Stable tumor vessel normalization with pO2 increase and endothelial PTEN activation by inositol trispyrophosphate brings novel tumor treatment

Tumor hypoxia is a characteristic of cancer cell growth and invasion, promoting angiogenesis, which facilitates metastasis. Oxygen delivery remains impaired because tumor vessels are anarchic and leaky, contributing to tumor cell dissemination. Counteracting hypoxia by normalizing tumor vessels in order to improve drug and radio therapy efficacy and avoid cancer stem-like cell selection is a highly challenging issue. We show here that inositol trispyrophosphate (ITPP) treatment stably increases oxygen tension and blood flow in melanoma and breast cancer syngeneic models. It suppresses hypoxia-inducible factors (HIFs) and proangiogenic/glycolysis genes and proteins cascade. It selectively activates the tumor suppressor phosphatase and tensin homolog (PTEN) in vitro and in vivo at the endothelial cell (EC) level thus inhibiting PI3K and reducing tumor AKT phosphorylation. These mechanisms normalize tumor vessels by EC reorganization, maturation, pericytes attraction, and lowering progenitor cells recruitment in the tumor. It strongly reduces vascular leakage, tumor growth, drug resistance, and metastasis. ITPP treatment avoids cancer stem-like cell selection, multidrug resistance (MDR) activation and efficiently enhances chemotherapeutic drugs activity. These data show that counteracting tumor hypoxia by stably restoring healthy vasculature is achieved by ITPP treatment, which opens new therapeutic options overcoming hypoxia-related limitations of antiangiogenesis-restricted therapies. By achieving long-term vessels normalization, ITPP should provide the adjuvant treatment required in order to overcome the subtle definition of therapeutic windows for in vivo treatments aimed by the current strategies against angiogenesis-dependent tumors.

Abstract Tumor hypoxia is a characteristic of cancer cell growth and invasion, promoting angiogenesis, which facilitates metastasis. Oxygen delivery remains impaired because tumor vessels are anarchic and leaky, contributing to tumor cell dissemination. Counteracting hypoxia by normalizing tumor vessels in order to improve drug and radio therapy efficacy and avoid cancer stem-like cell selection is a highly challenging issue. We show here that inositol trispyrophosphate (ITPP) treatment stably increases oxygen tension and blood flow in melanoma and breast cancer syngeneic models. It suppresses hypoxia-inducible factors (HIFs) and proangiogenic/glycolysis genes and proteins cascade. It selectively activates the tumor suppressor phosphatase and tensin homolog (PTEN) in vitro and in vivo at the endothelial cell (EC) level thus inhibiting PI3K and reducing tumor AKT phosphorylation. These mechanisms normalize tumor vessels by EC reorganization, maturation, pericytes attraction, and lowering progenitor cells recruitment in the tumor. It strongly reduces vascular leakage, tumor growth, drug resistance, and metastasis. ITPP C. Kieda and B. El Hafny-Rahbi contributed equally to this study.
Electronic supplementary material The online version of this article (doi:10.1007/s00109-013-0992-6) contains supplementary material, which is available to authorized users.

