AMG 479, a fully human anti–insulin-like growth factor receptor type I monoclonal antibody, inhibits the growth and survival of pancreatic carcinoma cells

Pancreatic carcinoma is one of the most deadly forms of cancer. Less than 4% of the 200,000 patients annually diagnosed with pancreatic carcinoma live more than 5 years, due largely to extensive, unresectable, metastatic disease present at the time of diagnosis (1). Gemcitabine, which was approved in 1996 based on a 1-month increase in median survival time, is the standard of care for patients with this disease (2). Recent clinical trials of a number of approved oncology therapeutics combined with gemcitabine have not shown improved clinical benefit against pancreatic carcinoma, underscoring the need for new therapeutic targets to treat patients with this disease (3).

One potential therapeutic target in pancreatic carcinoma is the insulin-like growth factor (IGF) type I receptor (IGF-IR), which regulates cell survival and cell cycle progression via the PI3K/Akt and extracellular signal–regulated kinase pathways through insulin receptor substrate (IRS) and Src homology and collagen adapter proteins (4). IGF-IR seems to be expressed in most human cancers, including pancreatic carcinoma, and deregulation of the IGF-IR signaling axis is common (5, 6). The two IGF-IR ligands, IGF-I and IGF-II, have been implicated in cancer initiation and progression (4). Tumor expression of IGF-I has been observed (7), and high levels of IGF-I in plasma have been correlated with an increased cancer risk. The expression of IGF-II is frequently up-regulated in tumor tissue due to loss of imprinting (8). The majority of circulating IGF-I and IGF-II is produced by hepatocytes, and the local abundance of these growth factors may explain why the liver is a preferred site for pancreatic cancer metastasis (9, 10).

IGF-IR is closely related to the insulin receptor (INSR) and differs from other receptor tyrosine kinases by existing as disulfide-linked heterotetrameric structures composed of two α and two β chains (11). IGF-IR and INSR have distinct roles in regulating growth and metabolism (12–14). However, the recent identification of hybrid IGF-IR/INSR receptors has raised questions about the separation of these two biological axes (13, 15). Although hybrid receptors seem to be active in tumor cells, their role in metabolic growth and survival is poorly understood (16, 17).

Anti–IGF-IR monoclonal antibodies (the first of these being α-IR3; ref. 18), kinase-selective small molecules, and antisense and dominant-negative receptors have shown that inhibition of IGF-IR signaling can generate antineoplastic effects in preclinical tumor models (19, 20). Recent early-phase clinical data suggest that anti–IGF-IR monoclonal antibodies may have a therapeutic benefit in several malignancies (7, 21).

AMG 479 is a fully human monoclonal antibody (IgG₁) isolated against the human IGF-IR. We have investigated the ability of AMG 479 to bind and inhibit the signaling of IGF-IR homodimers and IGF-IR/INSR hybrids in pancreatic carcinoma cell lines. Mechanisms of AMG 479 inhibition defined in vitro were validated in vivo using a pharmacodynamic assay based on stimulation of pancreatic carcinoma xenografts by IGF-I. The biological effects of AMG 479 were examined as a monotherapy and in combination with gemcitabine. Our findings suggest that AMG 479 displays the characteristics of an optimal IGF-IR–targeted clinical agent with the ability to (a) inhibit binding of IGF-I and IGF-II to IGF-IR; (b) block the activation of IGF-IR homodimers and IGF-IR/INSR hybrid receptors, but not INSR homodimers; (c) promote down-regulation of IGF-IR; and (d) inhibit murine models of cancer.

Four- to six-week-old, female, athymic nude mice (Harlan Sprague-Dawley Labs) were used in all experiments. The laboratory housing cages had a 12-h light/dark cycle and met all Association for Assessment and Accreditation of Laboratory Animal Care specifications. All experimental procedures were done in accordance with Institutional Animal Care and Use Committee and U.S. Department of Agriculture regulations. Water and food were supplied ad libitum.

Human pancreatic carcinoma cell lines, BxPC-3 and MiaPaCa2, were purchased from American Type Culture Collection and maintained in RPMI and DMEM, respectively, supplemented with 10% fetal bovine serum. 32D myeloid cells engineered to express human IGF-IR and IRS-1 (32D IGF-IR+IRS-1 cells; ref. 22) were maintained in RPMI supplemented with 10% fetal bovine serum, 10 ng/mL IL-3 (Amgen, Inc.), and 250 μg/mL G418. A human IgG₁ (hIgG₁) raised against streptavidin was used as a nonspecific control antibody. Gemcitabine was obtained from Ely Lilly. IGF-I, IGF-II, and insulin were obtained from Sigma.

