Antixenograft tumor activity of a humanized anti-insulin-like growth factor-I receptor monoclonal antibody is associated with decreased AKT activation and glucose uptake

The insulin-like growth factor (IGF) system plays crucial roles in regulating cellular proliferation, survival, and transformation (1–3). The central node of this system is IGF-I receptor (IGF-IR), a homodimeric receptor tyrosine kinase linked by disulfide bonds, with each monomer consisting of an extracellular α-subunit and a membrane-spanning β-subunit (4). The extracellular domain (ECD) of IGF-IR mediates ligand binding, which induces the activation of intracellular tyrosine kinase activity. Phosphatidylinositol 3-kinase, mitogen-activated protein kinase/extracellular signal-regulated kinase, and other downstream pathways are further recruited to amplify and execute prosurvival and proliferative signals that lead to stimulated growth (1).

Tumor cells can use the IGF system for their abnormal growth. Overexpression of IGF-I or IGF-II and/or IGF-IR and activation of this pathway are often observed in several prevalent cancer types (1). In particular, epigenetic changes, such as loss-of-imprinting at the IGF-II locus, frequently occur in colon and ovarian cancers as well as in several pediatric malignancies (5). High serum IGF-I concentrations correlate with increased risk for breast, prostate, and colon cancers in humans (6). Gain-of-function studies show that the IGF system can drive tumorigenesis in animal models (7, 8), whereas IGF-IR deficiency conversely inhibits cellular transformation (9). Taken together, these data provide a strong scientific rationale for developing targeted therapies to inhibit this pathway in human cancer.

Multiple approaches, such as anti-IGF-IR monoclonal antibodies and small-molecule receptor tyrosine kinase inhibitors, have been developed to inhibit the activity of this pathway in cancer (10–21), and many of them are being actively tested in phase I or II clinical trials (22). Because small-molecule inhibitors of the IGF-IR kinase often cross-inhibit insulin receptor, we have developed a novel humanized, affinity-matured IgG1κ monoclonal antibody (h10H5) against the IGF-IR ECD. We report that h10H5 effectively inhibits IGF-IR mediated signaling by blocking both IGF-I and IGF-II binding to IGF-IR and by inducing IGF-IR down-regulation through internalization and degradation while leaving insulin receptor unaffected. Furthermore, we show that degradation of the ECD and intracellular domain of IGF-IR is differentially affected by the lysosomal and proteasomal inhibitors, consistent with immunofluorescence data showing h10H5 uptake by clathrin-mediated endocytosis followed by transport to late endosomes and lysosomes for degradation. Functionally, h10H5 exhibits antiproliferative activity against cancer cell lines in vitro and in vivo, which can be enhanced by not only chemotherapy but also anti–vascular endothelial growth factor (VEGF) antibodies. Notably, the antitumor activity of h10H5 is associated with decreased AKT activation, a composite transcription profile with significant changes involving DNA metabolic and cell cycle machineries, and can also be monitored in vivo by 2-fluoro-2-deoxy-d-glucose (FDG)-positron emission tomography (PET) imaging. Taken together, these data not only support the clinical testing of h10H5 as a biotherapeutic for human tumors that rely on IGF-IR for proliferation and survival but also show the utility of a novel endpoint assay for h10H5 activity that can be used to assess antitumor activity noninvasively in vivo.

BALB/c mice (Charles River Laboratories) were hyperimmunized with human IGF-IR ECD (amino acids 1-902) in Ribi adjuvant (Ribi Immunochem Research). B cells from these mice, all of which showed high anti-IGF-IR antibody titers by direct ELISA and by fluorescence-activated cell sorting of IGF-IR-expressing MCF7 cells, were fused with mouse myeloma cells (X63.Ag8.653; American Type Culture Collection) based on described protocols (23). After 10-12 days, the supernatants were harvested and screened for antibody production by direct ELISA and fluorescence-activated cell sorting. Murine 10H5 was chosen from a series of positive clones for its strong binding to the purified IGF-IR ECD and cell-surface IGF-IR as well as its effective blockade of ligand-receptor interaction. Humanization of 10H5 was achieved by direct CDR grafts onto the acceptor human consensus (subtype III) framework of trastuzumab (24). Mutant h10H5 carrying a D265A mutation (based on Kabat numbering), which has diminished binding to all Fcγ receptors, was produced for antibody-dependent cell-mediated cytotoxicity studies. Biacore analysis, using the F(ab) fragment of h10H5 to avoid avidity effects, showed a dissociation equilibrium constant (K_d) of 217 and 159 pmol/L for binding to the recombinant human and cynomolgus monkey IGF-IR ECD, respectively (data not shown). h10H5 did not react with the closely related human insulin receptor or murine IGF-IR, thus showing strong binding selectivity (data not shown).