Introduction
Tumor hypoxia, decisive for cancer progression, upregulates the hypoxia-inducible factors (HIFs), O 2 sensors in animal cells. Hypoxic tumor cells become resistant to radiotherapy and chemotherapy, getting to be highly aggressive and metastatic [1]. HIF-1α is associated with increased vessel numbers, tumor grade severity, poor prognostic, and treatment failure. Hypoxia-induced tumor angiogenesis [2] to build new vessels for oxygen and nutrients supply to tumor cells is in fact inefficient. It leads to incomplete vessels that are permeable and allow metastatic spreading of tumor cells escaping through nonsealed endothelial cells (ECs) [3]. Antiangiogenic strategies aiming at inhibition of tumor neo-vascularization have not provided lasting benefits because, by increasing tumor hypoxia, they result in selection of drug-resistant, aggressive cancer stem-like cells [4]. Tumor vessel normalization [5], rather than destruction, is a promising approach to cancer therapy since vessel abnormalization is now recognized as a hallmark of cancer [6]. The challenge is to counteract the vicious circle of hypoxiainduced abnormal vessels and tumor hypoxia maintained because of vessel defects [7].
New strategies aim at regulating intratumor vessels by reducing the activity of hypoxia sensors like PHD1-3 enzymes (prolyl hydroxylases), which target HIFs for degradation [8]. Vessel normalization beneficial effects were confirmed by the double antiangiogenic protein [9] targeting both vascular endothelial growth factor (VEGF) A [10] and angiopoietins, which restored tumor vessels efficacy. Vessel normalization prevents tumor cell dissemination [11], allows efficient delivery of cytotoxic drugs, and increases efficacy of radiotherapy [5] through control of HIFs activity [12].
Such approaches have allowed treatment protocols at time lapses defined as therapeutic windows during which vessels are normalized [13]. The technical difficulties to set adequate therapeutic windows prompted the search for long term normalization as an alternative goal for cancer antiangiogenesis therapy [14].
Hypoxia-induced angiogenesis is inhibited when human microvascular endothelial cells are cultured under hypoxic and flow conditions in the presence of RBCs loaded in vitro with ITPP [15] that overcame low oxygen tension (pO 2 ). In vitro, ITPP was shown to act as an allosteric effector of haemoglobin, and it was observed to reduce HIF-1α [16]. Given the central role that hypoxia plays in initiation and progression of neoplasms, these findings suggested a high potential for "oxygen tension compensation" in cancer therapy [15,17] [18].
We tested here this hypothesis and the potential utility of ITPP to reach this challenge by treatment of melanoma and mammary cancer-bearing mice [19]. We show that ITPP treatment reduces tumor growth and eradicates lung metastasis. Biochemical changes in tumor and its microenvironment, upon ITPP treatment, appear to result predominantly, from the selective reversal of tumor hypoxia through vessel normalization. As an inositol phosphate derivative, ITPP molecule was tested for potential activation of endothelial phosphatase and tensin homolog (PTEN) and thus for its ability to bring a new tool to regulate angiogenesis independently of the cancer cell type. Indeed, drugs affecting the PTEN-regulated PI3K/AKT/mTOR pathway [20] that induces HIF-1α in a hypoxia-independent mechanism [21] may act on ECs to normalize tumor vessels. As PTEN is one of the most frequently mutated tumor suppressor, its inactivation leads to permanent AKT phosphorylation that maintains tumor growth. Consequently, this work was undertaken to check the hypothesis that ITPP treatment efficiently contributes to long-term vessel normalization through both oxygenation-dependent and oxygenationindependent control of HIF. We could show that angiogenesis regulation by ITPP treatment occurred through the downregulation of HIF-dependent proangiogenic genes. It regulated PTEN/AKT pathway by activating the endothelial PTEN that controls angiogenesis [22]. Resulting functional vessels were shown to facilitate access of chemotherapeutic agents to tumor cells. VEGF-induced leakiness was reduced, and invasive metastatic cell escape was abolished. Compensating intratumor hypoxia, treatment reduced the number of hypoxia-and multidrug-resistant as well as stemness-marker-positive tumor cells, reduced anaerobic glycolysis, and stopped the recruitment of HIF-mediated bone marrow derived CXCR4 + precursor cells [23]. This work deciphers some aspects of the potent multifactorial therapeutic effects of ITPP treatment on neoplastic angiogenesis, inhibition of tumor growth, and prevention of metastasis through maturation of tumor vasculature and opens to new angiogenesis-based therapies of hypoxia-inducing diseases. It shows that ITPP provides means to reach the goal of persistent angiogenesis normalization as an alternative to antiangiogenic therapies [14].

Cells
Endothelial cells are FVB mouse lung microvascular endothelial cells (MLMEC FVB) [24,25]. 4T1 murine breast cancer cells [19] were kindly provided by Professor Danuta Dus (IITD, PAN, Wroclaw, Poland). B16F10LucGFP are B16F10 murine melanoma cells, transduced with retroviral vectors containing a firefly luciferase complementary DNA (cDNA) driven by 5'LTR promoter followed by IRES sequence and enhanced green fluorescent protein cDNA (see Supplemental data). B16F10LucGFP cells were compared to B16F10 cells to validate their use in terms of similarity of growth and metastatic potential in vivo. Luciferase activity was shown not to be impaired by hypoxia, which could potentially affect the detection sensitivity (supplementary Fig. S1A).

Cell culture and oxygen regulation
The B16F10, B16F10LucGFP cell lines (see Supplemental data) and the 4T1 breast cancer cell line were cultured, respectively, in Dulbecco's modified Eagle medium and RPMI (Gibco) with 10 % fetal bovine serum (FBS) (PAA). The MLMEC-FVB endothelial cells were cultured in OptiMEM/2 % FBS. Cells were routinely cultured in a humidified incubator in 19.5 % oxygen and were oxygendeprived in a Biotronix incubator allowing p0 2 regulation (1 %) and time setting.
Mouse subcutaneous melanoma and breast cancer models C57BL6 mice and BALB/c mice were from Janvier Laboratory (France). Animal care and experimental procedures were approved by the CNREEA 03 Ethics Committee.
B16F10 or B16F10LucGFP cells were implanted in C57BL6 mice leg as subcutaneous tumors by injection of a plug constituted by 2×10 5 cells in 100 μl Matrigel™ (BD Biosciences) to help angiogenesis. 4T1 murine mammary carcinoma (10 4 cells in Matrigel) cells were injected in the mammary fat pad of BALB/c mice.
For the experimental metastases, see "Supplementary methods." ITPP treatment and chemotherapeutic protocols ITPP, prepared as described [16], was injected intraperitoneally (1.5 g/kg: in saline). Protocol consisting of serial treatments, over 4 weeks, was selected. It was started on day7 and repeated on day8 posttumor inoculation (day0). The following serial treatments were applied on days15 and 16, 21 and 22, and 28 and 29.

