Tumour angiogenesis normalized by myo‐inositol trispyrophosphate alleviates hypoxia in the microenvironment and promotes antitumor immune response

Abstract Pathologic angiogenesis directly responds to tumour hypoxia and controls the molecular/cellular composition of the tumour microenvironment, increasing both immune tolerance and stromal cooperation with tumour growth. Myo‐inositol‐trispyrophosphate (ITPP) provides a means to achieve stable normalization of angiogenesis. ITPP increases intratumour oxygen tension (pO2) and stabilizes vessel normalization through activation of endothelial Phosphatase‐and‐Tensin‐homologue (PTEN). Here, we show that the tumour reduction due to the ITPP‐induced modification of the tumour microenvironment by elevating pO2 affects the phenotype and properties of the immune infiltrate. Our main observations are as follows: a relative change in the M1 and M2 macrophage‐type proportions, increased proportions of NK and CD8+T cells, and a reduction in Tregs and Th2 cells. We also found, in vivo and in vitro, that the impaired access of PD1+NK cells to tumour cells is due to their adhesion to PD‐L1+/PD‐L2+ endothelial cells in hypoxia. ITPP treatment strongly reduced PD‐L1/PD‐L2 expression on CD45+/CD31+ cells, and PD1+ cells were more numerous in the tumour mass. CTLA‐4+ cell numbers were stable, but level of expression decreased. Similarly, CD47+ cells and expression were reduced. Consequently, angiogenesis normalization induced by ITPP is the mean to revert immunosuppression into an antitumor immune response. This brings a key adjuvant effect to improve the efficacy of chemo/radio/immunotherapeutic strategies for cancer treatment.


| INTRODUC TI ON
Tumour angiogenesis is a direct consequence of hypoxia in endothelial cells present in the tumour microenvironment. The endothelial cell response to hypoxia-induced signals in a tumour is called the angiogenic switch. 1 Vessel growth occurs through the recruitment of endothelial cells, by sprouting of pre-existing vicinal vessels, paracrine action on bone marrow endothelial cell precursors, or other processes, including intussusception, vascular co-option and mimicry. 2 All the above-cited mechanisms are responses to hypoxia, aiming to feed the tumour with nutrients and oxygen through vessel formation in a pathologic microenvironment. The resulting abnormal angiogenesis is a hallmark of cancer, and its functions are deeply impaired. 3 Inefficient tumour vessels do not ensure sufficient blood flow to increase the oxygen level inside the tumour upon oxygen release by erythrocytes. 4 Consequently, the tumour remains hypoxic, causing a vicious circle of tumour angiogenic activity and continuous stimulation of proangiogenic pathways. Characterized by the stabilization and transcription of HIFs, 5 the increased production of factors such as VEGFs, angiopoietin 2 and IL8 maintains a proangiogenic and pathologic microenvironment for tumour cells 3,6,7 and modulates the autocrine and paracrine effects of the cytokine response towards the tumour. 7 In this setting, the tumour develops permanent adaptations to the balance of secreted factors and cells. Typically, stromal cells, such as fibroblasts, change their phenotype and become activated, expressing molecules like podoplanin, 8 which help tumour development. 9 Cooperation of the tumour microenvironment with tumour cell growth operates through chemokine/chemokine receptor axes, such as CXCL12/CXCR4 and CCL21/CCR7, which respectively recruit immune cell populations which help the tumour growth 10 and cause the dissemination of metastases in an organ-specific manner. 11,12 The tumour microenvironment, therefore, provides optimal conditions for tumour growth, mainly by allowing tumour escape from the immune response. [13][14][15] Indeed, VEGF-A is directly produced by tumour cells during hypoxia and is proangiogenic and a potent immunosuppressor. As it circulates, VEGF-A participates in the recruitment of suppressor cells, such as myeloid-derived suppressor cells (MDSCs) and Tregs, to the tumour site, where they promote tumour development. Macrophages react to the tumour microenvironment by adopting an M2-polarized phenotype. M2 tumour-associated macrophages participate in immunosuppression and cooperate with pathologic angiogenesis. 16,17 As in hypoxia, tumour cells and immunosuppressor cells, such as MDSCs and tumour endothelial cells, express immune checkpoint molecules like PD-L and PD-L2, 18,19 which can neutralize NK cells and induce CTL tolerization. 20 Thus, the entry of NK cells and activated CTLs might be impaired, and the activity of these cells may be inhibited. 21,22 Considering its effects on tumour growth and dissemination, pathologic angiogenesis constitutes an important target for tumour treatment. However, antiangiogenic approaches destroy vessels and create deeper hypoxia and must be avoided as they induce the selection of resistant cancer stem cells. 23,24 Based on these observations, the 'non-antiangiogenesis treatment' concept arose, 25 and strategies employing this concept have been designed to normalize tumour vessels rather than eliminate them. Normalization strategies focus on producing new functional vessels, using (a) anticancer drugs, because drugs can more easily reach tumour cells thanks to the re-established blood flow, (b) radiotherapy, thanks to the elevation of oxygen tension, and (c) immunotherapy, synergistically boosting the immune response. 15 The main challenge to overcome is to neutralize hypoxia to counteract its most deleterious effects on the tumour microenvironment. 26 Vessel normalization not only increases treatment efficacy, but also reduces oedema and increases the effects of cytotoxic anticancer drugs. Thus, there are many benefits to considering normalization as an adjuvant strategy. 15 As opposed to adapted-angiogenesis-based therapies, which must take advantage of transiently appearing therapeutic windows, 27 the possibility of stably normalizing vessels using myo-inositol trispyrophosphate (ITPP) has to date been the most effective option. 28 As an allosteric effector of haemoglobin, 29 ITPP also stabilizes normalized vessels by interacting with and activating endothelial PTEN. 28,30 This tumour suppressor molecule is a primary regulator of vessel formation and modulates endothelial cell reactivity through the control of immune checkpoints, such as PD-L1, 31 through cross controlling NOTCH4. 32 Myo-inositol-trispyrophosphate is a non-cytotoxic molecule, 28,33 and its tumour-reducing effects 30,34 are suggested to operate through vessel normalization, leading to sustained increases in oxygen partial pressure (pO 2 ) which counteracts hypoxia inside the tumour and inverts its effects. This work shows that ITPP-induced stable normalization induces deep changes in the tumour microenvironment due to hypoxia compensation using two tumour models: melanoma and mammary carcinoma. ITPP-induced stable normalization further stops the HIF-dependent cytokinic and cellular responses. As directly demonstrated here, NK cell infiltration and its antitumour activity are impaired by PD1 binding to PD-L1 expressed on hypoxic endothelial cells in the tumour, and this effect is reduced by ITPP treatment. This treatment mainly inverts immunosuppression-the key reason why current therapies fail. Here, we report for the first time a strategy that fulfils the main conditions required to increase the efficacy of cancer treatment. Indeed, this approach addresses the present challenges facing cancer treatment, influencing the tumour microenvironment by permitting the recruitment of active immune cells, lowering immune checkpoint activity, and reducing PD-L1 expression at the endothelial cell level to allow the entry of immune cells into the tumour.

