VE-cadherin junction dynamics in initial lymphatic vessels promotes lymph node metastasis

The dermal lymphatic vasculature is unperturbed in Vegfr2^Y949F/Y949F mice

The blood vasculature in the Vegfr2^Y949F/Y949F C57Bl/6 model develops normally, but the BECs are resistant to the permeability enhancing effect of VEGFA (Li et al, 2016). As lymphatic endothelial cells (LECs) express considerable levels of VEGFR2 (Wirzenius et al, 2007), it is important to understand the consequence of loss of the VEGFR2 phosphosite Y949 and its downstream signaling on lymphatic vessel development and lymphatic function.

To verify that lymphatic vessels in the Vegfr2^Y949F/Y949F mouse develop normally, embryos at embryonic (E) day 14.5 were examined. Neuropilin2 (Nrp2)⁺ lymphatic tip cell numbers and lymphatic vessel density in the dorsal skin were similar in Vegfr2^Y949F/Y949F and WT littermates (Fig 1A–C). Moreover, analyses in 8–10-wk-old mice showed similar lymphatic vessel endothelial hyaluronan receptor 1 (LYVE1)⁺ vessel diameter in the ear dermis of Vegfr2^Y949F/Y949F and WT mice, whereas vessel density was increased in the Vegfr2^Y949F/Y949F genotype (Fig 1D–F). Expression levels of VEGFR2, VEGFR3, and Podoplanin in LECs isolated from postnatal (P) day 10 lungs were not affected by the Y949F mutation (Fig 1G and H). We conclude that elimination of the phosphosite Y949 in VEGFR2 by replacing tyrosine (Y) with phenylalanine (F) did not perturb lymphatic development or the expression of VEGF receptors in LECs.

Tumor growth and lymphatic metastasis in Vegfr2^Y949F/Y949F mice

The VEGFA-resistant vascular barrier in the Vegfr2^Y949F/Y949F mouse blood vasculature correlates with decreased hematological metastastatic spread from RipTag2 neuroendocrine tumors or B16F10 melanoma (Li et al, 2016). To explore the effect of the Vegfr2^Y949F/Y949F mutation on lymphatic vessel barrier function in cancer, DsRed-expressing B16F10 melanoma cells were engrafted intradermally in the ear and analyzed after 7 or 12 d of tumor growth. The ear dermis was chosen as the site of injection as it restricts local growth and promotes spread to sentinel lymph nodes (Li et al, 2016). A low inoculation volume (5 μl) was used to minimize tissue damage and forced metastasis. There was no difference in the growth rate or weight at day 12 of the primary B16F10 tumors between the Vegfr2^Y949F/Y949F and WT mice (Fig 2A and B). However, the weight of sentinel cervical lymph nodes was significantly lower in the Vegfr2^Y949F/Y949F mouse (Fig 2C). In accordance, expression of Tyrp1 (melanocyte-specific gene) was decreased in sentinel lymph nodes of the Vegfr2^Y949F/Y949F mice compared with WT, indicating reduced lymphatic tumor dissemination (Fig 2D).

A similar pattern was observed when challenging mice with EO771-CCR7-tdTomato mammary carcinoma, engrafted in the mammary fat pad of female mice. In this model, C-C chemokine receptor type 7 (CCR7) expression was introduced in the EO771 tumor cell line to promote lymph node metastasis, previously also shown to promote lymph node metastasis in other models of breast cancer (Cunningham et al, 2010). Primary tumor weight at day 20 was similar between Vegfr2^Y949F/Y949F and WT mice (Fig 2E). Although the weight of sentinel inguinal lymph nodes was unaffected (Fig 2F), flow cytometry analysis of tumor cells positive for tdTomato in the inguinal lymph nodes showed significantly fewer tumor cell counts in the Vegfr2^Y949F/Y949F compared with the WT lymph nodes (Fig 2G and H). Of note, immune cells, potentially positive for tdTomato by uptake of tumor debris, were excluded in the analysis.