Luciferase activity
Luciferase Assay System (Promega) was used as described by the manufacturer and luminescence quantified as relative light units per number of cells or per milligram of tissue with manual luminometer (Lumat LB9507).
In specified experiments, tumors were depleted from CD45 + and/ or CD31 + cells by magnetic separation (Easy Sep magnet, StemCell Technologies Inc). Cytofluorimetry analyses were conducted on a FACS Sort (Becton Dickinson, Sunnyvale, CA, USA). Data were acquired on 5× 10 4 to 10 5 cells and analyzed using CellQuest software (Becton Dickinson).
PTEN localization and activation was studied on murine endothelial cells from lung (MLNEC FVB) that have been treated by hypoxia (22 h) and/or reoxygenation (25 h) in the presence or absence of ITPP (25 mM).
Fluorescent microscopy detection was performed on a Zeiss 200M inverted fluorescence microscope (Le Pecq, France), video microscopy station with controlled temperature, hygrometry, and gaz composition. Analysis was done with the Axiovision, software. Tumor necrosis was analyzed after hematoxylin-eosin staining of tumor sections.
Quantitative real-time PCR Total RNA was extracted from whole tumor using RNeasy Plus RNA extraction kit (Quiagen). cDNA was made with Transcriptor First strand cDNA Synthesis kit (Roche), and quantitative real-time PCR was performed with the QuantiTect SYBR Green RT-PCR kit (Qiagen), using LightCycler 480 (Roche). The data were analyzed with LightCycler 480 Software. Primers were supplied by Qiagen.
Magnetic resonance imaging MR experiments on mice were performed on 9.4T horizontal magnet dedicated to small animal (94/21 USR Bruker Biospec, Wissembourg, France), equipped with a 950mT/m gradient set. Detailed method is given in "Supplementary methods." The animals were put under gaseous anesthesia during MRI exams (50 % N 2 O, 0.7 l/min-50 % O 2 , 0.7 l/min-Isoflurane, 1.5 %). Breathing rate was monitored by an air pillow placed on the mouse chest to adjust the anesthetic output. Magnetic resonance angiography-time of flight (MRA-TOF) experiment [26] allowed visualizing the 3D structure of the vascular tree of the tumor on the same animal. The total duration of the MRA experiment was 50 min. MRA sensitivity was increased by working at high field (9.4 T). Angiograms were produced using maximum intensity projections (MIPs) using Paravision 4.0 (Bruker). MRA-TOF is used for angiography, but this pulse sequence could also be seen as a "saturation recovery" pulse sequence (T1 sensitive). We have used it for detection of necrosis (Fig. 1b). Measurement of the tumor size was performed with a classical morphological spin-echo pulse sequence.

PO 2 and blood flow measurements
Mice were anesthetized by ketamine (Imalgène® 500, Rhone Mérieux, France) and xylazine (Rompun® 2 %, Bayer, France) (4:1 ratio) intraperitoneally injected (40 μl/20 g mouse) before the oxylite probe tips (Oxford optronics) were installed inside the tumor and oxygen pressure or blood flow recorded by the oxylite 2000E pO 2 or OxyFlow 2000 systems (Oxford Optronics) [27]. pO 2 was computed by determining the O 2 -dependent fluorescence lifetime of ruthenium chloride on the tip of an optical fiber probe. The fluorescence lifetime is inversely proportional to the pO 2 in the tissue. The OxyFlow, a microvascular perfusion monitor, uses laser Doppler flowmetry established method for the measurement of blood perfusion with probe coupled to the O 2 probe tip.
For each animal anesthetized by ketamine/xylazine, [ 18 F]-FMISO PET imaging was performed using a device dedicated to small animal (eXplore VISTA®, GE Healthcare, USA). The spatial resolution of this system is given as 1.4 mm full width at half maximum at the center of the field of view [28].
Whole body images were acquired 1.30 h after i.v. administration of [ 18 F]-FMISO (8.9±1.3 MBq/mouse) for a total acquisition time of 20 min, an energy window set at 250-700 keV, two bed positions, and 6 ns coincidencetiming window. Image reconstruction used an orderedsubset expectation maximization (FORE/2D OSEM) method including corrections for radioactive decay, scanner dead time, and scattered radiation. No correction was applied for partial volume or attenuation.
Quantitative analysis of scans were performed using eXplore VISTA® workspace. Volume of interest (VOI) was delineated over tumor, and whole body, by summing multiple two-dimensional regions of interest from consecutive tomographic planes encompassing the entire uptake volume. The VOI of reference tissue was drawn on paw muscle and considered as background.
For each VOI, total activity was obtained as counts per minute (cpm), as well as mean activity in cpm per pixel and  Luciferase was detected intracellularly by specific antibodies and labeled by PerCP-Cy7 antirabbit IgG confirming: the reduced growth of tumor cells in ITPP-treated mice (%) and counts by direct cytometry analysis. Cells were numbered on the basis of intracellular Luciferase detection (n=8; *p<0.05) from dot plots or inset from histogram analysis for quantification of B16F10LucGFP in the tumor converted to cpm per milliliter using a calibration constant (obtained by imaging a mouse-size cylindrical calibration phantom containing a known activity of 18 F). By assuming a tissue density of 1 g/ml, VOI activity in cpm per milliliter was converted to cpm per gram. Tumor uptake was calculated by dividing total tumor activity (in cpm) by total whole body activity (in cpm).

Tumor capillary leakiness
It was assessed by Evans blue dye extravasion to the tumor interstitium. The dye was extracted by formamide [29]. The concentration measured spectrophotometrically was correlated to tumor weight.
Elisa detection of circulating VEGF VEGF was assessed in serum (100 μl) by a typical sandwich ELISA kit for Mouse VEGF (Duo Set from R&D Systems). Assays were conducted according to the manufacturer's instructions.

