The angiopoietin/Tek system has become a target of growing interest in the development of cancer therapeutics [
]. Angiopoietin-1 (Angpt1) is an activator of tyrosine kinase receptor Tek (also called Tie2) expressed mainly on endothelial cells. Tek activation and phosphorylation results in downstream signaling promoting vascular maturity and endothelial cell survival [
]. Angiopoietin-2 (Angpt2) antagonizes Angpt1 binding and thus Tek signaling in endothelial cells [
], but can also act as a weak agonist in the absence of Angpt1 [
]. Angpt2 may also have an agonistic role in lymphatic vessels [
]. Angpt2 can activate integrins, leading to endothelial destabilization [
], an effect that may be increased in situations of low Tek expression [
]. In regards to vascular leakage, Angpt1 counteracts hyperpermeability induced by several leakage promoting stimuli [
] while Angpt2 weakens the vascular barrier [
It is known that Angpt2 is elevated in many human cancers [
] and preclinical studies of anti-Angpt2 agents often demonstrate additive anti-angiogenic effects on primary tumor growth when combined with inhibitors of the VEGF pathway (reviewed in [
]). Furthermore, Angpt2 blocking antibody has been shown to inhibit metastatic dissemination to the lung in part by enhancing endothelial cell-cell junction integrity [
]. Increased Tek activation has been shown to decrease metastasis in mice in different experimental models of cancer [
]. Wu et al. tested the anti-metastatic potential of vasculotide, a purported Angpt1 mimetic, in experimental metastasis. Tumor cells were injected directly into the venous circulation thus modelling the late stages of metastasis, i.e. extravasation and seeding into distant organ. Vasculotide reduced human breast cancer cell extravasation to the lung, but failed to inhibit human colon cancer extravasation to liver and lymphatics as well as human renal cancer cell extravasation to lung [
]. Interestingly, vasculotide also delayed dissemination of spontaneous lung metastases from orthotopic breast cancer xenographs without affecting primary tumors [
]. Goel et al. showed anti-metastatic effects in vivo utilizing an inhibitor of vascular endothelial protein tyrosine phosphatase (VE-PTP), AKB-9778 [
]. Normally, VE-PTP deactivates Tek, thus an inhibitor of VE-PTP would sustain Tek activation. VE-PTP inhibition delayed the early phase of mammary tumor growth and slowed growth of lung metastases. Park et al., recently showed that administration of an antibody, ABTAA (Angpt2-binding and Tek activating antibody), resulted in normalization of tumor vessels, reduced tumor growth, reduced metastasis, and enhanced drug delivery [
]. In contrast, adenoviral overexpression of Angpt1 in mice facilitated tumor cell dissemination and metastasis establishment [
In the clinic, the dual Angpt2 and Angpt1-neutralizing peptibody, trebananib (AMG386), recently failed to improve overall survival when combined with paclitaxel in patients with recurrent platinum-sensitive ovarian cancer in the phase III TRINOVA-1 trial, despite earlier improved progression-free survival [
]. Trebananib also failed in phase II trials involving metastatic gastro-esophageal [
], colorectal [
], and metastatic clear cell renal carcinomas [
]. Although Trebananib has a higher affinity for Angpt2 compared to Angpt1, the inhibition of Angpt1 is thought to contribute to its lack of effect. It is evident that more studies are needed to define how Angpt1, Angpt2 and Tek act in tumor growth and what their roles are in metastasis in order to develop better therapies.
To clarify the role of Angpt1 and Tek in tumor metastasis, we utilized doxycycline-inducible conditional Angpt1 and Tek knockout mice. We investigated how Angpt1 deficiency affected tumor growth and lung metastasis by crossing these mice to the MMTV-PyMT transgenic mice that develop mammary tumors and lung metastasis. To investigate the initial phase of metastasis, extravasation, we also performed intravenous injection of tumor cells in Angpt1 and Tek deficient mice to evaluate dissemination to the lung. Overall, we found that loss of Angpt1/Tek leads to increase distant metastasis without affecting primary tumor growth.
