Epitope-Specific Mechanisms of IGF1R Inhibition by Ganitumab

Background Therapeutic antibodies targeting the IGF1R have shown diverse efficacy and safety signals in oncology clinical trials. The success of these agents as future human therapeutics depends on understanding the specific mechanisms by which these antibodies target IGF1R signaling. Methodology/Principal Findings A panel of well-characterized assays was used to investigate the mechanisms by which ganitumab, a fully human anti-IGF1R antibody undergoing clinical testing, inhibits IGF1R activity. Epitope mapping using IGF1R subdomains localized the ganitumab binding site to the L2 domain. Binding of ganitumab inhibited the high-affinity interaction of IGF-1 and IGF-2 required to activate IGF1R in cells engineered for IGF1R hypersensitivity and in human cancer cell lines, resulting in complete blockade of ligand-induced cellular proliferation. Inhibition of IGF1R activity by ganitumab did not depend on endosomal sequestration, since efficient ligand blockade was obtained without evidence of receptor internalization and degradation. Clinically relevant concentrations of ganitumab also inhibited the activation of hybrid receptors by IGF-1 and IGF-2. Ganitumab was not an agonist of homodimeric IGF1R or hybrid receptors in MCF-7 and COLO 205 cells, but low-level IGF1R activation was detected in cells engineered for IGF1R hypersensitivity. This activation seems biologically irrelevant since ganitumab completely inhibited ligand-driven proliferation. The in vivo efficacy profile of ganitumab was equivalent or better than CR and FnIII-1 domain-specific antibodies, alone or in combination with irinotecan. CR domain-specific antibodies only blocked IGF-1 binding to IGF1R but were more potent than ganitumab at inducing homodimer and hybrid receptor downregulation in vitro, however this difference was less obvious in vivo. No inhibition of hybrid receptors was observed with the FnIII-1 domain antibodies, which were relatively strong homodimer and hybrid agonists. Conclusions/Significance The safety and efficacy profile of ganitumab and other anti-IGF1R antibodies may be explained by the distinct molecular mechanisms by which they inhibit receptor signaling.


Introduction
The type I insulin like growth factor receptor (IGF1R) is a heterotetrameric complex consisting of two disulfide-linked achains that bind IGF-1 and IGF-2 and two b-chains that include a transmembrane and a tyrosine kinase domain [1]. High-affinity binding of IGF-1 or IGF-2 to the IGF1R extracellular domain (ECD) induces a conformational change, which leads to autophosphorylation of intracellular tyrosine residues [2]. Four protein structural motifs in the IGF1R ECD have been shown to be involved in ligand binding and selectivity: L1, L2, CR, and FnIII-1 [1]. Activation of IGF1R leads to stimulation of the PI3K/ Akt and other signaling pathways resulting in pro-survival and proliferative signals [3].
IGF1R is closely related to the insulin receptor (INSR), there being 35% to 70% identity between their ECDs, depending on the structural motif [1]. The selectivity of IGF1R for IGF-1 and of INSR for insulin governs the specificity of each receptor in regulating metabolism and growth in mammals [4,5,6]. IGF-2 activates both IGF1R and the mammalian INSR-A variant, effectively integrating signaling from both receptors [7]. Integration of IGF1R and INSR signaling can also occur through hybrid receptors, which are preferentially activated by IGF-1 or IGF-2 [7]. The role of IGF1R/INSR hybrid signaling in normal physiology and disease is an active area of investigation.
The data implicating IGF-1 and IGF-2 in cancer risk and tumor progression have positioned IGF1R as a prime oncolgy therapeutic target, anticipated to have activity against a number of human malignancies [8]. Multiple anti-IGF1R antibodies have entered clinical development during the last decade, and the safety and efficacy of these agents vary for reasons that are not clear [9,10,11]. There is, therefore, interest in understanding how epitope-dependent mechanisms can contribute to clinical activity. Here, we have compared the physical and biological mechanisms of IGF1R inhibition by ganitumab, an investigational anti-IGF1R antibody currently being evaluated in clinical trials, with a number of anti-IGF1R antibodies that target distinct IGF1R epitopes. Our results identify epitopedependent mechanisms of receptor inhibition that may be important in distinguishing the clinical effects of ganitumab and other investigational anti-IGF1R antibodies.

