Background
Current angiogenic inhibitors targeting VEGF signaling show therapeutic efficacy in many aggressive tumors, but fail to provide enduring clinical benefit in most cases [
1‐
7].
In addition, the VEGF inhibitors with documented anti-tumor efficacy have been found in mouse models to be prone to elicit tumor adaptation and progression to stages of greater malignancy, with heightened invasiveness and in some cases increased lymphatic and distant metastasis [
8]. Therefore, there is a need to validate novel therapeutic targets alongside VEGF signaling inhibition. In this context, Dll4/Notch signaling appears as a promising candidate.
Acting downstream of VEGF signaling, Dll4/Notch signaling essentially contributes to proper vascular remodeling during embryonic vascular development [
9‐
11]. The endothelial ligand Dll4 interacts with Notch 1 receptors of adjacent endothelial cells, triggering γ-secretase proteolytic cleavage of Notch intracellular domain (NICD), which subsequently translocates to the nucleus as a complex with the recombination signal binding protein Jκ (RBP-Jκ) and activates effector genes including the members of the Hes and Hey families of basic helix-loop-helix transcription factors [
12] and EphrinB2 [
13]. In the postnatal period Dll4 has a low level of expression in quiescent blood vessels of normal tissue [
14]. However, its up-regulation is a hallmark of proliferating tumor vessels in both preclinical murine models [
10,
15‐
18] and different human malignancies [
14,
19‐
21]. Besides, Dll4 expression was identified in malignant cells of different types [
22], and the ligand was shown to have an impact on colon cancer stem cell frequency by suppressing apoptosis of tumor cells [
23,
24]. Nevertheless, vascular endothelium represents the most prominent and constant site of Dll4 expression within tumors.
Targeted
Dll4 allele deletion, local overexpression of Dll4/Notch-blockers, systemic application of soluble Dll4/Notch-inhibitors and DNA vaccination were found to result in a significant suppression of tumor growth in numerous preclinical models, including malignancies resistant to VEGF inhibitors [
15‐
17,
25‐
28]. Tumor growth retardation due to Dll4/Notch inhibition is associated with an apparently paradoxical increase of endothelial proliferation, migration and subsequent tumor vessel density, but also excessive branching with defective lumenization and impaired mural cell recruitment, both leading to non-functional vessel formation and subsequent tumor starvation. Nevertheless, the normalization capacity of these vascular aberrations remains undetermined as well as the persistence of beneficial effects attributed to the Dll4/Notch inhibition. Additionally, poor blood perfusion raises concerns that therapeutic Dll4/Notch inhibition may reduce the effectiveness of concomitant chemotherapy while hypoxia can contribute to more malignant cell selection [
29]. Besides, chronic Dll4 blockade was found to disrupt normal organ vascular homeostasis and induce vascular tumor formation [
15,
30,
31]. Thus, in certain contexts, the putative side effects of blocking Dll4/Notch pathway may come to limit its clinical usefulness.
The vessel defects induced by the Dll4/Notch inhibition closely resemble vascular abnormalities commonly observed in human malignancies; so increased Dll4 expression in tumor vasculature, that reduces endothelial sensitivity to VEGF [
32], may be considered a host defense mechanism serving to restrict tumor vascularization and malignant cell access to the bloodstream such as Angiopoietin-2 does in the case of vascular cooption [
33]. Thus, amplification of Dll4/Notch signals appears as a rational therapeutic option to be tested. Virus-transduced malignant cells that over-express Dll4 activate Notch signaling in co-cultured endothelial cells and restrict VEGF-induced endothelial cell growth [
17,
26]. When expressed in tumor cells, Dll4 also functions as a negative regulator of tumor angiogenesis; however, there is no consistent information on the effects of Dll4 over-expression on tumor kinetics. While it was reported to act as a negative driver of tumor expansion in tested malignant cell lines grafted in mice [
17,
26,
34], Dll4 expression in transduced human glioblastoma, prostate and gastric cancer cell xenografts was associated with promoted tumor growth, to some extent, due to a reduction of tumor hypoxia and apoptosis or increased secretion of matrix metalloproteinase-2 [
26,
35].
