Abstract
T-cell infiltration of solid tumors is associated with improved prognosis and favorable responses to immunotherapy. Mechanisms that enable tumor infiltration of CD8+ T cells have not been defined, nor have drugs that assist this process been discovered. Here we address these issues with a focus on VE-cadherin, a major endothelial cell–specific junctional protein that controls vascular integrity. A decrease in VE-cadherin expression is associated with tumor pathology. We developed an oligonucleotide-based inhibitor (CD5-2), which disrupted the interaction of VE-cadherin with its regulator miR-27a, resulting in increased VE-cadherin expression. Administration of CD5-2 in tumor-bearing mice enhanced expression of VE-cadherin in tumor endothelium, activating TIE-2 and tight junction pathways and normalizing vessel structure and function. CD5-2 administration also enhanced tumor-specific T-cell infiltration and spatially redistributed CD8+ T cells within the tumor parenchyma. Finally, CD5-2 treatment enhanced the efficacy of anti-PD-1 blocking antibody. Our work establishes a role for VE-cadherin in T-cell infiltration in tumors and offers a preclinical proof of concept for CD5-2 as a therapeutic modifier of cancer immunotherapy via effects on the tumor vasculature. Cancer Res; 77(16); 4434–47. ©2017 AACR.
Introduction
The microenvironment of solid tumors has increasingly been recognized as one of the major drivers of tumor progression and resistance to therapy (1). Various components of the microenvironment including increased oncotic pressure, acidosis, and hypoxia lead to immunologic and metabolic alterations preventing the normal function of tumor-resistance mechanisms (2–4). The alterations in the microenvironment are strongly linked with abnormalities in tumor blood vessels, which are often structurally disordered dilated and tortuous with low perivascular coverage. Functionally, the blood flow can be erratic and at times limiting and the vessels have disturbed barrier function, leading to a hyperpermeable state (5, 6). Abnormal tumor blood vessels also contribute to the immune suppressive characteristics of the tumor microenvironment (TME; ref. 7) through their influence on immune cell delivery to tumors, endogenous immune surveillance, and immune cell function (8). Indeed, human tumors that fail to respond to therapy suffer from the exclusion of T cells from the TME and especially in the hypoxic core of the tumors (9). As such the central infiltration of T cells is almost universally a favorable prognostic marker in humans (10). Given the major role that the vasculature plays in controlling immune cell infiltration, restoring the integrity of the tumor vasculature is essential for effective anticancer therapy.
Antiangiogenic therapy was originally developed as a mode of “starving” the tumor of oxygen and nutrients; however, such therapies have shown limited efficacy when given alone. Rather, when given in low doses, antiangiogenic agents alter vascular structure and function towards a more “normal” state (11). Vascular normalization is characterized by a more organized network of vessels, enhancing drug delivery, improving tumor oxygenation, and may induce a more immunosupportive environment (2, 12). Importantly, normalization is associated with improvement in barrier function of the vessels and reorganization of the junctional molecules, VE-cadherin, ZO-1 (13), ZO-2, and claudin-5 (14).
The central molecule ensuring the barrier function of endothelial cells (EC) is the type II, endothelial-specific cadherin, VE-cadherin (15). VE-cadherin mediates not only EC–EC adhesion through adherens junctions and formation of tight zipper like structures along sites of contact (16), but also coordinates the function of other molecules involved in barrier function. VE-cadherin influences VEGFR2 signaling (17), it interacts with the EC-specific phosphatase, VE-PTP that regulates the TIE-2 pathway and vessel stability (18) and VE-cadherin reinforces the expression of claudin-5, strengthening tight junctions (19). Thus, decreases in membrane levels of VE-cadherin can have major physiologic consequences, as besides the consequences of loss of adherens junctions integrity, there is loss of its coordinating and cooperative role with other barrier-function related proteins. Reduced levels of VE-cadherin result in leaky vessels (15). Genetic ablation of VE-cadherin is lethal (20), antibodies against VE-cadherin cause vascular leak syndrome (15), soluble VE-cadherin contributes to systemic inflammation and sepsis (21), and downregulation of VE-cadherin expression is associated with tumor growth (22). Indeed, data from Oncomine shows that in a variety of tumors from 445 patients, >75% had decreased VE-cadherin expression. Furthermore, VE-cadherin was decreased in all breast, lung, sarcoma, and melanomas. Thus, a drug that specifically enhances or restores the levels of VE-cadherin in solid tumor endothelium is likely to have major beneficial effects on both the structure and function of these abnormal vessels and in normalizing the TME.
The microRNA, miR-27 is highly expressed in many solid tumors and indeed has been identified as a tumor-associated miRNA (23–25). We have previously shown that miR-27a targets VE-cadherin (26), suggesting the possibility that inhibiting miR-27 maybe clinically beneficial. However, the use of antagomirs (a class of chemically engineered oligonucleotides, which silence endogenous microRNA) that bind to miR-27a will potentially affect all miR-27a targets. We have shown that targets of miR-27a are significantly enriched in EC junctional molecules, some of which are involved in disruption of EC junctions, as opposed to VE-cadherin, which strengthens EC junctions. To overcome the lack of target specificity of antagomirs we have developed an oligonucleotide-based inhibitor (CD5-2) that, rather than binding to miR-27a, binds to the miR-27a binding site in the 3′UTR of VE-cadherin and prevents (or Blocks), the miR-27a–dependent degradation of VE-cadherin and not other targets of miR-27a (26). The specificity of CD5-2 for VE-cadherin was previously verified experimentally because it has no effect on two other confirmed miR-27a targets in ECs, SEMA6A, and PPARγ (26). Furthermore, in silico screening on other potential binding sequences in the genome have shown that the closest miR-27a target sequence has a two nucleotide miss-match and therefore unlikely to bind (data not shown). CD5-2 transfection into ECs results in upregulation in the expression of endogenous VE-cadherin. Furthermore, in vivo it upregulates VE-cadherin and results in marked changes in barrier function and improvement in recovery from ischemic injury. Here we investigate the effects of CD5-2 therapy on the tumor endothelium, with a focus on the resulting changes in the tumor immune microenvironment and the impact of these changes on the efficacy of cancer immunotherapy.
