The tumor vessel targeting agent NGR-TNF controls the different stages of the tumorigenic process in transgenic mice by distinct mechanisms

Abstract

NGR-TNF is a vascular targeting agent in advanced clinical development, coupling tumor necrosis factor-α (TNF) with the CNGRCG peptide, which targets a CD13 isoform specifically expressed by angiogenic vessels. Antitumor efficacy of NGR-TNF has been described in different transplantation tumor models. Nevertheless, the mechanism underlying its activity is not fully understood. In the wild type and in the immunodeficient (RAG^−/−) RIP1-Tag2 models of multistage pancreatic carcinogenesis, we demonstrate that CD13 is highly expressed on endothelial cells of hyperplastic and angiogenic islets, whereas its expression is down regulated in tumors where it partially colocalize with pericytes. In vivo CNGRCG peptides coupled to fluorescent nanoparticles (quantum dots) bind to CD13 and colocalize with anti-CD31, in pancreatic islets. At early stage, low doses of NGR-murine (m)TNF have a direct cytotoxic effect inducing endothelial cell apoptosis, reducing vessel density and eventually inhibiting the development of angiogenic islets. At a later stage, NGR-mTNF is able to reduce tumor growth inducing vascular normalization, exclusively when treatment is carried out in the immunocompetent mice. Interestingly, NGR-mTNF-treated tumors from these mice are characterized by CD8⁺ T cell infiltration. At molecular level, overexpression of genes involved in vessels normalization was detected only in NGR-mTNF-treated tumors from immunocompetent mice.

These findings identified a new mechanism of action of NGR-mTNF, providing support for the development of new therapeutic strategies combining chemotherapy or active/adoptive immunotherapies to low dose NGR-TNF treatment.

Abbreviations

EC

endothelial cells

hrs

hours

sTNF-R

soluble TNF receptors

TNF

tumor necrosis factor-α.

Introduction

NGR-TNF is a recombinant protein derived from the fusion of CNGRCG tumor homing peptide with the N-terminal of TNF. The CNGRCG peptide selectively targets tumor blood vessels by binding to an isoform of the CD13 receptor (also named aminopeptidase N).¹ CD13 is highly expressed on the endothelial cells (EC) of neoangiogenic vessel during tumorigenesis and it is upregulated in response to hypoxia and angiogenic growth factors, conditions that are characteristic of the tumor microenvironment. Therefore, this metallo-aminopeptidase has been identified as a potential target for cancer therapy.² NGR-human (h)TNF is currently being tested in phase II and III clinical studies involving different types of tumors as monotherapy and in combination with chemotherapeutic agents.^3-9 It shows antitumor activity at low dose (0.8 µg/sqm) and has a very low toxicity profile rendering it suitable for long-term maintenance treatment.

NGR-TNF is able to increase vascular permeability, temporarily improving drug-penetration into tumors. Indeed, in vitro experiments using endothelial cell monolayer showed that NGR-murine (m)TNF at low doses (i.e. in the picogram range) induces the disassembly of VE-cadherin, a protein critical for adherents junctions, causing the formation of intercellular gaps and the alteration of the endothelial barrier function.¹⁰ Recently, other rapid and transient effects induced by NGR-mTNF have been described. In particular, targeting TNF to neoangiogenic vasculature, induces upregulation of ICAM-1/VCAM-2 leukocyte adhesion molecules on endothelial cells, and release of pro-inflammatory cytokines and chemokines in the tumor microenvironment. Remarkably, these modifications favor a rapid and long lasting infiltration of tumor-specific CD8⁺ T cells ¹¹ Similarly, tumor infiltration by CD8⁺ T cells, was induced by an RGR-targeted TNF (i.e. TNF-RGR administered at doses in the microgram range). In this system, lymphocytes infiltration was mediated by vessel normalization.¹²

Noteworthy, in agreement with these experimental results, in several NGR-hTNF phase II trials involving different tumor histotypes, it was observed a positive correlation of peripheral blood lymphocyte count with antitumor activity of NGR-hTNF administered in combination with chemotherapy.¹³

