Angiopoietin-1/Tie-2 activation contributes to vascular survival and tumor growth during VEGF blockade

Abstract

Approval of the anti-vascular endothelial growth factor (VEGF) antibody bevacizumab by the FDA in 2004 reflected the success of this vascular targeting strategy in extending survival in patients with advanced cancers. However, consistent with previous reports that experimental tumors can grow or recur during VEGF blockade, it has become clear that many patients treated with VEGF inhibitors will ultimately develop progressive disease. Previous studies have shown that disruption of VEGF signaling in tumors induces remodeling in surviving vessels, and link increased expression of angiopoietin-1 (Ang-1) with this process. However, overexpression of Ang-1 in different tumors has yielded divergent results, restricting angiogenesis in some systems while promoting it in others. These data raise the possibility that effects of Ang-1/Tie-2 may be context-dependent. Expression of an Ang-1 construct (Ang1*) did not significantly change tumor growth in our model prior to treatment, although vessels exhibited changes consistent with increased Tie-2 signaling. During inhibition of VEGF, however, both overexpression of Ang1* and administration of an engineered Ang-1 agonist (Bow-Ang1) strikingly protected tumors and vasculature from regression. In this context, Ang-1/Tie-2 activation limited tumor hypoxia, increased vessel caliber, and promoted recruitment of mural cells. Thus, these studies support a model in which activation of Tie-2 is important for tumor and vessel survival when VEGF-dependent vasculature is stressed. Understanding such mechanisms of adaptation to this validated form of therapy may be important in designing regimens that make the best use of this approach.

Figures

Figure 1. Tumor vasculature is altered by expression of Ang1*

(A) Schematic comparison of native Ang-1, Ang1*, and Bow-Ang1. The Ang1* construct differs from native Ang-1 in that the N-terminal 77 amino acids are replaced by the 73 N-terminal amino acids of Ang-2. The ability to activate Tie-2 resides in the C-terminal fibrinogen-like domain, and its bioactivity therefore resembles native Ang-1. BowAng1 is a tetrameric Ang-1 mimetic which takes advantage of the requirement for higher-order multimers for Tie-2 activation, and which potently stimulates this axis [19]. (B) Ang1* secreted by SKNEP-1 cells stimulates Tie-2 phosphorylation in cultured human endothelial cells. Recombinant human Ang-1 was used as a control. Stimulation was less pronounced but also present in murine endothelioma cells (BEnd3). (C) Tumor weights did not differ in Ang1*- and GFP-expressing SK-NEP1 xenografts after the initial 5.5 weeks of growth. (D) Vasculature is remodeled in Ang1* tumors as compared to GFP-expressing controls. Immunostaining for vascular basement membrane (Collagen IV) and endothelial cell (PECAM) markers demonstrates a decrease in fine branching in Ang1* tumors. There is a distinct increase in recruitment of cells expressing the nascent pericyte marker NG2(+)to vessels, without a corresponding increase in the marker for the more differentiated mural cells (alpha smooth muscle actin, ASMA). (E) Corresponding to the greater predominance of large caliber vessels in Ang1* tumors suggested by immunostaining, mean vascular radius is increased 124.4% versus GFP-transfected xenografts at Day 0 (P < 0.0001).

Figure 2. Tumors engineered to over-express an angiopoietin-1 construct are not regressed by VEGF blockade

Xenografts expressing Ang1* or GFP were allowed to grow for 6 weeks, then treated with 25 mg/kg VEGF Trap biweekly. Whereas GFP-expressing tumors were regressed by 80% from starting weights by day 36 of treatment with VEGF Trap (filled triangles; 5.6 ± 1.1 g at day 0 vs. 1.0 ± 0.3 g at day 36, P = 0.0003), Ang1*-expressing tumors were not regressed (filled circles). Fc-treated Ang1* controls and GFP controls grew progressively (empty circles, Ang1* Fc-treated; empty triangle, GFP controls Fc-treated).