IGF-IR and INSR extracellular domains (ECD), IGF-IR(ECD)-mFc and INSR(ECD)-mFc, contained human IGF-IR (amino acid residues 31–935) or human INSR (amino acid residues 28–956) fused to a murine IgG₁ Fc-coding region, following the methods of Bass et al. (23). The fusion proteins were expressed in Chinese hamster ovary cells and purified by recombinant Protein A-Sepharose (Amersham) affinity chromatography. Scintillation proximity assays showed that these soluble receptors retain physiologic ligand selectivity (data not shown).

The Biacore equilibrium method was used to determine the equilibrium-binding constant (K_D) of AMG 479 for IGF-IR or INSR. IGF-IR(ECD)-mFc or INSR(ECD)-mFc was incubated with AMG 479 in PBS, 0.005% P-20, and 0.1 mg/mL bovine serum albumin at room temperature for 4 h. Free receptor was detected using a CM5 surface (Biacore GE Healthcare) coated with AMG 479.

The ability of AMG 479 to inhibit binding of IGF-I, IGF-II, or insulin to either IGF-IR(ECD)-mFc or INSR(ECD)-mFc was assessed with scintillation proximity assays using the homologous ligand as a positive control and rituximab (Biogen Idec) as a negative IgG₁ control. Each assay contained PBS, 0.05% Tween 20 (Mallinckrodt); 0.1% bovine serum albumin, 50 ng IGF-IR(ECD)-mFc or INSR(ECD)-mFc; 500 μg SPA PVT antimouse IgG fluromicrospheres (Amersham); 0.64 nmol/L ¹²⁵I-labeled IGF-I, IGF-II, or insulin (Amersham); and unlabeled AMG 479, rituximab, IGF-I, IGF-II, or insulin (10⁻¹¹–10⁻⁶ mol/L). Binding was assessed after 2-h incubation at room temperature.

BxPC-3 and MiaPaCa2 cells were harvested and incubated with 1 μg phycoerythrin-conjugated antihuman IGF-IR or antihuman INSR monoclonal antibodies (BD Pharmingen) for 1.5 h at 4°C. Mean fluorescence levels were determined by flow cytometry and converted to absolute levels of IGF-IR and INSR using Quantum microbeads (Bangs Laboratories). Levels of IGF-IR were measured in 32D IGF-IR+IRS-1 cells using FITC-labeled AMG 479.

Hybrid receptors were detected by immunoprecipitation and Western blotting. BxPC-3 and MiaPaCa2 cells were harvested and lysed in modified radioimmunoprecipitation assay buffer (1× PBS, 1% Triton X-100, and 0.1% SDS). Homogenates (250–500 μg total Protein) were incubated with 4.5 μg of AMG 479 or anti-INSR antibody, 83-14 (QED Biosciences), overnight at 4°C. Antibody complexes were captured using immobilized Protein G-agarose (Pierce Biotechnology), separated by 10% Tris-glycine SDS-PAGE gel electrophoresis (Invitrogen), then transferred to an Immobilon-P transfer nitrocellulose membrane (Millipore), and probed with anti–IGF-IR C20 (Santa Cruz Biotechnology) or anti-INSR C19 antibodies (Santa Cruz Biotechnology). Receptor β-chain antibody complexes were detected using ECL Plus Western Blotting Detection System (Amersham) and the VersaDoc imaging system and software (Bio-Rad).

BxPC-3 and MiaPaCa2 cells were incubated in serum-free media in low-adherence culture plates with or without 200 nmol/L IGF-I for 24 h. Cells were then pulsed with 1× bromodeoxyuridine (BrdUrd) labeling reagent (Invitrogen) and fixed in 90% ice-cold methanol. For flow cytometry, cells were stained with 0.2 μg/μL BrdUrd-Alexa647 antibody (Invitrogen) and a 1:5 dilution of FITC-caspase-3 antibody (Invitrogen) followed by treatment with PI/RNase staining solution (BD Pharmingen).