Lymphangiogenesis occurs preferentially at the tumor periphery in response to tumor-secreted growth factors (Christiansen & Detmar, 2011). The Y949F mutation did not affect tumor-induced lymphangiogenesis, as shown by the unperturbed LYVE-1+ lymphatic vessel density at the tumor periphery in the B16F10-engrafted Vegfr2^Y949F/Y949F and WT mouse ears (Fig S1A and B). Moreover, there was a trend but no significant difference in expressions of Vegfa and Cdh5 in Vegfr2^Y949F/Y949F and WT B16F10 tumors (Fig S1C and D). We conclude that in two different orthotopic cancer models, breast cancer and melanoma, metastatic spread to sentinel lymph nodes was suppressed in mice lacking the VEGFR2 Y949 phosphosite.

Lymphatic drainage and intravasation of tumor cells in Vegfr2^Y949F/Y949F mice

The potential consequence of the Y949F mutation on lymphatic function was addressed by monitoring the clearance of the near-infrared tracer P20D800 by intravital imaging (Proulx et al, 2013) of the mouse ear. The movement of interstitial fluid into the lymphatic vessels (lymphatic clearance) is mediated by both interstitial fluid pressure and the pumping created by the smooth muscle cells surrounding the collecting vessels (Bazigou et al, 2014). Lymphatic clearance and half-life of the tracer was not significantly different between healthy mutant and WT mice (Fig 3A and B). However, after engraftment of B16F10 melanoma, lymphatic clearance of the fluorescent tracer was faster in Vegfr2^Y949F/Y949F mutant mice with reduced half-life compared with WT mice (Fig 3A and B).

Movement of tumor cells into initial lymphatic vessels is dependent on interstitial pressure gradients but also on chemokines produced in the microenvironment and the expression of adhesion molecules on LECs (Shields et al, 2007; Swartz & Lund, 2012). The presence of metastatic melanoma cells within the peritumoral lymphatic vessels was determined by counting the number of DsRed-expressing B16F10 cells, using orthogonal projection of confocal imaging. Tumor cells were more frequently found inside lymphatic capillaries of WT mice at day 7 than in Vegfr2^Y949F/Y949F mice (Fig 3C and D).

The similar lymphatic clearance rate by unchallenged WT and Vegfr2^Y949F/Y949F lymphatic vessels suggests no functional deviation because of the VEGFR2 mutation. When subjected to tumor challenge, the Vegfr2^Y949F/Y949F mutation led to improved lymphatic drainage and reduced transmigration of tumor cells into peritumoral lymphatic vessels, indicating an enhanced lymphatic barrier.

Characterization of lymphatic endothelial junctions in Vegfr2^Y949F/Y949F mice

We hypothesized that the improved interstitial fluid clearance in the melanoma-bearing Vegfr2^Y949F/Y949F ear and the reduced metastatic spread observed in this model may be related to stabilization of VE-cadherin in the lymphatic endothelial junctions. VE-cadherin junctions undergo “zippering” in response to VEGFA (Zhang et al, 2018; Zarkada et al, 2023), that is, conversion to a linear and continuous VE-cadherin immunostaining pattern. By analyzing the pattern of VE-cadherin–positive LEC junctions, it was evident that zippering of the lymphatic junctions was established at the tumor periphery in both WT and Vegfr2^Y949F/Y949F mutant mice (Fig 4A and B). The extent of zippering was slightly lower in the mutant mice, as shown by the extent of VE-cadherin coverage at the cell perimeter (Fig 4B). Increased lymphatic zippering in the tumor-proximal lymphatic vessels compared with vessels distal from the tumor was also evident from the decrease in the number of VE-cadherin fragments and increase in the fragment length (Fig 4C and D).