Human breast cancer cell lines MCF7 and SW527 and neuroblastoma cell line SK-N-AS were obtained from the American Type Culture Collection. Antibodies to phosphorylated and total IGF-IR β-chain, AKT, and extracellular signal-regulated kinase 1/2 were purchased from Cell Signaling Technology. Anti-IGF-IR α-chain monoclonal antibody 10F5 was generated in-house as for murine 10H5. Lysosomal protease inhibitors leupeptin and pepstatin A were from Roche Applied Science. Bortezomib was produced in-house and is the active ingredient of the proteasome inhibitor Velcade that has been approved for treating multiple myeloma (25). Recombinant IGF-I and IGF-II were purchased from R&D Systems.

Recombinant human IGF-IR ECD was coated onto 96-well plates at 200 ng/well by overnight incubation at 4°C. All subsequent steps were at room temperature: after washing with 0.05% Tween 20 in PBS, plates were blocked for 1 h with 5% bovine serum albumin in PBS, incubated for 1 h with serially diluted h10H5, anti-IGF-IR antibody αIR3 (Calbiochem), or control antibody anti-gp120 (to herpes simplex virus glycoprotein D). Biotinylated IGF-I or IGF-II (80 ng/mL) was subsequently added for 1 h, washed, and detected with streptavidin-horseradish peroxidase (1:6,000) for 20 min. The signals were developed with TMB substrate, and the reactions were stopped with 100 μL/well of 1 mol/L phosphoric acid before reading the absorbance at A_{450 nm}.

SK-N-AS and MCF7 cells grown on LabTek II slides were incubated with 5 μg/mL h10H5 for 5 min to 4 h at 37°C, 5% CO₂ in complete (10% fetal bovine serum) growth medium containing lysosomal protease inhibitors (5 μmol/L pepstatin A and 10 μg/mL leupeptin) and 10 μg/mL Alexa 488-transferrin. Cells were then chilled, washed five times in cell medium, fixed for 20 min with 3% paraformaldehyde, quenched for 10 min with 50 mmol/L NH₄Cl in PBS, and permeabilized with saponin buffer [0.4% (w/v) saponin, 2% fetal bovine serum, 1% bovine serum albumin]. Uptaken h10H5 was detected with Cy3 anti-human and where indicated cells were also stained with 1:1,000 mouse anti-LAMP1 (BD Biosciences) or 1:100 rabbit anti-IGF-IR β-subunit followed by FITC-conjugated secondaries (Jackson ImmunoResearch). Slides were coverslipped with 4′,6-diamidino-2-phenylindole–containing VectaShield (Vector Labs) and imaged by epifluorescence microscopy using the ×100 objective of a DeltaVision deconvolution microscope (Applied Precision) powered by SoftWorx (version 3.4.4). Figures were compiled using Adobe Photoshop CS.

Cells were lysed in lysis buffer [20 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 1% Triton X-100] with protease inhibitors and/or phosphatase inhibitor cocktail (Sigma) on ice for 20 min and cleared by centrifugation at 14,000 rpm for 15 min. Equal amount of cell lysates measured by the BCA assay (Pierce Biotechnology) were immunoprecipitated with anti-IGF-IR antibody 10F5 or directly Western blotted.

For cell viability assays, MCF7 cells were seeded at 2,000 per well and incubated overnight at 37°C in RPMI 1640 with 10% fetal bovine serum (Sigma) and 1% Glutamax (Promega). Cells were then starved in serum-free RPMI 1640 (phenol red–free) for 5 h followed by addition of serially diluted h10H5 for 1 h and then supplemented with 1% fetal bovine serum and 1 ng/mL IGF-I for 5 days. Cell viability was determined using the CellTiter-Glo Luminescent Cell Viability kit (Promega) according to the manufacturer's instructions.

For the FDG uptake assay, SK-N-AS cells were plated in 96-well plates at 10,000 per well. The following day, medium was removed and replaced with glucose-free DMEM containing 0.1% or 0% fetal bovine serum and a range of h10H5 concentrations for a further 48 h. To determine FDG uptake, 2 μCi/well [³H]FDG ([5,6-³H]; 0.74-2.22 TBq/mmol; 20-60 Ci/mmol; American Radiolabeled Chemicals) was added for the last 24 h and harvested onto Unifilter GF/C filter plates using a FilterMate Harvester (Perkin-Elmer). Microscint 20 scintillation fluid was added to all wells and radioactivity incorporated per well (counts/min) was determined using a Packard TopCount NXT microplate scintillation counter.