The effects of IGF-I, IGF-II, insulin, and AMG 479 on the activation of INSR, IGF-IR, and their downstream effectors (Akt, p70S6 kinase, IRS-1, and GSK3β) were monitored using Meso Scale Discovery (MSD) multiplex assays, in which levels of total and phosphorylated protein were determined separately. BxPC-3 or MiaPaCa2 cells were serum starved for 24 h and incubated with IGF-I, IGF-II (Sigma), or insulin (0–200 nmol/L; Amgen) for 20 min. To determine the effect of AMG 479 on ligand-induced activation, the experiments were repeated with fixed concentrations of ligands plus a range of AMG 479 concentrations (0–1 μmol/L).

32D IGF-IR+IRS-1 cells were seeded in 96-well microtitre plates at 3 × 10⁴ per well in 100 μL RPMI containing 5% fetal bovine serum. AMG 479 (0–1 μmol/L) and gemcitabine (0–1 μmol/L) were added to the cells in an equal volume of medium supplemented with 4 nmol/L IGF-I. After 2 d, the viable cell signal (absorbance at 450 nm) was determined using the WST-8 reagent (Dojindo Molecular Technologies) added per manufacturer's instructions 1 h before harvest.

BxPC-3 and MiaPaCa2 cells were seeded in low-adherence 96-well plates (Corning) at 2,000 per well (BxPC-3) and 5,000 per well (MiaPaCa2) in 100 μL serum-free media. Gemcitabine, AMG 479, and IGF-I were added in an equal volume of medium immediately after seeding. After 2 d, viable cell signal was quantified using ATPlite (Perkin-Elmer) and a luminescence plate reader (EnVision) per manufacturer's instructions.

Five million MiaPaCa2 or BxPC-3 cells were injected s.c. into female athymic nude mice. When the average tumor size reached ∼300 to 450 mm³, mice were randomly assigned into four groups (three mice per group). Two groups of mice were pretreated with 1-mg AMG 479 and two with 1-mg hIgG₁ by i.p. injection. After 6 h, one AMG 479 and one hIgG₁ group received human IGF-I (1–15 μg) by i.v. injection. Control groups received 1× PBS. Xenografts were collected 15 min after IGF-I challenge and snap frozen in liquid nitrogen. Samples were homogenized in 3 volumes of TBS [20 mmol/L Tris-HCl (pH 8.5), 0.15 mol/L NaCl] using a Polytron followed by the addition of 3 volumes of TBS containing 2% Triton X-100. Cell lysates were cleared by centrifugation at 14,000 rpm and analyzed using the MSD assays.

Detection of BrdUrd and caspase-3 in xenografts was done as previously described (24). Briefly, incorporation of BrdUrd in tumor sections was detected with a rat anti-BrdUrd antibody (Accurate), a biotin-labeled rabbit anti-rat IgG secondary antibody (Vector Laboratories), and a Vectastain Elite ABC detection kit (Vector Laboratories). Cleaved caspase-3 was detected with a rabbit anti–caspase-3 antibody (Cell Signaling) followed by a peroxidase-labeled goat anti-rabbit IgG antibody (Jackson Immunoresearch Laboratories). Slides were examined without knowledge of treatment group by routine light microscopy using a Nikon Eclipse 90i microscope (Nikon Instruments). Images were captured using a Nikon DXM1200F digital camera, 20× objective, and ACT-1 (version 2.63) software.

Mice bearing established (∼200 mm³) BxPC-3 or MiaPaCa2 xenografts were randomly assigned into four groups (10 mice per group) and treated i.p. twice per week with AMG 479 (30, 100, or 300 μg/dose), hIgG₁ (300 μg/dose), gemcitabine (80 mg/kg), or AMG 479 plus gemcitabine for the duration of the experiment. Tumor volumes were measured twice per week using calipers. An Inhibitory Emax model was used to calculate the AMG 479 ED₅₀ based on tumor volumes at the end of the experiment using WinNonlin version 5.1.1 (Pharsight).

Serum samples were collected from mice before the fifth dose of AMG 479 (trough level) and 2, 8, 24, and 48 h after the fifth dose (n = 2 per time point). AMG 479 levels were measured using a sandwich ELISA; a mouse anti-AMG 479 antibody was used to capture AMG 479 from the sample, and a horseradish peroxidase–labeled mouse anti-AMG 479 antibody was used for detection. The AMG 479 EC₅₀ was calculated from the simulated C_min based on steady-state levels.