Statistical analysis
Data represent mean±SD of 5 or 10 (when specified) representative experiments on 5≤n≤10 animals in each group. Statistical significance was calculated by Student's t test (p= 0.05; p=0.001; n=6 animals per group, one representative experiment out of N>10, p<0.001).

Results
ITPP treatment counteracts melanoma tumor growth and lung metastases and improves mice survival ITPP treatments produced a strong reduction of tumor growth assessed in terms of visible tumor size (Fig. 1b). The mean tumor weight at day23 was 2.5±0.5 g and reduced to 0.5±0.2 g in treated animals (n=10 per group and per experiment out of N≥5 separate experiments). This was precisely quantified by measurement of bioluminescence emission from B16F10Luc tumors measured at days 18 and 24, which increased 5.5-fold as compared to 2.8-fold in the treated animals, shown in Fig.1a (n=6 animals per group, one representative experiment out of N >10, p< 0.001). Tumor growth reduction was quite visible in Fig. 1b showing the maximal and minimal sizes reached by tumors and the treatment effect, which is expressed as mean values of the tumor volumes at day 23 on Fig. 1c. MRI calculation of the tumor volume is precised on Fig. 1d. This was confirmed by the numbers of tumor cells detected in the tumor mass ranging from 30±7 % in nontreated animals, down to 3±1 % after ITPP treatment (Fig. 1e). Moreover, the metastatic invasion was drastically diminished. Indeed, in the subcutaneous model, no visible invasion but a faint Luciferase activity was detectable in lungs of treated animals (Fig. 2a). This difference was confirmed by artificial metastases assessment, after i.v. injection of B16F10LucGFP cells, which weakly developed in ITPP-treated animals compared to controls (Fig. 2b) (n=20 in each group). These results corroborate the observed survival rate (Fig. 2d). Further on, no control animal was ethically allowed to survive longer than 30 days, as opposed to 100 % survival of ITPP-treated animals, which were killed on day 60 for further analyses. Endpoints were then fixed when tumors reached 2 cm 3 . Indeed, Fig. 2c indicates that the total body weight of control animals clearly decreased compared to ITPP-treated tumor-bearing animals, despite the tumor growth. ITPP itself did not change the weight of the mice compared to saline injected mice, daily (Fig. 2c, inset).
ITPP treatment selectively counteracts hypoxia in the tumor microenvironment and normalizes tumor vessels pO 2 was measured directly inside tumors before, during, and after ITPP treatment by intratumor assessment of the O 2 -dependent fluorescence quenching of ruthenium. Nontreated tumors were strongly hypoxic (pO 2 <2 mmHg). pO 2 measurement in real time inside the tumor showed that a first injection of ITPP caused a rapid pO 2 increase detectable after 30 min, reaching 40 mmHg within next 5 min (Fig. 3a).
After serial injections of ITPP, as indicated in "Materials and methods," we show that the pO 2 level was stabilized at a high level for, at least, 72 h (Fig.3b). Moreover, the pO 2 increase specifically concerned the hypoxic tumor site as, in the muscle of the contra lateral healthy leg of the same animal, no pO 2 change was detected by concomitant measurements (Fig. 3a). The validation of such hypoxia compensation by ITPP treatment was assessed and confirmed by performing the same experiments on murine breast cancer 4T1 model. Figure 3c gives typical data registered from one mouse out of 10 treated animals and showing a similar behaviour. A moderate increase was obtained after the first injection, and pO 2 strongly and stably increased after the second injection of ITPP. The reversal of intratumor hypoxia upon ITPP treatment was functionally paralleled by increased intratumor blood flow, measured concomitantly to pO 2 , by laser Doppler and reported for day 22 tumors (Fig. 3d). Blood flow change may contribute to the rapid pO 2 increase and indicate a normalization of the tumor vessel function. Chemical confirmation of this process was assessed by histochemical detection of hypoxic sites in tumor. Hypoxic areas were evidenced by pimonidazole adducts formation with reduced proteins. Existing blood vessels, detected by double labeling for CD31 did not insure proper tissue oxygenation (Fig. 2e), confirming the poor efficacy of tumor angiogenesis. ITPP treatment prevented the formation of hypoxic areas, influencing deeply blood vessels structure, size, and density (Fig. 4a). These data were confirmed in living mice injected with [ 18 F]-FMISO to quantify hypoxia in the tumor (Fig. 4b). ITPP-treated mice clearly displayed restricted tumor growth and lower intratumor hypoxia.
ITPP treatment of tumor-bearing animals normalizes structure and function of vessels in the tumor Vessel normalization was validated in live tumor-bearing animals upon ITPP treatment. Magnetic resonance imaging of tumor vasculature indicated strong structural changes. Typical chaotic tumor vessel architecture was observed by magnetic resonance angiography (MRA), while in ITPP-treated mice, vasculature appeared less dense but organized (Fig. 5a). Intratumor examination after treatment revealed CD31 labeling typical for endothelial cell, which delineates vessel-like structures after ITPP treatment as opposed to CD31 + aggregates in controls (Fig. 5a). Furthermore, in treated tumors, vessel-like structures appear at the tumor periphery, surrounded by pericytes, positive for smooth muscle antigen (SMA + ) (Fig. 5a). Comparison with the dispersed SMA + labeled cells in the nontreated tumor mass suggests vessel normalization upon ITPP treatment. Confirmation of ITPP-induced vessel normalization is brought by pericytes recruited and lining the CD31 + endothelial cells of the treated tumor vessels (Fig. 3a, d) as opposed to the random distribution of SMA + cells in nontreated tumor, shown by confocal microscopy (Fig. 5a). "Normalization" accompanied a strong reduction of tumor size (Figs. 1 and 5a). Normalization was confirmed in terms of vessel function, first by reduction of tumor vessels permeability. Evans blue leakage was significantly diminished after two sequential treatments by ITPP (Fig. 5b) correlating with the reduced concentration of circulating VEGF, the main vessel Fig. 2 Reduction of colonization and increased survival in cancer bearing mice treated by ITPP. a Reduced luciferase activity after ITPP treatment in lungs of mice bearing subcutaneously implanted melanoma, indicating reduced metastasis (n=8 ; p<0,001, one experiment out of 10). b Reduced lung colonization, after treatment, in artificial metastasis model where melanoma cells were injected intravenously. Representative samples out of 20 mice in each group. c Evolution of the animal body weights. Nontreated animals are losing weight compared to ITPPtreated animals (n=10 in each group, the number of experiments is N>10), inset shows the effect of daily ITPP injection compared to saline. d Survival curve showing the rescue of melanoma bearing mice treated by ITPP (n=10 in each group, one typical experiment out 5) permeant growth factor (Fig. 5c). ITPP-induced vessel maturation was shown by reduction of the invasive index [30]. CD105 + /CD31 + ratio measures endoglin (CD105) versus PECAM-1 (CD31) expressing endothelial cells and reflects tumor neo-angiogenic activity. CD105 + /CD31 + cells ratio, calculated among tumor CD45 − cells, was lowered upon ITPP action (Fig. 5d), mainly because CD31 + cell number increased (Fig. 5a, d). The strong enhancement of VEGF receptors 1 and 2 on nonleukocyte CD31 + CD45 − cells upon treatment (Fig. 5e) confirms maturation of blood vessels reflecting direct interactions of endothelial cells with mural pericytes (Fig. 5a). CD31 + cell numbers increased among the whole tumor population (Fig. 5f) and was accompanied by enhancement of the hypoxia-dependent, endothelial tyrosine kinase Tie-2 receptor for angiopoietins 1 and 2 ( Fig. 5f and Supplementary Fig. 2d), a marker of matured vessels [6,10].

