Constitutive Active Mutant TIE2 Induces Enlarged Vascular Lumen Formation with Loss of Apico-basal Polarity and Pericyte Recruitment

Abnormalities in controlling key aspects of angiogenesis including vascular cell migration, lumen formation and vessel maturation are hallmarks of vascular anomalies including venous malformation (VM). Gain-of-function mutations in the tyrosine kinase receptor TIE2 can cause VM and induce a ligand-independent hyperactivation of TIE2. Despite these important findings, the TIE2-dependent mechanisms triggering enlarged vascular lesions are not well understood. Herein we studied TIE2 p.L914F, the most frequent mutation identified in VM patients. We report that endothelial cells harboring a TIE2-L914F mutation display abnormal cell migration due to a loss of front-rear polarity as demonstrated by a non-polarized Golgi apparatus. Utilizing a three-dimensional fibrin-matrix based model we show that TIE2-L914F mutant cells form enlarged lumens mimicking vascular lesions present in VM patients, independently of exogenous growth factors. Moreover, these abnormal vascular channels demonstrate a dysregulated expression pattern of apico-basal polarity markers Podocalyxin and Collagen IV. Furthermore, in this system we recapitulated another pathological feature of VM, the paucity of pericytes around ectatic veins. The presented data emphasize the value of this in vitro model as a powerful tool for the discovery of cellular and molecular signals contributing to abnormal vascular development and subsequent identification of novel therapeutic approaches.

Angiogenesis is the process of establishing new blood vessels from the preexisting vasculature. This process is finely controlled by the balance between activators which stimulate blood vessel growth, and inhibitors, which prevent vascular proliferation 1 . Angiogenesis is characterized by a series of temporally distinct events. First, activated endothelial cells (EC) produce proteolytic enzymes (e.g. MT1-MMP) that degrade the extracellular matrix to allow EC to migrate and proliferate to form cellular sprouts and lumens 2 . Also, these cellular protrusions undergo lumen formation by achieving apico-basal membrane polarization [3][4][5][6] . Finally, the last step of the angiogenic process is termed blood vessel maturation and consists of deposition of basement membrane and recruitment of pericytes that in turn stabilizes the vasculature while preventing further EC proliferation 1,[7][8][9] .
Improper angiogenic signaling results in malformed vessels that can proliferate excessively generating instability of poorly developed vascular structures with reduced pericyte recruitment. This dysfunctional process is a hallmark of several vascular anomalies and tumor angiogenesis 10,11 . The hyperactivation of tyrosine kinase receptors on EC can lead to pathogenic angiogenesis by driving vascular dysmorphogenesis that resembles human vascular malformations [12][13][14] .
The endothelial tyrosine kinase receptor TIE2 is a crucial player in the vascular proliferation and maturation processes during angiogenesis by binding to the ligands ANGPT2 (Angiopoietin-2) and ANGPT1 (Angiopoietin-1), respectively.

Results
Constitutive active mutant TIE2 increases wound-induced migration speed with loss of frontrear polarity. To investigate the angiogenic properties of EC expressing a constitutive active form of the TIE2 receptor we utilized HUVEC engineered to express TIE2-L914F (HUVEC-TIE2-L914F), the most frequent mutation found in VM patients 17,22 . Proliferation and migration are the first events leading to new vessel formation from pre-existing ones 26,27 . HUVEC-TIE2-L914F exhibited growth advantage compared to HUVEC-TIE2-WT (wild-type) and normal HUVEC (Supplemental Fig. S1), as we and others have previously reported 20,28 . Next, we investigated the migration ability of HUVEC-TIE2-L914F compared to HUVEC-TIE2-WT and normal HUVEC and found that HUVEC-TIE2-L914F migrated through a scratch/wound faster than the control cells (Fig. 1A,B). To determine if increased motility is an intrinsic property of HUVEC-TIE2-L914F, we tracked the cell movement trajectories over a 2-hour period. When cells were seeded in monolayer there was no detectable difference between the cell pace of HUVEC-TIE2-L914F and HUVEC (Fig. 1C,D and Supplemental Video 1). Conversely, the migration speed in response to scratch/wound was significantly increased in the TIE2-mutant EC (Fig. 1E,F and Supplemental Video 2). The hallmark of wound migration is re-orientation of the Golgi complex in the direction of the cell migration 29 . To investigate the orientation of the EC during the migration process, EC were fixed 2 hours after performing the scratch/wound and stained with GM130, a marker of the Golgi apparatus 30 . Compared to normal HUVEC that moved perpendicular to the wound, the majority of the HUVEC-TIE2-L914F at the migrating front displayed a non-polarized Golgi apparatus (Fig. 1G,H). These results reveal that expression of the constitutive active mutant TIE2, TIE2-L914F, in EC confers growth advantage and induces migration in aberrant directions due to loss of cellular front-rear polarity.