For in vivo efficacy studies, repeated measures ANOVA was used to compare reduction in tumor volume in mice treated with AMG 479 plus gemcitabine versus either agent alone. Dose-response experiments were analyzed by repeated measures ANOVA followed by post hoc Scheffe. Changes in phosphorylated (p)IGF-IR, pINSR, and pAkt in the pharmacodynamic assay were compared using Student's t test.

AMG 479 bound to IGF-IR(ECD)-mFc with an apparent K_D of 0.33 ± 0.14 nmol/L without cross-reacting with the INSR (Fig. 1A and B). The AMG 479 half-life for binding to IGF-IR(ECD)-mFc was estimated to be 9.7 h using the association rate constant (k_a = 6 × 10⁴ [1/s]) derived from a Biacore kinetic assay and the K_D obtained by the equilibrium method.

AMG 479 inhibited binding of IGF-I and IGF-II to IGF-IR(ECD)-mFc (IC₅₀ = 0.53 nmol/L for IGF-I and 0.59 nmol/L for IGF-II) but did not inhibit binding of ¹²⁵I-insulin to INSR(ECD)-mFc (Fig. 1C). Binding of ¹²⁵I-labeled growth factors to their receptors was completely inhibited by the respective unlabeled homologous growth factors (Fig. 1C). As expected, rituximab, the control IgG₁ antibody, did not interfere with ligand binding (Fig. 1C).

Using quantitative flow cytometry, we determined that the number of cell-surface IGF-IR per cell was similar in MiaPaCa2 (mean: 38,000 per cell) and BxPC-3 (mean: 32,000 per cell) cells, whereas the number of INSR was significantly different in MiaPaCa2 (mean: 12,000 per cell) compared with BxPC-3 (mean: 500 per cell) cells.

IGF-IR autophosphorylation was induced by IGF-I and IGF-II in a concentration-dependent manner in MiaPaCa2 and BxPC-3 cells; low-level IGF-IR activation by insulin was observed but only at high insulin concentrations (Fig. 2A). AMG 479 potently inhibited IGF-I– and IGF-II–induced IGF-IR autophosphorylation (IC₅₀ = 1–3 nmol/L); there was no evidence of IGF-IR agonism by AMG 479 (Fig. 2A).

Because BxPC-3 cells express very low levels of cell-surface INSR, no growth factor–induced autophosphorylation of INSR was observed in this cell line (Fig. 2B). The level of INSR autophosphorylation in MiaPaCa2 cells was similar with each growth factor (Fig. 2B). AMG 479 inhibited IGF-I–induced INSR phosphorylation in MiaPaCa2 cells (albeit at higher concentrations, EC₅₀ = 110 nmol/L). The antibody partially inhibited IGF-II–induced INSR phosphorylation and had no effect on insulin-induced INSR phosphorylation (Fig. 2B).

The IGF-I–induced phosphorylation of Akt in BxPC-3 and MiaPaCa2 cells paralleled IGF-I–induced activation of IGF-IR (Fig. 2C). MiaPaCa-2 cells were more sensitive (reduced IC₅₀s) to IGF-II– and insulin-dependent phosphorylation of Akt than in BxPC-3 cells, probably due to the presence of higher levels of INSR (homodimers and hybrids). AMG 479 effectively inhibited IGF-I– and IGF-II–induced Akt phosphorylation, showing full inhibition in the BxPC-3 cell line (IC₅₀ = 2.0–4.0 nmol/L) and almost complete inhibition in the MiaPaCa2 cell line (IC₅₀ = 50–60 nmol/L; Fig. 2C). As expected, AMG 479 fully inhibited insulin-induced Akt phosphorylation in BxPC-3 cells but had no effect in MiaPaCa2 cells. AMG 479 also inhibited IGF-I–, IGF-II–, and insulin-induced phosphorylation of three other downstream markers: IRS-1, GSK3β, and p70S6 kinase (Supplementary Table S1).⁶

The presence of IGF-IR/INSR hybrids in MiaPaCa2 cells and the ability of AMG 479 to capture these hybrid receptors were confirmed by immunoprecipitation studies using AMG 479 or 83-14. Hybrid receptors were detected in MiaPaCa2 cells and to a much lower extent in BxPC-3 cells. In MiaPaCa2 cells, about 45% of the IGF-IR seemed to form hybrid receptors compared with only 8% in BxPC-3 cells (Fig. 2D).