Mice & breeding
Floxed Angpt1 and floxed Tek mice, were crossed with a ROSA-rtTA/tetO-Cre bitransgenic mice to generate inducible whole body knockout of
) upon administration of doxycycline in the drinking water as previously described [
]. In short, knockout was induced at embryonic day 16.5 by administration of doxycycline as above in the pregnant dam’s drinking water until weaning. Controls (WT) were Angpt1
/w or Tek
/w littermates with ROSA-rtTA, tetO-Cre. All mice received doxycycline.
Transgenic mice expressing polyomavirus middle T (PyMT) oncogene under control of the mouse mammary tumor virus long terminal repeat (MMTV) [
] were crossed with
, all mice were backcrossed >6 generations to FVB. Both male and female (50/50) mice on a mixed background and 8–12 weeks old were used for all other experiments. Littermate mice negative for floxed alleles were used as controls (WT). All animal experiments were approved by the local ethics review committees of Mount Sinai Hospital (Toronto, ON, Canada), Northwestern University (Chicago, IL), and Uppsala University (Sweden, ethical # C122412/13 and C99/15).
Mice were genotyped by PCR using the following primer pairs; Angpt1 flox (for 5′-CAATGCCAGAGGTTCTTGTGAA and rev 5′-TCAAAGCAACATATCATGTGCA, Angpt1 wt 233 bp, flox 328 bp), Angpt1 del (for 5′-CAATGCCAGAGGTTCTTGTGAA and rev 5′-TGTGAGCAAAACCCCTTTC, 481 bp), ROSA-rtTA (for 5′-GAGTTCTCTGCTGCCTCCTG and rev 5′-AGCTCTAATGCGCTGTTAAT), general Cre allele (for 5′-ATGTCCAATTTACTGACCG and rev 5′-CGCCGCATAACCAAGTGAA, 673 bp), Tek wt (for 5′- TCCTTGCCGCCAACTTGTAAAC and rev 5′- AGCAAGCTGACTCCACAGAGAAC, 175 bp), Tek flox (for 5′- TCCTTGCCGCCAACTTGTAAAC and rev 5′- AGCAAGCTGACTCCACAGAGAAC, 604 bp) and PyMT (for 5′- GGAAGCAAGTACTTCACAAGGG and rev 5′- GGAAAGTCACTAGGAGCAGGG, 530 bp).
Tumor growth and metastasis in MMTV-PyMT-transgenic mice
Female 6-week old MMTV-PyMT-
n = 4), MMTV-PyMT-WT (
n = 5),
n = 7) and WT (
n = 10) mice were used in the study. All MMTV- PyMT mice were heterozygous for MMTV-PyMT. Bodyweight was recorded weekly and mice were euthanized at 16 weeks of age. Weight and volume of individual mammary tumors were measured. Tumor volume was calculated using the formula V = (L x W x W)/2. Tumors and lungs were fixed in 10% formalin for 4 h and then embedded in paraffin. Paraffin sections from 4 levels of the lungs were stained with rat-anti-PyMT (Santa Cruz) and scanned using NanoZoomer (Hamamatsu). Lung tumors were counted and the tumor area was calculated from the measured diameter of individual tumors and compared to total lung area using NanoZoomer software (Hamamatsu).
B16F10 melanoma transfection and characterization
B16F10 melanoma cells (ATCC) were maintained in DMEM (10% FCS, 2 mM L-glutamine, 100 U/ml penicillin/streptomycin) at 37°C in a humidified 5% CO
2 incubator. For the generation of stable EGFP expressing B16F10 melanoma cell, 1 × 10
6 cells were plated per well (9.6 cm
2/well) in a 6-well plate and transfected 16 h after with 45 μg of PB-EGFP along with 1 μg PBase using ExGen500 (R0511, Fermentas) according to the manufacturer’s protocol. Stable clones were derived after 2 weeks of selection using 1 μg/ml puromycin. Cells were then trypsinized and sorted by FACS to collect the 10% of cells with highest GFP signal; these cells were further expanded for experiments. Cells were positively identified as melanoma cells by their deposits of melanin in vitro and in vivo. Cells were tested negative for
Mycoplasma, and cells were typically used in experiments within 4 passages after thawing.