Materials and Methods
Cell lines and reagents
Human umbilical vein ECs (HUVEC) were isolated by collagenase and cultured in gelatin-coated 25 cm2 flasks in HUVEC medium (M199 with Earle's Salts, 20 mmol/L HEPES, 20% FCS, sodium bicarbonate, 2 mmol/L glutamine, 1 mmol/L sodium pyruvate, nonessential amino acids, penicillin, and gentamicin). HUVECs were authenticated by flow cytometry to express VE-cadherin or CD31 and used between passages 2 and 4. B16F10, B16F10-luc-G5 mouse melanoma cells (derived by co-author J. Holst in 2013), and MC38 colon cancer cells (obtained by co-author M.J. Smyth in 2015) were tested as being mycoplasma free by using the Lonza MycoAlert Mycoplasma Detection Kit, and authenticated by examination of morphology and consistent in vivo performance. All tumor cells were cultured in DMEM supplemented with 10% FCS, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37°C in a humidified chamber with 5% carbon dioxide. Tumor cells were stored in liquid nitrogen and thawed and used within 15 passages for in vivo experiments. Fluorescent microspheres R50 was purchased from Duke Scientific Corporation, FITC-lectin was from Vector Laboratories, and Hypoxyprobe-1 Kit was obtained from Chemicon. All antibodies used in this study were listed in Supplementary Table S1.
In vivo mouse tumor models
All animal experiments were performed according to the protocols approved by the NSW Local Sydney Health District Ethics Review Committee. B16F10 melanoma model and MC38 colon cancer model were developed with subcutaneously implanted tumor cells in C57BL/6 mice. Control or CD5-2 was administrated intravenously when the tumors became palpable. Control Ig (2A3) or purified anti-mouse PD1 mAb (RMP1-14) was intraperitoneally given to the mice on day 8, 12, and 16 in MC38 tumor model and on day 8, 11, 14, and 17 in the B16F10 model. Tumor-bearing RIP1-Tag5 mice (spontaneous pancreatic cancer model) were treated with control or CD5-2 at week 27 (see Supplementary Information).
Evaluation of tumor vascular morphology using scanning electron microscope
The tumor tissue was harvested from the mouse and rinsed in saline to remove blood and debris on the surface of tissue. Scanning electron microscope (SEM) fixative (2.5% SEM grade glutaraldehyde, 2% formaldehyde pH 7.4, 2 mmol/L calcium chloride, 2% sucrose, and 0.1 M Cac buffer pH 7.4) was then taken up into a syringe and directly injected into the tumor until it was hard (needle fixation). Care was taken to keep the injecting pressure low to avoid destroying the blood vessels. The tumors were cut into smaller pieces over SEM fixative and incubated in SEM fixative for 72 h at 4°C. Specimens were further prepared for scanning electron microscopy as previously described (27) and tumor vasculature was examined on a JEOL 6380 JSM at up to ×3,000 magnification. Four control and four CD5-2 vessel samples were examined.
Measurement of tumor vascular permeability
Fluorescent 50 nm polymer microspheres R50 (250 μL/kg) were diluted in 0.9% NaCl to a volume of 100 μL and injected into the tumor-bearing C57BL6 mouse via the tail vein. The fluorescent microspheres were allowed to circulate for 6 hours before the tumors were harvested, embedded in optimal cutting temperature (OCT) compound, and 8 μm frozen sections were cut in a cryostat (Leica). Specimens were examined with a confocal fluorescence microscope (Leica SP5) and quantified with Image J software (National Institute of Mental Health, Bethesda, MD).
Evaluation of tumor vascular perfusion
Tumor-bearing mice were injected with 150 μL of 2 mg/mL fluorescein isothiocyanate-conjugated tomato (Lycoper-siconesculentum) lectin (Vector Laboratories) diluted in 0.9% NaCl intravenously into the tail vein. After FITC-lectin was allowed to circulate for 5 minutes, the tumors were excised, embedded in OCT compound, and 8 μm frozen sections were cut in a cryostat. The frozen tumor sections were fixed, blocked, incubated with CD31 antibody, and then incubated with Alexa Fluor 647 goat anti-rat secondary antibody. Specimens were examined with a confocal fluorescence microscope (Leica SP5) and a perfusion index was quantified as the percentage of lectin-positive vessels per CD31+ vessel in each confocal fluorescent microscopic field.
Assessment of tumor hypoxia
Tumor-bearing mice were injected intravenously with 60 mg/kg of Hypoxyprobe-1 (HP2-100; Chemicon) that had been resuspended at a concentration of 30 mg/mL in 0.9% sterile saline. The solution was allowed to circulate for 90 minutes before the tumors were removed, embedded in OCT compound, and 8 μm frozen sections were cut in a cryostat. Next, the frozen tumor sections were fixed, blocked, and stained with the CD31 primary antibodies and mouse anti-pimonidazole (Chemicon), followed by incubating with Alexa Fluor 647 goat anti-mouse and Alexa Fluor 488 goat anti-rabbit secondary antibodies using a Mouse-on-Mouse Staining Kit (Vector Laboratories). Six random photographs were taken of each tissue and an average of three mice per group were used to quantify hypoxia area.
In situ hybridization
To detect the localization of Blockmir CD5-2 in tumor parenchyma, control or CD5-2 was intravenously injected into the C57BL/6 mice bearing tumors. Six hours following the injection, tumors were harvested and fixed overnight in 4% paraformaldehyde/20% sucrose. Fixed tumors embedded in OCT and cut into 50 μm sections. Fixed sections of tumors were permeabilized using 0.1% Tween-20 in 4% paraformaldehyde for 10 minutes. Prehybridization was done overnight at 68°C in hybridization mix (50% deionized formamide, 5× saline-sodium citrate; SSC, 5× Denhardt's, 0.5 mg/mL Herringsperm DNA, and 0.2 mg/mL baker's yeast RNA). A final concentration of 5 nmol/L carboxyfluorescein (FAM)-oligonucleotide probe was used for hybridization. All hybridizations were performed at 68°C for overnight in dark humidified chamber. After hybridization, the tumor sections were washed consecutively with 2× SSC, followed by 0.5× SSC and finally 0.2× SSC at the hybridization temperature. Next, the tumor was immunostained for CD31 with the same method for immunofluorescence staining.