In BECs, phosphorylation of VE-cadherin at Y685 marks vessels susceptible to increased permeability when appropriately stimulated with agonists such as VEGFA (Claesson-Welsh et al, 2021). Immunostaining for pY685 VE-cadherin using a validated antibody (Jin et al, 2022) showed robust reactivity in dermal lymphatic vessels; however, there was no difference in signal strength between the WT and Vegfr2^Y949F/Y949F ear dermis (Fig S2A and B).

Uptake of interstitial fluid correlates with the dynamic turnover and distribution of VE-cadherin (Trzewik et al, 2001; Schmid-Schönbein, 2003; Baluk et al, 2007), which is reflected in the morphology of VE-cadherin as visualized by immunofluorescent staining. To follow VE-cadherin dynamics in a more sensitive, unbiased manner throughout the cell, we developed an automated classification of the shapes of VE-cadherin fragments based on the aspect ratio and circularity. Four categories of VE-cadherin shapes were defined: category 1—fragments with an aspect ratio above the upper quartile (Q3) and circularity between 0–0.25; category 2—aspect ratio Q2–Q3 and circularity 0.25–0.5; category 3—aspect ratio Q1–Q2 and circularity 0.5–0.75; category 4—aspect ratio below Q1 and circularity 0.75–1 (Fig 4E). The short round shapes in categories 3–4 may result from dynamic turnover of VE-cadherin, that is, internalization (Bentley et al, 2014).

Applying this classifier to the healthy ear dermal LYVE1⁺/VE-cadherin⁺ lymphatic vessels showed that there was no difference in the frequency of VE-cadherin⁺ categories when comparing unchallenged WT and Vegfr2^Y949F/Y949F dermis (Fig 4F). However, the frequency of category 4 VE-cadherin shapes increased, whereas category 3 shapes decreased, in the melanoma-proximal lymphatic vessels compared with healthy tissue in the WT mouse ear (Fig 4G). The more elongated VE-cadherin shapes (categoryies 1 and 2) remained unaffected. In contrast, the VE-cadherin shapes were similarly distributed irrespective of the presence of a tumor in the Vegfr2^Y949F/Y949F mouse (Fig 4H). Consequently, comparing VE-cadherin fragment distribution between the genotypes showed that melanoma-proximal WT lymphatic vessels had an increased proportion of category 4 VE-cadherin shapes compared with Vegfr2^Y949F/Y949F (Fig 4I).

Together, our data identified zippering of button junctions in the tumor-proximal lymphatics. Junctional zippering was accompanied by an increase in pan-cellular VE-cadherin fragments with high circularity and a low aspect ratio in the WT mouse lymphatic endothelium but not in mice expressing the Y949F mutant VEGFR2. These data indicate increased VE-cadherin dynamics in the tumor-challenged WT but not in the Vegfr2^Y949F/Y949F initial lymphatics.

Effect of VEGFA and VEGFC on dermal lymphatic vessels

Hypoxia-induced increase in VEGFA levels in the tumor microenvironment correlates with disintegration of VE-cadherin junctions and dismantling of the blood vascular barrier (Weis et al, 2004; Li et al, 2016). In lacteal and dermal lymphatic vessels, VEGFA has been shown to induce zippering of junctions through activations of VEGFR2 (Zhang et al, 2018; Zarkada et al, 2023). We hypothesized that the zippering of dermal lymphatic junctions and the shape shift of VE-cadherin fragments in the tumor periphery may be related to production of VEGFA and/or VEGFC, which both bind to VEGFR2 (Simons et al, 2016), in the tumor microenvironment. VEGFA secreted by malignant B16F10 melanoma cells vary from picogram to nanogram levels dependent on culture conditions (Collet et al, 2014; Jean et al, 2014; Palazon et al, 2017). Furthermore, serum VEGFA levels in melanoma patients vary from 100–500 pg/ml, values dependent on the disease stage (Ugurel et al, 2001; Osella-Abate et al, 2002; Pelletier et al, 2005; Palmer et al, 2011). Injection of 0.25 ng of VEGFA in 5 μl in the WT ear dermis led to zippering of lymphatic junctions with a reduced number of VE-cadherin fragments and increased fragment length (Fig 5A–D). In Vegfr2^Y949F/Y949Fmice, lymphatic junctions were more zippered than in the WT in basal conditions (Fig 5B), zippering increased further in both genotypes by VEGFA injection (Fig 5A and B), accompanied by a decrease in the number of VE-cadherin fragments/vessel length (Fig 5C and D).