Total RNA was prepared from SK-N-AS xenograft tumors using the RNeasy kit (Qiagen). Affymetrix human genome U133 Plus 2.0 arrays were employed for expression profiling according to the manufacturer's instructions. Probe set analysis was done using Affymetrix MAS5.0. Quantitative real-time reverse transcription-PCR was done in triplicates for each sample using the TaqMan Gold reagent kit (Applied Biosystems) and Stratagene Mx4000 according to the manufacturer's instructions (standard curve method: user bulletin no. 2, ABI PRISM 7700 SDS). For normalization, the mean expression level of target genes was divided by the mean of the expression level of the housekeeping gene RPL19. The sequences of the primer/probe sets (5′-FAM reporter dye and 3′-TAMRA quencher labeled) are listed in Supplementary Table S1.¹

Female nu/nu and C.B-17 SCID beige mice were inoculated s.c. with 10 × 10⁶ SK-N-AS and 5 × 10⁶ SW527 tumor cells. Once tumors reached a mean volume of 130 to 260 mm³, mice were then randomized into groups of 8 to 10 mice and treated with antibody and/or chemotherapy. Percent tumor growth inhibition (%TGI) was calculated as 100 × [(C − T) / C], where C and T are the mean tumor volume of the control and treated groups, respectively. Data analysis and generation of P values were done using the Dunnett's t test for tumor volumes, or a log-rank test for doubling time, defined as the number of days for a tumor to double its size measured at the day of randomization, with JMP software version 6.0 (SAS Institute).

A separate SK-N-AS xenograft study was conducted to evaluate FDG-PET imaging as a measure of drug response. Dynamic PET scans were done on tumors following caliper measurements on day 0 before treatment and again on days 3, 7, 10, and 14. At the beginning of a 30-min dynamic PET scan, ∼250 μCi F18-FDG was injected into the lateral tail vein of each mouse. PET data were processed into a time series of images to allow quantification of the tumor relative to the blood pool using the liver signal as a proxy for blood (26) and the region of interest analysis was done (Siemens Preclinical Solutions). Numerical integration and calculations for the conversion of the raw data into Patlak plots were done with the PET Parser program. The slope of the linear portion of the Patlak plot is equal to Ki, the FDG uptake rate constant (units/s). Treatment responses were assessed as the percentage change in Ki relative to the pretreatment value, with a negative change representing a decrease in FDG uptake. These difference data were used for the t tests comparing vehicle and treatment groups at the imaging time points. Data collation and statistical analysis were done using Microsoft Excel using the built-in Student's t test in a two-tailed comparison.

h10H5 is a humanized anti-IGF-IR monoclonal antibody with high binding affinity and specificity as described in Materials and Methods. In competitive solid-phase binding assays, h10H5 inhibited IGF-I and IGF-II binding to IGF-IR with an IC₅₀ of 3.4 nmol/L (Supplementary Figs. S1A and B).¹ To investigate the effects of h10H5 on IGF-IR signaling, serum-starved human breast cancer MCF7 cells were treated with antibody for 20 min before IGF-I or IGF-II stimulation. In comparison with untreated cells (Supplementary Fig. S2A,¹ lane 1), IGF-I induced robust and sustained phosphorylation of IGF-IR and AKT from 10 to 120 min, but only transient extracellular signal-regulated kinase 1/2 phosphorylation, which peaked at 10 min and reduced at later time points (Supplementary Fig. S2A,¹ lanes 2-6). Exposure of cells to h10H5 strongly inhibited these effects of IGF-I throughout the entire time course (Supplementary Fig. S2A,¹ lanes 7-11), with the greatest effect at 1 to 10 μg/mL h10H5 (Supplementary Fig. S2B and C,¹ lanes 4-6). In contrast, the control (anti-gp120) antibody failed to inhibit IGF-I- or IGF-II-mediated signaling at the same concentrations (Supplementary Fig. S2A,¹ lanes 12-14; Supplementary Fig. S2B and C,¹ lane 3). In the absence of ligand stimulation, h10H5 (up to 100 μg/mL) did not induce any observable phosphorylation of IGF-IR, AKT, or extracellular signal-regulated kinase 1/2 (data not shown). Taken together, these data indicate that h10H5 is a potent antagonist of IGF-I- and IGF-II-mediated IGF-IR signaling.

One common mechanism of action of anti-IGF-IR antibodies is the induction of receptor down-regulation (12–14, 17–19). Similarly, h10H5 led to IGF-IR depletion in MCF7 cells after 1 h (Supplementary Fig. S2A,¹ lanes 9-11). We expanded our analysis to a different cell line, SK-N-AS, by examining IGF-IR levels in a more prolonged time course with various doses of h10H5 treatment (Fig. 1A). As in MCF7 cells, IGF-IR was already partially down-regulated at 1 h, reaching maximal reduction at 4 to 8 h and not diminishing further at 24 h with ≥1 μg/mL h10H5 (data not shown). Down-regulation also occurred to a similar extent at 0.1 μg/mL but was slower at early time points. In contrast, neither IGF-I nor anti-gp120 antibody treatment affected IGF-1R levels (Fig. 1A), consistent with prior observation showing minimal effects of IGF-I on IGF-IR degradation (14). Although both h10H5 and IGF-ligands bind IGF-IR with high affinity, only h10H5 possesses bivalent binding capacity. Down-regulation of IGF-IR likely involves antibody-mediated cross-linking, because F(ab) fragments do not elicit this effect as shown previously using anti-IGF-IR antibody 19D12 (19). h10H5-mediated IGF-IR down-regulation was specific, because insulin receptor levels were unaffected by h10H5 treatment (Fig. 1A).