The efficacy of the NGR-mTNF was demonstrated in a number of transplantable tumors ^14,15 but the mechanism of action of the drug in preventing tumor progression in spontaneous tumor models is still unknown. To investigate this issue, we use the RIP1-Tag2 transgenic mice, which express the SV40 large T antigen (Tag) under the control of the insulin gene promoter and develops tumors through temporally and histologically distinct stages.¹⁶ The pancreatic islets of RIP1-Tag2 puppies have a normal anatomical and histological feature (“normal stage”). Beginning at 3–5 weeks of age a hyperproliferative switch occurs, which results in the appearance of “hyperplastic islets”. Then, a fraction of these islets undergoes an angiogenic switch, leading to the appearance of hemorrhagic islets, which develop into solid tumors at 12 weeks of age. Induction of angiogenesis is a discrete and necessary step in tumor development, thus rendering this model an useful tool for studying angiogenesis inhibitors.^17-20 We assessed NGR-mTNF activity at two distinct stages of the pancreatic islet carcinogenesis: (i) early treatment at the hyperplastic stage to affect the angiogenic switch (prevention trial), (ii) treatment of mice bearing small tumors to determine whether tumor expansive growth and progression to invasive carcinoma could be stopped (intervention trial). Additionally, the role of the immune system in the NGR-mTNF activity was studied in the RIP1-Tag2 RAG1^−/− immunodeficient mice.²¹ The results presented here clearly demonstrate that early treatment of RIP1-Tag2 mice with NGR-mTNF results in a significant reduction of angiogenic islets in both mouse models. Moreover, histological analysis shows a decrease of vascular surface area and an improvement in vascular functionality. In the intervention trial, NGR-mTNF is able to control tumor growth inducing vascular normalization, exclusively when the treatment is carried out in the immunocompetent mice. These data, strongly support the hypothesis that in this mouse model, NGR-mTNF exerts a direct cytotoxic effect affecting the angiogenic switch in the initial phases of tumor formation, and promotes vessel normalization and adaptive immune responses during tumor development.

Results

Identification and localization of CD13⁺ cells in pancreas from RIP1-Tag2 mice

NGR-mTNF is targeted to the tumor vasculature by the CNGRCG motif,¹ which specifically interacts with the CD13 (aminopeptidase N) expressed by angiogenic vessels. To monitor the distribution of CD13 during tumor development, we analyzed pancreatic cryosections of RIP1-Tag2 mice at different stages (hyperplastic islets, angiogenic islets and tumors). As illustrated in Fig. 1, there are few CD13 positive cells in hyperplastic islets compared with angiogenic islets and tumors, consistent with a differential expression of this marker during tumorigenesis. Moreover, CD13 is predominantly expressed by CD31⁺ endothelial cells (Fig. 1A) but not by pericytes (Figs. 1B, C, D) in the vasculature of hyperplastic and angiogenic islets, whereas in tumors, CD13 expression partially overlaps with the pericyte markers NG2, CD140b or αSMA, but does not colocalize with CD31 (Figs. 1B, C, D). To confirm the stage specific expression of CD13 on endothelial cells, we isolated angiogenic islets and solid tumors from 6–8 to 10–13 week old mice and analyzed them by flow cytometry.²² In agreement with the immunofluorescence microscopy results, we detected the expression of CD13 in a large percentage of endothelial cells (electronically gated as CD45⁻ CD31⁺ cells) from angiogenic islets (Fig. 2A), whereas a significant lower number of endothelial cells from tumors were positive for CD13.

Figure 1.

Identification and localization of CD13⁺ cells. Co-staining of hyperplastic islets (left panels), angiogenic islets (middle panels) and tumors (right panels) of RIP1-Tag2 mice with specific antibodies: (A) CD31 (red) and CD13 (green). (B) NG2 (red) and CD13 (green). (C) CD140b (red) and CD13 (green). D) SMA (green) and CD13 (red). Scale bar 100 µm. A magnification of the selected area is reported in the upper right corner of each panel.

Figure 2.

Stage-specific CD13 expression in RIP1-Tag2 mice. (A) Quantification of CD45⁻CD31⁺CD13⁺ cells in RIP1-Tag2 pancreatic islets and tumors isolated at different stages of carcinogenesis (mean ±SE n = 4−6). (B) RIP1-Tag2 mice (9 weeks) were injected with Quantum dots (QD) alone (noneQD) or with NGR-QD in combination with anti-CD13 mAb, isotype control mAb or unlabeled NGR peptide. Pancreas were excised and processed for whole mount staining. The arrow indicates one of the several vessels stained by NGR-QD. Both anti-CD13 Abs and NGR peptide compete with NGR-QD for the binding to islets vessel. QD (red) CD31 (green). Scale bar 200 µm.

Taken together, these results indicate that CD13 expression on tumor vessels changes during the different phases of tumor development being mainly restricted to endothelial cells in the early stage of carcinogenesis. In the late stages of tumor growth CD13 expression is reduced on endothelial cells and seems to increase on pericytes. This pattern of expression may due to the fact that during tumor progression endothelial cells and pericytes are differently exposed or susceptible to the angiogenic factors known to regulate CD13.²³

To investigate the interaction between the CNGRCG peptide and CD13 in vivo, the CNGRCGVRSSSRTPSDKY peptide, corresponding to the 17 N-terminal amino acids of the NGR-TNF, was conjugated to quantum dots (hereafter reported as NGR-Qd) ²⁴ and injected in 9 weeks old RIP1-Tag2 mice, during the angiogenic switch stage. Specimens of fresh pancreas were then analyzed by ex vivo whole mount histology. NGR-Qd binding was detected along angiogenic islets-associated vessels (Fig. 2B, upper right panel). To confirm the specificity of the NGR-CD13 interaction, NGR-Qd were injected along with anti-CD13 antibody or unlabeled CNGRCGVRSSSRTPSDKY peptides. As expected, the NGR-Qd binding was specifically competed (Fig. 2B, lower panels), demonstrating that NGR-mTNF is indeed able to target the angiogenic vessels through the engagement of CD13 in RIP1-Tag2 mice.