Gene expression analysis was done on B16F10 cells to investigate if they express components of the Angpt/Tek system and compared to expression in lung of adult wildtype mice. Trizol (Invitrogen) was used to extract mRNA according to the manufacturer’s protocol, followed by cDNA synthesis using iScript reverse transcription supermix (BioRad). Real time PCR was performed using 100 ng of cDNA with iTaq universal SYBR Green supermix (BioRad) and appropriate primers on a CFX-96 Real Time system (BioRad). Expression results were normalized to endogenous control
and relative quantification was done using the Livak method (2
]. The following primer pairs were used for analysis;
(for 5′- GGCTATAAGTTCTTTGCTGACCTG and rev 5′- AACTTTTATGTCCCCCGTTGA),
(for 5′- GGGGGAGGTTGGACAGTAA and rev 5′- CATCAGCTCAATCCTCAGC),
(for 5′- GATCTTCCTCCAGCCCCTAC and rev 5′- TTTGTGCTGCTGTCTGGTTC),
(for 5′- TGGAGTCAGCTTGCTCCTTT and rev 5′- ACCTCCAGTGGATCTTGGTG),
(for 5′- CAGGCTGCTGTAACGATGAA and rev 5′- CTATGTGCTGGCTTTGGTGA) and
(for 5′- TGAGTGGCTGTCTTTTGACG and rev 5′- CGCACACAGCAGTTCTTCTC).
To investigate if Angpt1 could affect migration and behavior of B16F10 melanoma cells we performed in vitro studies. B16F10 cells (40,000 and 20,000 cells) were seeded in a 96-well plate and allowed to grow O/N at the same conditions as above. A scratch was made in each well using Woundmaker (Essen Bioscience), and 2000 ng/ml of Angpt1 (130–06, Peprotech) was used to stimulate half of the wells. Wells were imaged every 20 min for 24 h using Incucyte Zoom (Essen Bioscience) and migration was calculated using the manufacturer’s software.
B16F10 Melanoma tumor growth and metastasis
To study primary tumor growth, 1 × 10
6 B16F10 cells was injected subcutaneously (s.c.) at 2 sites on the flank of 7
Δ/Δ mice and 5 WT mice. Tumors were measured with calipers to calculate tumor volume and mice were euthanized 15 days after injection.
To study tumor metastasis, 1 × 10
5 B16F10 cells were injected in the dorsal tail vein of 20
Δ/Δ mice and 19 WT mice. Mice were euthanized 21 days after injection, and lungs were fixed and cut in 1 mm sections were tumors were counted using a dissection microscope. A similar experiment was done in 14
Δ/Δ mice with 16 WT C57 mice as controls. Earlier time points were also used where 2 × 10
6 B16F10 cells were injected and tissue harvested after 24 h in
Δ/Δ and WT mice (
n = 6 for both groups). At the 24 h time point lungs were viewed at 40× using a fluorescence stereo zoom microscope (AZ100, Olympus). GFP-positive cells were quantified from images using Elements software.
Vascular leakage was evaluated using cadaverine (~1 kD) conjugated to Alexa Fluor 555 (Thermo Fisher Scientific). Cadaverine (0.1 mg) was injected via the tail vein in two groups of
Δ/Δ mice and two groups of WT mice (
n = 4 for each group). Cadaverine was allowed to circulate for 10 min before injection of 1 × 10
6 B16F10 cells in one
Δ/Δ group and one WT group. After 4 h, all 4 groups of mice were perfused with HBSS to clear out cadaverine in blood vessels. One lobe was weighted and homogenized in PBS, followed by measurement of fluorescent signal in a plate reader. Other parts of the lungs were fixed and imaged in a dissection microscope at 60× followed by quantification of GFP-positive pixels using Adobe Photoshop.
Integrin signaling was investigated in another set of experiments. Lungs from WT and
Δ/Δ at baseline and 4 h after tail vein injection of 1 × 10
6 B16F10 cells were dissected and studied. Protein from lungs was extracted by homogenizing tissue in RIPA buffer (Thermo Fisher Scientific) containing protease and phosphatase inhibitors (PhosSTOP and Complete, Roche). Following incubation at 4°C and centrifugation, supernatant was collected and measured for protein concentration using a BCA assay (Pierce), aliquoted and stored at −80°C. For Western blot analysis, 20 μg denatured protein samples were separated on 4–20% MiniProtean gels (BioRad) and then transferred to PVDF membranes. Blots were blocked with 5% BSA and incubated with primary antibodies, rabbit anti-pFAK(397) (44-625G, Thermo Fisher Scientific) and rabbit-anti-β-actin (4967, Cell Signaling). After washing and incubation with anti-rabbit HRP-conjugated secondary antibody, proteins were visualized using ECLplus detection reagents (GE, Uppsala, Sweden).