Animals
Female CD1 nude mice (Charles River Laboratories, Wilmington, MA) or athymic nude mice (Harlan Laboratories, Inc., Placentia, CA) were housed in sterile cages, five per cage. The laboratory housing the cages met all Association for Assessment and Accreditation of Laboratory Animal Care International specifications. All experimental procedures were approved by the Institutional Animal Care and Use Committee of Amgen Inc. and performed in accordance with Amgen's Institutional Animal Care and Use Committee and the United States Department of Agriculture rules and regulations. All efforts were made to minimize animal suffering.

Antibodies
The anti-IGF1R antibodies evaluated in our study are shown in Table 1. The fully human anti-IGF1R L2 domain monoclonal antibodies, ganitumab, L2-A, L2-B, and L2-C, were isolated using a Target Quest TM human phage-displayed antibody library (Dyax Corp., Cambridge, MA). The fully human anti-IGF1R FnIII-1 domain antibodies, F1-A, F1-B, and F1-C, were isolated using the BioSite TM antibody platform (Biosite, San Diego, CA). Human (h)IGF1R(ECD)-mFc was the capture agent, and all antibodies that cross reacted with hINSR(ECD)-mFc were excluded. Murine anti-IGF1R monoclonal antibodies were obtained from a variety of sources: aIR3 from EMD Chemicals (Gibbstown, NJ), Mab 391 from R&D Systems (Minneapolis, MN), 1H7 and 26-3 from Santa Cruz Biotechnology (Santa Cruz, CA), and 24-57 from Thermo Scientific (Waltham, MA). The anti-INSR antibody 47-9 was obtained from R&D systems (Minneapolis, MN) and was used as a positive control in INSR binding experiments.
Fusion proteins (50-100 mg per 3.5610 5 biotin beads) were mixed with 1 mg anti-IGF1R domain-specific monoclonal antibody in 1 mL of PBS/0.5% BSA and incubated for 1 hour at room temperature. The antibody/fusion protein/biotin bead complexes were then incubated with 0.5 mg/mL FITC-labeled mouse F(ab')2 anti-human or anti mouse secondary antibodies (Jackson ImmunoResearch Laboratories Inc., West Grove, PA), The IGF1R binding domain recognized by each antibody was determined using avidin-fusion proteins. Antibody binding affinity (K D ) was determined using hIGF1R(ECD) and the Biacore TM equilibrium method. Values for ganitumab binding affinity have been reported previously [12]. The number of independent determinations is shown in parentheses. doi:10.1371/journal.pone.0055135.t001 FITC-labeled anti-avidin antibody, and 0.5 mg/mL PE-labeled goat anti-human or anti-mouse secondary in 1 mL BPBS for 1 to 2 hours. FITC and/or PE were detected with a FACScan (Beckton, Dickson and Company, Franklin Lakes, NJ).
For competition experiments, biotin beads bound to fusion proteins containing the complete ECD of IGF1R were preincubated with 10-50 mg/mL unlabeled competitor antibody followed by incubation with FITC-labeled domain-specific antibody.
The effects of each anti-IGF1R antibody on ligand binding were further investigated by generating binding curves with increasing concentrations of Ru-IGF-1 or Ru-IGF-2 in the presence of excess (1 mM) antibody. Each binding assay used 75 ng hIGF1R(ECD)-mFc preloaded on anti-mouse IgG-coated MA 6000 96-well plates (Mesoscale Discovery, Gaithersburg, MD) as recommended by the manufacturer. An MSD6000 analyzer (Mesoscale Discovery) was used to detect bound Ru-labeled ligand.
The Biocore equilibrium method was used to determine antibody binding affinity as reported previously [12].