The in vivo effects of AMG 479 were studied using a xenograft pharmacodynamic assay. Administration of IGF-I (15 μg) i.v. in the presence of the control antibody (hIgG₁) to BxPC-3 tumor–bearing mice led to 37- and 15-fold stimulation of IGF-IR and Akt phosphorylation, respectively (Fig. 3A). Pretreatment of mice with AMG 479 led to a 90% inhibition of IGF-I–induced IGF-IR phosphorylation (P = 0.0031) and basal IGF-IR phosphorylation (P = 0.031). AMG 479 pretreatment also led to approximately 60% inhibition of IGF-I–induced Akt phosphorylation (P = 0.0045) and 50% inhibition of basal Akt phosphorylation (not significant, P = 0.13). Although BxPC-3 cells express low levels of cell-surface INSR, IGF-I administration led to a 23-fold increase in INSR phosphorylation, and ∼20% of INSR activation was inhibited by AMG 479 (not significant, P = 0.19). Basal INSR seemed to be similarly inhibited (not significant, P = 0.29), although the significance of this observation is unclear given the low absolute signal.

In vivo signaling studies in MiaPaCa2 xenografts were done using a wide range of IGF-I doses, allowing for a more quantitative analysis of IGF-I–induced signaling in cells expressing INSR and hybrid receptors (Fig. 3B). As expected, IGF-IR, Akt, and INSR were phosphorylated by IGF-I with maximum stimulation observed with 5, 10, and 15 μg IGF-I, respectively. The level of IGF-IR and Akt phosphorylation in MiaPaCa2 xenografts was comparable to that observed in BxPC-3 xenografts; however, the level of INSR phosphorylation (15 μg IGF-I) was ∼10-fold higher than observed in BxPC-3 cells. AMG 479 administration led to almost complete inhibition (90–97%) of IGF-I–induced (1–15 μg) phosphorylation of IGF-IR (P < 0.016 for all IGF-I doses), whereas basal IGF-IR activation was reduced by ∼80% (P = 0.018). Stimulated and basal Akt phosphorylation were also reduced by AMG 479 treatment, although the effect diminished at saturating IGF-I doses (67–76% at 0–1 μg IGF-I, P < 0.026; 41–42% at 5–15 μg IGF-I). Inhibition of INSR remained similar with all IGF-I doses as hybrid receptors were activated, saturating at ∼60% (5–15 μg of IGF-I, P < 0.036 for all IGF-I doses).

To assess the downstream consequences of IGF-IR and Akt inhibition by AMG 479, we studied the changes in apoptosis and proliferation by measuring cleaved caspase-3 expression and BrdUrd incorporation in pancreatic carcinoma xenografts. Immunohistochemical analysis of BxPC-3 xenografts showed that AMG 479 treatment increased cleaved caspase-3 expression at 6 h but did not affect BrdUrd incorporation (Fig. 4A). In MiaPaCa2 cells, AMG 479 treatment did not affect caspase-3 expression but led to a marked decrease in BrdUrd staining at 24 h (Fig. 4B). The effects of IGF-I on BxPC-3 and MiaPaCa2 cells cultured under low-adherence, serum-free conditions were remarkably consistent with those of AMG 479 in tumor sections by immunohistochemistry: In BxPC-3 cells, IGF-I prevented apoptosis and reduced caspase-3 expression from 69% to 26% with no effect on BrdUrd incorporation, whereas in MiaPaCa2 cells, IGF-I led to little or no change in apoptotic fraction or caspase-3 expression while increasing BrdUrd incorporation from 2.9% to 19% (Supplementary Table S2).⁶

The effect of AMG 479 in combination with gemcitabine on viable cell numbers was first examined using murine 32D IGF-IR+IRS-1 cells, which express ∼55,000 IGF-IR per cell. This cell line is dependent on IGF-I for growth and survival, making it a sensitive model to evaluate AMG 479 interactions with cytotoxic drugs (22). The IC₅₀s for AMG 479 and gemcitabine were 3.0 and 5.1 nmol/L, respectively. The combination of AMG 479 with gemcitabine did not significantly alter the IC₅₀ of either drug, and the inhibitory effects of the drugs seemed to be additive (Fig. 5A).