HUVEC-TIE2-L914F form massively enlarged VM-like vascular channels in a 3D fibrin gel.
When injected in vivo, in a xenograft model, HUVEC-TIE2-L914F generated massively enlarged vascular channels, mimicking important aspects of the VM pathogenesis 20,23 . We aimed at generating an in vitro 3D (three-dimensional) system as a tool to study VM lumen morphogenesis. When HUVEC were embedded in fibrin gels topped with fibroblasts, they formed regularly shaped lumenized longitudinal vessels ( Fig. 2A), as previously reported 25 . Conversely, HUVEC-TIE2-L914F generated ectatic, hollow cyst-like channels that expanded over time invading the gel, while HUVEC-TIE2-WT formed mildly enlarged lumens ( Fig. 2A and Supplemental  Fig. S2). To further show the effects of the constitutive active TIE2 receptor on the lumen formation, we infected HUVEC with lentivirus expressing a doxycycline-inducible TIE2-L914F (pInducer21-TIE2-L914F). When cells were subjected to doxycycline treatment for 48 hours, they exhibited constitutive TIE2 phosphorylation and activation of downstream effectors AKT and ABL 21 (Fig. 2B). We embedded the HUVEC-pInducer-TIE2-L914F cells in fibrin gel. Without doxycycline administration, lumens formed as in normal HUVEC. At day 7 of lumen formation, doxycycline 1 μg/ml was added to the gel and analysis at days 9, 12 and 14 (2, 5 and 7 days of doxycycline treatment) revealed lumen/vascular area and diameter enlargement as a result of the TIE2-L914F expression (Fig. 2C,D). These data show that HUVEC-TIE2-L914F in the 3D fibrin gel can phenocopy enlarged lumen formation as detected in VM patients.
The TIE2-L914F mutation is sufficient to induce ectatic vascular channel formation in absence of fibroblasts. In the 3D fibrin gel system employed, fibroblasts are an essential component as they release growth factors that enable EC to sprout and subsequently form lumens 31 . Exogenous growth factors such as ANGPT1 (Angiopoietin-1) and bFGF (basic Fibroblast Growth Factor) cannot rescue the lumen formation in the lack of fibroblasts 31 . To determine the potential of HUVEC-TIE2-L914F to form ectatic lumens in stringent conditions, we generated fibrin gels without fibroblasts and/or growth factors. HUVEC and also HUVEC-TIE2-WT failed to sprout and form lumenized structures in the absence of both fibroblasts and growth factors while HUVEC-TIE2-L914F formed massively enlarged channels (Fig. 3A,B). The addition of growth factors in gels without fibroblasts did not rescue the phenotype in the control EC, while HUVEC-TIE2-L914F ectatic channel formation ability was unperturbed (Fig. 3A,B). The presence of fibroblasts in starvation conditions (no growth factors, EBM2) could still support the sprouting and lumen formation ability of control EC (Supplemental Fig. S3). In our assay, HUVEC-TIE2-WT formed mildly enlarged lumens in presence of fibroblasts and growth  EC expressing constitutive active mutant TIE2 form channels with scarce recruitment and contact with pericytes, akin to patients' VM lesions. VM vascular channels are characterized by scarce and irregular pericyte recruitment 18,20,37 . We sought to investigate if the fibrin gel system could mimic this important aspect of the VM aberrant vascular formation. Fibrin gels were assembled with Cytodex ® beads coated with a mixture of EC www.nature.com/scientificreports www.nature.com/scientificreports/ and human retinal pericytes. HUVEC-derived vascular lumens showed significant pericyte coverage, while ectatic channels formed by the HUVEC-TIE2-L914F were contacted only by few, sparse pericytes (P < 0.001) (Fig. 5A,B). Next, pericyte cultures were treated for 4 days with conditioned medium from HUVEC or HUVEC-TIE2-L914F, or with EGM2 as control growth medium. There was no difference in the pericyte proliferation curve treated with HUVEC and HUVEC-TIE2-L914F conditioned medium (Fig. 5C), suggesting pericytes' proliferation is not affected by HUVEC-TIE2-L914F. Next, we investigated if cytokines expressed by HUVEC-TIE2-L914F could affect pericyte coverage in HUVEC-derived vascular lumens. Fibrin gels were set up with Cytodex ® beads coated with HUVEC-GFP and pericytes-BFP, and gels were treated with conditioned medium from HUVEC or HUVEC-TIE2-L914F. This experiment revealed a significant difference (P < 0.05) in the number of pericytes adhering to HUVEC-derived vascular channels treated with different conditioned medium (Fig. 5D,E).