Proliferation Assays
Serum-starved (24 hours) Balb/C 3T3 hIGF1R or 32D hIGF1R/IRS-1 cells (in triplicate), were pretreated with a range of ganitumab or control hIgG1 concentrations (1 pM to 1 mM) for 1 hour, and incubated in the presence or absence of human IGF-1 (2 nM) or IGF-2 (8 nM) for 30 minutes prior to the addition of 3 H-thymidine (1 mCi; Perkin Elmer, Waltham, MA). Incorporation of 3 H-thymidine was measured 24 hours later. To obtain growth curves, COLO 205 and MCF-7 cells were cultured in fresh RPMI with 10% FBS in a 96-well format (5 replicates) in the presence of control hIgG1 or anti-IGF1R antibody added at the time of cell seeding. The increase in cell confluence was monitored with an IncuCyte TM instrument (Essen Instruments, Lancaster, PA).

IGF1R and INSR Phosphorylation
A rapid challenge assay, whereby ligand and antibodies were added simultaneously to cells for 5 minutes, was used to assess the phosphorylation status of IGF1R and INSR. COLO 205, MCF-7, and Balb/C 3T3 hIGF1R cells were starved overnight before antibody treatments in the presence of either 2 nM IGF-1 or 8 nM IGF-2. 32D hIGF1R/IRS-1 cells were starved in DMEM supplemented with 10 ng/mL IL-3 before antibody treatment with either 4 nM IGF-1 or 16 nM IGF-2. Human IGF1R was immunoprecipitated from cell lysates by incubating with 4.5 mg antibody F1-B for 2 hours, and immune complexes were captured with Protein-G agarose beads (Thermo Scientific, Rockford, MD). IGF1R was separated by 10% SDS PAGE and electroblotted onto PDVF membranes. Total and phosphorylated IGF1R were detected with the C-20 (Santa Cruz Biotechnology Inc., Santa Cruz, CA) and pY1158 (Life Technologies Inc, Grand Island, NY) antibodies, respectively, followed by an anti-goat HRP conjugate.
Mesoscale TM multiplex assays were used to quantify total and phosphorylated IGF1R, INSR, and Akt as described by the manufacturer. Analysis of IGF1R and INSR from cells treated with CR domain antibodies was performed using IgG1 high-bind plates (Mesoscale Discovery); FnIII-B was used as the capture antibody in insulin multiplex assay plates. It should be noted that hybrid IGF1R/INSR phosphorylation signals were recovered in both the INSR and IGF1R channels in the MSD IGF1R multiplex assay.

IGF1R and INSR Levels and Receptor Internalization
Cell-surface receptor levels were determined by quantitative flow cytometry as described previously [12]. Total IGF1R or INSR levels were analyzed by western blotting using the C-20 and C-19 antibodies (Santa Cruz Biotechnology), respectively.
An anti-b-tubulin antibody, D-10 (Santa Cruz Biotechnology), was included as a loading control. For in vivo studies, tumors were snap frozen in liquid nitrogen, homogenized in RIPA buffer, and cleared by centrifugation before analysis by western blotting. Total INSR was also detected using an insulin signaling multiplex assay (Mesoscale Discovery).

In Vivo Efficacy Studies
CD1 nu/nu mice or female athymic nude mice (4-6 weeks old) were used in all experiments. 32D hIGF1R/IRS-1 cells were selected for in vivo growth by passaging the cells subcutaneously twice. Mice were injected subcutaneously with 5610 6 cells (32D IGF1R/IRS-1 or COLO 205) in Matrigel. Animals with tumors of approximately 200 mm 3 were randomly assigned to treatment groups (10 per/group). The animals were treated intraperitoneally (IP) with anti-IGF1R antibodies or control antibodies at the indicated doses, twice weekly for the duration of the experiment. Tumor volumes and body weights were monitored twice weekly using calipers and an analytical scale, respectively. For the analysis of receptor levels and downstream signaling, the tumors were snap frozen in liquid nitrogen and processed as described previously [12].

Statistical Analysis
Data obtained from the ligand-binding and cell-based assays were analyzed with GraphPad (Prizm TM , La Jolla, CA). Differences in tumor volume were analyzed with repeated measures analysis of variance (RMANOVA) and Scheffe's Post Hoc analysis using StatView (5.0.1, SAS Institute Inc., Cary, NC).