NGR-mTNF treatment controls the development of angiogenic islets and improves vascular functionality

In RIP1-Tag2 mice, hyperplastic and dysplastic islets begin to appear between 3 and 5 weeks of age. Successively the “angiogenic switch” results in the appearance of angiogenic red islets characterized by dilated blood vessels and microhemorrhages. To investigate the effects of NGR-mTNF on the angiogenic switch, we treated RIP1-Tag2 mice at 5 weeks of age, one or three times a week for 3 weeks. Mice treated with NGR-mTNF three times a week (Fig. 3A) display a significantly reduced number of angiogenic red islets (Fig. 3B). We then analyzed cellular suspensions from angiogenic islets for the presence of viable endothelial cells, identified as CD45⁻ CD31⁺ cells, by flow cytometry. As shown in Fig. 3C, we observed a significant reduction of viable endothelial cells. In agreement, NGR-mTNF treatment significantly reduced the vascular surface area, measured by immunofluorescence staining of CD31⁺ blood vessels in hyperplastic and angiogenic islets (Fig. 3D). Since it has been demonstrated that NGR-mTNF induces apoptosis of tumor and endothelial cells in vivo,²⁵ we evaluated the number of apoptotic cells in angiogenic islets after the last treatment, by immunofluorescence microscopy with anti-cleaved-caspase3 mAb.²⁵ As expected, we observed a significant increase of apoptotic cells (Fig. S1). In addition, NGR-mTNF treatment increased vascular perfusion measured by delivery of FITC-conjugated lectin to tumor vessels (Fig. 3E).

Figure 3.

In prevention trials, NGR-mTNF treatment reduces the number of angiogenic islets and vessel density, and improves vascular functionality. (A) RIP1-Tag2 mice were treated with NGR-mTNF or saline three times a week (3 t/week) from 5 to 9 weeks of age. Pancreas were processed for islet isolation or embedded in OCT for immunohistochemical analysis. (B) NGR-mTNF treatment reduces the number of red angiogenic islets. The columns indicate mean value ±SE (n >12 samples from 3 independent experiments). (C) NGR-mTNF treatments reduces CD45⁻CD31⁺ endothelial cells. The columns indicate mean value ±SE (n > 8 samples from 2 independent experiments). (D) Vascular surface area (CD31⁺ area normalized on DAPI area) is reduced both in hyperplastic and angiogenic islets in NGR-mTNF treated mice. The columns indicate mean value ±SE (n = 5–7 animals from 3 independent experiments). (E) NGR-mTNF treatments increase vascular functionality (lectin perfused area normalized on CD31⁺ area) in hyperplastic and residual angiogenic islets. The columns indicate mean value ±SE (n = 5–7 animals from 3 independent experiments).

Similar results were obtained when RIP1-Tag2 mice were treated once a week (Fig. S2A). NGR-mTNF induces a significant reduction of angiogenic red islets and vascular surface area (Fig. S2B and D), as well as an increase of vascular perfusion (Fig. S2E) that was still detectable 96 hrs after the last treatment (data not shown), whereas the amount of viable endothelial cells was only slightly decreased (Fig. S2C).

In conclusion, both schedules tested indicated a positive effect of the drug in reducing the number of angiogenic islets and, in particular, more frequent treatments seem to be more effective.

NGR-mTNF treatment reduces tumor formation and growth

To evaluate the effects of NGR-mTNF on tumor formation and tumor growth, we started treatments in 10-week-old mice and continued the treatment until 13 weeks of age (i.e. intervention trial). We tested the two different schedules of treatments already used during the angiogenic phase. As shown in Fig. S3, a 3 d pulse of NGR-mTNF treatment (Fig. S3A) fails to reduce both tumor formation (Fig. S3B) and growth (Fig. S3C). Moreover, in situ analysis of pancreatic tissues did not reveal any reduction in vessel density (Fig. S3D) and in vascular functionality (Fig. S3E). Consequently, we concluded that in intervention trials, the dose-dense schedule of treatment was not effective on vessel perfusion and tumor growth in contrast to what observed in the early stages of angiogenic switch. Surprisingly, when we treated mice according to the schedule of less frequent treatments (Fig. 4A), tumor formation and tumor growth were reduced (Figs. 4B and C) and, although the vascular surface area was not reduced, vascular normalization, indeed, did improve (Figs. 4D and E).

Figure 4.

For figure legend, see page 7.

