The strategy of targeting angiogenesis to inhibit cancer progression has received increasing attention in recent years. Despite the recent approval of targeted therapies in this area, optimizing the use of anti-angiogenic drugs in the clinic has been difficult. Challenges that face anti-angiogenic agents that are currently under development include choosing disease areas that might benefit most, optimizing combination strategies with existing standards of care and defining patient populations that might respond best to therapy. Preclinical models of disease provide the best opportunity for addressing these issues, therefore appropriate reagents for use in these systems are essential for driving drugs through development.
Volociximab has been shown to inhibit the growth of new blood vessels in preclinical models of ocular angiogenesis [
21]. This effect was found to translate into decreased tumor growth in the rabbit VX2 carcinoma model [
22]. These experiments provided a strong proof of concept demonstration of volociximab activity
in vivo and defined a novel mechanism of action for angiogenesis inhibition. However, the VX2 model is limited in that it represents a very aggressive tumor, must be passaged
in vivo, is carried out in immunocompetent animals (resulting in antibody clearance) and requires large amounts of antibody. To further define volociximab mechanism of action and identify appropriate settings for its use in tractable animal models of cancer, it was therefore imperative a similar reagent with activity in mouse be generated.
A number of antibodies against mouse α5β1 are available commercially. We have found that although some of these antibodies inhibit binding of α5β1 to fibronectin, none inhibited other biological functions, such as migration,
in vitro angiogenesis or tumor growth
in vivo (unpublished observations; [
23,
24]). However, the α5 knockout mouse is embryonically lethal due to gross defects in vascular architecture [
28], suggesting that in mice, as in humans, α5β1 is important for blood vessel formation and/or integrity. The new panel of reagents described herein represents a number of α5- and β1-specific antibodies. Of note, Fc-fusion protein-based immunizations resulted in a higher proportion of α5-specific antibodies, whereas placenta-based immunization resulted in a higher proportion of heterodimer-specific antibodies, including 339.1 (data not shown). As the overall number of antibodies produced by each method was similar, this suggests that the purified material may have resulted in similar immunogenicity while maintaining a more native quaternary structure
in vivo. In either case, many of the antibodies that bound α5 or were specific for α5β1 heterodimer blocked binding to fibronectin and competed, at least in part, with one another in ELISA or FACS assays (data not shown). Of these antibodies, one group cross-reacted with human integrin, while another did not, suggesting that at least two distinct epitopes were represented. This implies that inhibition of binding to fibronectin can be achieved through blocking at multiple sites, possibly through steric hindrance. Importantly, not all antibodies that block binding to fibronectin have equivalent biological function
in vitro or
in vivo. 517-2 and 339.1, for example, each bind with high affinity (0.21 nM and 0.59 nM, respectively) block binding to fibronectin and inhibit migration. Moreover, both antibodies have rat IgG1 constant regions, which like volociximab, a human IgG
4, would be predicted to lack significant effector activity. However, only 339.1, which does not cross-react with human α5β1, elicits significant cell death
in vitro and inhibits angiogenesis and tumor progression
in vivo. This finding suggests that although these antibodies have similar biological functionality and similar affinities, initiation of the cell death program requires binding to a highly specific epitope. This result also suggests that 339.1 binds to the murine cognate of the epitope recognized by volociximab, which would be predicted to be non-homologous between mouse and human α5β1, since volociximab does not cross-react with mouse integrin. A corollary of this hypothesis is that an antibody that recognizes both human and mouse integrin would not bind this important epitope, and therefore might not elicit cell death, as is the case with 517-2.
339.1 inhibits tumor growth in an A673 rhabdosarcoma model. This model was chosen to evaluate anti-α5β1 activity because it was reported to be sensitive to the mouse parent antibody of bevacizumab, A.4.6.1, suggesting that its growth is highly dependent on angiogenesis [
8]. However, A.4.6.1 and bevacizumab do not inhibit tumor growth in other xenograft models to the same extent as in the A673 model [
29,
30]. The reasons for this are not fully understood; 339.1 is currently being evaluated in additional xenograft models to determine if similar differences in sensitivities are observed with this antibody. Comparing xenograft models that respond to 339.1 to varying degrees may reveal molecular mechanisms that will help stratify patients and illuminate combination strategies that might best slow disease progression. In addition, the combination of 339.1 and volociximab is currently being assessed in xenograft models to determine the effect of targeting both tumor and host α5β1
in vivo.