Macrophages in the lung were evaluated by FACS 3 weeks after tail vein injection of 1 × 10
5 B16F10 cell in
Δ/Δ and WT mice. Lungs were dissected and digested into single cell suspension by incubation in Hank’s Balanced Salt Solution (HBSS) containing 1 mg/ml Collagenase I, 1 mg/ml Collagenase IV, 1 mg/ml Collagenase V, and 1 U/ml DNAse I for 45 min at 37°C. The cells were then washed 3 times by centrifugation at 500 x g for 5 min and exchange of buffer, HBSS with 3% BSA and 1 mM EDTA. Lung suspensions were stained with anti-mouse CD36 (BD Bioscience) for macrophages and anti-mouse CD206 (BD Bioscience) to identify changes in macrophage polarization between
Δ/Δ and WT mice in a BD FACS Calibur. In addition, staining with anti-Tek (124,007, Biolegend) was done. Mean fluorescence intensity (MFI) was calculated using BD FlowJo software (BD Bioscience). Quantification of macrophages labelled with Isolectin B4 was done on vibratome sections of lung using a Leica SP8 confocal microscope.
Electron microscopy and Microsphere experiments
To investigate if the diameter of lung capillaries were different in
Δ/Δ mice at baseline we utilized performed measurements on micrographs and studied the distribution of microspheres in WT and
For electron microscopy, lungs from 4
Δ/Δ and 4 WT mice were harvested, cut in 1 mm cubes, and immersion fixed in Karnovsky’s fixative (2.5% paraformaldehyde and 2% glutaraldehyde in 0.05 M Na-cacodylate buffer pH 7.2). Tissue were post fixed in 1% OsO4 for 1 h, dehydrated in alcohol and embedded in epoxy resin and heat-cured. Ultrathin sections (~50 nm) were contrasted with lead citrate and uranyl acetate and examined in a Tecnai G2 electron microscope. Micrographs were taken at 4200× and capillaries defined as a vessel containing 1 red blood cell. At least 50 micrographs were taken from each animal and the cross sectional area of capillaries was measured using ImageJ. An average was calculated from each animal which was then used to calculate the group average.
To investigate microsphere distribution, 1 × 10
5 fluorescent microspheres (FluoSpheres, Invitrogen) were injected in the dorsal vein of 5 WT and 5
Δ/Δ mice. The microspheres had a diameter of ~15 μm which is similar to B16F10 melanoma cells. Mice were euthanized 4 h after injection, and lungs were fixed O/N at 4 °C in 10% formalin. Vibratome sections of 100 μm thicknesses were counterstained, mounted and imaged in a confocal microscope (SP8, Leica). The number of microspheres was counted on tile scan images of whole lung sections using ImageJ and then expresses as microspheres/area lung tissue.
Attachment assay of B16F10 melanoma cells
To investigate if the attachment of B16F10 cells to endothelial cells could be affected by Angpt1 we performed attachment assays. Human umbilical vein endothelial cells (HUVECs; passage 5–7) were routinely cultured in gelatin-coated tissue culture flasks in EGM-MV medium. For experiments, HUVECs were seeded in 24-well plates (50,000 cells/well) and allowed to reach >90 confluency. HUVECs were pre-incubated with 1000 ng/ml Angpt1 (ALX-201-314-C050, Enzo Life Sciences) for 4 h before adding B16F10 melanoma cells to wells (10,000 cells/well). The control experiments (Angpt1-) were performed without Angpt1. Cells were then incubated at 37°C with occasional movement of the plate for 10 min. Cells were then washed carefully twice with PBS to remove unattached B16F10 cells, fixed in 10% formalin and stained with Hoechst 33,258. Each well was imaged and GFP positive cells (B16F10 cells) were quantified. Experiments were performed 8 times with four replicates for both conditions.