Inhibition of Ligand Binding
The effect of each anti-IGF1R antibody on IGF-1 and IGF-2 binding to IGF1R was determined with purified hIGF1R(ECD)-mFc. A logarithmic-scale on the y-axis was used to highlight the differences in ligand binding obtained in the antibody survey. Antibody titrations against ,0.25 nM Ru-labeled ligand showed that ganitumab, L2-(A-C), and F1-(A-C) inhibited IGF-1 and IGF-2 binding (Fig. 1A, B; Table 1). 1H7 and 24-57 also inhibited IGF-1 and IGF-2 binding, as reported previously [13,14]. aIR3 and Mab 391 inhibited IGF-1 binding, but not IGF-2 binding; Ab 26-3 increased IGF-1 and IGF-2 binding. Antibody competition, when observed, reached a plateau of about 20% of ligand binding relative to control. The extent of competition obtained with unlabeled IGF-1 was significantly greater (,2% of control). This difference between ligand and antibody saturation was not observed with L2 and FnIII-1 domain antibodies and IGF-2.
The effects of each anti-IGF1R antibody on ligand binding were further investigated by generating binding curves with Ru-IGF-1 or Ru-IGF-2 in the presence of excess antibody. All of the inhibitory IGF1R antibodies that we examined blocked the high-affinity binding observed between IGF1R and IGF-1 (at ,1 nM). However, the low-affinity interaction of IGF-1 with IGF1R that was observed with L2 and FnIII-1 domain antibodies was at least 10-fold greater than observed with CR domain antibodies. The differences obtained in the antibody survey were more easily visualized in a double logarithmic plot of the binding data (Fig. 1C, D). The results obtained with IGF-2 were similar to IGF-1 except that CR domain antibodies did not block binding of IGF-2.

Crossreactivity With INSR
None of the IGF1R antibodies used in our study cross reacted with purified hINSR(ECD)-mFc or prevented binding of insulin to hINSR(ECD)-mFc (data not shown). Activation of INSR by IGF-1, IGF-2, and insulin was not inhibited by up to 1 mM ganitumab, Mab 391, or F1-B (Fig. S2).
control antibody at each time point. (D). Mice with established (200-300 mm 3 ) subcutaneous tumors were treated with ganitumab, Mab 391, or F1-B (300 mg/dose; IP). At the indicated time points, 3 animals were sacrificed and IGF1R levels were determined. % control is the signal obtained for an individual animal divided by the mean for the control antibody multiplied by 100 for each treatment group. Total INSR level was determined in the same cell (E) and tumor (F) extracts used for the long-term analysis of IGF1R (C, D). doi:10.1371/journal.pone.0055135.g003     Table S1. Across cell lines, activation of IGF1R by IGF-1 occurred within a narrow range of concentrations (EC 50 : 0.2 to 2.7 nM). A similar pattern was observed with IGF-2, but the concentration range was shifted 10-fold (EC 50 :2.1 to 33 nM). The effects of IGF-1 and IGF-2 on INSR were only assessed in the COLO 205 and MCF-7 cell lines. For INSR, the EC 50 for IGF-1 was 20-fold lower than the EC 50 for IGF-2. The EC 50 for INS in COLO 205 and MCF-7 cells was 2 and 3.5 times higher than the EC 50 for INS in CHO cells engineered to overexpress hINSR.
Ganitumab and Mab 391 treatment inhibited the proliferation of COLO 205 and MCF-7 cells in 10% FBS growth media 1.2-to 1.8-fold. In contrast, F1-B increased the growth of COLO 205 cells and had no effect on the growth of MCF-7 cells (Fig. S3).