Statistical analysis
Statistical analyses using a two-tailed Student t test were performed with Graph-Pad Prism 5.0 (GraphPad software). Data are presented as mean ± SEM. Differences were considered statistically significant at P < 0.05.
Results
CD5-2 retards the growth of tumors
We initially used the B16F10 melanoma model to assess the effect of CD5-2 on tumor growth. CD5-2, at 30 mg/kg was delivered intravenously at the time the tumor became palpable (normally day 5 following initial inoculation). The results (Fig. 1A and B) indicate that CD5-2 is able to inhibit the growth of the tumor, even when given as a single dose in the absence of other antitumor agents. The experiments were repeated with three independently synthesized batches of CD5-2, and in a double-blind experimental controlled situation, and the results show a significant inhibition of tumor growth (42% decrease in the tumor volume, n = 4; Fig. 1C). The decrease was also evident when measured as luciferase activity, using a B16F10-cell line containing a luciferase reporter construct (Fig. 1D and E). In addition, CD5-2 resulted in a similar inhibition in the growth of the Hepa1-6 liver carcinoma (Fig. 1F). CD5-2 inhibited B16F10 tumor growth when delivered at 30 and 60 mg/kg but not at 3, 7.5, and 15 mg/kg (Supplementary Figs. S1A and S1B). Multiple injections of CD5-2 (three injections per week) resulted in further inhibition of tumor growth (Fig. 1G). At the time when the B16F10 growth was highly aggressive in control drug (days 15 onwards), CD5-2 resulted in stabilization of tumor growth. Importantly, there was no observable weight loss, nor diarrhea in the mice receiving multiple injections of CD5-2, suggesting no major toxicity problems (Supplementary Fig. S1C). Given the significant effects on tumor growth with just one application of CD5-2, all subsequent experiments were performed with a single dose, as proof of principle of its effects.
CD5-2 retards the growth of tumors. A, Growth curve of B16F10 tumors given 30 mg/kg of control or CD5-2 intravenously into C57BL/6 mice. Data represent mean ± SEM. *, P < 0.05, n = 5 mice per group, four independent experiments, paired t test. B, Representative picture of tumors harvested from mice treated with control (top) and CD5-2 (bottom). C, Quantification of tumor volume of mice 7 days following the administration of 30 mg/kg of control or CD5-2. Data represent mean ± SEM. *, P < 0.05, paired t test, n = 5 mice per group, four independent experiments using three different batches of CD5-2 and control. D, Luciferase-tagged B16F10 melanoma cells were subcutaneously injected into the right flank of albino C57BL/6 mice. Control or CD5-2 was intravenously injected into the mice 9 days following the cell injection. Bioluminescent imaging of tumor growth 15 days following the injection is shown. E, Quantification of bioluminescence on day 15. Data represents mean ± SEM. **, P < 0.01, n = 5 mice per group, paired t test. F, Growth curve of primary Hepa1-6 liver tumors in control and CD5-2-treated mice. Data represent mean ± SEM. *, P < 0.05, n = 5 mice per group, paired t test. G, Growth curve of B16F10 tumors given three injections of control or CD5-2 intravenously into C57BL/6 mice (from day 10 to day 17). Data represent mean ± SEM. *, P < 0.05, n = 8 mice per group, paired t test.
CD5-2 penetrates into tumor vessels and normalizes the structure of tumor vasculature
Because our previous work has shown that CD5-2 has profound effects on ECs, we investigated whether the inhibitor localized to ECs with the tumor. This was performed using in situ hybridization with a 6-carboxyfluorescein (6-FAM)-labeled antisense probe for CD5-2. The conditions were established such that the binding of the FAM-probe to endogenous miR-27a was minimized (data not shown). In vitro experiments showed the probe detected CD5-2 transfected EC, but not control transfected cells (Supplementary Fig. S2A). In vivo, 6 hours after intravenous administration, CD5-2 localized in the EC within the vasculature of the tumor (Fig. 2A).
CD5-2 penetrates into tumor vessels and normalizes the structure of tumor vasculature. A, The detection of CD5-2 localization in control and CD5-2-treated mice using in situ hybridization. The localization of CD5-2 (green) is shown within the tumor vasculature (red). Scale bar, 25 μm. B, Representative images of B16F10 melanoma sections (day 5 following the injection of control or CD5-2) stained for CD31 (red) to visualize tumor vessels. Scale bar, 50 μm. Quantification of the number of vessels and the average vessel volume per field. Data represent mean ± SEM. *, P < 0.05, n = 5 mice per group, paired t test. C, Representative scanning electron micrographs of tumor vessels from control or CD5-2-treated animals. Scale bar, 25 μm. D, Representative images of VE-cadherin expression (green) in tumor vessels (red). Scale bar, 25 μm. The ratio of the fluorescence intensity of VE-cadherin to CD31 was determined and is presented as relative values. Data represent mean ± SEM. n = 5 mice per group. *, P < 0.05, paired t test. E, Pericytes. F, SMC coverage. Top row, endothelium and associated pericytes or SMC were visualized by CD31 (red) and NG2 (green) or αSMC immunofluorescence staining, respectively. Scale bar, 50 μm. Bottom row, high magnification of selected area in top row. Scale bar, 25 μm. Pericyte/SMC coverage was quantified by calculating the percent fraction of vessel length that overlapped with NG2/αSMC staining to determine contact between the two cell types. Data represent mean ± SEM. *, P < 0.05, n = 8 mice per group, paired t test. G, Endothelium and associated basement membrane were visualized by CD31 (red) and collagen IV (green). Scale bar, 50 μm. Basement membrane support of tumor vasculature was quantified by calculating the percent fraction of vessel length that overlapped with collagen IV staining to determine contact between ECs and basement membrane. Data represent mean ± SEM. **, P <0.01, n = 5 mice per group, paired t test.