Shape classification of pan-endothelial VE-cadherin fragments was performed to analyze the effect of VEGFA or VEGFC on the dynamics of lymphatic junctions in the healthy tissue. Injection of VEGFA (0.25 ng in 5 μl) in the WT ear dermis led to an increased detection of category 4 VE-cadherin shapes and concomitantly, reduced category 3 shapes at 15 min postinjection, whereas elongated fragments (categories 1 and 2) remained unaffected (Fig 5E). The distribution of VE-cadherin shape categories was restored to basal 30 min after injection of VEGFA (Fig 5F), indicating transient dynamics. In accordance with the observation in the tumor-bearing mice (Fig 4H), VE-cadherin shapes were not altered in response to VEGFA injection in the Vegfr2^Y949F/Y949F dermis (Fig 5G and H). Injection of VEGFC (25 ng in 5 μl) also induced increased frequency of category 4 VE-cadherin shapes in the WT but not in the Vegfr2^Y949F/Y949F dermis (Fig S3A–C).

These results show that similar to the zippering of VE-cadherin junctions in tumor-proximal lymphatics, injection of VEGFA induced zippering of button junctions in dermal lymphatic vessels. The lymphatic vessels in Vegfr2^Y949F/Y949F mice were resistant to VEGFA/VEGFC–induced increase in circular VE-cadherin shapes. Together, these data indicate that the Y949 site in VEGFR2 is dispensable for VEGFA-induced zippering of lymphatic junctions which is in agreement with a recent study (Zarkada et al, 2023). However, importantly, signaling downstream of Y949 in VEGFR2 is required for VEGFA/VEGFC–induced internalization and turnover of VE-cadherin.

VEGFA-mediated VE-cadherin shape change is dependent on Src

Signaling pathways involved in VEGFA-mediated zippering of lymphatic junctions have been revealed in recent studies (Zhang et al, 2018; Zarkada et al, 2023). However, VEGFA-induced signaling regulating VE-cadherin dynamics in lymphatic vessels has not been identified. SFKs have been implicated in VEGFR2-induced modulation of blood vascular VE-cadherin junctions, by phosphorylating VE-cadherin, leading to disruption of VE-cadherin homophilic interactions followed by internalization. The role of c-Src in the regulation of lymphatic junctions was tested in vivo using a c-Src^fl/fl; Cdh5CreERT2 mouse model (Src iECKO) (Schimmel et al, 2020; Jin et al, 2022), in which tamoxifen treatment results in an endothelial-specific deletion of c-Src. No effect of c-Src deletion on lymphatic vessel morphology was observed within the period of observation, in agreement with the c-Src–deficient blood vasculature morphology appears normal (Jin et al, 2022). The Src iECKO mouse failed to respond to VEGFA with an increase in category 4 VE-cadherin fragments, which was established in the Cre negative control ear dermis (Fig 6A and B). Interestingly, with VEGFA treatment, the percentage of elongated VE-cadherin fragment (category 2) was increased in the Src iECKO genotype, compared with the control (Fig 6C), in accordance with more stable lymphatic endothelial junctions when Src activity was suppressed. We conclude that Src expression/activity is required for VEGFA-induced increase of the circular VE-cadherin shape (category 4) in capillary lymphatic vessels.