In summary, this 3D in vitro model of VM channel formation recapitulates the paucity of pericytes in the ectatic TIE2 mutant lumens and suggests cytokines released by HUVEC-TIE2-L914F could also influence pericyte coverage in normal, HUVEC-derived vascular structures.

Discussion
In this study we show that fundamental steps during angiogenesis including migration, lumen formation and maturation are largely aberrant in EC expressing a constitutive active mutant form of the TIE2 receptor. We focused on the mutation TIE2 p.L914F, which is the most frequent mutation found in patients affected by VM. Modeling the VM ectatic lumen formation in a 3D fibrin gel system revealed that the mutation p.L914F is sufficient to induce formation of dilated ectatic channels instead of conventional tubular structures. The www.nature.com/scientificreports www.nature.com/scientificreports/ HUVEC-TIE2-L914F-derived channels displayed disrupted apico-basal polarity and scant recruitment and contact with pericytes.
HUVEC-TIE2-L914F exhibited faster wound closure migration compared to HUVEC and HUVEC-TIE2-WT. During this process, HUVEC-TIE2-L914F lost the front-rear polarity and migrated in random directions. Conversely, we showed that cell movement in a non-confluent monolayer was similar in mutant and control EC. www.nature.com/scientificreports www.nature.com/scientificreports/ Previous studies reported that EC expressing the constitutive active TIE2 mutants p.Y897F + p.RL915L exhibit random cell migration movements in monolayer and fail to rearrange in a cobblestone layer, similarly to our results 23

. Other authors reported decreased EC migration in response to VEGF-A (Vascular Endothelial Growth
Factor-A) in cells expressing TIE2 p.R849W 38 . The difference in these results could be explained by the fact that p.R849W is a milder activating mutation found in familial cases of VM 18 .
Unpolarized random migration could lead to aberrant lumen formation 4,39 . In VM, vascular lumens are massively enlarged. We modeled the assembly of TIE2 mutant EC into vascular channels akin to VM histology in a 3D morphogenesis assay. While HUVEC organized into a network of tubular structures, HUVEC-TIE2-L914F generated ectatic hollow cyst-like channels. HUVEC-TIE2-L914F generated these dilated vascular structures even in stringent conditions (no fibroblasts and no growth factors) that impeded HUVEC and HUVEC-TIE2-WT from forming sprouts. It is important to note that in this 3D fibrin gel assay HUVEC-TIE2-WT can form mildly enlarged lumens in presence of fibroblasts, we speculate this is due to the overexpression of the TIE2 receptor that responds to the fibroblast and growth factor stimulation. When the cytokines were removed, overexpression of TIE-WT was not sufficient to drive lumen formation. Comparable to our findings, in an EC spheroid sprouting assay, authors reported enlargement of the cell spheroid during a 16-hour period of time when EC expressed the hyperactivating TIE2 mutations: p.Y897F + p.RL915L, p.L914F or p.R849W. HUVEC-TIE2-L914F generate spheroids bigger in size compared to HUVEC-TIE2-R849W 23 , suggesting the levels of TIE2 activation are correlated to lumen size/enlargement.