CD5-2 treatment resulted in a significant reduction in the mean vessel volume, although there was an increase in the number of vessels (Fig. 2B). Thus, CD5-2 affected vessel morphology without vessel pruning. Consistent with this, CD5-2-treated ECs in vitro showed no changes in angiogenic-associated characteristics including proliferation (Supplementary Fig. S2B), senescence (Supplementary Fig. S2C), and migration (Supplementary Figs. S2D and S2E). When analyzed under SEM the ECs of the tumor vessels displayed properties of a non-quiescent, hyperactive endothelium appearing loosely connected and detached from each other, exhibiting obvious intercellular gaps in the vessel wall and showing weak characteristics of cell-cell contact (Fig. 2C). In contrast, the ECs in the tumor blood vessels treated with CD5-2 showed more organization into a flattened single monolayer with cobblestone appearance indicative of a quiescent and less active endothelium (Fig. 2C). CD5-2 also induced an increase in VE-cadherin expression in these tumor-associated vessels (Fig. 2D) and an increase in pericyte and smooth muscle cell (SMC) coverage as defined by NG2 (Fig. 2E) and αSMA expression (Fig. 2F), respectively. Moreover, the extracellular matrix around the vessels was altered showing an increase in the amount of collagen IV intravenously following CD5-2 delivery (Fig. 2G). Together, these structural changes suggested that the vessels were “normalized” by CD5-2 treatment.
CD5-2 enhances the function of tumor vasculature
Because CD5-2 fortified the structural nature of the tumor vessels, we assessed whether their function was also improved. CD5-2 treatment reduced the vessel permeability as measured by the number of R50 fluorescent microspheres within the tumor parenchyma (Fig. 3A). Consistent with this decreased vascular permeability (28) there was a decrease in the extent of fibrinogen deposited into the matrix (Fig. 3B). In addition, as shown in Fig. 3C, the percentage of perfused vessels (yellow) was dramatically enhanced following delivery of CD5-2.
CD5-2 enhances the function of tumor vasculature. A, Representative images of tumor vessel leakiness in the treatment of control or CD5-2. R50 fluorescent microspheres were injected intravenously into C57BL/6 mice bearing B16F10 tumors. The extravasated 50 nm fluorescent microspheres (white) from tumor vessels stained for CD31 (red) are shown. Scale bar, 50 μm. To quantify the level of leakage and standardize it for tumor vessel area, the ratio of the number of microspheres to CD31 area was determined and is presented as relative values. Data represent mean ± SEM. **, P < 0.01, n = 6 mice per group, paired t test. B, Representative images of fibrinogen deposition in the treatment of control or CD5-2 are shown. Scale bar, 50 μm. To quantify the level of fibrinogen deposition and standardize it for tumor vessel area, the ratio of the area of fibrinogen to vessel was determined and is presented as relative values. Data represent mean ± SEM. *, P < 0.05, n = 6 mice per group, paired t test. C, Representative images of tumor vascular perfusion in the treatment of control or CD5-2. FITC-conjugated lectin was injected intravenously into C57BL/6 mice bearing B16F10 tumors. Double positive staining for FITC-conjugated lectin (green) and CD31 (red) was used to evaluate the perfused tumor vessels. Scale bar, 50 μm. To quantify the percentage of perfused vessels (yellow), the ratio of the number of perfused vessels to total vessels was determined and is presented as relative values. Data represent mean ± SEM. *, P < 0.05, n = 6 mice per group, paired t test. D, Representative images of tumor hypoxia in the treatment of control or CD5-2. Scale bar, 50 μm. Double positive staining for pimonidazole (green) and CD31 (red) was used to evaluate the level of tumor hypoxia and hypoxic area was quantified. Data represent mean ± SEM. *, P < 0.05, n = 6 mice per group, paired t test.
The hypoxic microenvironment of a tumor is chiefly related to an abnormal vessel network and the consequently abnormal perfusion. Because enhanced tumor vascular perfusion was found in the treatment of CD5-2, we explored whether CD5-2 has any effects on tumor hypoxia. The hypoxic area in the tumors, as detected by pimonidazole staining of the hypoxia probe Hypoxyprobe-1, was significantly diminished by CD5-2 treatment compared to the control (Fig. 3D).
CD5-2 facilitates CD8+ T-cell penetration into the tumor parenchyma and induces tumor apoptosis
CD5-2 had no direct effects on the tumor cells themselves as evidenced by the unaltered proliferative capacity of the B16F10 cells, measured by Ki67 and PCNA staining in vivo (Supplementary Figs. S3A and S3B), or by the lack of effect of CD5-2 on the proliferation of B16F10 tumor cells in vitro (Supplementary Fig. S3C). Necrotic areas in the tumors were not changed by CD5-2 compared with control (Supplementary Fig. S3D). Furthermore, CD5-2 did not induce cellular senescence determined by either p21 (Supplementary Fig. S3E) or p16 staining (Supplementary Fig. S3F). However, CD5-2 resulted in a significant increase in apoptosis within the tumor mass as measured by TUNEL positive cells (Fig. 4A).
CD5-2 facilitates CD8+ T-cell penetration into the tumor parenchyma and induces tumor apoptosis. A, Representative confocal images of B16F10 melanoma sections stained for TUNEL (green) to visualize the apoptotic cells. Scale bar, 50 μm. The percentage of TUNEL positive cells was quantified. Data represent mean ± SEM. n = 6 mice per group, paired t test. B, Representative composite images of B16F10 tumors. DAPI (blue, nuclei), CD8 (green, cytotoxic T cells), and CD31 (red, endothelium). The images indicate that more T cells infiltrate into the middle of tumor parenchyma in the CD5-2-treated mice compared with that in the control-treated mice. Scale bar, 50 μm. The distance between leading edge of invasive CD8+ T cells and edge of tumor section (as indicated by line) was quantified. Data represent mean ± SEM. **, P < 0.01, n = 5 mice per group, two independent experiments, paired t test. C, Growth curve of primary B16F10 tumors given 30 mg/kg of control or CD5-2 intravenously into nude mice. Data represent mean ± SEM, seven mice from three independent experiments, paired t test. D, Representative images of pericyte coverage in nude mice treated with control or CD5-2. Scale bar, 25 μm. Pericyte coverage was quantified as previously. Data represent mean ± SEM. *, P < 0.05, n = 3 mice per group, three independent experiments, paired t test. E, Representative images of SMC coverage in nude mice treated with control or CD5-2. Scale bar, 50 μm. SMC coverage was quantified as previously. Data represent mean ± SEM. *, P < 0.05, n = 3 mice per group, three independent experiments, paired t test.