The quality and quantity of pericyte coverage are important parameters of vessel maturation and functionality. In RIP1-Tag2 mice, tumor vessels are lined with immature CD140b⁺ pericytes and, to a lesser extent with more mature αSMA-expressing cells that are located near vessels but mostly detached.²⁶ We analyzed tumors 72 hrs after the last treatment, by immunofluorescence with markers of mature pericytes (i.e. αSMA, NG2 proteoglycan) and of pericyte progenitors (i.e. CD140b). NGR-mTNF increases the area of αSMA⁺ and NG2⁺ pericytes per tumor, whereas the CD140b positive area did not change (Fig. 4F). These data indicate that mice treated with NGR-mTNF are characterized by more mature and stabilized vessels as compared with those of untreated mice. The significant increase of mature pericyte area and the higher vessel functionality in NGR-mTNF treated tumors support the hypothesis that this molecule is able to induce remodeling and normalization of the tumor vasculature, which delay tumor growth.

NGR-mTNF treatment upregulates mRNA levels of IFNγ, proinflammatory cytokines and vascular stabilizing factor in tumors

To investigate the mechanism underlying NGR-mTNF-mediated normalization of tumor vessels, we analyzed the expression of a panel of genes involved in the modulation of angiogenesis including endothelial cell markers, angiogenic/antiangiogenic factors, and immune modulators. In tumor lysates obtained from 13 week-old RIP1-Tag2 mice treated with NGR-mTNF we observed, by quantitative RT-PCR, a significant modulation of genes involved in vessel normalization (Fig. 5A). We found a significant increased expression of VE-cadherin, which is a marker of vessel stabilization.^27,28 Interestingly, we also detected the upregulation of sphingosine-1-phosphate receptor 1 (S1pr1) whose activation by an agonist of the natural ligand has been reported to improve VE-cadherin organization and junction integrity between endothelial cells, reducing vascular leakiness.²⁹ The other three genes that are modulated in NGR-mTNF-treated tumors encode proinflammatory cytokines: IL-6, IL-1β, and IFNγ. IL-6 and IL-1β, which are secreted by many tumor stroma cells, ³⁰ including M1-like macrophages,^31-33 contribute to the activation of the protective adaptive immunity.^34,35 In agreement, mRNA for IFNγ, a cytokine produced by Th1 and CD8⁺ T cells, was strongly upregulated by NGR-mTNF treatment, supporting the idea of an increased infiltration and/or activation of proinflammatory lymphocytes. The expression level of several genes involved in proangiogenesis and vessel abnormalization ^28,36 was not altered (Fig. 5A).

Figure 5.

NGR-mTNF treatment increases the expression of genes involved in vessels normalization and the number of tumor infiltrating CD8⁺ cells. RIP1-Tag2 mice were treated with NGR-mTNF or saline once a week from 10 to 13 weeks of age as illustrated in Fig. 5. Tumors were excised from pancreas and processed for gene expression profiling or cytofluorimetric analysis. (A) the expression of genes involved in angiogenesis was examined by quantitative RT-PCR performed on the entire tumor mass. Only the genes differentially expressed between NGR-mTNF and saline treated tumors (*, p < 0.05; **, p < 0.01) and some genes relevant for proangiogenesis and vessel abnormalization are shown (mean fold change ±SE of n = 8 tumors for each group, from 3 independent experiments). (B) NGR-mTNF treatments (black bars) increase the percentage of CD8⁺ T cells in tumors over saline (gray bars), whereas they do not affect the number of CD4⁺ cells. The column indicate mean value ±SE (n > 11 samples from 4 independent experiments).

Altogether, these results support the hypothesis that NGR-mTNF treatment induces tumor vascular normalization and a favorable inflammatory tumor microenvironment, eventually activating antitumor immune responses.

NGR-mTNF treatments promote infiltration of CD8⁺ T cells in tumors

Tumor vessel normalization induced by different agents has been associated with an increased infiltration of naturally occurring as well as adoptively transferred T cells.^12,28,37,38 To test the hypothesis that NGR-mTNF-mediated vessel normalization could increase the number of extravasating T cells, tumors were collected 72 hrs after the last weekly treatment, cellularized and analyzed by FACS to detect CD4⁺ and CD8⁺ T cells. Little infiltration by either CD4⁺ or CD8⁺ T lymphocytes was observed in tumors from untreated control mice. In contrast, NGR-mTNF-treated tumors showed a significantly higher percentage of CD8⁺, but not CD4⁺ cells (Fig. 5B). As expected, the CD8⁺ infiltrating lymphocytes displayed an effector phenotype (CD44⁺CD62L⁻) (data not shown). These data are in agreement with the higher level of IFNγ mRNA expression observed in NGR-mTNF treated mice as compared to the untreated, and support the hypothesis that NGR-mTNF is able to stimulate the recruitment of CD8⁺ T lymphocyte, which could play a role in the control of tumor growth.