| Ethics statement
All animal-related experiments were conducted in accordance with approved guidelines and regulations. Experimental protocols were approved by the French Ethics Committee for Animal Experimentation, CNRS Orleans campus CNREEA 03 Ethics Committee, authorization number CLE CCO 2010-004.

| Cell lines and cell culture
Murine brain (MBrMEC) and bone marrow (MBMMEC)-derived organospecific microvascular endothelial cells (ECs) were established as described 34

| Hypoxia treatment
Endothelial cells were seeded in 24-well Primaria plates in duplicate, with 10 × 103 of MBrMEC or 12.5 × 103 of MBMMEC cells per well in 500 μL of OptiMEM medium, and allowed to adhere to the culture surface for 24 hours in a humidified incubator at 37°C with 5% CO 2 .

| Preparation of single-cell suspensions
Tumour samples were immediately transferred into PBS on ice.
Biopsies were cut into small pieces, then filtered through a cell strainer after collagenase/dispase (Gibco) dissociation. Erythrocytes were eliminated by lysis buffer (eBiosciences).
When specified, tumours were depleted of CD45+ and/or CD31+ cells by magnetic separation (Easy Sep magnet, StemCell Technologies Inc).

| Cell staining for flow cytometry
Single cell suspensions were stained with monoclonal antibodies for 1 hour at 4°C. One to four-color flow cytometry was performed (FACS LSR, Becton Dickinson). Data were acquired using the CellQuest software (Becton Dickinson) using at least 100 000 events gated for live cells.

