Integrin ␣v3 targeted therapy for Kaposi’s sarcoma with an in vitro evolved antibody1 CHRISTOPH RADER,2 MIKHAIL POPKOV, JOHN A. NEVES, AND CARLOS F. BARBAS III2 Department of Molecular Biology and The Skaggs Institute for Chemical Biology, The Scripps Research Institute, La Jolla, California, USA SPECIFIC AIMS The goal of this study was to evaluate integrin ␣v3 as a molecular target for antibody therapy for Kaposi’s sarcoma (KS). We used in vitro evolution based on phage display to generate a humanized and affinitymatured antibody directed to integrin ␣v3 to evaluate tumor targeting and tumor growth inhibition in a mouse model of KS. PRINCIPAL FINDINGS 1. In vitro evolution of antibody JC-7U directed to integrin ␣v3 Using a phage display strategy that preserves only the original CDR3 sequences of light and heavy chain while subjecting the remaining sequence to selection from naive human V gene libraries, we previously reported the full humanization of mouse monoclonal antibody (mAb) LM609 directed to human integrin ␣v3. Here we further improved the affinity of humanized LM609 using a phage display strategy for sequential and parallel optimization of three complementarity determining regions (LCDR1, HCDR1, and HCDR3) of the antibody molecule. The evolved Fab, designated JC-7U, had an affinity of 150 pM and was converted into IgG1 using a new mammalian expression vector (Fig. 1). 2. Tumor targeting of antibody JC-7U in a mouse model of Kaposi’s sarcoma A previously described nude mouse model of KS based on human KS cell line SLK was used for our studies. As an isotype control for JC-7U IgG1, we used human b12 IgG1, which binds to gp120 of HIV-1. Flow cytometry revealed strong binding of JC-7U IgG1 to SLK cells, whereas no binding was detected for b12 IgG1. For in vivo studies, tumor induction was performed by subcutaneous injection of 5 ⫻ 106 SLK cells in the right flank of nude mice. We first analyzed tumor targeting by injecting 100 g of JC-7U IgG1 or b12 IgG1 into the tail vein of mice with an established SLK xenograft. After 24 h, tumor, lung, liver, and spleen were removed and processed for the detection of bound human IgG1 by immunohistochemistry. As shown in Fig. 2, JC-7U IgG1 2000 was found to have intensely penetrated the tumor tissue whereas control antibody b12 IgG1 revealed only slight nonspecific uptake. By contrast, neither JC-7U IgG1 nor b12 IgG1 were found in normal tissue. 3. Antibody JC-7U inhibits tumor growth in a mouse model of Kaposi’s sarcoma With the selective tumor targeting of JC-7U IgG1 established, we next analyzed whether JC-7U IgG1 can inhibit tumor growth. Three different groups of five animals each were treated on days 1, 4, 7, 10, and 13 after tumor induction. Each treatment involved a 200 l tail vein injection of 1) 0.5 mg/mL JC-7U IgG1 in PBS, 2) 0.5 mg/mL b12 IgG1 in PBS, or 3) PBS alone. Thus, animals in the antibody groups received five doses of 100 g or 5 mg/kg. For comparison, clinical studies with Vitaxin, another humanized derivative of LM609, have used doses up to 4 mg/kg with little or no toxicity, which suggests that our dosing regimen is in a clinically relevant range. Study of animals treated with JC-7U IgG1 revealed no obvious signs of toxicity for this protein, as indicated by no weight loss or behavioral change during the course of therapy. Tumor volumes of treated animals were measured every third day starting on day 9 and ending on day 54 (Fig. 3). Tumor growth in animals treated with PBS or b12 IgG1 revealed no difference compared with tumor growth in untreated animals, resulting in an average tumor volume of 650 and 685 mm3, respectively, tumor growth in animals treated with JC-7U IgG1 was notably slower (Fig. 3). The difference in tumor growth was statistically significant (P⬍0.05) from day 27 until the end of the experiment 27 days later. On day 54, average tumor volume in the animals treated with JC-7U IgG1 was 292 mm3, which is ⬃45% of the average tumor volume observed in the control groups. As an additional measurement, all tumors were surgically removed on day 54 and weighed. The average tumor weight in the three treatment groups confirmed the results obtained for 1 To read