Mice

The Vegfr2^Y949F/Y949F strain (Li et al, 2016) on the /C57BL/6J background was created using VelociGene technology (Valenzuela et al, 2003) (Regeneron Pharmaceuticals) and maintained by crossing heterozygotes. Cdh5-CreERT2 mice were provided by Ralf Adams (Max-PIanck Institute, Münster, Germany) (Kogata et al, 2006; Wang et al, 2010). c-Src-floxed mice were purchased from the NiceMice National Resource Center for Mutant Mice, Model Animal Research Center, China, and crossed with the Cdh5-CreERT2 mice. Cre activity and gene deletion were induced by intraperitoneal injections to mice with 2 mg tamoxifen (Sigma-Aldrich) for 5 d, and mice were used for experiment on day 7 when the deletion efficiency was about 80% (Jin et al, 2022). All animal experiments were repeated at least three independent times with age-matched WT and mutant mice. Cohorts were chosen to ensure reproducibility and to allow stringent statistical analysis. For the different studies, embryos, pups, and young adults (4–8 wk) were used.

Ethics statement

Animal experiments were carried out in accordance with the ethical permit approved by the Committee on the Ethics of Animal Experiments of the University of Uppsala (permit no. C119/13) and conform to the guidelines from Directive 2010/63/EU of the European Parliament. Anesthesia of the mice in experiments was achieved by inhalation of 2–3% isoflurane. Anesthetic depth was assessed by toe pinch before the procedures. Mice were closely monitored during and after anesthesia. For tissue collection, euthanasia was performed by cervical dislocation.

Antibodies and growth factors

All antibodies used in this study, their origin, and working concentrations are listed in Table 1. Recombinant human VEGFA₁₆₅ (PeproTech) and human VEGFC (Sigma-Aldrich) were used for in vivo intradermal injections and in vitro experiments.

Intradermal B16F10 primary tumor and metastasis model

Lentivirus-transduced DsRed-expressing B16F10 cells (kindly provided by Professor David D Schlaepfer, Department of Pathology, La Jolla, CA) were cultured in DMEM GlutaMAX medium (Gibco) containing 10% FCS (Sigma-Aldrich), washed, and resuspended in Matrigel (Becton Dickinson; BD) for inoculation. Mice were anesthetized with 3% isoflurane (Isoba), and cells injected in the mouse left ear dermis (0.4 × 10⁵ in 5 μl growth factor–depleted Matrigel; BD) using a 30G insulin syringe (Terumo). Tumor volume was measured with a caliper every other day. At day 7 (D7) after inoculation, ears were collected and used for wholemount immunostaining. At D12, the cervical nodes were collected, weighed, and used for immunostaining or quantitative PCR.

EO771-CCR7-tdTomato mammary tumor model analysis

EO771-CCR7-tdTomato cells expressing the C-C chemokine receptor type 7 (CCR7) and the fluorescent protein tdTomato were generated by retroviral transduction. Cells were cultured in RPMI-1640 Hepes (Gibco) supplemented with 10% FBS, 1% L-glutamine, and 1% penicillin-streptomycin (all Gibco) at 37°C, 5% CO₂. Tumors were inoculated by injection of 1 × 10⁵ EO771-CCR7-tdTomato cells in 5 μl PBS (Gibco) into the fourth mammary fat pad of female mice. Mice were euthanized, and tumor-draining lymph nodes were harvested when tumors reached 10–12 mm in one dimension (size restriction based on ethical permit). Single-cell suspensions of tumor-draining lymph nodes were obtained after digestion in RPMI supplemented with 0.2 mg/ml collagenase P (Sigma-Aldrich), 0.8 mg/ml dispase II (Sigma-Aldrich), and 0.05 mg/ml DNase (Sigma-Aldrich) and incubated with the anti-CD16/CD32 (Invitrogen) antibody to block unspecific binding. Surface antigens were stained with the following specific antibodies: BV421 anti-Ter119 (563998, cloneTER-119; BD Biosciences), PerCp-Cy5.5 anti-CD45 (45-0451-80, clone 30-F11; Thermo Fisher Scientific), PerCp-Cy5.5 anti-CD11b (45-0112-82, clone M1/70; Thermo Fisher Scientific). Dead cells were excluded by SYTOX Blue dead cell staining (Invitrogen). CountBright Absolute Counting Beads (Life Technologies) were added to each sample to calculate total cell numbers. Data were acquired on a BD FACSAriaIII flow cytometer (BD Biosciences) and analyzed with FlowJo software (version 10.6.1; FlowJo, LLC).