The effect of IGF-I and gemcitabine on viable cell numbers was determined in BxPC-3 and MiaPaCa2 cells grown as adherent or nonadherent cultures in media containing 0%, 1%, or 10% serum. IGF-I did not increase cell numbers in the presence of 10% serum; however, the positive effects of IGF-I emerged in low-adherence plates as the serum concentration was reduced (Supplementary Fig. S1).⁶ In contrast, the potency of gemcitabine was greatest in adherent cells in the presence of serum (Supplementary Fig. S1). Treatment of BxPC-3 and MiaPaCa2 cells with increasing concentrations of IGF-I resulted in a respective 6- to 7-fold and 2-fold increase in cell viability (Fig. 5B). IGF-I concentrations in excess of 1 nmol/L were required for inhibition by gemcitabine: maximum inhibition was ∼25% to 50% of the greatest positive effect of IGF-I on viable cell signal. When AMG 479 was combined with gemcitabine in the presence of a fixed concentration of IGF-I (10 nmol/L), the positive effect of IGF-I was inhibited and the activity of gemcitabine was reduced (Fig. 5C).

To determine if gemcitabine can directly inhibit IGF-IR signaling, adherent, serum-starved BxPC-3 and MiaPaCa2 cells were stimulated with IGF-I for 20 minutes, and levels of phosphorylated IGF-IR and Akt were measured using MSD multiplex assays (Supplementary Fig. S2).⁶ Gemcitabine inhibited IGF-IR autophosphorylation in BxPC-3 cells treated with ≥8 nmol/L IGF-I in a dose-dependent manner. However, gemcitabine did not inhibit IGF-IR autophosphorylation in MiaPaCa2 cells or Akt phosphorylation in either cell line.

AMG 479 significantly inhibited the growth of BxPC-3 and MiaPaCa2 xenografts in a dose-dependent manner. Treatment of BxPC-3 tumor–bearing mice with 30, 100, and 300 μg of AMG 479 twice per week resulted in statistically significant tumor growth inhibition of 42% (P = 0.041), 69% (P = 0.0001), and 80% (P < 0.0001), respectively, compared with the hIgG₁ control group. Treatment of MiaPaCa2 tumor–bearing mice also led to statistically significant tumor growth inhibition: 46%, 57%, and 78%, respectively (P < 0.0001 for all three doses; Fig. 6A).

The efficacy of gemcitabine as monotherapy or in combination with AMG 479 was tested against established BxPC-3 and MiaPaCa2 xenografts. Gemcitabine alone (80 mg/kg) significantly inhibited tumor growth (∼50%) in both models (Fig. 6B). Analysis by repeated measures ANOVA confirmed that the efficacy achieved by the combination of agents was significantly better than either agent alone: by 30% (P = 0.0005) in the BxPC-3 model and by 22% (P = 0.0052) in the MiaPaCa2 xenograft model (Fig. 6B).

Administration of four doses of 30, 100, and 300 μg AMG 479 to mice bearing BxPC-3 xenografts led to steady-state AMG 479 serum levels of 4.7, 22, and 45 μg/mL, respectively (Fig. 6C). The AMG 479 ED₅₀ and serum EC₅₀ calculated from the tumor growth inhibition of BxPC-3 xenografts and the trough levels of AMG 479 were 32 μg/dose and 3.05 μg/mL, respectively (Fig. 6C). Similar measurements were derived from the MiaPaCa-2 xenograft study (data not shown). Serum IC₅₀ was ∼30-fold greater than the IC₅₀ obtained in vitro for AMG 479 inhibition of IGF-IR signaling.

The levels of basal and total phosphorylated IGF-IR were determined for peak and trough levels of AMG 479 at the end of the BxPC-3 xenograft efficacy study. We observed significant inhibition of IGF-IR phosphorylation in all AMG 479 treatment groups with maximum inhibition (∼82%) observed with 300 μg AMG 479. The level of total IGF-IR was also substantially reduced (60%) by treatment with AMG 479 (Fig. 6D).

AMG 479 is a fully human monoclonal antibody against IGF-IR under investigation as an oncology therapeutic. We used purified IGF-IR extracellular domains to show that AMG 479 inhibits the high-affinity binding of IGF-I and IGF-II to IGF-IR. The ability of AMG 479 to inhibit IGF-I as well as IGF-II binding to IGF-IR may be clinically important because humans have high circulating levels of both ligands. In addition, overexpression of IGF-II as a result of loss of imprinting has been well documented as an important autocrine/paracrine factor in human carcinogenesis (4). We observed that AMG 479 blocks ligand binding to IGF-IR with no evidence of cross-reactivity to INSR or interference with insulin binding. The inability of AMG 479 to block insulin binding to INSR may also be clinically important: it may minimize the interference with glucose metabolism that has been reported with other IGF-IR antagonists (7).