Internalization and Degradation of IGF1R and INSR
Treatment with CR and FnIII-1 domain antibodies was consistently associated with a $70% reduction in total IGF1R (Fig. 3A). Internalization occurred within 15 minutes of antibody application as determined by confocal microscopy with FITC-labeled antibodies (data not shown), and receptor degradation was detected after 4 hours of treatment (data not shown). The dose dependence of IGF1R downregulation was examined for selected antibodies in COLO 205 cells (Fig. 3B). Mab 391 and F1-B exhibited the highest potency (IC 50 ,0.1 nM) followed by L2-B (IC 50 0.1-1 nM). Ganitumab and L2-C downregulation of IGF1R only occurred at antibody concentrations .100 nM (Fig. 3B). Extending ganitumab treatment to 14 days did not reduce IGF1R levels, rather levels tended to increase (Fig. 3C). Similar data were obtained in the MCF-7 cell line (Fig. S4A). The ability of ganitumab, Mab 391, and F1-B to downregulate IGF1R expression was also evaluated in mice bearing established COLO 205 tumors treated twice weekly with 300 mg antibody for up to 14 days (Fig. 3D). In vivo, ganitumab, Mab 391, and F1-B reduced total IGF1R by 50% to 60%. Similar results were obtained with MCF-7 cells, which express a higher ratio of IGF1R to INSR than COLO 205 cells (Fig. S4B).

IGF1R and INSR Activation
To distinguish antibody effects on IGF1R homodimer and hybrid receptors, inhibition of ligand-induced activation of IGF1R and INSR was characterized in four cell lines that contained different levels of IGF1R and INSR (Table S1). Each anti-IGF1R antibody was an effective inhibitor of IGF-1-induced activation of IGF1R regardless of the epitope, although differences in IC 50 were observed ( Table 3). The L2 and FnIII-1 domain antibodies were more effective inhibitors of IGF-2-induced activation of IGF1R than the CR domain antibodies (Table 3  To further characterize the observed inhibitory effects, a rapid ligand challenge assay was developed to minimize receptor inhibition by internalization. Balb/C 3T3 hIGF1R cells were treated with increasing IGF-1 or IGF-2 in the presence of 1 mM ganitumab, Mab 391, or F1-B, and the incubation time was reduced to 5 minutes. Under these conditions, ganitumab inhibited IGF-1-and IGF-2-induced activation of IGF1R at all ligand concentrations (Fig. 4D). Mab 391 and F1-B inhibited IGF-1-induced, but not IGF-2-induced activation of IGF1R (Fig. 4D).

Agonistic Potential of Anti-IGF1R Domain Antibodies
The agonistic potential of anti-IGF1R domain antibodies against IGF1R, INSR, and hybrid receptors in the absence of ligands was investigated (Fig. 5). The FnIII-1 domain antibodies Figure 6. Inhibition of the growth of 32D hIGF1R/IRS-1 tumor xenografts. (A). TGI by ganitumab, Mab 391, or 1H7. Mice bearing 32D hIGF1R/IRS-1 xenografts (,200 mm 3 ) were randomly assigned into treatment groups (n = 10) and treated with antibody (300 mg, twice per week, IP) for the duration of the experiment. Tumor volumes and body weights were measured twice per week using calipers and an analytical scale, respectively. Data are presented as mean tumor volume 6 standard error of the mean. (B). Relationship between ganitumab serum concentration and TGI. Ganitumab serum concentration (mg/mL) was determined 2 hours after the final dose and plotted as the mean 6 standard deviation. (C-F). Inhibition of IGF1R signaling in vivo. Mice bearing 32D hIGF1R/IRS-1 xenografts (,400 mm 3 ) were randomly assigned into treatment groups (n = 6) and treated with ganitumab or IgG1 control for 6 hours. Three animals from each group were then challenged with 5 mg IGF-1 or PBS IV for 15 minutes. Total and phosphorylated IGF1R, IRS-1, Akt, and p70S6K in tumor extracts (100 mg) were determined with MSD multiplex assays. No significant changes in GSK3b were observed (data not shown). *p,.0003. doi:10.1371/journal.pone.0055135.g006 (F1-B, 1H7) exhibited a consistently high level of IGF1R agonism (4-to12-fold times over baseline) in each cell line examined (Fig. 5A-D). Ganitumab and L2-C did not significantly stimulate IGF1R phosphorylation in COLO 205 and MCF-7 cells (Fig. 5A,  B). However, partial agonism with L2 antibodies was observed in Balb/C 3T3 hIGF1R cells (4-fold) and 32D hIGF1R/IRS-1 cells (4-to 12-fold) (Fig. 5C-D). IGF1R stimulation (3-fold) was obtained with aIR3 in COLO 205 cells (Fig. 5A), and both of the CR domain antibodies exhibited partial agonism in 32D hIGF1R/ IRS-1 cells (Fig. 5D). In general, IGF1R agonism was maximal at an antibody concentration of approximately 10 nM.