Intravital imaging of lymphatic clearance

Lymphatic vessel drainage function was assessed by measuring the clearance over time after intradermal skin injection of the infrared probe P20D800 (Karaman et al, 2015). Mice were anesthetized with isoflurane (2%), and 3 μl of 3 μM P20D800 was injected intradermally in the ears with a 30G insulin syringe (Terumo). For mice with D7 tumors, injection was performed in the peritumoral area. The mice were then positioned in a whole-animal fluorescence imaging system (NightOWL II; Berthold Technologies), and images were acquired with the following imaging settings: λ_ex: 745 nm, λ_em: 800 nm, and an exposure time of 4 s. Subsequent images were acquired of the ears at 1, 2, 3, 4, 6, and 24 h after injection. Mice were allowed to wake up and move freely between imaging time points. Fluorescence signal intensities were adjusted to baseline ear signals before injection of tracers to calculate tissue enhancement values. The fluorescence intensity values over time were fit to a one-phase exponential decay model in GraphPad Prism 7.0 software with lymphatic clearance expressed as decay constant k (expressed in h⁻¹) or as half-life (expressed in h) using the following equations:Normalized Fluorescence Intensity=e−kt(1)HalfLife=In(2)/k(2)

Quantitative PCR

Cervical nodes from B16F10 tumor-bearing mice at D12 were dissected to remove the fat/connective tissues and weighed. RNA was extracted and purified from the nodes using an RNeasy mini kit (Qiagen). RNA concentrations were measured in a NanoDrop spectrophotometer (Thermo Fisher Scientific) and adjusted to equal concentrations, followed by reverse transcription using SuperScript III (Thermo Fisher Scientific). Quantitative real-time PCR (qRT-PCR) was performed on a Bio-Rad CFX96 real-time PCR machine using SsoAdvanced SYBR Green Supermix (Bio-Rad). The housekeeping gene Gapdh (Mm99999915_g1; Thermo Fisher Scientific) was used as an internal control. The comparative Ct method was used to calculate fold difference in gene expression. For further analysis and data visualization, basal gene expression in samples was expressed as the fold-change in gene expression as compared with the unaffected inguinal node. The primer sequences used for Tyrp1 were as follows:

In vivo VEGFA/C injections

Mice were anesthetized with isoflurane (2%), and 5 μl of 50 pg/μl VEGFA or 5 μl of 5 ng/μl VEGFC were injected into the left ear dermis using a 30G insulin syringe (Terumo). An equal volume of PBS was injected into the right ear as a control. The injection site was visually marked using a marker pen. Ears were collected after 15 and 30 min, followed by fixation in 4% PFA and whole-mount immunostaining.