RNA sequencing data
Extraction of RNA-seq data was done for two different datasets to investigate gene expression. A published study was utilized to look at expression levels in late stage carcinoma tumors of mammary glands from MMTV-PyMT mice compared to FVB mammary gland [
]. The data was downloaded from NCBI GEO database (accession number: GSE76772).
We also extracted data from RNA-seq experiments from the lungs of adult WT (
n = 3) and
n = 4) mice (unpublished data, Jeansson lab). In these experiments, a cDNA library was made using SMARTer Stranded Total RNA Sample Prep Kit (Clontech). Sequencing was performed on an Illumina HiSeq 2500.
Data are expressed as mean ± SEM unless otherwise stated. Statistical analysis was performed using 2-tailed Student’s t-test to analyze statistically significant differences between groups. Logarithmic values were used in the case of a skewed distribution. A
p < 0.05 is considered to be statistically significant.
In the current study we investigate both primary tumor growth and metastasis in Angpt1 deficiency. We found that primary tumor growth is not affected in Angpt1 deficient mice utilizing both the spontaneous MMTV-PyMT mammary tumor model and subcutaneous flank injections of B16F10 melanoma cells.
The metastatic process comprises several events, including tumor invasion, intravasation of tumor cells, their circulation and arrest in capillary beds, extravasation into the distant organ and colonization [
]. In the current study, experiments with MMTV-PyMT mice allowed for auditing of tumor growth and metastasis in a spontaneous model. Angpt1 knockout MMTV-PyMT mice showed a significant increase in metastasis to the lung compared to control MMTV-PyMT mice, without affecting primary tumor growth (Fig.
). The MMTV-PyMT model provides a reliable model of human disease and progression from noninvasive to invasive cancer [
]. Although a good model, there are some challenges when breeding it to a transgenic system for conditional knockout. Firstly, the onset of tumors is strain and gender dependent, thus the Angpt1 knockout mice was backcrossed to FVB for >6 generations and only females were used for the experiments. In this model, it is also difficult to investigate which events in the metastatic process that are affected in Angpt1 knockout leading to increased metastasis.
To study in more detail the later part of metastasis, i.e. attachment, rolling and extravasation, we performed tail vein injections of B16F10 melanoma cells and investigated their seeding to the lung 4 h, 24 h and 3 weeks after injection. We found that Angpt1 deficient mice had significantly more tumor cells in the lung at all time points (Figs.
In the experimental metastasis model, the wildtype B16F10 cells were injected in Angpt1 knockout mice. To rule out that the increased seeding of B16F10 cells in lungs of Angpt1 knockout mice depended on a change in the tumor cells themselves we performed some characterization studies of B16F10 cells. In vitro assays of B16F10 cells showed that were unresponsive to changes in Angpt1 concentration (Fig.
). In addition, B16F10 cells do not appear to express
Angpt1 has several well-known anti-inflammatory properties [
] while Angpt2 is pro-inflammatory and has been shown to activate Tek-positive tumor associated macrophages [
]. We therefore investigated if changes in macrophages contributed to increased metastasis to the lung in Angpt1 deficient mice. We found no differences in macrophage polarization or differences in Tek-positive macrophage populations after tail vein injection of B16F10 melanoma cells (Fig.
). The total number of macrophages was also similar; however, other inflammatory cells were not investigated.
To investigate the mechanism for the increased seeding to the lung in Angpt1 knockout mice we performed a number of experiments utilizing B16F10 cells. One explanation for an increase in lung seeding could be increased vascular leakage. We therefore investigated if blood vessels were leaky at baseline and after injection of B16F10 melanoma cells utilizing a small fluorescently labelled tracer, Cadaverin (1 kDa). Cadaverin is often used to investigate vascular leakage [
]. No increase in leakage was seen which is in accordance with previous studies at baseline in Angpt1 deficient mice [
]. There are some limitations to this leakage method in our B16F10 injection model. The cadaverin is injected and allowed to circulate before tumor cells are injected. To remove intraluminal cadaverin the mouse is perfused with PBS followed by fixation. If tumor cells are clogging capillaries this would decrease the removal of intraluminal cadaverin and thus give a higher fluorescence signal for cadaverin in lung homogenates.