In this study, Dll4 was amplified for the first time in the tumor endothelium, its predominant site of expression, and the consequences were analyzed in both grafted and in more representative autochthonous tumor models that better reflect host-tumor interaction and wherein the lesions arise and develop resembling the human disease. We consistently found that endothelial Dll4 overexpression reduces the growth of Lewis Lung Carcinoma (LLC) grafts, chemically-induced murine skin tumors as well as transgenic RIP1-Tag2 (RT2) mouse insulinomas, due to decreased vascular proliferation by modifying the activity of angiogenesis regulators. Importantly, we show that Dll4 overexpression reduces vascular responsiveness to VEGF seeming indicated for concomitant application with VEGF-inhibitors and stabilizes tumor circulation allowing for more efficient chemotherapy delivery while at the same time reducing the formation of distant-site metastasis.
Methods
Experimental animals
Double heterozygous Tie2-rtTA-M2 TetO7-Dll4 mice were generated as described [
36] and used as hosts for tumor xenografts and chemically induced skin tumors. The RIP1-Tag2 (RT2) mice were kindly provided by Dr. Oriol Casanovas (Catalan Institute of Oncology) and used for breeding with Tie2-rtTA-M2 TetO7-Dll4 line for generation of triple mutants (RT2 Tie2-rtTA-M2 TetO7-Dll4) capable of developing pancreatic insulinomas and overexpress endothelial Dll4 after tetracycline or doxycycline-induction.
Restricted to the endothelium by the Tie-2 promoter, Dll4 overexpression was activated in inducible Tie2-rtTA-M2 TetO7-Dll4 (xenograft and skin tumorigenesis experiments) and RT2 Tie2-rtTA-M2 TetO7-Dll4 mutants (autochthonous insulinoma experiment) by administration per os of the tetracycline analogue doxycycline (2 mg/ml in drinking water ad libitum; Dll4 over-expression mice, D4OE). Non-induced Tie2-rtTA-M2 TetO7-Dll4 or RT2 Tie2-rtTA-M2 TetO7-Dll4 littermates, receiving just water, served as Dll4 basic expression controls (D4BE). Non-induced TetO7-Dll4 littermates, receiving 2 mg/ml doxycycline in drinking water ad libitum, served as Doxycycline control mice (D4BE + Doxy).
The animals were housed in ventilated propylene cages with sawdust bedding, in room with temperature between 22 °C and 25 °C and a 12-hours-light/12-hours-dark cycle. The mice were fed standard laboratory diet. From 12 weeks of age, RT2 and RT2 Tie2- rtTA TetO7-Dll4 mice received 5% sugar in drinking water to relieve hypoglycemia. All animal-involving procedures of this study were approved by the Faculty of Veterinary Medicine of Lisbon Ethics and Animal Welfare Committee (Approval ID PTDC/CVT71084/2012).
Xenograft mouse model
Male, 8-week old Tie2-rtTA TetO7-Dll4 mice (n = 12) were distributed into two equal groups that respectively continued receiving water (control Dll4-basic expression group, D4BE) or started receiving 2 mg/ml doxycycline in drinking water (Dll4 over-expression group, D4OE). A week later, Lewis lung carcinoma cells (LLC - ATCC® CRL- 1642TM, 106) were implanted subcutaneously into the flank of each mouse. Injected sites were monitored daily. Once palpable, tumor largest (a) and smallest (b) diameters were measured and tumor volumes were calculated using the formula: V = a × b2 × 0.52. Two weeks after the LLC injection, when the largest tumors approached the prescribed maximal, xenografts were dissected and processed for histological studies. For the metastasization experiment the basic protocol described above was used but the experiment was run for 6 weeks after the LLC injection.