We then performed CD8+ staining to evaluate the infiltration of cytotoxic T cells, which may influence the degree of apoptosis. There was no significant difference in the number of infiltrated CD8+ T cells in tumor parenchyma between control and CD5-2-treated groups by immunofluorescence staining (Supplementary Fig. S3G). However, when the positions of the CD8+ cells were investigated, it was found that in 8/10 mice analyzed, the CD8+ T cells in CD5-2-treated tumors were positioned centrally, whereas in the control treated tumors the CD8+ T cells were mainly located in the marginal area of the tumors (10/10 mice; Fig. 4B). This effect was specific for CD8+ T cells as no significant change in the number or localization of CD4+ T cells (data not shown), CD45+ lymphocytes (Supplementary Fig. S3H), or in F4/80+ monocytes (Supplementary Fig. S3I) was observed. The change in position of the CD8+ T cells was unlikely influenced by the necrosis as there was no difference in the necrotic area of the tumor after CD5-2 treatment (Supplementary Fig. S3D). The position of the CD8+ T cells and the enhanced degree of apoptosis following CD5-2 treatment suggested that the retardation of tumor growth is likely to be immune mediated. To confirm this possibility immunocompromised nude mice were used. The CD5-2 had no effect on the tumor growth (Fig. 4C) in nude mice. However, it did alter the vasculature in these mice as both pericyte (Fig. 4D) and SMC (Fig. 4E) coverage were enhanced following the delivery of CD5-2.
CD5-2 enhances immunotherapy
The effects of CD5-2 to normalize the vasculature of the tumor-associated vessels and the enhancement of CD8+ T cells into the central regions of the microenvironment suggested that it might function to promote immunotherapy. To test this we used three models, the RIP-Tag5 pancreatic neuroendocrine tumor model as a model for adaptive T-cell therapy (8), the colon carcinoma MC38 model, which is sensitive to check-point blockade and B16F10, which is insensitive to PD-1 blockade (29).
In the RIP-Tag5 model, CD5-2 enhanced the perfusion of the vessels, confirming an effect on the vasculature, similar to that seen in the B16F10 model (Fig. 5A). Tumor-specific CD8+ T cells were activated ex vivo and these cells were then adoptively transferred into tumor-bearing RIP-Tag5 mice that had previously been given either CD5-2 or control. CD5-2 treatment resulted in a significant enhancement in the infiltration of the activated tumor specific CD8+ T cells (Fig. 5B). Again, in this model, the insulinomas are small and do not have a necrotic or hemorrhagic center, suggesting that the enhanced T-cell infiltrate was not a result of attraction to these areas.
CD5-2 enhances immunotherapeutic effects. A, Representative images of tumor vascular perfusion in RIP-Tag5 pancreatic tumors. Double positive staining for FITC-conjugated lectin (green) and CD31 (red) was used to evaluate the perfused tumor vessels. Scale bar, 100 μm. To quantify the percentage of perfused vessels (yellow), the ratio of the number of perfused vessels to total vessels was determined and is presented as relative values. Data represent mean ± SEM. *, P < 0.05, n = 3–5 mice per group, unpaired t test. B, Adoptive transfer of CD8+ T cells in RIP-Tag5 pancreatic tumors. CD8+ surface area (%) was quantified. Scale bar, 100 μm. Data represent mean ± SEM. *, P < 0.05, n = 3–5 mice per group, unpaired t test. C, Growth curve of MC38 tumors. Anti-PD-1 or control IgG given intraperitoneally on day 7, 11, and 14. Control or CD5-2 was given intravenously on day 7. Data represent mean ± SEM. *, P < 0.05, ***, P < 0.001; n = 5 mice per group, two-way ANOVA test. D, Representative images of VE-cadherin expression. Scale bar, 50 μm. Fluorescent intensity of VE-cadherin was determined by Fiji software. Data represent mean ± SEM. n = 8 mice per group. *, P < 0.05. E, Representative images of SMC coverage. Scale bar, 50 μm. Data represent mean ± SEM. ***, P < 0.001, n = 8 mice per group, paired t test. F, Representative images of fibrinogen deposition in the treatment of control or CD5-2 are shown. Scale bar, 50 μm. Data represent mean ± SEM. **, P < 0.01, n = 8 mice per group, paired t test. G, CD8+/Gr1 hi ratio of MC38 tumors. Data represent mean ± SEM. *, P < 0.05; **, P < 0.01, n = 7–8 mice per group as indicated by individual spot, unpaired t test. H, The percentage of CD8+ T cells that are granzyme B positive. Data represent mean ± SEM. *, P < 0.05; **, P < 0.01, n = 7–8 mice per group, unpaired t test. I, Quantification of mRNA of PD-L1 in ECs 24 hours after transfection with control or CD5-2. Data represent mean ± SEM of three independent experiments. **, P < 0.01. J, Growth curve of B16F10 tumors. Anti-PD-1 or control IgG given intraperitoneally on day 8, 11, 14, and 17. Control or CD5-2 was given intravenously on day 8. Data represent mean ± SEM. ***, P < 0.001, n = 5 mice per group, two-way ANOVA test.
In the MC38 model, anti-PD-1 treatment, delivered three times over the course of 18 days (29) and a single treatment of CD5-2 both showed significant inhibition of tumor growth (Fig. 5C). Strikingly, the combination resulted in a further inhibition of growth, more than either treatment alone. Consistently, immunofluorescent staining showed that the direct target, VE-cadherin, was significantly upregulated following the delivery of CD5-2 (Fig. 5D). More interestingly, in line with previous report (30) that indicates SMCs gradually disappear with the growth of MC38 tumors, blood vessels treated with control exhibited rare SMC coverage. In contrast, CD5-2 was able to maintain robust SMC coverage (Fig. 5E). Also, basement membrane support was improved by CD5-2 even in the context of high collagen IV deposition around the tumor blood vessels in control group (Supplementary Fig. S4A). Less fibrinogen deposit was observed in the presence of CD5-2 (Fig. 5F), suggesting vascular permeability was improved. Importantly, with the enhancement of vascular structure and function by CD5-2, the infiltration of CD8+ T cells into the central region in tumors was significantly elevated (Supplementary Fig. S4B).