Antitumor efficacy of NGR-mTNF is lost in immunodeficient mice

To verify whether the immune system has a relevant role in mediating NGR-mTNF antitumor activity, we took advantage of the immunodeficient variant of RIP1-Tag2 mice (RIP1-Tag2 RAG1^−/−) developed by Casanovas et al.²¹ RIP1-Tag2 RAG1^−/− mice are completely deficient in adaptive immunity (B and T cells), but they have the same inflammatory infiltration of innate immune cells (macrophages and granulocytes) detected in RIP1-Tag2 mice, and display the same multistage tumor progression pattern.²¹ We analyzed CD13 expression in angiogenic islets from 6–8 to 10–13 weeks old mice and in tumors from 10–13 weeks old mice (Fig. S4). As observed in immunocompetent mice, CD13 is significantly more expressed in endothelial cells (electronically gated as CD45⁻CD31⁺) from vessels of angiogenic islets than from tumor vessels. Unlike the RIP1-Tag2 animals, which are characterized by a peak of CD13 expression at 8 weeks (Fig. 2A), in RIP1-Tag2 RAG1^−/− mice the peak was observed slightly before. This effect could be due to minor difference in the genetic background of the two strains.

To investigate effects of NGR-mTNF on the angiogenic switch occurring in an immunodeficient environment, we treated RIP1-Tag2 RAG1^−/− mice at 5 weeks of age, three times a week for 3 weeks (Fig. 6A). NGR-mTNF significantly reduced both the number of angiogenic red islets (Fig. 6B) and the vascular surface area (Fig. 6C), and improved the vessel functionality (Fig. 6D), thus excluding a role for the immune system in this early stage of tumor development. Different results were obtained when we evaluate the effects of NGR-mTNF on tumor formation and growth starting a weekly treatment at week 10 of age (Fig. 7A). In the absence of the adaptive immune system, no significant differences in tumor formation and growth were observed in NGR-mTNF treated mice compared to the untreated ones (Figs. 7B and C). Moreover, microscopic analysis of pancreatic tissues fails to reveal reduction of vessel density and increased vascular functionality (Figs. 7D and E).

Figure 6.

In RIP1-Tag2 RAG1^−/−, NGR-mTNF treatment reduces the number of angiogenic islets and vessel density, and improves vascular functionality. (A) RIP1-Tag2 RAG1^−/− mice were treated with NGR-mTNF or saline three times a week (3 t/week) from 5 to 9 weeks of age. Pancreas were processed for islets isolation or embedded in OCT for immunohistochemical analysis. (B) NGR-mTNF treatment reduces the number of red angiogenic islets. The columns indicate mean value ±SE (n>15 samples from 4 independent experiments). (C) vascular surface area (CD31⁺ area normalized on DAPI area) is reduced both in hyperplastic and angiogenic islets in NGR-mTNF treated mice. The columns indicate mean value ±SE (n = 4 animals from 2 independent experiments). (D) NGR-mTNF treatments increase vascular functionality (lectin perfused area normalized on CD31⁺ area) in hyperplastic and residual angiogenic islets. The columns indicate mean value ±SE (n = 4 animals from 2 independent experiments).

Figure 7.

In RIP1-Tag2 RAG1^−/− mice NGR-mTNF treatment fails to reduce tumor number, tumor growth, and vessel density, and to increase vascular functionality, and alter the expression of genes involved in vessel normalization. (A) RIP1-Tag2 RAG1^−/− mice were treated with NGR-mTNF or saline, once a week from 10 to 13 weeks of age (intervention trial) as illustrated in Fig. 5. NGR-mTNF treatment fails to reduce (B) tumor formation, as well as (C) total tumor burden per animal. The columns indicate mean value ±SE (n >10 samples from 3 independent experiments). (D) Vascular surface area (CD31⁺ area normalized on DAPI area), as well as (E) vascular functionality (lectin perfused area normalized on CD31⁺ area) evaluated in tumors from NGR-mTNF and saline treated mice are comparable. The columns indicate mean value ±SE (n = 4 animals from 2 independent experiments). (F) Gene expression, assessed by quantitative RT-PCR in whole tumor lysate from NGR-mTNF treated RIP1-Tag2 RAG1^−/− mice, is expressed as mean fold change ± SE over saline treatment (n = 5 tumors for each group, from 2 independent experiments). The genes shown are the same presented in Fig. 6. (*p < 0.05).

We then examined the expression of genes involved in angiogenesis modulation in tumors from control and NGR-mTNF-treated mice (Fig. 7F). IFNγ expression was undetectable, as expected for the absence of T cells. Moreover, all markers of vessel normalization, such as VE-cadherin, S1pr1, and proinflammatory cytokines were not upregulated. On the contrary, the expression of two proangiogenic factors, EGF and FGF1 was significantly increased. These data correlate with the absence of vessel normalization observed in RIP1-Tag2 RAG1^−/− tumors after NGR-mTNF treatment (Fig. 7E).

Altogether these results provide support for the hypothesis that the presence of an adaptive immune system is critical for NGR-mTNF antitumor activity.