Chemically-induced skin tumorigenesis model
Male, 8-week old Tie2-rtTA TetO7-Dll4 mice (n = 12) were separated into two equal groups and continued receiving respectively water (control Dll4 basic expression group, D4BE) or started receiving 2 mg/ml doxycycline in drinking water (Dll4 overexpression group, D4OE). A week later, all mice were shaved and treated with a single dose of 25 μg of 7,12-dimethylbenz[a]anthracene (DMBA; Sigma, St. Luis, MO) in 200 μL acetone per mouse applied to the dorsal skin. Beginning a week after DMBA-induction, tumor onset and growth was promoted by treating mice twice a week for 19 weeks with 4 μg of 12-O-tetradecanoylphorbol-13-acetate (TPA; Sigma, St. Luis, MO) in 100 μL of dimethyl sulfoxide (DMSO) per mouse. The appearance of skin lesion was monitored and recorded weekly. Mouse weight and tumor sizes (diameters) were periodically measured and lesion diameters were converted to tumor volume using the following formula: V = length × width × height × 0.52. Tumor burden of each individual mouse was calculated as the sum of individual tumor volumes. Twenty weeks after the DMBA-initiation, mice were euthanized and skin tumors were dissected and processed for histological and molecular analyses.
RIP1-Tag2 (RT2) insulinoma model
RT2 Tie2-rtTA TetO7-Dll4 mice were generated as described above. Approaching the RT2 tumor stage conventionally used for therapeutic intervention assessments, 9.5-week old RT2 Tie2-rtTA TetO7-Dll4 mice (n = 16) were distributed into two equal groups that respectively started receiving water (control RT2 Dll4 basic expression group, RT2 D4BE) or 2 mg/ml doxycycline in drinking water (RT2 Dll4 overexpression group, RT2 D4OE). The mice were sacrificed at the age of 13.5 weeks just before most RT2 progenitors die due to tumor burden and hypoglicemia. The pancreata were dissected and macroscopic tumors (≥1 × 1 mm) were excised. Tumor volumes were calculated using the formula: V = length × width × height × 0.52. The volumes of all tumors from each mouse were added to give the overall tumor burden per animal. Subsequently, insulinoma samples were processed for histological analyses.
Tumor tissue preparation, histopathology and immunohistochemistry
When dissected and measured as described above, tumors (LLC xenografts, skin lesions and RT2 insulinomas developed in D4BE and D4OE mice) were fixed in 4% paraformaldehyde (PFA) solution at 4 °C for 1 h, cryoprotected in 15% sucrose, embedded in 7.5% gelatin, frozen in liquid nitrogen and cryosectioned at 20 μm. For the metastasization experiment, lungs were dissected and macro-metastasis were observed under a dissection microscope. The lungs were resected for fixation in Bouin’s fixative for posterior metastasis histological confirmation. Skin tumor and insulinoma tissue sections were stained with Hematoxylin and Eosin (H&E) and subjected to review by a pathologist. Simultaneously, double fluorescent immunostaining to platelet endothelial cell adhesion molecule (PECAM) and pericyte marker alpha smooth muscle actin (α-SMA) was performed on xenograft and autochthonous skin and pancreatic tumor sections to examine tumor vascular density and vessel maturity. Rat monoclonal anti-mouse PECAM (BD Pharmingen, San Jose, CA) and rabbit polyclonal anti-mouse α-SMA (Abcam, Cambridge, UK) were used as primary antibodies and species-specific conjugated with Alexa Fluor 488 and 555 (Invitrogen, Carlsbad, CA) were used as secondary antibodies. Tissue sections were incubated with primary antibody overnight at 4 °C and appropriate secondary antibody for 1 h at room temperature. Nuclei were counterstained with 4´,6-diamidino-2-phenylindole dihydrochloride hydrate (DAPI; Molecular Probes, Eugene, OR). Stained sections were examined under a Leica DMRA2 fluorescence microscope with Leica HC PL Fluotar 10 and 20X/0.5 NA dry objective, captured using Photometrics CoolSNAP HQ,(Photometrics, Friedland, Denmark), and processed with Metamorph 4.6-5 (Molecular Devices, Sunnyvale, CA). The NIH ImageJ 1.37v program was used for morphometric analyses. Vessel density was estimated as the percentage of each tumor section field occupied by a PECAM-positive signal. Mural cell recruitment was assessed by quantitating the percentage of PECAM-positive structures lined by α-SMA-positive coverage.