The difference in IGF-IR and INSR expression profiles between the two pancreatic cell lines, BxPC-3 and MiaPaCa2, allowed us to characterize the activity of AMG 479 against the tumor growth-promoting activity of IGF-IR and IGF-IR/INSR hybrids. The near complete inhibition of IGF-I– and IGF-II–induced activation of IGF-IR by AMG 479 observed in BxPC-3 and MiaPaCa2 cell-based assays is consistent with our AMG 479 ligand-blocking results. Our immunoprecipitation experiments suggest that IGF-IR/INSR hybrid receptors are enriched in MiaPaCa2 cells. The molecular nature of this interaction is not yet clear; however, its functional significance is supported by the enhanced activation of INSR and Akt by IGF-I and IGF-II in MiaPaCa2 cells and the inhibition of this effect by AMG 479 treatment. The complete inhibition of insulin-induced Akt activation by AMG 479 in BxPC-3 cells strongly suggests that most of the INSR present in this cell line (in vitro) are present as hybrid receptors, rendering them sensitive to AMG 479 inhibition.

The results of our in vivo pharmacodynamic assays also strongly suggest that AMG 479 can inhibit signaling through IGF-IR/INSR hybrids as wells as IGF-IR homodimers. AMG 479 seemed to significantly reduce INSR activation in response to exogenous IGF-I (all doses) in MiaPaCa2 tumors; however, because AMG 479 does not inhibit INSR homodimers, this inhibitory effect probably reflects the targeting of hybrid receptors by the antibody. In addition, the level of IGF-IR inhibition (>95%) was sufficient to include all the IGF-IR homodimers and IGF-IR/INSR hybrids. These results are consistent with observations that hybrid receptors are responsive (IC₅₀ = 3.8 nmol/L) to IGF-I. The inability of AMG 479 to completely block both INSR and Akt activation is probably the result of the use of supraphysiologic concentrations of IGF-I, which could directly activate INSR homodimers (17, 25). Except for the IGF-I stimulation of INSR seen in BxPC-3 xenografts but absent in BxPC-3 cells (even at high ligand concentrations), the activity of IGF-I and AMG 479 obtained in vivo was consistent with the results obtained in adherent serum-free cultures. The induction of INSR phosphorylation by IGF-I in BxPC-3 xenografts was most likely responsible for the lack of complete Akt inhibition observed (compared with culture cells) and may reflect a higher prevalence of INSR homodimers when these cells are grown in vivo. Together these results suggest that AMG 479 can effectively block hybrid receptor activation and signaling without affecting INSR homodimers. The isolation of AMG 479 and other IGF-IR antibodies for clinical development occurred before the importance of IGF-IR/INSR hybrids was fully appreciated. We believe that showing that AMG 479 can inhibit activation of hybrid receptors is an important aspect of the preclinical validation of this clinical agent.

Our characterizations of IGF-I and AMG 479 activity in vitro and in vivo have enabled us to estimate the level of IGF-I exposure necessary for the growth of BxPC-3 and MiaPaCa2 tumors. The levels of activated IGF-IR observed in xenografts (3% BxPC-3, 4% MiaPaCa2) were far below the value expected based on the free plasma IGF-I concentration. The reported free concentration (0.74 nmol/L) of plasma IGF-I in mice is close to the EC₅₀ that we determined for BxPC-3 and MiaPaCa2 cells in culture (26). The actual tumor exposure was estimated to be 0.03 to 0.04 nmol/L IGF-I using the basal IGF-IR signals (expressed as a percent of maximum potential IGF-I stimulation) and the EC₅₀ determined for IGF-IR activation in vitro. These calculations imply that only relatively low levels of IGF-IR activation are needed to drive the growth of BxPC-3 and MiaPaCa2 xenografts. This conclusion is also supported by the fact that AMG 479 inhibited basal IGF-IR activity in these xenografts while showing potent antitumor efficacy (80% tumor growth inhibition) when used as a single agent. These results would not be expected in MiaPaCa2 xenografts if hepatic IGF-I was able to activate INSR homodimers (shown to be resistant to AMG 479) or if AMG 479 was inefficient at inhibiting IGF-IR/INSR hybrid receptors. In addition, the ability of AMG 479 to completely inhibit xenograft IGF-IR signaling when challenged with supraphysiologic concentrations of IGF-I suggests that this antibody should readily cope with the increase in plasma IGF-I in response to pituitary IGF-IR blockade (27) and the wide human variation in plasma IGF-I and IGF-II (28).