Cancer cells have been shown to slow down and arrest in capillaries of similar diameter as that of the tumor cells, suggesting that they first become physically restricted before forming stable attachments [
]. The Angpt1/Tek system has been implicated in vessel diameter regulation. Goel et al. recently showed that inhibition of VE-PTP increased Tek activity and inhibited several stages of tumor progression and metastasis [
]. Apart from structural and functional normalization of tumor vessels, VE-PTP inhibition resulted in an increase in vessel diameter leading to improved tumor perfusion and reduced hypoxia [
]. Several studies in mice have shown an increase in vessel diameter when activating of Angpt1/Tek signaling through different methods [
]. Also, venous malformations are linked to a gain of function mutation in Tek [
]. Hence, it is possible that the loss of Angpt1 in our model would decrease capillary diameter and change the interaction between vascular endothelial cells and B16F10 cells, thus increasing attachment and extravasation. We therefore investigated cross-sectional area in electron micrographs of lung capillaries and found it to be similar in Angpt1 deficient mice and control mice (Fig.
). To further study if a smaller capillary diameter could explain the increased seeding of tumor cells to the lung we injected microspheres of similar size as the tumor cells (15 μm) intravenously. Microspheres showed a similar distribution in controls and Angpt1 deficient mice (Fig.
Angpt2 has been shown to induce endothelial destabilization through binding to β1-integrin in situations of low Tek expressions [
]. Induced endothelial expression of Angpt2 leads to increased lung metastasis through reduced junctional Tek localization and increase β1-integrin signaling [
]. It should be noted that Angpt2 binds with a significantly higher affinity to Tek compared to integrins, perhaps explaining why low Tek may be required for Angpt2-integrin signaling to occur. In the current study, we could not find any changes in the expression of Angpt2 and Tek in
mice, or any changes of FAK phosphorylation at Tyr397 at baseline or after tail vein injection of B16F10 cells (Table
). This suggests that integrin signaling is not changed in
mice, hence not the mechanism for the increase in lung seeding of B16F10 cells in.
Angpt2 promotes inflammation by inducing vascular leakage and by increasing the expression the adhesion molecules Icam1 and Vcam1 [
], while Angpt1 reduces leukocyte adhesion by reducing the same factors [
]. Experiments with HUVECs and attachment of B16F10 cells in vitro showed that the presence of Angpt1 decreased attachment of B16F10 cells to the endothelial cells (Fig.
). However, we could not find any differences in
expression in lungs of Angpt1 knockout mice compared to WT mice (Table
). Future studies are needed to characterize the expression of these molecules in the endothelial compartment of the lungs. If other adhesion factors could be important for attachment of B16F10 cells to the endothelium in regards to the Angpt/Tek system remains to be investigated.
We thank K. Harpal (Lunenfeld-Tanenbaum Research Institute, Toronto) for histologic staining, D. White (University of Toronto) and A. Bang (Lunenfeld-Tanenbaum Research Institute, Toronto) for FACS, Anders Ahlander at SciLife Lab BioVis at the Rudbeck laboratory (Uppsala University) for electron microscopy, and Liqun He (Uppsala University) for bioinformatics analysis.
M.J is funded by the Swedish Research Council grant 521–2012-865, a Young Investigator grant from Department of Immunology, Genetics and Pathology, the Åke Wiberg Foundation, the Magnus Bergvall Foundation and Venture Sinai Women 2. S.E.Q holds the Charles Mayo Chair of Medicine at the Feinberg School of Medicine and a Finnish Distinguished Professorship at the Oulu Biocenter. S.E.Q is funded by NIH, NHLBI grant HL1241200, CIHR grants M0P62931 and M0P77756, E-rare Joint Translation Call (JTC 2011) for European Research Projects on Rare Diseases, and TF grant 016002. O.V is funded by NIH grants R01CA172669 and R24EY022883. I.P. Michael was supported by a fellowship from the Canadian Institutes for Health Research (CIHR). These funding sources provided support for the conduct of research; they played no role in study design, collection, analysis and interpretation of data, preparation of manuscript, or decision to submit the article for publication.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. RNA-seq data from carcinoma tumors of mammary glands from MMTV-PyMT mice compared to FVB mammary gland was analyzed from a published study accessed downloaded from NCBI GEO database (accession number: GSE76772). RNA-seq data from lung of WT and
Δ/Δ mice can be obtained by contacting the corresponding author.