Flow cytometric analysis of the immune infiltrate in MC38 tumors showed that CD5-2+ anti-PD-1 treatment had no effects on the percentage of CD4+ T cells (Supplementary Fig. S4C) or NKp46+ NK cells (Supplementary Fig. S4D) in CD45.2+ cells, but resulted in a significant increase in the CD8+/CD11b/Gr1 hi ratio (Fig. 5G). The nature of the other CD45+ population of cells that are not altered by CD5-2 treatment await identification. Furthermore, the percentage of CD8+ T cells that are granzyme B positive, as an indication of activation, was increased by CD5-2 and also by anti-PD-1 (Fig. 5H) though no significant effect was observed in the frequency of INFγ+CD8+ T cells (Supplementary Fig. S4E). As a possible mechanism for the increase in activation of the CD8+ T cells, CD5-2 treatment of EC monolayers in vitro results in a decrease in the PD-1 ligand, PD-L1 (Fig. 5I). To demonstrate the significant effect of normalization of the vasculature on augmenting check-point blockade responses, we turned again to the B16F10 melanoma model, which is unresponsive to anti-PD-1 treatment. Although CD5-2 by itself inhibited tumor growth, in combination with anti-PD-1 treatment there was a highly significant further increase in the retardation of tumor growth (Fig. 5J). Thus, CD5-2 has the capacity to convert the refractory B16F10 tumor to become responsive to anti-PD-1 treatment.
CD5-2 results in activation of multiple signaling pathways in the endothelium
VE-cadherin is the direct target for CD5-2 as we have shown previously, resulting in a 30% to 50% increase in protein expression (26). However, this relatively small change is giving large functional changes, suggesting multiple pathways of activation as a result of the increase in VE-cadherin. CD5-2 transfection into ECs not only results in the increase in VE-cadherin protein expression and mRNA levels (Supplementary Figs. S5A and S5B) but also a decrease in the phosphorylation of the VE-cadherin on tyrosine 658 (Y658; Fig. 6A). This phosphorylation site is associated with VE-cadherin endocytosis and permeability control (31). Thus, these decreases in phosphorylation are consistent with the increased expression of VE-cadherin and the decrease in vascular leak seen in the tumor-bearing mice given CD5-2 (Fig. 3). Interestingly, CD5-2 had no effect on the basal phosphorylation of Y731 (Fig. 6A), a site that is dephosphorylated on leukocyte adhesion (31).
CD5-2 results in activation of multiple signaling pathways in the endothelium. A, Phosphorylation of VE-cadherin at Y658 and Y731 residues and total VE-cadherin in HUVEC lysates 24 hours after control and CD5-2 transfection. β-Actin was used as a loading control. Quantification of densitometric ratio for VE-cadherin activity at Y658 or at Y731. B, The expression of claudin-5, ZO-1, and ZO-2 in HUVEC lysates 24 hours after control and CD5-2 transfection. β-Actin was used as a loading control. Changes in the levels of claudin-5, ZO-1, ZO-2, and JAM-C were measured as pixel density and normalized to β-actin. C, Phosphorylation of FOXO1, AKT (Thr308), and total FOXO1, AKT in HUVEC lysates 24 hours after control and CD5-2 transfection. β-Actin was used as a loading control. Densitometric ratios for FOXO1 and AKT (Thr308) activities were quantified. D, Phosphorylation of AKT, total AKT, and the expression of claudin-5 in HUVEC lysates after control and CD5-2 treatment in the absence or presence of LY294002 (PI3K/AKT inhibitor, 24 hours of incubation). β-Actin was used as a loading control. Claudin-5 expression in the absence or presence of LY294002 (24 hours) was quantified. All data are presented as the normalized expression of mean of at least three independent HUVEC lines ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001, paired t test.
VE-cadherin modulates claudin-5 (19) recognized as both a tight junction (TJ)-associated molecule and a central mediator for connecting AJs and TJs. Claudin-5 was significantly enhanced with CD5-2 treatment as were the TJ-associated proteins ZO-1 and ZO-2 (Fig. 6B). However, JAM-C, an inducer of vascular permeability (32) was not changed, at least at the total protein level (Fig. 6B), suggesting the effects of CD5-2 are selective. AKT (Thr308) and FOXO1 are part of the signaling pathway downstream of claudin-5. These were also increased by CD5-2 treatment (Fig. 6C). Upregulation of claudin-5 has been reported to be mediated through the AKT (Thr308)/FOXO1/claudin-5 signaling pathway (19). CD5-2-mediated effect on this signaling pathway was confirmed by the use of the inhibitor LY29004. The CD5-2 induced increase in FOXO1 phosphorylation was inhibited after 2 hours treatment with the inhibitor, although the protein level of Claudin-5 expression was not affected (Supplementary Fig. S5C). However, when the cells were treated with LY29004 for 24 hours, both the CD5-2-mediated phosphorylation of AKT and increase in claudin-5 expression were reversed (Fig. 6D).
TIE-2 signaling is associated with correct organization and maturation of newly formed vessels (33). CD5-2 treatment of ECs resulted in an increase in TIE-2 phosphorylation following Ang1 stimulation, together with increased AKT phosphorylation at Ser473 and both basal and phosphorylated eNOS (Thr495 and Ser1177), known downstream signaling molecules (Fig. 7A). The effects of CD5-2 on the multiple pathways involved in vessel stabilization were dependent on VE-cadherin as demonstrated by the lack of effects of CD5-2 in VE-cadherin null ECs, where CD5-2 failed to activate either the TIE-2/AKT (Ser473)/eNOS pathway or the AKT (Thr308)/FOXO1/claudin-5 pathway (Fig. 7B). In addition, the CD5-2 effects were dependent on the 3′UTR of VE-cadherin as ECs, transfected with VE-cadherin that lacked the 3′UTR (Fig. 7C), were also unresponsive. However, in normal mouse brain ECs, as in human ECs, CD5-2 resulted in an increase in VE-cadherin (Supplementary Fig. S5D).