Perfusion study
To visualize the blood-perfused, i.e. functional, portion of the tumor circulation, mice were anesthetized and biotin-conjugated lectin from Lycopersicon esculentum (100 μg in 100 μl of PBS; Sigma, St. Luis, MO) was injected via caudal vein and allowed to circulate for 5 min before the animal vasculature was perfused transcardially with 4% PFA in PBS for 3 min. Tumor samples were collected and processed as presented above.
Tissue sections (20 μm) were stained with rat monoclonal anti-mouse PECAM antibody (BD Pharmingen, San Jose, CA), followed by Alexa 555 goat anti-rat IgG (Invitrogen, Carlsbad, CA). Biotinylated lectin was visualized with Strepatavidin-Alexa 488 (Invitrogen, Carlsbad, CA). The images were obtained and processed as described above. Tumor perfusion was quantified by determining the percentage of PECAM-positive structures that were co-localized with Alexa 488 signals corresponding to lectin-perfused vessels. In all cases, signal positive areas per microscopic field (n = 20 per mouse sample) were determined by the percentage of black pixels per field after transforming the RGB images into binary files.
Quantitative transcriptional analysis
Using a SuperScript III FirstStrand Synthesis Supermix qRTPCR (Invitrogen, Carlsbad, CA), first-strand cDNA was synthesized from total RNA previously isolated with RNeasy Mini Kit (Qiagen, Valencia, CA) from skin tumors developed in D4BE and D4OE mice (n = 10 for each group). Real-time PCR analysis was performed as described (Trindade, Kumar et al. [37]) using specific primers for β-actin, Pecam, Dll4, Hey2, Vegf-a, Vegfr1, Vegfr2, Vegf-c, Vegfr3, Pdgf-β
,
EphrinB2 and Tie2. Gene expression was normalized to β-actin. Primer pair sequences are available upon request.
Total tumor doxorubicin quantitation
Total tumor doxorubicin was quantified using a method similar to Mayer et al [
37]. At endpoint, 10% w/v tumor homogenates were prepared in tissue lysis buffer. Samples of the homogenate (200 uL) were placed in 2-mL microcentrifuge tubes, and 100 uL of 10% (v/v) Triton X-100, 200 uL of water, and 1500 uL of acidified isopropanol (0.75 N HCl) were added. After vortexing, doxorubicin was extracted overnight at -20 °C. The next day, the tubes were first vortexed and then centrifuged at 15,000 g for 20 min, and stored at -80 °C until analysis. Doxorubicin was quantified fluorometrically (kex = 470 nm, kem = 590 nm). To correct for nonspecific background fluorescence, the samples were compared with a standard curve made from the fluorescent emission of known amounts of doxorubicin added to acidified isopropanol extracts of homogenized tumor tissue from untreated mice. The data are expressed as microequivalents of doxorubicin/g tissue.
Statistical analyses
Data processing was carried out by engaging Statistical Package for the Social Sciences version 15.0 (SPSS v. 15.0; Chicago, IL). Statistical analyses were performed using Mann-Whitney-Wilcoxon test. The results are presented as mean ± SEM or mean ± SD when more appropriate. P-values < 0.05 and <0.01 were considered significant (indicated in the figures with *) and highly significant (indicated with **), respectively.