the full text of this article, go to http://www. fasebj.org/cgi/doi/10.1096/fj.02– 0281fje; to cite this article, use FASEB J. (October 18, 2002) 10.1096/fj.02– 0281fje 2 Correspondence: Department of Molecular Biology, BCC-526, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA. E-mail: crader@scripps.edu and carlos@scripps.edu 0892-6638/02/0016-2000 © FASEB Figure 1. Tumor growth inhibition in a nude mouse model of KS in the presence of JC-7U IgG1. JC-7U IgG1 (black) or control antibody b12 IgG (white circles) was given in five doses of 5 mg/kg every third day starting on day 1 after tumor induction. Statistically significant differences in tumor volume were observed 14 days after the last treatment until the end of the experiment. Shown are average tumor volumes ⫾ sd (n⫽5 for both groups). the measurement of the tumor volume. We did not observe growth inhibition in vitro at or beyond the administered in vivo concentration of JC-7U IgG1, Figure 3. Tumor targeting of JC-7U IgG1. JC-7U IgG1 (A, E) and control antibody b12 IgG1 (B, F ) were injected into the tail vein of nude mice with an established human KS tumor. After 24 h, tumor (A, B) and lung tissue (E, F ) were removed and processed for the detection of bound human IgG using goat anti-human IgG secondary antibodies. C, D) Mouse endothelial cells that infiltrate the human KS tumor were detected with a rat anti-mouse CD31 mAb, followed by rabbit anti-rat IgG secondary antibodies. Note that the sites of CD31 expression and JC-7U IgG1 tumor penetration overlap. Scale bar ⫽ 50 m for panels A–F. suggesting that antibody-dependent cellular cytotoxicity (ADCC) is the main mechanism of tumor growth inhibition of human JC-7U IgG1 in our nude mouse model of KS. Like its parental antibody, mouse mAb LM609, JC-7U binds to human but not mouse integrin ␣v3. In the SLK xenograft, JC-7U recognizes human integrin ␣v3 expressed by tumor cells but not mouse integrin ␣v3 expressed by angiogenic endothelial cells. Thus, the reduction of KS tumor growth in the presence of JC-7U is not an antiangiogenic effect but can be attributed to a specific, direct response against the tumor, which allows us to dissect out the antitumor effect of this antibody therapy. 4. Antibody JC-7U partially blocks the binding of HIV-1 Tat protein to integrin ␣v3 Figure 2. Schematic diagram illustrating the in vitro evolution of antibody JC-7U. Because of its high affinity and its high degree of humanization, JC-7U IgG1 is an excellent drug candidate for therapeutic applications that involve integrin ␣v3 as the molecular target. Of particular interest is therapy for KS, breast cancer, melanoma, and other cancers in which integrin ␣v3 is expressed on both angiogenic endothelial cells and tumor cells, a dual antiangiogenic and antitumor strike with a single drug. Mouse sequences are shown in yellow, human sequences in blue, optimized sequences in red. INTEGRIN ␣V3 TARGETED THERAPY OF KAPOSI’S SARCOMA Based on published evidence that the HIV-1 Tat protein can act as a growth factor for KS by binding to RGD receptors, including integrin ␣v3, that are expressed on the surface of KS cells, we investigated whether JC-7U can interfere with this interaction. We established a binding assay in which a mixture of human integrin ␣v3 and synthetic HIV-1 Tat-biotin was incubated in solution in the presence or absence of JC-7U Fab. Complexes of integrin ␣v3 and Tat-biotin were then captured with immobilized streptavidin and were detected by using the anti-3 mouse mAb AP-3, which does not interfere with JC-7U binding, followed by goat anti-mouse secondary antibodies. Background signals were determined by running the same assay in the absence of Tat-biotin. In the absence of JC-7U, a strong signal above background was observed, which revealed 2001 an interaction of integrin ␣v3 and Tat-biotin. In the presence of a threefold molar excess of JC-7U Fab over Tat-biotin or at equimolar concentrations, the signal was reduced to ⬃45%. The same reduction was observed in the presence of JC-7U IgG1. Thus, JC-7U partially blocks the binding of HIV-1 Tat protein to human integrin ␣v3. CONCLUSIONS AND SIGNIFICANCE In view of the fact that KS tumors