CD5-2 results in activation of multiple signaling pathways in the endothelium. A, Phosphorylation of Tie2, AKT (Ser473), eNOS (Thr495 and Ser1177), and total Tie2, AKT, eNOS in HUVEC lysates 24 hours after control and CD5-2 transfection in the absence or presence of Ang1. β-Actin was used as a loading control. Densitometric ratios for Tie2, AKT (Ser473), and eNOS (Thr495 and Ser1177) activities were quantified. Data are presented as the normalized expression of mean of at least three independent HUVEC lines ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001, paired t test. B, Phosphorylation of AKT (Ser473 and Thr308), eNOS (Thr495), and total AKT, eNOS, claudin-5 in VE-cadherin null EC lysates 24 hours after control and CD5-2 treatment. β-Actin was used as a loading control. Densitometric ratios for AKT (Ser473 and Thr308) and eNOS (Thr495) activities and claudin-5 expression in VE-cadherin null cells were quantified. C, Phosphorylation of AKT (Ser473 and Thr308), eNOS (Thr495), and total AKT, eNOS, claudin-5 in ECs lacking 3′UTR of VE-cadherin after control and CD5-2 transfection. β-Actin was used as a loading control. Densitometric ratios for AKT (Ser473 and Thr308) and eNOS (Thr495) activities and claudin-5 expression in VE-cadherin null cells were quantified. Data are presented as the normalized expression of mean of at least three independent experiments ± SEM.
Discussion
We have developed a first-in-class drug, CD5-2, that is unique in its design (an oligonucleotide that inhibits the interaction between a microRNA and its binding site in a single target to prevent a translational inhibition of the endogenous gene, hence a Blockmir). The use of oligonucleotide-based drugs has had a checked history. However, the recent FDA approval for one such drug for spinal muscular dystrophy suggests that this family of drugs may herald a major development. The Blockmir CD5-2 is directed against VE-cadherin, the central master controller of vascular integrity. Increases in VE-cadherin expression and the major associated functional changes in the endothelium, is likely to have benefit in immunotherapy for cancer (Supplementary Fig. S6).
CD5-2 treatment has profound effects on two aspects of tumors. First, it normalizes the tumor-associated endothelium improving mural cell coverage (pericyte and SMC), decreasing vessel volume and enhancing stabilization of the ECM components. Functionally, the vessels are better perfused and show a decrease in leakiness. Second, CD5-2 treatment results in significant changes in the TME. There is a decrease in the hypoxic microenvironment, an enhanced tumor oxygenation and most strikingly, more effective penetration of CD8+ T cells into tumors. The CD5-2-mediated inhibition of tumor growth is immune cell mediated since, although CD5-2 treatment leads to changes in the vasculature in B16F10 tumors grown in nude mice (increased pericyte and SMC coverage) there was no effect on the rate of tumor growth. In normal mice, CD5-2 treatment resulted in more T cells within the central core of B16F10 tumors. In the human setting, the degree of T-cell infiltrate (rather than surrounding the tumor) is a good prognostic indicator of the response to immuno- and chemotherapy (34). Interestingly, CD5-2 alone is able to inhibit tumor growth as shown in the B16F10 melanoma model, the Hepa1-6 hepatocarcinoma model, and also in the MC38 colon carcinoma model. In these models, a spontaneous antitumor response is generated. Normalization of the vasculature by CD5-2 thus leads to improved endogenous T-cell infiltration. In contrast, in the RIP-Tag5 model, there is a failure to generate a spontaneous antitumor immunologic response, although CD5-2 treatment normalized the vasculature. However, CD5-2 did allow a significantly more intense penetration of adoptively transferred tumor specific CD8+ T cells into the tumor. CD5-2 treatment enhanced the effectiveness of the checkpoint inhibitor anti-PD-1, both in the anti-PD-1–sensitive MC38 tumor but significantly also in the B16F10 melanoma model, which is insensitive to anti-PD-1 treatment. Analysis of the cellular content of the MC38 tumor was remarkable for the increased ratio of CD8+ T cells to CD11b/Gr-1hi cells after CD5-2 treatment and the T cells were more activated. The CD11b/Gr-1hi cells are likely to be neutrophils, although the possibility that they may also be resident dendritic cells cannot be ruled out. Interestingly, Gentles and colleagues (10) recently showed that in a large cohort of 25 different tumor types, neutrophils are indicative of a bad prognosis whereas the levels of CD8+ T cells is a feature that is favorable. Taken together, the studies in these models show CD5-2 maybe an effective adjunct to CAR T cell and to immune check-point therapy where there is a pressing need to find agents that enhance the tumor-infiltrating capacity of specific antitumor T cells and also CD8+ T cells with enhanced immune function (35, 36). Although formal toxicity studies have not been undertaken, it is noteworthy that we have not seen any adverse effects (such as weight loss, diarrhea) in the animals that received multiple injections of CD5-2. Furthermore, as we have previously reported the drug had no noticeable effects on normal vasculature (26), likely because regulation of miR-27a is associated with activated endothelium and thus may not be part of the normal pathway for regulation of VE-cadherin during homeostasis.
