Hye Paik and coworkers have shown that blocking N-cadherin binding in in vitro co-cultures abrogated the interaction of endothelial cells and pericytes 23 . Similar results in vivo in chick brain suggested that N-cadherin is essential for the interaction of endothelial cells and pericytes and the resulting stabilization of blood vessels 24 . Furthermore, genetic inactivation of N-cadherin resulted in impaired pericyte coverage of endothelial sprouts in vitro 25 .
In our experiments, we observed a non-autonomous phenotype in the pericytes of the developing mouse retina. As N-cadherin and AmotL1 are expressed in both endothelial cells and pericytes, it is compelling to speculate that ablation of AmotL1 in the endothelial compartment will affect the pericytes in a non-autonomous way. Indeed, in amotL1 ec−/ec− retinas, the pericyte coverage was decreased as the pericytes changed shape and bulged out from the vessels.
Extensive data show that endothelial-derived PDGF-B is essential for the recruitment of pericytes in newly formed vessels 26 . Impaired access to PDGF-B through genetic manipulation alters the retinal vascular density and increases tumor blood vessel diameter 27 . This is quite similar to what we have observed in AmotL1 deficiency in endothelial cells during normal and tumor blood vessels development. AmotL1, however, appeared not to be essential for pericyte recruitment as there were no apparent change in the number of pericytes/vessels in the amotl1 ec/ec mice.
TGF-β has long been implicated in the maturation of blood vessels by stimulation of basal membrane components 28 . Recently, Li and coworkers showed that inactivation of the TGF-beta effector protein smad4 in the central nervous system affected the pericytes-endothelial coupling of brain vessels. Interestingly, Smad4 together with Notch signaling affected the expression of N-cadherin. Cerebral capillaries with decreased levels of N-cadherin  could maintain the ability to recruit pericytes but failed to form proper adhesions resulting in pericytes detachments 29 . We propose a model by which AmotL1 is recruited to N-cadherin localized to peg-sockets in order to mechanically couple endothelial cells to pericytes via formation of actin filaments the functionally integrating cell-cell junctions with actomyosin contractility. This would be similar to the mechanisms of action of AmotL2, which in complex with VE-cadherin exerts force between endothelial cells via para-cellular actin filaments. We hypothesize that both AmotL1 and AmotL2 functionally integrate mechanical signals via distinct adhesion molecules to modulate vascular morphogenesis.
These findings raise the question whether AmotL1 is a crucial mediator of other cellular processes. Epithelial-mesenchymal transition involves the downregulation of E-cadherin and upregulation of N-cadherin. Expression of N-cadherin on the tumor cells surface facilitates the adhesion to endothelial cells and subsequently extravasation and metastasis 30 . Indeed, a recent study has shown that AmotL1 is upregulated and promotes invasion in breast cancer 31 . Further studies may shed light whether the tumor: endothelial interactions are dependent on mechanical forces relayed by AmotL1.
Western blot. Cells were lysed in the following buffer (50 mM Tris-HCL pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% Nonident P40), 1 x protease inhibitor (Roche, 04693159001). Lysates were prepared with SDS sample buffer (Novex, 1225644) containing 10% sample reducing agent (Novex, 1176192). Proteins were fractionated in a polyacrylamide Bis-Tris 4-12% gradient precast gel (Novex, NP0322BOX). Afterwards, proteins were transferred to a nitrocellulose membrane (Whatman, 10401396). The membrane was blocked in 5% non-fat milk PBS and 0.1% Tween 20 and incubated with the primary antibody overnight at 4 °C. The membrane was thereafter incubated for 1 hour at R.T. with an adequate horseradish peroxidase-conjugated secondary antibody. Labeled proteins were detected with chemiluminescence (ECL; Amersham, RPN2232). Immunoprecipitation analysis. Cell lysates were prepared in the buffer (50 mM Tris-HCL pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% Nonident P40, 1 x protease inhibitor (Roche, 04693159001)). Lysates were incubated with protein G sepharose beads (GE, 17-0618-01) for 1.5 hours at 4 °C as pre-cleaning. Afterwards antibodies against AmotL1 or control rabbit immunoglobulins added in the lysates overnight under rotation at 4 °C. Protein G sepharose beads were added for additional 2 hours. Beads were washed 5 times with lyses buffer and heated for 10 min at 95 °C in 2 x LDS sample buffer (Novex, 1225644) containing 10% sample reducing agent (Novex, 1176192). The samples were fractionated by SDS PAGE and subsequently western blotted. Fractions of whole cell lysates were western blotted for evaluation of IP protein input level.
Retina angiogenesis assay. Mouse retina angiogenesis assay were performed as previously published 12,32 .
In brief, tamoxifen (Sigma) was injected intraperitoneally into newborn pups at the dose of 50 μ g/pup/day from postnatal day P1 to P3. Eyes were dissected at P6 and fixed in 4% paraformaldehyde for whole mount immunostaining. Branching and vessel density was analyzed using the Angiotools software (https://ccrod.cancer.gov/ confluence/display/ROB2/Quick+ Guide).
Immunofluorescence staining. Fixed mouse retinas were dissected out from eyeball under microscope.

Xenograft Tumor Model.
Tamoxifen was injected intraperitoneally to wild-type as well as amotL1 flox/flox male mice (≥ 8 mice per group) at age of 6-week-old (1 mg/mouse/day) for 5 days in a row. Approximately, 0.5 × 10 6 LLC tumor cells were subcutaneously injected into the right flank of the mouse on the last day of tamoxifen injection. Tumor growth was palpated twice per week and tumor volumes were calculated according to the formula: 0.52 × length × width × width. One week before sacrifice, same dosage of tamoxifen was injected Scientific RepoRts | 6:30622 | DOI: 10.1038/srep30622 intraperitoneally into the tumor bearing mice for 5 days. When tumors reached the diameter of 10 mm, mice were sacrificed and tumors were resected and analyzed by cryohistology. siRNA transfection. For siRNA transfections cells were seeded the day before transfection on glass slides (BD Falcon Cultureslides BD Biosciences) coated with 1% BD Matrigel Basement Membrane Matrix (BD Biosciences) in growth medium without antibiotics. Just before transfection, growth medium was exchanged for OPTI-MEM I Reduced Serum Media (Invitrogen). Smartpool siGENOME siRNAs (Dharmacon/Thermo Scientific) against VE-cadherin siRNA from Dharmacon (SMARTpool: ON-TARGETplus Cdh5 siRNA, L-041968-01-0005) as well as Non-Targeting siRNA Pool #2 (D-001206-14) were transfected into the cells with Oligofectamine Transfection Reagent (Invitrogen) according to the manufacturers protocol. siRNAs were used at the final concentrations of 0.32M for both control and AmotL2 siRNAs. Four hours after transfection serum was added to a final concentration of 20%. Cells were allowed to grow for 72 hours prior to evaluation by immunofluorescence staining or western blot. Statistical analysis. Statistical analysis of in vivo results based on at least four animals and four images from each sample tissue. Comparisons between different groups were performed using the standard two-tailed Student t-test. A value of P < 0.05 was considered as statistically significant ( * P < 0.05, ** P < 0.01 and *** P < 0.001).