are of endothelial origin and characterized by intense angiogenesis, it is logical that many of the recently developed antiangiogenic drugs are being studied for the treatment of KS. With the emerging picture of KS pathogenesis and angiogenesis, molecular targets are being revealed that may provide effective targeting of drugs to both the KS vasculature and the KS tumor itself. We hypothesized that integrin ␣v3 might fulfill these criteria. Integrin ␣v3 is up-regulated on endothelial cells in response to angiogenic growth factors and has been established as a target for antiangiogenic therapy. Soluble antagonists of integrin ␣v3 such as mAb LM609, RGD peptides, or peptidomimetics, initiate endothelial cell apoptosis and thereby inhibit angiogenesis. In addition to its expression on the surface of angiogenic endothelial cells, integrin ␣v3 is expressed on the surface of tumor cells in a variety of cancers. In 2002 Vol. 16 December 2002 melanoma and breast cancer, for example, tumor progression and metastasis correlate with integrin ␣v3 expression. Thus, independent of its role in tumor angiogenesis, integrin ␣v3 is likely to be functionally implicated in the pathogenesis of a variety of cancers. This is particularly apparent for KS, where a central position of integrin ␣v3 within angiogenesis and KS pathogenesis has emerged, suggesting that integrin ␣v3 may become a prime molecular target for KS therapy. The hypervascular morphology of KS makes it highly accessible to therapeutic macromolecules, such as antibodies, whose general applicability for cancer therapy is limited by poor tumor penetration. With JC-7U, we developed a highly advanced drug candidate for therapeutic applications that involve integrin ␣v3 as the molecular target. JC-7U IgG1 was found to selectively target tumors in a mouse model of KS and inhibit tumor growth at a therapeutically relevant dose. We anticipate that the tumor growth-inhibiting activity of JC-7U will be even more effective in KS patients as compared with mouse models of KS. First, in the presence of human angiogenic endothelial cells, JC-7U is expected to exert an antiangiogenic effect based on initiation of endothelial cell apoptosis and ADCC in addition to an antitumor effect. Second, in AIDS-associated KS, JC-7U will partially block the binding of HIV-1 Tat protein to human integrin ␣v3 and, thus, potentially diminish its tumor growth-promoting activity. The FASEB Journal RADER ET AL. Table 1 Binding parameters of Fab directed to human integrin αvβ3 Fab Mouse LM609 b kon/104 (M−1 s−1) 8.6 a koff/10−4 (s−1) Kd (nM) 8.6 10 Humanized LM609-7 23 (18 b) 5.9 (5.4 b) 2.6 (3.0 b) JC-7U 24 0.35 0.15 a Binding kinetics were determined by using surface plasmon resonance. Human integrin αvβ3 was immobilized on the sensor chip. Kd was calculated from koff/kon. All Fab fragments were produced by using E. coli expression. bData from Rader et al. (10). Fig. 1 Figure 1. Affinity maturation of humanized LM609 by CDR walking. CDR walking involves the sequential or parallel optimization of CDRs by focused mutagenesis and subsequent selection by phage display. Shown is a combination of sequential and parallel CDR optimization used for the affinity maturation of humanized LM609. Randomized sequences are hatched; optimized sequences are black. CDR3 of the light chain was optimized first, followed by a parallel optimization of CDR1 and CDR3 of the heavy chain. Optimized CDR1 and CDR3 of the heavy chain were combined in the last step. This focused mutagenesis strategy involved the randomization of four hypervariable codons in each targeted CDR encoding sequence. LC, light chain; HC, heavy chain. Fig. 2 Figure 2. Randomized and selected sequences of LCDR3, HCDR1, and HCDR3 for one of the selected clones, designated JC-7U. The top and middle panels show the four codons of each CDR that were subjected to randomization by NNK (nnk in the figure) doping (n = a, c, g, or t; k = g or t). The bottom panel shows the selected NNK codons and their corresponding amino acids (single-letter amino acid code). As shown in bold type, 10 of 12 codons but only 5 of 12 amino acids differed between the original and the selected sequence. No