Isolation of LECs

A modified protocol for LEC isolation was followed (Frye et al, 2018). Mouse embryos were collected at E18.5 and placed in ice-cold DMEM. The dorsal skin was dissected and transferred to ice-cold HBSS supplemented with 10 mM Hepes and 5% FBS. The dissected skin was digested in a mixture of DMEM, 20% FBS, 10 mM Hepes (all from Gibco), 2.5 mg/ml collagenase II, 2.5 mg/ml collagenase IV, and 1 mg/ml deoxyribonuclease (all from Worthington) for 30 min at 37°C, pipetting with a wide-bore transfer pipette every 5 min to assist tissue dissociation. Skin cell suspensions were filtered through a 40-μm disposable cell strainer (BD Falcon), and cells were centrifuged at 300g for 10 min, suspended in cold PBS with 0.5% BSA, 2 mM EDTA, and incubated with mouse CD45 MicroBeads (MACS; Miltenyi Biotec) for 15 min at 4°C. Labeled cells were magnetically separated (MACS separator), and CD45-negative cells were collected. The cells were then washed once in PBS with 0.5% BSA, 2 mM EDTA, and first incubated with a primary rabbit anti-mouse LYVE-1 antibody (#11-034; AngioBio) for 40 min at 4°C. Subsequently, the cells were labeled with Anti-rabbit IgG MicroBeads (MACS; Miltenyi Biotec) for 15 min at 4°C and magnetically separated (MACS separator). The isolated LYVE-1–positive cells were lysed in RIPA buffer, and proteins were separated by SDS–PAGE (4–12% gradient gel) (Thermo Fisher Scientific), transferred to nitrocellulose membranes (GE Healthcare), and incubated sequentially with primary and HRP-conjugated secondary antibodies (Table 1). Signals were detected using enhanced chemiluminescence (Cytiva), and images retrieved using Bio-Rad ChemiDocMP and were analyzed using Image Lab software.

Immunostaining

For adult tissues, ears were split in half, and the inner part of the ear (without cartilage) was used for analyses. For ears with tumors at D7, after splitting the ears and clearing cartilage, the tumor was removed, and ears were fixed in 4% PFA and permeabilized with 0.3% Triton-X/100 for 10 min. After incubation overnight at 4°C in 5% nonfat dry milk/0.3% Triton-X/100 in PBS, tissues were incubated with primary antibodies overnight at 4°C and with secondary antibodies for 2 h at RT before mounting on glass slides. For pY685 VE-cadherin staining, ears were first fixed in 2% PFA for 1 h at RT before the staining procedure. E14.5 embryo back skins were dissected and fixed in 4% PFA. Immunostaining was proceeded as for the adult skin. For lymph node staining, tissues were fixed in 4% PFA overnight at 4°C and dehydrated subsequently in 30% and 15% sucrose at 4°C. Tissues were sectioned in 5-μm sections and immunostained.

Imaging, image analysis

Microscopy was done with a Leica TCS SP8 Confocal Microscope with PMT-HyD detectors and LAS X Navigator software version 3.5.2.18963. Images for vessel density and diameter were acquired using a 10x dry/NA 0.3 objective, whereas all other images were obtained using a 63x oil/NA1.3 objective with a further zoom of 2x for capturing metastatic cells and junctions. At least 5–7 fields of view (FOV) per sample were obtained, and all image analyses were performed using ImageJ2 software from NIH. AngioTool (Zudaire et al, 2011) was used to calculate vessel density, which was normalized to the vessel area/FOV, whereas vessel diameter was manually measured using the line tool (ImageJ). Protein intensity measurements were normalized to area of measurement.

Shape analysis of VE-cadherin fragments

For VE-cadherin junctional quantification, the images were cropped into 20 × 20-μm blocks, and each block was subjected to segmentation. Objects identified post segmentation were measured for values of circularity and the aspect ratio. The data were exported into MATLAB for further analysis. Using a binning algorithm, in which the boundaries of the bins are defined by the circularity and aspect ratio values, the objects were classified into the four categories, and the percentage of each junctional class was calculated on the total number of junctions per FOV. Significance was calculated using the one-tailed Wilcoxon rank-sum test.

Statistical analysis

A t test or Mann–Whitney test was performed to compare means between two groups. Two-way ANOVA was used to compare tumor growth between WT and mutant mice. Two-way ANOVAs with the Tukey’s post hoc test was performed when two factors were involved, for example, treatment and junction type. For nonparametric comparison, the Wilcoxon sign-rank test was used. GraphPad Prism 10.1.0 was used for statistical analyses. All quantifications are represented as mean ± SEM. The threshold for statistical significance was set at 0.05.