Inhibition of tumor growth by AMG 479 was associated with reduced IGF-IR autophosphorylation and a reduction in total IGF-IR expression, with both effects seeming to contribute equally in long-term (2–3 week) efficacy experiments. The mechanism by which total receptor is reduced has not yet been determined. However, in contrast to other anti–IGF-IR antibodies, AMG 479 was also able to fully inhibit ligand-induced activation of IGF-IR without affecting receptor turnover (29–31). The potent and efficient blockade of IGF-IR observed with BxPC-3 and MiaPaCa2 cells in vitro was obtained with the simultaneous addition of AMG 479 and ligand (IGF-I or IGF-II) for a brief incubation period (20 minutes) without evidence of receptor degradation (or internalization; data not shown). It is possible that receptor turnover induced by AMG 479 in vivo could explain why the antibody seemed to inhibit IGF-IR/INSR hybrids and IGF-IR homodimers with more equivalent potency than observed with relatively brief in vitro antibody treatment. Efficacy against BxPC-3 or MiaPaCa2 xenografts was the result of increased apoptosis or decreased cellular proliferation, respectively. The distinct mechanisms of action directly correlated with the effects of IGF-I as a survival factor in BxPC-3 cells and a mitogen in MiaPaCa2 cells when these cells were cultured with serum-free growth medium in low-adherence plates. This observation suggests that these culture conditions, which have been shown to be ideal to study IGF-I activity, best resemble those found in the xenograft microenvironment. The variables that govern these two distinct biological outcomes are not yet known, but may include preferential signaling through IGF-IR/INSR hybrids or the expression of mutant k-ras in MiaPaCa2 cells.

One of the most pressing issues in the clinical development of IGF-IR–targeted therapeutics is their successful combination with cytotoxic drugs. The combination of AMG 479 and gemcitabine exhibited strong additive inhibition of rapidly proliferating IGF-I–dependent 32D cells, while also showing a significant additive effect against pancreatic carcinoma cells in serum-free, low-adherence cultures. The clinical relevance of these in vitro observations was supported by similar findings when AMG 479 and gemcitabine were combined in the treatment of mice bearing BxPC-3 or MiaPaCa2 tumors. Our data suggest that the combination of AMG 479 and gemcitabine may provide more benefit against pancreatic cancer than either agent alone. A phase II study of AMG 479 in combination with gemcitabine in patients with pancreatic carcinoma is ongoing to test this hypothesis.

In summary, our characterization of AMG 479, a novel fully human anti–IGF-IR monoclonal antibody, suggests that AMG 479 displays the characteristics of an optimal IGF-IR–targeted clinical agent with the ability to (a) inhibit binding of IGF-I and IGF-II to IGF-IR, (b) block the activation of IGF-IR homodimers and IGF-IR/INSR hybrids, (c) trigger IGF-IR down-regulation in vivo, (d) inhibit tumor growth through induction of apoptosis or inhibition of cel lular proliferation, and (e) cooperate with standard chemotherapeutics to enhance antitumor efficacy. These unique properties may translate to important clinical benefits in malignancies where IGF-IR and hybrid receptors influence tumor growth, as well as those resistant to chemotherapeutic agents.

Pedro Beltran, Petia Mitchell, Young-A Chung, Elaina Cajulis, John Lu, Brian Belmontes, Joanne Ho, Mei Mei Tsai, Min Zhu, Steve Vonderfecht, Robert Radinsky, and Frank Calzone are employees of and own stock in Amgen, Inc. Renato Baserga has previously received payments as an Amgen consultant. No other potential conflicts of interest were disclosed.

The order of the authors, Petia Mitchell, Young-A Chung, and Elaina Cajulis, who contributed equally to the major content of this manuscript, was randomly determined. We thank Grace Chung and Larry Daugherty for assistance with flow cytometry, Robert Ortiz and Jennifer Tsoi for assistance with the pharmacokinetic analysis, Barbara Felder and Efrain Pacheco for assistance with immunohistochemistry, and Kathryn Boorer for editorial assistance.