Discussion

Based on the observation that anti-VEGF therapy has limited effects as monotherapy, but enhances the efficacy of systemic chemotherapy, Jain and colleagues postulated the “vascular normalization” hypothesis, stating that antiangiogenic therapies can revert the grossly abnormal structure and function of the tumor vasculature toward a more normalized state.³⁹ A large number of pre-clinical and clinical studies have provided evidence supporting the presence of a transient normalization window favoring delivery of chemical compounds and effector cells into the tumor mass. ²⁸

NGR-mTNF is able to improve doxorubicin and melphalan delivery in tumors when administered 2 hrs in advance. This effect has been associated to the capacity of TNF to alter endothelial barrier functions, increasing permeability and lowering interstitial tumor pressure.^14,15 Nevertheless, NGR-mTNF mechanism of action is still poorly understood especially if we consider the long-term effects on vessels and/or the involvement of the immune system. In this study, we addressed the effects of NGR-mTNF administered in the RIP1-Tag2 mouse model of pancreatic islet carcinogenesis,¹⁶ which recapitulates the tumor progression multistep process and has largely been used to study the local effects and the mechanism of action of angiogenesis inhibitors.^20,21,40

In the early phase of carcinogenesis, we observed that NGR-mTNF reduces the number of hemorrhagic islets, probably by preventing the angiogenic switch, and induces vessel normalization in the few surviving islets. Notably, in this phase we detected a decrease of the vascular surface area most likely related to the high level of CD13 expression on endothelial cells and the cytotoxic activity of NGR-mTNF.⁴¹ In support of the hypothesis of a direct effect of NGR-mTNF on endothelial cells, we observed that the three-weekly treatment was more effective than the weekly administration in reducing their percentage. Moreover, the same effects were observed when NGR-mTNF was administered in the prevention trial to immunodeficient RIP1-Tag2 RAG1^−/− mice. Later on in the intervention trial, administration of the NGR-mTNF once a week, does not cause a decrease in tumor vessel density but a clear normalization of the vascular network. This may be due to the reduced CD13 expression by the endothelial cells in tumors, as well as to the to the high proliferation rate of tumor cells, which may hide the cytotoxic effect of NGR-mTNF on the vascular surface area.

Tumor microenvironment plays an important role in angiogenesis and tumor growth; the presence of specific cytokines, chemokines and other factors could skew TAM polarization away from the M2- to a tumor-inhibiting M1-like phenotype, promote antitumor immune responses and vessel normalization, which eventually decreases tumor growth and metastasis and enhances chemotherapy effects.²⁷ In particular, IFNγ has been shown to re-educate human TAM toward M1-polarized immunostimulatory macrophages, in vitro.⁴² Depending on the local concentration, TNFα has been shown to induce either tumor vessel destruction or vessel normalization.¹² In patients undergoing isolated limb perfusion, high-dose TNFα initially causes endothelial cell activation and redistribution of junctional and cytoskeletal molecules, then the activated tumor vessels are progressively destroyed.⁴³ On the contrary, low-dose (microgram range) of RGR-targeted TNFα modulate tumor-resident macrophages to induce vessel stabilization and to reduce vascular leakiness, enhancing their functionality and immune-mediated tumor rejection.¹² Our results, showing reduced tumor burden and increased vessel functionality in treated mice, support the hypothesis that the low dose of NGR-mTNF used in this study, could alleviate immunosuppression improving tumor vascularization. This hypothesis is further supported by the gene expression analyses performed on total RNA from growing tumors showing the up-regulation of proinflammatory cytokines and vascular stabilizing factors in NGR-mTNF treated mice. Differently from our experimental setting, vessel cell death has never been detected in tumors from TNF-RGR treated mice,¹² whereas early treatment of RIP1-Tag2 mice with NGR-TNF results in a significant decrease of the vascular surface area. This discrepancy, may be due to the different stage of tumor progression studied or to the different receptor engaged (the RGR motif binds as putative receptor, the platelet-derived growth factor receptor ⁴⁴). Indeed, we have experimental evidence showing that CD13 is not a mere homing receptor, but directly contributes to the activation of the signaling pathways elicited by NGR-TNF (manuscript in preparation).

Normalization of the tumor vasculature has been associated to an increase lymphocytes infiltration in several preclinical models. Angiogenesis inhibitors like anginex, endostatin, angiostatin have been shown to induce vessel remodeling, leading to leukocytes infiltration in tumors.²⁸ Additionally, in RIP1-Tag2 mice both the radiation-induced local release ^37,38 and the RGR-mediated tumor targeting of TNF,¹² have been shown to stabilize tumor vessels and allow T cell adoptively transferred to extravasate and to destroy tumors. In agreement with this hypothesis, we detected an increased infiltration of activated CD8⁺ T cells in the tumors from NGR-TNF-treated mice. Moreover, when we analyzed the effect of NGR-mTNF on tumors growing in RIP1-Tag2 RAG1^−/−, we fail to detect any effect in vascular normalization and tumor control, as well as in the upregulation of genes encoding proinflammatory cytokines and vascular stabilizing factors.