Antitumor Activity of Anti-IGF1R Domain Antibodies
The antitumor activity of the domain-specific antibodies was determined using 32D hIGF1R/IRS-1 subcutaneous xenografts (Table 4). Potent tumor growth inhibition (TGI) ranging from 75% TGI to tumor regression was achieved with L2, CR, and FnIII-1 domain antibodies in this model. Ganitumab and L2-C treatment led to tumor regressions (43% and 52%) at the maximum dose tested (300 mg, IP, twice weekly; Table 4, Fig. 6A). Using a range of ganitumab doses (3 to 300 mg), the in vivo ED 50 for 32D hIGF1R/IRS-1 TGI was determined to be 1.5 mg/mL (Fig. 6B). Total or almost total TGI was achieved with the CR-domain antibodies Mab 391 and aIR3 and the FnIII-1 domain antibodies F1-A and F1-B (Table 4, Fig. 6A).
To determine the effect of ganitumab on IGF1R pathway signaling in vivo, established 32D hIGF1R/IRS-1 tumors that were pretreated for 6 hours with ganitumab or control antibody were challenged with IGF-1 for 15 minutes. Pretreatment with ganitumab led to almost complete inhibition of IGF-1-induced activation of IGF1R, IRS-1, Akt, and p70S6K (Fig. 6C-F). Ganitumab treatment without IGF-1 stimulation did not show any evidence of IGF1R agonism (Fig. 6C). Mechanism-of-action studies on ganitumab-treated, established 32D hIGF1R/IRS-1 tumors showed complete inhibition of Ki67 labeling (data not shown). This observation is consistent with the complete inhibition of 3 H-thymidine incorporation demonstrated in vitro ( Fig. 2A).

Antitumor Activity of Anti-IGF1R Domain Antibodies in Combination with Irinotecan
COLO 205 xenografts were used to evaluate the effect of combining domain-specific IGF1R antibodies with irinotecan. Ganitumab (300 mg, IP, twice weekly) and irinotecan (35 mg/kg, IV, once per week) treatment significantly inhibited tumor growth by approximately 75% as single agents. Statistically significantly better TGI (p,0.0003 vs either agent) resulting in regressions was achieved by the combination of ganitumab and irinotecan (Fig. 7A). Similar results were observed with L2-B (Fig. 7B) and Mab 391 (Fig. 7C). The addition of F1-B to irinotecan did not lead to statistically significantly better efficacy than irinotecan alone (Fig. 7D); similar results were obtained with the other FnIII-1 domain antibodies in combination with irinotecan (data not shown). The level of total IGF1R in COLO 205 tumors harvested at the end of treatment was consistently reduced ($70%) regardless of the IGF1R domain targeted (data not shown).