| Immunohistochemistry
Tumour tissues were embedded in tissue freezing medium (Tissue-Tek; Sakura) and snap-frozen in liquid nitrogen. Tumour cryosections were fixed and stained with mouse anti-CD31 (rat monoclonal

| ELISA assay for VEGF production
VEGF was measured in RenCa cell culture supernatants (after a 72 hours exposure to hypoxia/normoxia) using a commercially available enzyme-linked immunosorbent assay that recognizes natural and recombinant mouse VEGF (ELISA, R&D Systems, Minneapolis, USA). The results were normalized to 10 6 cells.

| Quantitative real-time PCR
Cellular mRNA was extracted using the RNeasy Plus mini kit (Qiagen). mRNA was eluted in RNase-free water. Absorption spectra were measured on an ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) before storage at −80°C. RNA was reverse-transcribed to cDNA using the 'Maxima First Strand cDNA Synthesis kit for RT-qPCR' (Fermentas); 3 µg of RNA was used for each sample. The obtained cDNA was stored at −20°C before qPCR.
Real-time PCR was performed on a LightCycler 480 (Roche) using

SYBR Premix Ex Taq (Perfect Real Time) (Takara) and QuantiTect
Primer Assay (Qiagen) in white 96-well optical microtiter plates (Roche). All reactions were performed in triplicate and reported as average values. For reference, seven housekeeping genes were tested. Means and standard deviations were calculated, and the gene which had the lowest standard deviation was chosen as the reference. For each target gene, the mean and standard deviation were calculated and then normalized by the corresponding value for the reference gene (ppia), to obtain the ΔCp.

| Real-time PCR for vegf-a and vhl
Total RNA was extracted from cells using TRIzol and the RNA extrac-

| Oxymetry measurement of the tumour and effect of treatment with ITPP
Oxymetry imaging was done by the photoacoustic method at day 7 after 4T1 tumour implantation (5 × 10 5 cells in the mammary fat pad), on a locally shaved area using depilatory cream and anaesthetic. Isoflurane (Isovet, Virbac) was used at 3.5% for 3-60 seconds, then 1.5% along the imaging process adjusted to a breathing frequency of 50/min. Mice were then placed on the heating pad of the VEVOLAZR Imaging Station (FUJIFILM VisualSonics) for in vivo imaging. Images were recorded in the OxyHemo mode, which collected data at 750 nm (Hb-tot) and 850 nm (Hb-O 2 ) to create and display a parametric map of estimated oxygen saturation using the LZ400 probe at 30 MHz. The saturation of oxygen (SO 2 ) in the tumour was followed before and after injection of ITPP. The saturation values corresponded to an average of sO2 values on the 3D tumour volume using a step size of 0.102 mm.
Blood flow was assessed by Laser Doppler (Oxy Flow).

| Statistical analysis
All data are expressed as means ± SEM. In vitro cell data are expressed as means + SEM. All statistical analyses were performed using GraphPad Prism 7.0 software. In adhesion and flow cytometry in vitro, the experiment data represent the means of three biological replicates in triplicate, *P < .05 using the Mann-Whitney U test.
Flow cytometry experiments data are represented by the means ± SEM of N representative experiments. Statistical significance was calculated by Student's t test (Microsoft Excel). P values were determined by Student's t test. A P value of less than 0.05 was considered statistically significant. *Indicates statistically significant differences (P < .05), ** indicates P < .01 and *** indicates P < .005.

| Tumour vessel normalization through ITPP treatment and tumour pO 2 status
When tumour-bearing animals were treated with ITPP under conditions that have been previously shown to normalize vessels, 30 a reduction in tumour size was observed ( Figure 1).
Typical examples are shown for both normal and nude mice for B16F10 melanoma ( Figure 1A-C) and 4T1 mammary carcinoma ( Figure 1D) cells. ITPP was shown to be non-toxic for both animal types 30 and cells treated separately. 30 Applying the protocol that we have described previously, 30 our hypothesis regarding the influence of pO 2 changes in the tumour microenvironment was assessed. Figure 1E shows the reduction of stemness associated SOX2+ cells 35 upon ITPP treatment, together with a deep change in the vesselassociated structures before treatment which appear in a distinct localisation and as regular vessel structures after ITPP treatment. showing that ITPP treatment following the previously defined protocol was more efficient than 20 Gy irradiation at increasing blood perfusion. 37 The immune response 38,39 was tested, and we found a deep influence of pO 2 changes on the cytokine composition of the tumour microenvironment. 30 The immune reaction was studied when ITPP-treated tumours were half the size of the controls.
This point was reached 23 days after injection with 10 4 B16F10 cells, both in C57BL/6 and Rj:NMRI-nu nude mice. It was reached after 31 days with 10 5 injected 4T1 cells in BALb/c-by mice.
Treatments were stopped on day 21 for 4T1 cells, and 50% tumour growth was observed at day 23 for melanoma cells. An equivalent reduction was not observed in mammary carcinoma cells until after the arrest of treatment at day 31 ( Figure 1F).