In both EC and epithelial cells the establishment of apico-basal polarity is necessary for proper lumen formation [40][41][42] . Apico-basal polarity is regulated by tight and adherens junction organization and is affected by GTPase activity 43,44 . Podocalyxin is a sialomucin that re-localizes in the apical/luminal side of the EC, while basement membrane Collagen IV is a marker expressed in the basal side of the EC [32][33][34][35]45 .
In our study, the subcellular localization of these markers followed the above-mentioned pattern in HUVEC-derived lumens but was completely disorganized in HUVEC-TIE2-L914F-vascular channels where Podocalyxin and Collagen IV colocalized in some cells and/or were randomly distributed. A similar phenotype was reported in an in vitro model of CCM (Cerebral Cavernous Malformation) based on HUVEC knocked-down for the CCM1/KRIT (Krev-interaction trapped protein 1) gene 34 . CCM is a vascular malformation also characterized by enlarged venous channels 46 . Our data further showed upregulation of Podocalyxin in HUVEC-TIE2-L914F compared to control EC. Germline deletion of Podxl in mouse leads to developmental defects and delayed opening of the aortic vascular lumen 5 , while EC-specific deletion results in blood brain barrier disruption during acute inflammation 47 . Podocalyxin overexpression is associated with multiple tumor types and correlates with poor outcome in breast cancer patients 48,49 . In epithelial cells Podocalyxin upregulation is responsible for the recruitment of actin to the apical membrane promoting its expansion 50 . Combined with our data, this suggests that the upregulation of Podocalyxin in the HUVEC-TIE2-L914F has a role in the aberrant apico-basal polarity marker distribution, and that this contributes to the massive lumen expansion.
Pericyte recruitment is essential during angiogenesis to prevent vascular regression and promote vessel maturation 1,7,8 . In VM, the ectatic vascular channels present scarce and irregular perivascular cells 18,28 . When injected in mouse, HUVEC-TIE2-L914F can form enlarged channels lacking pericyte/smooth muscle cell coverage 20 , akin to the human VM. Paucity of perivascular coverage is also found in tumor associated vasculature and in vascular anomalies such as CCM and hereditary hemorrhagic telangiectasia (HHT) 46,[51][52][53][54] . The lack of Pdgfb (Platelet-derived growth factor b) or its receptor Pdgfrβ in mice result in loss of perivascular cell coverage, with dilated vascular channels and EC hyperplasia [55][56][57][58] . In VM, TIE2-mutant EC express low levels of the FOXO1 (Forkhead box O1) target gene PDGF-BB and this could partly explain the decreased pericyte recruitment around the patients' lesions 28 . Here, we show that even in a system where pericytes are co-seeded with EC on the fibrin gel, pericytes do not move along or migrate to wrap the vascular structures formed by the TIE2-mutant EC, despite HUVEC-TIE2-L914F do not affect the pericyte proliferative ability.
Clinical studies postulate that in VM the lack of pericytes allows for the expansion/dilation of the vascular channels by hemodynamic factors and stretching of the walls of the venous spaces 59,60 . Our data suggest that the expansion of the TIE2-mutated vascular channels is independent of the pericyte contribution as our system does not include flow establishment and HUVEC-TIE2-L914F formed hollow channels that expanded in size with time.
The constitutive activation of the TIE2 receptor induced by the VM patient mutations stimulates both the PI3K/AKT and MAPK/ERK1/2 downstream signaling 16,22,23 . PIK3CA (Phosphatidylinositol-4,5-Bisphosphate 3-Kinase Catalytic Subunit Alpha) gene activating mutations have been found in 20-25% of VM patients 16,22,61 . These mutations result in constitutive active AKT signaling, but do not perturb the MAPK signaling pathway, this would suggest that AKT hyperactivation is sufficient to drive the VM pathogenesis. Of interest, in vascular anomalies a number of mutations in different genes could all be grouped by their downstream effectors being the PI3K and/or MAPK pathway 24 . In a recent study, EC expressing a constitutive active mutant HRAS V12 form enlarged sheet-like structures and fail to assemble in elongated tubes in the same 3D fibrin gel assay we utilized 62 . The HRAS-mutant EC phenotype was normalized by a PIK3CA inhibitor BYL719 (Alpelisib) while ERK inhibition result in a failure to undergo lumen morphogenesis and proliferation. This would suggest that in HRAS V12 both pathways can contribute to the phenotype. In VM, future studies are needed to deepen our understanding of the specific contribution of the PI3K and MAPK signaling and to translate this into patient-specific therapeutic options.