Discussion
Safety and efficacy data emerging from phase II clinical studies have highlighted the need to better elucidate the individual characteristics of the multiple anti-IGF1R antibodies currently in clinical development. In this study, we used a panel of wellcharacterized in vitro and in vivo assays to investigate the mechanisms by which ganitumab inhibits IGF1R. We observed notable differences among domain specific anti-IGF1R antibodies with regards to their ability to 1) inhibit ligand binding and ligandinduced activation of IGF1R and INSR, 2) downregulate IGF1R and INSR expression, 3) inhibit cell proliferation and tumor growth, and 4) activate IGF1R and INSR in the absence of ligand.
Previous studies have shown that IGF1R antibodies can vary significantly in their ability to block ligand binding [14,15]. Ganitumab and other L2 domain antibodies appear to be the most efficient at inhibiting both IGF-1 and IGF-2 from binding to IGF1R. This is an important therapeutic criterion as all human cancers are exposed to IGF-1 and IGF-2 due to hepatic endocrine synthesis, tumor stroma expression, and autocrine activation [16]. In addition, IGF-2 is a well-described transforming growth factor whose elevated expression as a result of loss of imprinting can be found in a considerable percentage of human cancers [17]. Our data on inhibition of ligand binding appears to have revealed a difference in ligand binding affinity to ligand-binding sites 1 and 2 on IGF1R. We used the De Meyts model for negative cooperative ligand binding or 'alternative bivalent crosslinking' [2,18] to interpret the results of our ligand-binding assays. The model defines a pair of ligand binding surfaces, site 1 and site 2, within each IGF1R a-chain that are held in an antiparallel arrangement within the receptor dimer. High-affinity ligand binding occurs when the a-chains are crosslinked by the occupation of one subsite pair (site 1 of a-chain 1 and site 2 of a-chain 2) with a bivalent IGF-1 or IGF-2 monomer. Antibody competition and ligand titration curves both indicate that ganitumab inhibits the highaffinity interaction (,0.2 nM K D ) of IGF-1 and IGF-2 associated with negative-cooperative IGF1R ligand binding [18]. However, the ligand titrations also revealed that preloading with excess (1 mM) ganitumab permits IGF-1 and IGF-2 binding at a reduced affinity site (1-100 nM) that is within the range described for negative cooperative ligand binding [2]. The binding data are consistent with the idea that ganitumab holds IGF1R in a configuration that prevents alternative bivalent crosslinking, but that is nonetheless open to the non-cooperative, low-affinity binding of multiple ligand molecules. This low-affinity binding in the ganitumab-bound state is functionally irrelevant since IGF1R activation was not detected with IGF-1 and IGF-2 (at concentrations #100 nM) after antibody preloading in the rapid challenge assay, designed to minimize receptor inhibition by internalization. The surfaces that remain accessible to ligand binding in the ganitumab-bound receptor remain to be determined, but it is clear that these include contacts shared by IGF-1 and IGF-2.
IGF1R/INSR hybrids have been identified in multiple human cancers. Their expression and activity may be, at least in some tumor types, even more relevant than that of IGF1R homodimers [7,19]. Therefore, it is critical to ensure that hybrid receptor activity is also inhibited with IGF1R therapeutic antibodies at clinically meaningful concentrations. Ganitumab and the other L2 domain antibodies were effective inhibitors of IGF-1-and IGF-2induced activation of hybrid receptors, albeit at a greater IC 50 compared with IGF1R homodimers (,50-fold). The difference in IC 50 may indicate that ganitumab binds hybrid receptors with reduced affinity. A previously published study of ganitumab activity in the MiaPaCa2 cell line supports our present data and demonstrate that ganitumab can effectively drive tumor growth inhibition through blockade of hybrid receptor signaling [12]. In addition, published clinical pharmacokinetic data show that concentrations of ganitumab achieved on a 12 or 20 mg/kg (0.2-2 mM) every-2-week dosing schedule are well within the IC 50 range required to inhibit hybrid receptors in vitro [20,21]. These data suggest that hybrid receptor inhibition is highly probable in patients treated with clinically relevant doses of ganitumab.
Antibody-mediated receptor downregulation or receptor activation can significantly alter the activity of monoclonal antibodies against tyrosine kinase receptors [22,23,24]. Confocal analysis showed that IGF1R downregulation observed with some anti-IGF1R antibodies in vitro was due to rapid receptor internalization and enhanced degradation. Ganitumab, however, did not depend on IGF1R endosomal sequestration, since efficient IGF-1 and IGF-2 blockade was obtained in vitro without receptor internalization and degradation. The ability to inhibit IGF1R signaling without receptor downregulation is an important characteristic since it has been shown that internalization and downregulation of target:antibody complexes can be inefficient in some tumor cell types, depends on clathrin/caveolin