| Effects of ITPP treatment on the tumour microenvironment-NK cell response
The immune cell infiltrate was analysed by immunocytochemical labelling of tumours extracted from the animals on day 23 after tumour cell implantation. NK cells were detected by anti-CD49b labelling, and endothelial cells by anti-CD31 labelling (Figure 2A,B). NK cells remained in the vessels of non-treated tumours (Figure 2A), whereas they  Figure 2B). This is confirmed in Figure 2C, which displays the distribution of CD49b+  Figure S3).
This result was confirmed in the mammary carcinoma model with 4T1 cell bearing mice, in which ITPP treatment induced higher recruitment of activated NK cells ( Figure 2I).

| Intra-tumour evolution of myeloid-derived suppressor cells and the phenotype of infiltrating macrophages upon ITPP treatment
The tumour microenvironment is characterized by the presence  Figure S4).

The proportion of macrophages displaying an M2
(CD45+CD11c+CD206+)-polarized phenotype, which aids tumour immune suppression, was reduced upon treatment with ITPP. This was observed in melanoma ( Figure 3C), as well as in mammary carcinoma ( Figure 3D).

| Evolution of the T cell populations infiltrating the tumour upon ITPP treatment
The proportion of Th2 cells reflects the inflammatory state and is known to affect tumour progression. Th2 cells, characterized as CD45+CD4+CCR4+, were decreased at the tumour site upon ITPP treatment ( Figure 4A). Regulatory T cells, which cooperate in tumour development and growth, also clearly responded to ITPP treatment,  Figure 4B). The effect of ITPP treatment on Tregs was confirmed in 4T1 mammary carcinoma-bearing BALB/c mice ( Figure 4C).

| The expression of immune checkpoint molecules in the tumour microenvironment is modulated by ITPP treatment
The   Figure 6 shows that when ITPP is used to treat B16F10Luc melanoma-bearing mice, the proportion of CD31+ endothelial cells is higher than in non-treated tumours ( Figure 6A). CD31 is more highly expressed on endothelial cells in normoxia than in hypoxia and is a junction molecule that strengthens vessels and reduces their permeability. 38,42 CD31 expression is indicative of vessel normalization. [43][44][45] Among CD31+ endothelial cells, PD-L1 was observed to be strongly expressed before ITPP treatment and was considerably reduced in treated tumour CD31+ endothelial cells ( Figure 6B). The second PD1 ligand, PD-L2, was also expressed, although to a lesser extent than PD-L1, and was reduced by hypoxia alleviation/vessel normalization after treatment with ITPP ( Figure 6C). Upon reduction in the PD1 ligands, ITPP treatment may also be accompanied by an increase of immunocompetent cells expressing PD1. As above, the proportion of CD45+ cells was increased by ITPP treatment ( Figure 2E). Reduction in the PD1 ligands, especially on endothelial cells, suggests a possible control of the mechanism by which NK cells enter the tumour. Figure 6D,E shows an increased number of CD45+ cells inside tumours that were treated by ITPP. This corroborates the data reported in Figure 2F-I, pointing to an increased proportion of NK cells inside the tumour upon ITPP treatment and increased activated NK cells (CD49b+CD226+), 46 as well as invasion of the tumour, as opposed to non-treated tumours where NK cells were found in the vessels (Figure 2A-D). Hypoxia induction of PD-L1 on endothelial cells was reduced upon ITPP treatment ( Figure 6B,C).
The Treg CD45+PD1+ cells 47 proportion was considerably reduced among the whole tumour population, both in melanoma ( Figure 6D) and mammary carcinoma ( Figure 6E).
The effect of ITPP on other immune checkpoints was also assessed, including CTLA4 expression among CD45+ cells. No significant change was observed in terms of cell numbers, but the level of CTLA4 expression was lowered with ITPP treatment ( Figure 6F). The intensity of HIF-1-regulated CD47 expression 27,48 was found to be reduced in tumour B16F10 melanoma cells ( Figure 6F). This stem cell marker protects tumours against cytotoxic immune cells, promotes evasion from phagocytosis and maintains cancer stem cells.