These results underline the importance of the immune system in NGR-mTNF anticancer activity. A role for adaptive immune system has previously been suggested by studies showing a synergism between NGR-TNF and doxorubicin in immunocompetent but not in nude or IFNγ^−/− mice ⁴⁵ and by the development of tumor-specific memory responses in mice previously cured by NGR-TNF.¹ In this context, our results confirm and extend these observations demonstrating a clear requirement of immune cells for the establishment of a normalized vascular network and the control of tumor growth by NGR-mTNF. Indeed, in treated tumors from immunodeficient RIP1-Tag2 RAG1^−/− mice, we did not observe increased vascular functionality, or upregulation of any of the markers of vessel normalization, whereas, we observed a significant increase of EGF and FGF1, two angiogenic factors associated to vessel abnormalization.^28,36 Of note, in our system the increase expression of these factors does not reflect into a more aggressive tumor phenotype compared to the untreated controls.

In intervention trial, NGR-mTNF administered three times a week did not induce any responses in contrast with what observed in the weekly schedule (Fig. S2). We speculated that the different effects may be correlated to counterregulatory mechanisms elicited by the multiple infusions of NGR-mTNF. The release of soluble TNF receptors (sTNF-R), a well-known TNF counterregulatory mechanism,⁴⁶ was not involved in the phenomenon, because the concentration of sTNF-R was not affected by the repeated treatments (data not shown). Since NGR-mTNF induces IFNγ up-regulation, as demonstrated by gene expression analysis, an alternative explanation is the onset of IFNγ-mediated counterregulatory mechanisms. Indeed, it has been demonstrated that frequent administrations of IFNγ-NGR increase the activity of indoleamine 2,3-dioxygenase (IDO) that may down-regulate T cells, eventually inhibiting antitumor immune responses.⁴⁷

In summary, we hypothesize that, in the very initial phase of tumor development, NGR-mTNF may exert a direct cytotoxic effects eliminating vessel endothelial cells of hyperplastic and angiogenic islets, which at this stage over express CD13. Later on, in tumors, when the expression of CD13 is down-modulated and preferentially located on pericytes, the antitumor effects of NGR-TNF are mainly mediated by different mechanisms involving the immune system. In particular, our results suggest that NGR-mTNF, by its direct effects on tumor associated vessels on tumor stroma cells (e.g. macrophages) determine vessel activation and extravasation of immune cells. The IFNγ released in tumors by activated T cells may synergize with the TNF inflammatory effects in the skewing to M1-like TAM, leading to the vessel normalization observed up to 72 hrs after the last treatment. Normalization of the tumor vasculature enhances tumor perfusion and infiltration of immune effectors, and maintains the polarization of TAM toward the immunostimulatory M1-like phenotype, thus favoring the establishment of an immunosupportive microenvironment, eventually affecting tumor growth.

Besides unveiling an important role for the adaptive immune system in tumor growth control and vessel normalization induced by NGR-mTNF, the findings described in this study provide support for the design of new treatment protocols that might improve the efficacy of chemo and immunotherapies by taking advantage of the vessel normalization window.

Materials and Methods

Tumor models and treatments

RIP1-Tag2 ¹⁶ and RIP-Tag2 Rag1^−/− mice ²¹ were maintained in the C57BL/6N background (Charles River). From 12 weeks of age all mice received 50% sugar food (Mucedola) and 5% sugar water to relieve hypoglycemia induced by insulin secreting tumors. Preparation and characterization of NGR-mTNF was previously described.²⁵ NGR-mTNF was administered intraperitoneally (i.p.) at 5 pg/gr per mouse once or three times a week. All mouse experiments were performed in accordance with the guidelines of the Ethical Committee of the San Raffaele Scientific Institute.

Determination of number of angiogenic islets and tumor burden

Angiogenic islets, identified as those that exhibited a red hemorrhagic appearance,²⁰ were isolated by retrograde perfusion following collagenase V (Sigma) digestion, and counted under a dissecting microscope. Tumors were micro-dissected from freshly excised pancreas. Images were captured under a dissecting microscope with a DP50 Olympus digital camera using Viewfinder Lite software (Pixera Corporation). Tumors were analyzed with ImageJ and the formula [volume = 0.52 × (width)2 × length] was applied to determine the tumor volume (mm³). Tumor burden per mouse was calculated by adding up the volume of every tumor. The index of tumorigenesis represents the percentage of tumors with a diameter ≥ 2 mm on the total number of islets that underwent angiogenic switch.