Discussion
The inhibition of Dll4/Notch signaling was demonstrated in preclinical models to induce immature and non-functional vessel proliferation on an accelerated rate and result in poor blood supply and consistent growth inhibition of different mouse, rat and human tumors [
15‐
17,
25]. Nevertheless, Dll4/Notch signaling-blockade remains a strategy with unpredictable clinical usefulness. Basically, vascular defects self-repair and reperfusion over long-term Dll4/Notch-suppression may revert tumor growth, particularly in association with increasing malignant cell invasiveness, previously documented as a consequence of antiangiogenic-induced hypoxia [
29]. Besides, reduced vessel competence due to Dll4/Notch-inhibition can be expected to limit concomitant chemotherapy effectiveness. Therefore, we examined the effects of the opposite strategy, based on endothelial Dll4 overexpression, which was anticipated to reduce vascular response within tumors and suppress their expansion. In addition, Dll4 overexpression was expected to promote vessel maturation and stabilize the tumour vasculature by reducing its remodeling capacity and, in this way, the risk of development of therapy resistance and also improve tumor drug delivery.
Our results demonstrate that endothelial Dll4 overexpression reduces the growth of LLC xenografts, autochthonous chemically-induced skin papillomas and RT2 insulinomas. In all three models, remarkable tumor burden reduction due to Dll4 overexpression was consistently associated with decreased endothelial density and presumably reduced overall tumor blood supply. In contrast, vascular maturity and functionality were improved as evidenced by the formation of larger branches, increased vessel network perfusion and increased mural cell recruitment. However, improved vessel competence was not found to be predominant over the beneficial effects caused by restricted vessel proliferation but could result in better cytostatic or other drug delivery at the tumor site. In addition, the enhanced vessel wall maturation seen in endothelial Dll4 overexpressing mice may help to prevent the penetration of the malignant cells into the circulation and subsequent metastasization.
The comparative gene expression analysis of skin tumors developed by wild-type vs. Dll4 overexpression mice confirmed that Dll4/Notch signaling restricts VEGF dependent neoangiogenesis. Although
Vegf-a was found to be up-regulated under the conditions of amplified Dll4/Notch signaling, which is likely to be due to increased hypoxia revealed by elevated
Hif1-a transcription, reduced vascular sensitivity to VEGF-A was achieved by reduced
Vegfr2 transcription and simultaneous up-regulation of
Vegfr1, which lacks significant signaling activity in endothelial cells. Explaining, at least partially, the molecular mechanisms leading to reduced vascular response in Dll4 overexpressing vs. control tumors, these findings also point out the potential capacity of endothelial Dll4 overexpression to increase the efficacy of currently available VEGF signaling-inhibitors whose clinical success has been limited by development of tumor-resistance. Similarly to
Vegf-a,
Vegf-c, a positive driver of normal lymphangiogenesis and an additional tumor pro-angiogenic factor [
41], was also found up-regulated possibly due to more pronounced hypoxia while Dll4 overexpression in tumor endothelial cells decreased receptor Vegfr3 levels and, thereby, tumor vascular sensitivity to VEGF-C.
Concerning the improved perivascular cell recruitment, we found Dll4 overexpression to influence the transcription of both
Ephrin-B2 and platelet-derived growth factor receptor beta (
PDGFRβ). In several developing tissues, binding of Ephrin-B2 to its receptor, EphB4,modulates adjacent endothelial cell interactions, while Ephrin-B2/EphB4 signaling between endothelial and mural cells controls mural cell motility and adhesion [
15,
42]. In parallel, high levels of platelet-derived growth factor B (PDGFB) in proliferating endothelial cells promote the recruitment of pericytes that express the PDGFRβ [
42]. Our evidence suggests that Dll4/Notch signaling amplification stabilizes tumor vessels by enhancing EphrinB2/EphB4 and PDGF/ PDGFRβ signaling and, therefore, promoting vascular maturation. Induced
Tie2 transcription can be considered complementary since Tie2, when activated by angiopoietin1, is essential to maintain the endothelium in the quiescent state [
43] by promoting mural cell recruitment [
44].