The mechanisms underlying the promotion of T-cell infiltration and activation within the TME following CD5-2 treatment are not understood at present. The enhancement may result from the decreased hypoxia within the TME, because hypoxia is known to limit CD8+ T-cell accumulation in the region and also to inhibit their activation (37). A reduction in the interstitial pressure may also promote T-cell accumulation (38). The increase in VE-cadherin may also be structurally important to improve the overall architecture of the EC junctions rendering them more permissive for leukocyte passage. Junctional integrity is crucial for effective leukocyte transendothelial cell migration (39). The increased number and activation status of the CD8+ T cells may also result from changes in the adhesion molecules and cytokines elaborated in the CD5-2-normalized endothelium. Indeed, we have shown there is a decrease in the one of the ligands for PD-1 namely PDL-1 on CD5-2 treated ECs in vitro. Together, our results, using three different tumor models and in vitro data, show CD5-2 has profound effects on the TME, decreasing the immune suppressive TME (hypoxia, PDL-1 expression) and promoting the immune stimulatory TME (CD8+ T cells, granzyme B levels). Given the direct target of CD5-2 is specifically expressed on ECs, we can draw the conclusion that CD5-2 treatment results in a return to a structurally more organized network of vessels with profound changes in the functional status of the endothelium. This results in a TME that supports antitumor immunity. In this regard, the endothelium, during an inflammatory response is crucial in selectively regulating the timing, the magnitude and the types of immune cells that adhere and extravasate, likely through the expression of distinct adhesion molecules and chemokines (40). Furthermore, adhesion molecules on the endothelium are known to be key activators of immune function (41). We are currently investigating the immune profile of CD5-2-treated endothelium in order to gain a handle on possible mechanisms for the effects of CD5-2 on inflammatory cell infiltrates into the tumor. Preliminary data suggest that it is likely to result from combinatorial changes in the expression of key adhesion and cytokines elaborated by the normalized endothelium.
Treatment of ECs with CD5-2 results in modest, but robust increases in VE-cadherin expression. In keeping with the strong signaling and coordinating role of VE-cadherin, the “hub” of EC-cell junctional molecules, the CD5-2 effects on other junctional proteins are marked, thus showing a major multiplier effect. CD5-2 treatment of ECs increased the tight junction pathway and also activated the Ang-1–TIE-2 pathway. Paradoxically, the Ang-1-TIE-2 interaction can both stabilize vessel integrity (42) and induce angiogenesis (43). The Ang-1-TIE-2 stabilization pathway acts in the context of cell–cell junctions (33). In addition, the EC-specific phosphatase, VE-PTP, associates and dephosphorylates VE-cadherin to stabilize junctions, and associates with TIE-2 to inhibit its activation and promote vessel instability (44). CD5-2 mediated increase in VE-cadherin may alter this ratio toward VE-PTP/VE-cadherin and thus to favor vessel stabilization. Of note, there are drugs being developed for cancer therapy that are antagonists of the TIE-2 pathway (45). These may act by targeting the TIE-2 pathway in non-ECs (e.g., monocytes) or through inhibition directed to TIE-2 based on synergy with anti-VEGF in preventing angiogenesis (46).
CD5-2 appears to offer some class-specific advantages to current therapies aimed at the vasculature. It decreases hypoxia, as opposed to increased hypoxia seen with anti-VEGF agents (47) and so-called “vascular disrupting agents” (48) and in contrast with anti-VEGF it maintains the vasculature rather than pruning the vessels. Although a number of studies have shown anti-VEGF results in vascular normalization within a certain therapeutic window, it can also result in an increase in the rate of tumor growth due to aggregated hypoxia (47, 49). Hypoxia is recognized as a barrier to immunotherapy, and hypoxic areas within tumors resist infiltration of T cells even in the situation of plentiful infiltration of T cells in normoxic regions of the same tumor (50). CD5-2 effects are in stark contrast to anti-VEGF, where the normalization of the tumor vasculature results in decreased hypoxia and an amplification of the infiltration of T cells. Furthermore, CD5-2 facilitates immunotherapy and its nature being a single stranded oligonucleotide with a discrete target offers potential synergies with other biological or small-molecule anticancer agents. In general, increasing the expression of proteins is more challenging than promoting their inhibition. By its design CD5-2 increases the expression of VE-cadherin and this increase is to levels seen in resting ECs, a level we suggest is within the physiological range with significant functional improvements. Of further note, CD5-2 also has potential in other diseases where vascular leak is central in their pathogenesis.
Disclosure of Potential Conflicts of Interest
S. Goel reports receiving a commercial research grant from Eli Lilly; and is a consultant/advisory board member for Eli Lilly. M.J. Smyth reports receiving a commercial research grant from Bristol Myers Squibb, Corvus Pharmaceuticals, Aduro Biotech, and is a consultant/advisory board member for F-star, Bristol Myers Squibb, KymAb, and Arcus Biosciences. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: Y. Zhao, J. Li, T. Muller, G. McCaughan, M.A. Vadas, J.R. Gamble
Development of methodology: Y. Zhao, J. Li, V.C. Cogger, J. Chen, S.F. Ngiow, J. Holst, G. Grau, T. Muller, M.A. Vadas, J.R. Gamble
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Zhao, K.K. Ting, J. Li, V.C. Cogger, J. Chen, S.F. Ngiow, G. Grau, M.J. Smyth, R. Ganss
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Zhao, K.K. Ting, J. Li, J. Chen, A. Johansson-Percival, S.F. Ngiow, S. Goel, G. McCaughan, M.J. Smyth, R. Ganss, J.R. Gamble
Writing, review, and/or revision of the manuscript: Y. Zhao, K.K. Ting, J. Li, V.C. Cogger, J. Chen, S.F. Ngiow, J. Holst, G. Grau, S. Goel, E. Dejana, G. McCaughan, M.J. Smyth, M.A. Vadas, J.R. Gamble
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Holst, T. Muller, J.R. Gamble
Study supervision: M.J. Smyth, M.A. Vadas, J.R. Gamble
Grant Support
G. McCaughan, M.A. Vadas, and J.R. Gamble were supported by the National Health & Medical Research Council of Australia, Program Grant No. 571408, M.A. Vadas, J.R. Gamble by the National Heart Foundation of Australia, Grant Nos. G10S5140 and G11S5855 and the Avner Pancreatic Cancer Foundation. J.R. Gamble holds the Wenkart Chair of the Endothelium, University of Sydney at the Centenary Institute. M.J. Smyth is supported by a NH&MRC Senior Principal Research Fellowship and J. Li is supported by an NHMRC/Gustav Nossal Postgraduate Scholarship.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Acknowledgments
We thank the staff at the Imaging Facility of the Centenary Institute for their assistance with confocal imaging.
Footnotes
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received November 16, 2016.
- Revision received January 12, 2017.
- Accepted June 14, 2017.
- Published first June 27, 2017.
- Corrected online August 22, 2018.
- ©2017 American Association for Cancer Research.
References
Volume 77, Issue 16