ITPP treatment induces tumor vessels maturation by regulating hypoxia-sensitive molecules
Hypoxia-sensitive genes turning on tumor angiogenesis displayed drastic changes upon ITPP treatment of tumorbearing animals. Since the ITPP effect was associated with pO 2 changes in the tumor, the levels of HIF-1 and HIF-2, crucial for cell response to oxygen, were analyzed by quantitative PCR. Figure 6a shows the strong downregulation of messenger RNA (mRNAs) for HIF1 − and HIF2, corroborating the reduction of HIF1 and HIF2 protein expression ( Supplementary Fig. S2 and Fig. 6a, respectively) and indicating that regulation occurred at the transcriptional level. HIFs mRNA and O 2 -dependent molecules like VHL and the tumor protective and proangiogenic enzyme HO-1 mRNAs were considerably reduced by ITPP (Fig 6a). Main oxygen sensors in angiogenesis, PHD-1, PHD-2, and PHD-3, are considerably underexpressed in treated animals, indicating a direct regulation by pO 2 , as described for VHL mRNA induction by hypoxia.
Increase in CD31, VEGFR1, and VEGFR2 mRNAs (Fig. 6a) confirmed the enhancement of corresponding protein-expressing cell numbers and the maturation effect observed among CD45 − cells (Fig. 5d) with the general increase in Tie2 + cell numbers (Fig.5e). mRNA for osteopontin, a key molecule of the tumor stroma, decisive for tumor invasion and known to be regulated by the PTE-N/AKT pathway in melanoma, was strikingly reduced by ITPP treatment (Fig. 6a).
In accordance with the finding that ITPP treatment did not influence endothelial cells growth in vitro [15] but controlled the angiogenic process, its in vivo effect might implicate activation of PTEN that controls both hypoxiadependent and hypoxia-independent mechanisms of tumor angiogenesis. We observed a clear reduction of the SDF1/CXCR4-dependent recruitment of endothelial precursor cells from the bone marrow. They cooperate to tumor angiogenesis by integrating neovessels in a PTEN/PI3-K/AKT/eNOS dependent process. Figure 6b shows the drastic reduction of the proportion of CXCR4 + , CD34 + , and CD45 − endothelial precursors cells, recruited inside the tumor upon ITPP treatment.