Flow cytometry analysis of RIP1-Tag2 tumors

Angiogenic islets and solid tumors were digested with 1 mg/mL collagenase B (Roche), 2mg/mL dispase (Sigma), 0.1 mg/mL DNasi I (Roche) for 1 h at 37°C and then filtered on a 70-μm nylon cell strainers. The cells were suspended in PBS with 2% FCS and, after Fc blocking (anti mouse CD16/32, clone 2.4G2 BD Bioscience), were labeled with mAbs against the following antigen: CD45.2 (104 BD Bioscience), CD31 (MEC 13.3 BD Bioscience), CD13 (R363 Serotec), CD8 (53–6.7 BD Bioscience), CD4 (GK 1.5 BD Bioscience). Live cells were identified using Live/Dead Fixable Far Red Dead Cell Stain Kit (Invitrogen). At least 100,000 cells per sample were acquired with a FACSCanto system (BD-Biosciences). Fluorescence minus one (FMO) stained samples were used to gate marker-positive cells. Percentages of stained cells were determined using the FACSDiva software (BD-Biosciences).

Lectin injection and fixation by vascular perfusion

At the end of treatments, anesthetized mice were i.v. injected with 100 µL of FITC-labeled L. Esculentum Lectin (1 mg/mL in 0.9% NaCl; Vector Laboratories), 10 min prior to the perfusion. Heart perfusion was performed with 10 mL of fixative (4% paraformaldehyde in PBS, pH 7.4) followed by 10 mL of PBS. Pancreas was harvested, incubated over night at 4°C in 30% sucrose and then frozen in OCT compound and processed for immunohistochemistry

Immunohistochemistry

Cryostat sections of pancreas (10 µm thickness) were dried on Superfrost Plus slides (Fisher Scientific), fixed with 4% paraformaldehyde and then incubated with 0.2% Triton X-100, 10% normal goat serum to block nonspecific antibody binding. Sections were then incubated for 15–18 hrs with primary antibodies at 4°C. The following antibodies were used: rat mAb anti-CD31 (PECAM-1, cl.MEC 13.3, BD Bioscience) or rabbit anti-CD31 (AbCam), anti-αSMA FITC (Sigma-Aldrich), rat mAb anti-CD140b (eBioscience), rabbit anti-NG2 (Millipore), rat mAb anti-CD13 (cl.R363, Serotec) or rat mAb anti-CD13-PE (cl.R3242, Serotec). Apoptosis was assessed by a polyclonal rabbit anti-cleaved-caspase3 (Cell Signaling). Secondary antibodies were Goat anti rat Alexa488 /Alexa568 and/or Goat anti rabbit Alexa488/Alexa568 (Invitrogen).

Fluorescence imaging and analysis

Tissue sections were examined with a Olympus BX61 fluorescence microscope equipped with single, dual and triple fluorescence filters, and a Colorview camera (Olympus). Digital images of all lesions visible in 3–4 cryostat sections cut at multiple levels of each pancreas were captured. Islets and tumors were analyzed with public domain ImageJ software. Tumor size was quantified by measuring tumor longer diameter. On the basis of the diameter size, lesions were divided in two groups: dysplastic-angiogenic islets (diameter between 150 and 1000 µm) and tumors (diameter ≥ 1000 µm). Islets and tumor border were outlined by freehand tool to identify the area of interest/analysis and a threshold was applied, excluding the region outside the lesions. Vessel surface area (VSA) and lectin expression were assessed on threshold-applied images by measuring the total tumor area occupied by pixel positive for CD31 and lectin fluorescence. Vascular functionality represents the percentage of lectin area normalized on CD31 area. Apoptotic area was obtained by measuring the total tumor area occupied by cleaved caspase-3, and the same protocol was used for αSma, NG2 and CD140b markers to asses pericyte area.

Gene expression analysis

From tumors isolated as previously described, total RNA was extracted with RNeasy Microarray Tissue Mini Kit (Qiagen), and cDNA synthesis was performed using the RT2 First Strand Kit (Qiagen). The transcriptional pattern was investigated with the RT2 Profiler PCR Array for mouse angiogenesis (Cat. No. 330231 PAMM-024Z; Qiagen), following manufacturer's instructions. The results of the expression of 84 key genes were analyzed by the SDS 2.3 program and the ΔΔCt method. Gene expression was normalized to three housekeeping genes (β-actin, β-glucoronidase, and heat shock protein 90α), and variation in gene expression was expressed as fold change ± standard error (SE) over the mean of control samples.

Statistical Analysis

Statistical analysis was performed by Student's t test. At least three independent experiments, unless otherwise indicated, were combined and data are reported as mean ± SE. Values of p < 0.05 were considered significant.

Disclosure of Potential Conflicts of Interest

Porcellini S, Asperti C, Valentinis B, Tiziano E, Mangia P, Bordignon C, Rizzardi GP and Traversari C are Molmed employees.

Acknowledgments

The authors thank Dr Vincenzo Russo, Unit of Immuno-Biotherapy of Melanoma and Solid Tumors, Scientific Institute San Raffaele, Milan, Italy for helpful discussions and Dr Douglas Hanahan, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland for providing us with reagents.

Supplemental Material

Supplemental data for this article can be accessed on the publisher's website.