Epithelial homeostasis was also revealed to be changed from endothelial Dll4 overexpression. Epithelial marker
E-cadherin transcription was downregulated while EMT markers
Snail-1,
Twist and
Slug were all upregulated. This is indicative of a higher pressure towards the metastatic phenotype in D4OE mice. Both the potential benefit of Dll4 overexpression in increasing chemotherapy effectiveness and its influence on metastasis formation were also evaluated in this study by use of a metastasizing LLC xenograft protocol. We understand there are advantages to using an orthotopic mouse model of metastisation [
22] instead of a xenograft protocol, especially because the tumor microenvironment is so different but the protocol we chose allowed us to explore the lung tropism of the LLC cells to more effectively restrict and direct circulating tumor cells to the lungs. Results revealed that endothelial Dll4 overexpression and the independent administration of doxorubicin, a common chemotherapeutic drug, were equally effective in reducing tumor burden and the formation of distant-site metastasis. It is also worth making note that results were always found to be tendentially better for the D4OE group but never significantly different from independent administration of doxorubicin. However, the administration of doxorubicin to D4OE mice resulted in the highest reduction of tumor growth and endpoint tumor burden, with no detectable metastasis found in the lungs of test mice. Evaluation of primary tumor drug accumulation revealed that doxorubicin accumulation was increased by 60% when endothelial cells were overexpressing Dll4. The highly significant decrease in metastasis formation in D4OE mice contrasts with the increase in EMT markers previously observed. This could be indicative that tumor cells that become malignant could be failing to effectively intravasate the highly smooth muscle cell covered neovasculature of the primary tumor and become trapped.
The results presented here represent the opposite to those described by Dll4/Notch genetic or pharmacologic inhibition when we look at the endothelial or smooth muscle cell layer phenotypes. Nevertheless, in both cases we observe a reduction of tumor growth. Probably because in both cases, despite opposing vascular phenotypes, tumor hypoxia is increased. Something similar was previously reported by us in a wound healing setting [
38]. Also in that case, both endothelial Dll4 loss- and gain-of-function resulted in impaired wound regeneration despite having opposing vascular phenotypes. As in that case, tissue dynamics depend more on vascular function than morphology.
Transduced malignant cells that over-express membrane Dll4 (entire molecule or functional extra-cellular portion) were previously found in subcutaneous tumor grafts to result in reduced tumor vessel density and produce wide, straight and less branched vessels [
17,
26,
34]. Although conflicting data were obtained regarding these neovessel functional capacities and repercussions on tumor expansion, a significant number of tumor lines responded with regression to Dll4 overexpression while quite minimal Notch activation was noted in several other tumors characterized as unresponsive to Dll4/Notch activation [
34]. This study focused on endothelial Dll4 overexpression, since Dll4/Notch predominantly mediates EC-EC rather than malignant cell-EC communication, as simulated in previous Dll4 overexpression experiments, even though the Dll4 molecule does appear in a wide range of human malignant cells [
22]. In addition, considering the accessibility of the tumor endothelium, therapeutic Dll4 delivery to the neoplastic cells seems much more complex to implement than endothelial targeting. More importantly, as the generalized Notch1/4 activation by a systemic agonist could produce several side-effects caused by the perturbation of different Notch-dependent physiological processes, selective
Dll4 genetic sequence delivery, e.g. using endothelial-specific liposomes, might restrict Notch over-activation to sites of active angiogenesis.
Acknowledgements
We thank Dr Oriol Casanovas for the RT2 mouse line, Dr Urban Deutsch for the Tie2-rtTA mouse line and MSc Patrícia Rodrigues for expert animal care.