In our recently published study, we successfully used this 3D fibrin system to establish the efficacy of mTOR and c-ABL inhibitors in preventing HUVEC-TIE2-L914F derived lumen formation and enlargement 21 . These data further support the use of this system as a tool to study the mechanisms leading to enlarged, ectatic VM lumen and to screen for drugs that could normalize the VM phenotype.

Materials and Methods
Cell culture. HUVEC and retrovirally-transfected HUVEC expressing full-length TIE2-WT or TIE2-L914F were a gift from Dr. Lauri Eklund and Miikka Vikkula, and were previously described 20 . All the experiments with human cells were performed in accordance with the Institutional Biosafety Committee guidelines and were approved by the Cincinnati Children's Hospital Medical Center (CCHMC).

Cell migration and movement analysis.
For the wound migration assay HUVEC, HUVEC-TIE2-WT and HUVEC-TIE2-L914F were grown to confluence in a six-wells plate in EGM2/10%FBS at 37 °C. Upon reaching confluence the cell monolayer was treated with 2 mM hydroxyurea for 4 hours (to prevent proliferation) and scratch/wounds were performed using a sterile pipette tip. After washing off released cells and cell debris, EBM2/0.5%FBS was added. Time-lapse images were taken over a 5 h period using the Nikon Ti-2 Spectra microscope. The wound healing velocity was calculated after collection of sequential time-lapse images and the relative cell migration distance was measured with ImageJ software (NIH).
For the analysis of individual cell movement, cell velocity in non-confluent conditions (30% confluency) and in response to scratch/wound was measured by imaging with a Nikon Ti-2 SpectraX Inverted Microscope and quantified with Nikon NIS-Elements and Image J. The analysis was performed using sequential time-lapse images over a period of 2 hours. In each wound or non-confluent monolayer 10-12 cells were tracked and all experiments were performed in quadruplicates (n = 4).
3D fibrin gel assay. 3D fibrin gel assay was performed as previously described 25 . Briefly, Cytodex ® 3 microcarrier beads (Sigma) were incubated with HUVEC, HUVEC-TIE2-WT or HUVEC-TIE2-L914F at a concentration of 400 cells/ bead for 4 h at 37 °C. The following day, coated beads were resuspended in 2 mg/mL of fibrinogen (Sigma) solution containing 0.15 U/mL of aprotinin (Sigma) at a concentration of 500 beads/ml. Then 0.625 U/ mL of thrombin (Sigma) and 0.5 ml beads/fibrinogen suspension were added per well of a 24-well plate and incubated at 37 °C to allow fibrin clotting. The gels were overlaid with human lung fibroblasts at 2 × 10 4 cells/well and medium was replaced every other day. Where indicated fibroblasts were omitted and EGM2/10% FBS medium substituted with EBM2 (Endothelial basal medium) which does not contain growth factors. In assays conduced with pericytes the Cytodex ® beads were seeded with EC (HUVEC or HUVEC-TIE2-L914F) and pericytes at a 10:1 ratio. Images were acquired with the Nikon A1R LUN-V Inverted Microscope and EVOS cell imaging system (Invitrogen) and analyzed with Nikon NIS-Elements and ImageJ.
Cell proliferation assay. Cells were seeded at 6000 cells per well in gelatin-coated 96-well plates and cultured in EGM-2 supplemented with 10% FBS. Cell proliferation was measured by sulforhodamine B (SRB) assay 63 and the optical density (OD) value was read at 540 nm using SpectraMax i3x Multi-Mode Detection Platform (Molecular Devices).
Immunofluorescence staining. Immunofluorescence staining was performed on cell monolayer, 2 hours after the scratch/wounds was performed. Cells were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) at room temperature for 15 minutes and permeabilized in 0.1% Triton X-100 (Sigma) for 5 minutes.