amino acid mutations in nonrandomized CDR and framework regions were found. Fig. 3 Figure 3. Overlayed Biacore 1000 sensorgrams obtained from the binding of humanized LM609-7 Fab and JC-7U Fab to immobilized human integrin αvβ3. For association, Fab fragments were injected at 80 nM between t = 125 s and t = 370 s, with a flow rate of 10 µl/min. For dissociation, the flow rate was increased to 50 µl/min. RU, resonance units. Fig. 4 Figure 4. Flow cytometry histograms demonstrating the binding of JC-7U Fab and humanized LM609-7 Fab to the human KS cell line SLK. For indirect immunofluorescence staining, cells were incubated with 2 µg/ml JC-7U Fab (bold line) or 2 µg/ml humanized LM609-7 (fine line) or buffer alone (dotted line), followed by FITC-conjugated secondary antibodies. The y axis gives the number of events in linear scale, the x axis the fluorescence intensity in logarithmic scale. Note that JC-7U Fab revealed a slightly stronger binding than did humanized LM609-7. This difference was confirmed at Fab concentrations of 200 and 20 ng/ml and was also seen when human melanoma M21 cells were used instead of SLK cells. Fig. 5 Figure 5. Vector PIGG designed for antibody expression in mammalian cells. The 9-kb vector contained both heavy- and light-chain expression cassettes driven by a bidirectional CMV promoter construct. Cloning sites for the variable domain encoding sequences were designed for compatibility with pComb3 phage display vectors. By using PIGG, Fab selected by phage display can be readily converted into IgG. Fig. 6 Figure 6. JC-7U IgG1 did not inhibit the growth of human KS cell line SLK in vitro. Shown is a proliferation assay based on [3H]thymidine incorporation during DNA synthesis. None of the tested concentrations in a range of 0.1 nM to 1 µM revealed a significant difference between JC-7U IgG1 and the control antibody b12 IgG1. Background signals were determined by running the same assay in the absence of antibody and were defined as 0% inhibition (n = 3). Fig. 7 Figure 7. Tumor targeting of JC-7U IgG1. JC-7U IgG1 (A and E) and control antibody b12 IgG1 (B and F) were injected into the tail vein of nude mice with an established human KS tumor. After 24 h, tumor (A and B) and lung tissue (E and F) were removed and processed for the detection of bound human IgG by using goat anti-human IgG secondary antibodies. C, D) Mouse endothelial cells that infiltrate the human KS tumor were detected with a rat anti-mouse CD31 mAb followed by rabbit anti-rat IgG secondary antibodies. Note that the sites of CD31 expression and JC-7U IgG1 tumor penetration overlap. Scale bar = 50 µm for A-F. Fig. 8 Figure 8. Tumor growth inhibition in a nude mouse model of KS in the presence of JC-7U IgG1. JC-7U IgG1 (black circles) or control antibody b12 IgG1 (white circles) was given in five doses of 5 mg/kg every third day starting on day 1 after tumor induction. Statistically significant differences in tumor volume were observed 14 days after the last treatment until the end of the experiment. Shown are average tumor volumes ± SD (n = 5 for both groups). Fig. 9 Figure 9. Tumor growth inhibition in a nude mouse model of KS in the presence of JC-7U IgG1. Shown are tumor weights after treatment with JC-7U IgG1, b12 IgG1, or PBS. Tumors were removed surgically on day 54 after tumor initiation and weighed. Tumor weights are averages ± SD (n = 5 for JC-7U and b12; n = 4 for PBS [1 animal in this PBS group died before day 54]). The differences between treatment with JC-7U IgG1 and b12 IgG1 (P=0.0019) or PBS (P=0.0004) were statistically significant. Fig. 10 Figure 10. JC-7U Fab partially blocked the binding of HIV-1 Tat protein to human integrin αvβ3. A mixture of human integrin αvβ3 and synthetic HIV-1 Tat-biotin was incubated in solution in the presence or absence of JC-7U Fab. Complexes of integrin αvβ3 and Tat-biotin were then captured with immobilized streptavidin and were detected by using an independent mouse mAb directed to integrin subunit β3 followed by goat anti-mouse secondary antibodies. Background signals were determined by running the same assay in the absence of Tat-biotin. Background-depleted signals are shown (n = 3; P=0.0058).