endocytosis, and is rarely complete [25,26]. Interestingly, effective inhibition of IGF1R activation by ganitumab in vivo was associated with partial downregulation (50% to60%) of total tumor IGF1R. This result suggests that, in human cancer models, total IGF1R downregulation may add to the ganitumab efficacy obtained by ligand inhibition. It is unclear what mechanisms are involved in triggering receptor downregulation in vivo. In vitro crosslinking studies using protein G (Amgen data on file) and prolonged in vitro treatments of cancer cells with ganitumab (Fig. S4) failed to reproduce the downregulation findings observed in vivo. The lowlevel activation induced by ganitumab in the Balb/C 3T3 hIGF1R and 32D hIGF1R/IRS-1 cells appeared to be specific to cells that overexpress IGF1R, since it could not be reproduced in other human cancer cell lines. In addition, it appeared to lack biological significance, since ganitumab completely inhibited proliferative responses in the presence of growth factors.
The in vivo properties of ganitumab were generally consistent with our in vitro data, but some differences were observed. The complete growth inhibition and regression of IGF1R hypersensitive tumors (32D hIGF1R/IRS-1) by ganitumab fully parallels the in vitro observations. The ED 50 obtained for 32D hIGF1R/IRS-1 TGI is consistent with the value obtained with pancreatic tumor models that express both IGF1R homodimer and hybrid receptors [12]. The 5-to10-fold increase in ED 50 relative to the EC 50 obtained for 32D hIGF1R/IRS-1 cells in vitro is expected given tumor resistance to antibody penetration [27]. This ED 50 compares well to the pharmacodynamic results obtained in a phase I clinical study of ganitumab in patients with advanced solid tumors [20]. The alignment between mouse and human data is supported by the fact that ganitumab binds and inhibits murine and human IGF1R with equal affinity and potency [28]. In contrast, the effects of ganitumab against COLO 205 and MCF-7 cells and xenografts were inconsistent and may be directly related to the dependency of these models on IGF-1 and other growth or hormonal factors for in vivo growth. For example, MCF-7 xenografts are dependent on estrogen, which is administered to female mice. The presence of estrogen as a driver might downplay the role of IGF-1 in this model and lead to ganitumab resistance in vivo. Similarly, the presence of multiple growth factors in growth media might supply COLO 205 cells with alternative pro-growth and survival signals. Access to these factors might become more limited in vivo leading to dependence on IGF-1 for growth and survival.
Combining IGF1R inhibitors with cytotoxic agents has been used as an approach to potentiate cellular apoptosis [29]. The results we have obtained with ganitumab and irinotecan using the COLO 205 tumor model are consistent with this therapeutic concept and show that IGF1R inhibition is compatible with cell cycle (S-phase) inhibition.
We have demonstrated that ganitumab and three new L2 domain antibodies that bind a similar epitope are distinct from antibodies that bind other IGF1R structural domains. The CR domain antibodies did not inhibit IGF-2 binding or IGF-2mediated IGF1R activation but were more potent than L2 domain antibodies in inducing IGF1R homodimer and hybrid receptor downregulation in vitro. The efficacy profile of ganitumab (and L2-[A-C]) was generally equivalent to or better than the CR domain controls. However, tumor exposure was biased in favor of the CR domain antibodies due to their selectivity for human IGF1R and superior pharmacokinetic properties associated with their murine IgG1 backbones. The significance of ganitumab's potency against IGF-2 cannot be assessed in our in vivo studies, since circulating IGF-2 is not detected in adult mouse models. The FnIII-1 domain antibodies were also mechanistically distinct; characterized by high agonistic activity, potent receptor downregulation, inefficient inhibition of IGF1R activation by IGF-2, and inefficient inhibition IGF1R/INSR hybrid activation by IGF-1 and IGF-2. It seems likely that receptor downregulation was primarily responsible for the inhibition of IGF-2 proliferation responses observed with the CR and FnIII-1 domain antibodies in cell-based assays. FnIII-1 antibodies were the least effective anti-IGF1R antibodies both in vitro and in vivo. These findings are consistent with the data available for antibody 1H7 [30]. However, a more recent IGF1R antibody screen has identified FnIII-1 domain-specific antibodies with low agonistic activity [31] suggesting potential mechanistic diversity within this domain.
The epitope-specific mechanisms we have described for ganitumab and other domain-specific anti-IGF1R antibodies are consistent with those obtained with murine anti-IGF1R monoclonal antibodies [13,14,15,30,32]. Our study suggests that in addition to the structural and pharmacokinetic distinctions that exist among the IGF1R antibodies in clinical development, mechanistic differences with important biological consequences may exist. Understanding these differences will be required to ensure optimal clinical development of this class of anti-cancer therapeutics.