Figure 3. Tissue hypoxia is diminished in Ang1*-expressing tumors during VEGF Trap treatment

(A) Using pimonidazole staining, GFP-expressing controls were found to be extensively necrotic and hypoxic at day 36, whereas Ang1*-expressing xenografts were minimally hypoxic at the same time point, consistent with comparatively robust perfusion. (B) VEGF Trap treatment stimulated expression of the hypoxia-regulated factors VEGF and CXCL12 (SDF-1) in GFP-transfected tumors, whereas expression was blunted in Ang1* expressing tumors, consistent with decreased hypoxia.

Figure 4. Remodeled vasculature in Ang1* tumors is resistant to destabilization by VEGF Trap

(A) L. esculentum perfusion studies demonstrate large vessels with an erratic network of fine branches in Ang1* tumors prior to treatment (day 0). These larger vessels persist despite ablation of the irregular small vessels after 1 and 5 days of VEGF Trap treatment, suggesting that this second population is VEGF-dependent. These alterations were strikingly different from those previously reported in perfusion studies of control SK-NEP1 xenografts during days 1 and 5 of VEGF Trap administration, in which the bulk of vasculature is eradicated by day 5 [21]. Similar loss of fine branches was observed on PECAM immunostaining. (B) We quantitated these findings using computerized image analysis of lectin perfusion studies. Whereas vessel number decreased significantly by day 5 of treatment (51% of day 0 values; P = 0.005), mean vessel density and vessel length were not significantly changed (83% of day 0 values; P=NS). Thus, vascular ablation is significantly blunted in Ang1*-expressing tumors during VEGF Trap administration, as compared with SK-NEP-1 controls [18].

Figure 5. Injection of an Ang-1 agonist prevents regression and permits tumor growth during VEGF Trap treatment

In order to confirm the ability of Ang-1/Tie-2 signaling to protect tumors from regression induced by VEGF blockade, we injected an engineered angiopoietin tetramer which activates this axis (Ang-F1-Fc-F1, BowAng-1 [19]; Regeneron Pharmaceuticals, Tarrytown, NY). SK-NEP-1 xenografts were allowed to grow for 29 days. Mice with palpable tumors were then randomly divided into three groups, receiving (1) injections of BowAng-1 (25 mg/kg delivered subcutaneously three times per week); (2) human IgG1 Fc protein (Fc; 25 mg/kg biweekly); and (3) no treatment. Untreated mice were sacrificed 35 days after implantation (day (−7)) to establish baseline vessel architecture and mean tumor weights (2.6 ± 0.9 g). After 13 days of pre-treatment (day 0), tumor weights in both BowAng1 and Fc pre-treatment groups were similar to each other (4.0 ± 1.3 g and 3.6 ± 1.1 g, respectively). At this point, Fc pre-treated mice were randomly subdivided into two treatment groups that either continued to receive Fc (empty triangles) or Fc + VEGF Trap injections (filled triangles). BowAng1 pre-treated mice remained on the same regimen and additionally started to receive VEGF Trap injections (filled circles). Mice were killed on treatment day 5 or day 27 for tumor weight evaluation (N=4 for each treatment group at each timepoint). At day 5, VEGF Trap caused xenograft growth delay in both Fc and BowAng1 pre-treated mice as compared to Fc treated controls (3.8 ± 1.2 g and 4.2 ± 2.2 g respectively, vs. 5.9 ± 0.9 g). By day 27, tumors in VEGF Trap treated mice showed partial regression (solid circles, 2.7 ± 0.5 g, P = 0.043, as compared to Fc treated mice, unfilled circles) consistent with our previous data [18]. Strikingly, BowAng1 + VEGF Trap-treated xenografts were comparable to the Fc-treated controls (7.8 ± 1.9 g vs. 9.1 ± 1.6 g, solid triangles), indicating that sustained delivery of Ang-1 prevents regression by VEGF Trap.