ITPP-induced tumor vessel normalization regulates energetic metabolism molecules linked to oxidative stress
The number of cells expressing stress-and metastasisrelated markers not only HIF2 but also lysyl oxydase (LOX) was assessed. Hypoxia-regulated LOX, involved in invasiveness [31] and responsible for H 2 O 2 production that inactivates PTEN [32], was drastically reduced by ITPP in the whole tumor (Fig. 6c).
Moreover, upon ITPP treatment, LOX and HIFs were both efficiently downregulated in their mRNA expression ( Supplementary Fig. S2). In parallel, the endotheliumrestricted enzyme, inducible NO synthase, produces NO that induces vessel dilatation and VEGF production responsible for permeabilization. Here, the number of cells expressing INOS [33] was significantly reduced (Fig. 6c), similarly to HO-1. The artificial metastasis model confirmed the beneficial effects of ITPP treatment ( Supplementary Fig. S2).
Tumor cells resist to poor oxygen supply using the anaerobic glycolysis as source of energy. This rescue pathway starts with enhanced Glut-1 receptor expression ending with lactate release, activating glycolytic pathway enzymes, such as lactate dehydrogenase (LDH) [34] and carbonic anhydrase IX (CAIX), a key enzyme allowing tumor cell survival in hypoxia and acidic pH [35]. Numbers of cells expressing Glut-1 receptors, LDH and CAIX, were drastically reduced upon ITPP treatment (Fig. 6d), indicating a significant reversal of hypoxia-induced resistance.

ITPP-treatment-induced tumor vessel maturation involves activation of endothelial PTEN
As the above data point to PTEN-mediated controls of angiogenesis upon ITPP treatment, PTEN activation was first studied in situ. In tumors, endothelial cells displayed a Fig. 4 Reduction of hypoxia in cancer bearing mice treated by ITPP. a Images showing CD31 + blood vessels (green) and hypoxic sites detected by pimonidazole staining (red), in tumor sections. Nuclei are in blue. Pimonidazole was i.p. injected, 1 h before killing then staining by anti pimonidazole antibodies. After ITPP treatment, no staining for hypoxia was visible in tumors and weakly detected by image analysis estimating the fluorescence intensities and distribution in the green and red channels. One representative picture out N>10 experiments (n =10 animals per group). b Quantification by PET imaging of hypoxia by [ 18 F]-FMISO fixation in melanoma bearing mice 14 days after subcutaneous implantation of the tumor cells (left, control) and upon serial ITPP treatments (right). The tumor radioactivity incorporation was quantified and expressed as: % of tumor activity=(total tumor activity/total whole body activity)×100. After ITPP treatment, the tumor activity decreased from 14.58±0.52 % (controls) to 7.6±0.6 % (n=8 animals in each group) showing the reversal of tumor hypoxia (N=4). Representative PET imaging of control and treated animals, normalized with the same color scaling (0-21 on both images) clear redistribution of PTEN getting distinct from CD31 labeling (Fig. 7a) while they colocalized before treatment (Fig. 7a). Concomitantly, a strong general decrease in AKT phosphorylation (Fig. 7b) in endothelial cells and the tumor tissue confirmed PTEN control of angiogenesis [22] and the efficient reversal of tumor hypoxia (Fig. 4a).
As PTEN activity requires its relocation from the cytoplasm towards the membrane [36], we attempted to decipher with CD31 + endothelial cells (red) in the organized and normalized vessels, after ITPP treatment (n=10) compared to nontreated tumorbearing mice (n=10). Scale bars represent 50 μm. b Reduction of tumor vessel leakiness measured by permeability to Evans Blue diffusion in tumor (n=8; *p=0,001). c Assessment of circulating VEGF by ELISA (n =8; *, p = 0,001). d Angiogenesis-associated Endoglin (CD105) related to CD31 endothelial cell marker, quantified by flow cytometry among CD45-depleted tumor population, predicting endothelial cells lower activity and motility after treatment (n=8; **p= 0.001). e Increased CD31 + and VEGF-Rs + cells, in CD45 depleted tumor population after ITPP treatment (n=10/group; *p=0.05). f Flow cytometry analysis showing endothelial cells maturation markers: enhanced CD31 and Tie-2 (CD202) expressing cell numbers (n=10/ group; *p=0.05) in vitro the direct effect of ITPP on PTEN activation in endothelial cells, by hypoxia/reoxygenation experiments conducted in the presence and/or absence of ITPP. Organo-specific murine lung endothelial cells showed a reorganisation of PTEN in the presence of ITPP (Fig. 8). PTEN first detected in the whole cytoplasm, colocalizing mainly with CD31 (Fig. 8a) migrated upon treatment with ITPP, towards the plasma membrane more efficiently in hypoxia (Figs. 5 and 7b) than in normoxia (Fig. 8a). This effect of ITPP was clearly enhanced in experiments involving hypoxia/reoxygenation (Fig. 8c) performed to mimic the in vivo sequence of events that occur during angiogenesis as shown by the preferential relocation in elongated endothelial cells (Fig. 8c insets).