| Mechanism of NK cell to endothelial cell recognition
The   Figure 7F).
The molecular mechanism regulating the adhesion between activated PD1+NK cells and hypoxic PD-L1+ endothelial cells ( Figure 7J,K) was tested for the involvement of immune checkpoint molecules in the adhesion process as quantified by flow cytometry. 45 Blocking with an anti-PD-L1 neutralizing antibody significantly reduced the adhesion of EL4IL2 to MBMMEC, by 35% ( Figure 7K).
This partial inhibition was obtained when the adhesion experiment was performed at 4°C, whereas at room temperature (20°C) no effect was observed. The bivalent IgG1 antibody, although devoid of active Fc, may activate the energy-dependent adhesion process ( Figure 7J), inducing an adhesion molecule cascade and masking the role of individual partners.

| Modulation of the tumour microenvironment: chemokine receptor expression in response to vessel normalization by ITPP
The intercellular interactions between immune and non-immune stromal cells at the tumour site occur through a fundamental crosstalk ensured by chemokines and their receptors. The metastatic process has largely been demonstrated for CXCL12 and its receptor CXCR4, and CCL21 and its receptor CCR7. A radical effect of ITPP treatment on CCR5 expression was observed on tumour cells ( Figure 8A), corroborating the inverse effect observed on the CD45+ population ( Figure 8C) and the M1 macrophage repolarization effect. 16 The CCR7/CCL21 axis is responsible for melanoma cell metastases into lymph nodes 11,16,49 and promotes tumorigenesis by stem cells. 50 A total reduction in CCR7 expression was observed on tumour cells ( Figure 7A), whereas no significant expression of this receptor was detected in either immune or endothelial cells ( Figure 8C,D). 51 The proportion of B16F10 cells expressing CCR10, the receptor for CCL27 (cutaneous T-cell attracting chemokine), was diminished upon ITPP treatment, whereas the proportion of immune cells (R = 10) ( Figure 8C) and the endothelial cell-enriched population (R = 2) ( Figure 8D) increased. This axis may control the recruitment of immunocompetent cells. 52 Tumour cells expressing the receptor for CXCL12 (CXCR4), which is involved in metastasis, 53 were strongly reduced ( Figure 8A), reflecting the effect of hypoxia on tumour cells, 54 whereas CXCR4 expression was not affected in CD45+ cells ( Figure 8C) or in the endothelial cell-enriched population ( Figure 8D).
Moreover, the endothelial enriched cell-population selectively displayed an induced expression of the fractalkine receptor, CX3CR1, upon ITPP treatment ( Figure 8D), which corresponds with the reoxygenation effect and vessel restoration. 54 Comparison of the expression of chemokine and chemokine receptors at the mRNA level in tumour cells in hypoxia and normoxia ( Figure 8B) indicated a strong effect of hypoxia on the chemokine components of the microenvironment. This is highly significant for chemokines, as CCL17, which promotes cancer, is dependent on hypoxia. 55 The major changes appear in the induction of mRNAs for chemokines CXCL12 and CCL21b, confirming the role of hypoxia in the metastatic process of melanoma cells.

| D ISCUSS I ON
The The main goal is thus to obtain treatments capable of compensating for hypoxia in the tumour. 3 Based on the rationale that a nonhypoxic tumour will no longer sustain pathologic angiogenesis and allow vessels to be normalized and functional, new treatment approaches are attempting to reach stabilization of normalized angiogenesis. 60 As previously described, pO 2 elevation creates important modifications in the tumour microenvironment, namely in parameters responsible for tumour immunosuppression: the main reason for treatment pitfalls.
Our two main approaches, through which we demonstrated that it is possible to normalize vessels in tumours, have also shown the ability to stably maintain normalization. Our approaches are complementary, and based on the restoration of oxygen tension. Counteracting the excess VEGF produced in hypoxia, which acts as an efficient chemoattractant for immunosuppressive cells, 43,61 our hypoxia-conditioned cell-targeted gene therapy allows the fine-tuning of therapeutic angiogenesisbased treatment. 45 The treatment efficiently reduced tumour size in both melanoma and mammary carcinoma. 43,45 The second strategy, based on direct modulation of O2 release, uses myoinositol trispyrophosphate, which directly allows for higher delivery of oxygen by red blood cells. 28   writing -review and editing (lead).

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.