ITPP-induced tumor vessels normalization prevents resistant cancer stem-like cells formation
In the ITPP-treated animals, reduction of p-glycoprotein expression among cells in the tumor (Fig. 9a) suggests that Fig. 6 Phenotypic effect of ITPP treatment on tumor metabolism. a RT-PCR analysis revealing: downregulation of hypoxia/oxygen sensing genes and prometastasis genes; upregulation of genes implicated in endothelial cells maturation. Results are percent of non treated samples level (n=8 animals, five separate experiments; **p =0.001; ***p=0.0001). b Reduction of the number of CXCR4 + CD34 + CD45 − precursor cells among tumor cells upon ITPP treatment. Quantification by flow cytometry from separate tumor samples (n=10/ group p<0.001). c Flow cytometry quantitative analysis showing a dramatic decrease in the number of cells expressing hypoxia-, stress-, and metastasis-related markers in primary tumors (day22) after ITPP treatments as described in "Materials and methods" (n=8/ group; 5 separate experiments, **p=0.001; ***p=0.0001). d Flow cytometry quantification analysis showing a dramatic decrease in the number of cells expressing markers of high energetic metabolism in primary tumors (day22) after ITPP treatments as described in "Materials and methods" (n=8/ group; five separate experiments; ***p=0.0001) hypoxia-induced loss of sensitivity to drugs, due to multidrug efflux pumps (MDRs), could be reversed by tumor reoxygenation. This is confirmed by the reduction upon ITPP treatment of the number of cells positive for ABCG-2 [35], which is a drug exclusion pump typical for stem cells, as well as other stemness markers, i.e., CD133 and Oct 3-4 that were detected in highly positive tumor cell subpopulations before treatment (Fig. 9b).

ITPP-induced tumor vessels normalization favors chemotherapy
As ITPP treatment improves O 2 delivery to hypoxic tissues and normalizes vessels, we studied its effect on melanoma treatment by drugs such as paclitaxel and cisplatin. Combined ITPP and drug treatments acted positively and led to Fig. 7 Effect of ITPP treatment on activation of endothelial PTEN and loss of tumor AKT phosphorylation. a PTEN, P-AKT (Ser473) and CD31 immunostainings. PTEN was expressed (red arrows) and colocalized with CD31 + endothelial cells (green arrows and green/red channels analysis of the label distribution, by image analysis) in nontreated tumor-bearing animals (left panel, n=10/group). Markers separately localized after ITPP treatment (right panel, n=10/group). The red/green channels display separate distribution by image analysis. b P-AKT distribution over the tumor (red arrows and red curve of the image analysis) observed in tumor stroma and endothelial cells, colocalized with CD31 staining (green arrows and green curve) in nontreated tumors. Expression of P-AKT was strongly reduced to punctual sites upon ITPP treatment. Image analysis point the separate localization with CD31 (right four panels, n=10/group)  Fig. 9c. The CD31 + microvessels density was reduced when animals were treated by ITPP/drugs as compared to numerous and poorly structured microvessels, CD31 + endothelial cells in controls (Fig. 8d). pO 2 and vessel normalization preceding drug treatment favored drugs cytoxicity, as indicated by necrotic areas corresponding to diffuse CD31 positivity and delineated by H&E staining (Fig. 9d). These data stress the potential of ITPP in combined therapies.

Discussion
When pO 2 in tumor microenvironment is brought to normal levels, tumor cells do not invade surrounding tissues and do not metastasize. This work shows that this effect is due to normalization of tumor angiogenesis into matured vessels resulting from selective compensation of hypoxia and control of PTEN/AKT pathway through endothelial cell membrane PTEN activation by ITPP treatment.
Intratumor neo-vessel strengthening may explain the reduction of metastatic cells escape from primary tumors. ITPP treatment indeed resulted in vessels normalization, through endothelial cells acquisition of a matured phenotype and reorganization of the vessel tumor microenvironment. Here, vessels strengthening, shown by pericyte alignment and maturation was confirmed by the induction of VEGF receptors in response to the pericyte/EC cross-talk [37]. Moreover, Tie-2, a specific endothelial tyrosine kinase receptor reduced in hypoxia and essential for normal blood vessel maturation by attracting pericytes upon binding of angiopoietin-1, was increased, corroborating the ITPP effect on vessel maturation. This was confirmed by the reduced invasive index reporting that the tumor angiogenic activity [30] through CD31-positive endothelial cell increases, while  Fig. 6 were histochemically labeled (day 22, ITPP treatments as described in "Materials and methods" (n=8/group; five separate experiments). Scale bars=50 μm. b Quantification by flow cytometry showing the reduction of cells positive for precursor and stem cell-associated markers (CD133, Oct3-4, ABCG-2) after ITPP treatment. CD133 + immunostaining corroborated the reduction visible on frozen section staining of primary tumors as in a. Scale bars=50 μm. c Lung metastasis is suppressed by chemotherapeutic drugs (Paclitaxel and Cisplatin), when treatment is preceded by ITPP injection. Tumor cells are detected by their Lucifease activity in the lungs of animals from control, ITPP, CisPt plus Paclitaxel and combined treatments ITPP + drugs as described in "Materials and methods." Data are reported for day22 (n=10/group; 5 experiments; ***; p=0.001). d CD31 staining of endothelial cells (green) and eosin/hematoxylin staining obtained in primary tumor frozen sections from experiment described in c. Efficient tissue necrosis was obtained when chemotherapeutic treatment is preceded by ITPP injection as described in "Materials and methods." Scale bars=50 μm CD31, ensuring endothelial cell junction and vessel efficacy, is reduced by hypoxia [38]. ITPP-treatment effect on the number and function of CD31 + cells confirms vessel normalization and differs deeply from CD31 + cell reduction, which results from antiangiogenic treatments.
The O 2 -dependent HIFs pathway is regulated by ITPP treatment. PHD/HIF regulatory axis is described as a promising therapeutic target to disable tumor capacity to adjust to hypoxia and control cell survival. Inhibition of residual PHDs shown here would avoid feedback protection of HIF and reduce tumor resistance to hypoxia [39].
Moreover, similarly to HIF-1α mRNA [40], other O 2dependent mRNAs like VHL mRNA, regulated by the O 2sensitive angiomiR 92-1 [41] and the tumor protective and proangiogenic enzyme HO-1 [42], were considerably reduced by ITPP treatment. Our data on PHDs and VHL mRNA reduction in long-term-treated tumors might reflect the whole tumor stroma response to stable reoxygenation and to reduced level of HIFs mRNAs, downregulating their regulatory proteins mRNA as described [43].
Vessel strengthening control of tumor cell escape is accompanied by a remarkable reduction of mRNA for osteopontin. Disappearance of such key molecule of tumor stroma helps explain the reduction of tumor cell dissemination. Osteopontin is indeed decisive for tumor invasion [44] and known to be regulated by the PTE-N/AKT pathway in melanoma [45]. This strong modification of the tumor stroma reaction upon ITPP treatment was linked to the PTEN activation at the endothelial cell level, thus independent of the PTEN status of the tumor cells.
Contributing to control cell dissemination, activation of PTEN is favored here by downregulation of LOX expression. Indeed, local production of H 2 O 2 by LOX would inactivate the tumor suppressor-PTEN-explaining the positive regulation loop between LOX and HIF in cancer development [32]. As PTEN activation was shown at the endothelial cell level, it implies that ITPP treatment has for main target the vascular cell biology. It restores the PTENmediated control on tumor angiogenesis due to activated AKT through the PDK/PI3K/AKT/mTOR pathway [22]. Effect on endothelial cells was indeed associated with a strong general decrease in AKT phosphorylation in the tumor mass confirming PTEN control of normal vs pathological angiogenesis and the efficient reversal of tumor hypoxia [46]. Fig. 10 Schematic outline of the proposed action of ITPP on HIFs regulation. 1 ITPP regulates angiogenesis by activating PTEN that inhibits PI3K action, AkT phosphorylation and mTOR actitivy towards HIF at the endothelial cell level. This regulates the vessels and increases the blood flow, while 2 hypoxia mediated O 2 delivery by ITPP, allosteric effector of hemoglobin [15][16][17][18], increases the intra tumor pO 2 , which also acts to destabilize HIF1α and down regulates VEGFs and VEGFs-related gene cascade. This directly reduces the endothelial cell mobilization, activation and growth, thus regulating the tumor VEGF-mediated pathological angiogenesis ITPP treatment is clearly targeting the vasculature. It is thus applicable to angiogenesis-dependent pathologies independently of the PTEN tumor suppressor mutations that occur in the majority of tumors.
The biological significance of the above described effect on PTEN activation is further illustrated by its effect on the recruitment of endothelial precursors, the second main mechanism by which tumors build angiogenesis [47]. Among the bone marrow derived cells that are recruited by the tumor to help its progression, endothelial precursors are mobilized to integrate tumor forming neovessels mainly by the SDF1/CXCR4 axis. This work shows the drastic reduction of the proportion of CXCR4 + , CD34 + , CD45 − cells, and endothelial precursors recruited in response to tumor SDF-1 chemoattraction, which is PTEN/AKT dependent [48].
Consequently, ITPP treatment contributes to restore vessel wall integrity and efficient blood supply by counteracting both hypoxia-dependent and hypoxia-independent HIF induction as summarized on Fig. 10. ITPP action appears to contribute in control of both mechanisms. The part played by PTEN vs O 2 delivery could not be directly shown using endothelial cell-specific mutation of Pten (Tie2CrePten) in mice. Tie2CrePtenflox/+ mice only being viable [22]. Such normalization is known to result in improved chemotherapeutics delivery by efficient and, as shown here, nonpermeable mature vessels. It also reverted stem-like resistant invasive phenotype of tumor cells, prevented activation of glycolysis pathway shown by the numbers of cells expressing Glut-1 receptors, LDH and CAIX, which were drastically reduced upon ITPP treatment. Indicating an efficient reversal of hypoxia-induced resistance towards drug and cancer stem-like cells selection ITPP treatment contributed to the efficiency of chemotherapy [33].
Indeed, ITPP-induced vessel normalization was accompanied by the reduction of drug efflux pumps thus counteracting chemo-resistance built by MDRs [34]. It also reduced drastically the number of cancer stem cells as opposed to their selection operated by the strong intratumor hypoxia, which results from antiangiogenic therapies using monoclonal antibodies-bevacuzimab (Avastin) against VEGF-A or VEGFR2 inhibitors (Sunitinib) as documented [4,49]. Our data help to explain why antiangiogenic cancer therapies provide poor results and why drug-induced improvement of vascular health correlates with better cancer prognosis. This work shows the strength of such an approach allowing stable vessel normalization. This important effect of ITPP should overcome the problem of adequate therapeutic windows for future therapies.
Our data stress the potential of ITPP in combined therapies. ITPP should provide the adjuvant needed to chemoand radiotherapy efficacy providing enhanced O 2 supply and vessel normalization [5,12], an alternative to antiangiogenic strategies [14].
Collectively, these findings highlight the multifactor and potent therapeutic use of ITPP and demonstrate its fundamental interest for advancing therapy of hypoxia and angiogenesis-dependent pathologies.