Previously, the lysosomal inhibitor methylamine but not the proteasome inhibitor MG115 was shown to reduce the down-regulation of the β-subunit of IGF-IR induced by the anti-IGF-IR antibody A12 (Imclone; ref. 14). However, IGF-IR degradation was still evident at 24 h after methylamine treatment, suggesting that additional pathways are likely employed in A12-mediated IGF-IR down-regulation. We therefore investigated whether similar mechanisms are involved in h10H5-induced IGF-IR down-regulation. Because the ECD and intracellular domain of IGF-IR are exposed to different cellular environments during the internalization and trafficking processes, we examined whether they are differently regulated. Antibodies specific to the ECD α-subunit and the β-subunit intracellular domain were used to follow their levels before and after h10H5 treatment using bortezomib to inhibit the proteasome (25) and leupeptin and pepstatin A to suppress lysosomal protease activities. The α- and β-subunits of IGF-IR exhibited similar kinetics of h10H5-induced down-regulation in SK-N-AS cells in the absence of these inhibitors (DMSO control; Fig. 1B and C). Treatment with both classes of inhibitors delayed h10H5-induced IGF-IR down-regulation (Fig. 1B and C), but bortezomib was more effective at preventing β-subunit down-regulation (Fig. 1C), whereas the lysosomal inhibitors preferentially inhibited α-subunit degradation (Fig. 1B). These results are consistent with the possibility that the ECD (α-subunit) is degraded in the lysosomes, whereas the intracellular domain is more readily accessible to the cytoplasmic proteasome, in accordance with their membrane topologies. The difference between our data and previously published results (14) could be due to that bortezomib is a more potent inhibitor of proteasome than MG115.

Anti-IGF-IR antibodies have been reported to induce efficient IGF-IR internalization (13, 18), but the detailed intracellular trafficking processes are not well characterized. We therefore did immunofluorescence microscopy to monitor the kinetics and localization of h10H5 internalization. IGF-IR is normally localized to the plasma membrane of MCF7 cells (data not shown) but rapidly internalizes within 5 min on h10H5 addition most likely via clathrin-coated vesicles as shown by colocalization with transferrin (Fig. 2A-C), which is known to internalize via this pathway. By 20 min, much less of the h10H5 signals remained colocalized with transferrin, indicating divergence from the transferrin recycling pathway (Fig. 2D-F). After 60 min, it was detectable within a subset of late endosomes and lysosomes as shown by colocalization with LAMP1 (Fig. 2H-I). The total h10H5 signal was weaker by the 4-h time point, indicating some degradation of the antibody, despite of the presence of lysosomal protease inhibitors, but was still detectable within the lysosomal lumen, surrounded by LAMP1 on the limiting membrane (Fig. 2J-L). Similar findings were observed in SK-N-AS cells (data not shown). The trafficking route of h10H5 confirms the above lysosomal inhibitor data by showing that h10H5 directs IGF-IR to lysosomes, where degradation of luminally exposed ECD occurs. Indeed, further immunofluorescence analysis revealed that the α-subunit is more stable than the β-subunit proteasomally degraded intracellular domain following h10H5 treatment in the presence of lysosomal inhibitors (data not shown).

The observed inhibitory effect of h10H5 on IGF-IR signaling prompted us to examine whether this antibody could also affect cell growth in vitro. Although MCF7 cells grew 4- to 5-fold in the presence of a control (B-cell-specific) antibody, h10H5 significantly inhibited their IGF-I-dependent growth during the 5-day continuous treatment (Fig. 3). This inhibition was dose dependent, with an IC₅₀ between 34 and 57 ng/mL, and was similarly observed in human SK-N-AS neuroblastoma and SW527 breast cancer cells (data not shown).

To assess the effect of h10H5 on the growth of tumor xenografts, nude mice bearing s.c. SK-N-AS tumors were given a single i.v. dose from 0.5 to 200 mg/kg (Fig. 4A). TGI was observed in all h10H5 dose groups compared with the vehicle control (Fig. 4A), with 51% TGI (P = 0.0003) as low as 0.5 mg/kg and 66% to 74% (P < 0.0001) at 2 to 200 mg/kg, suggesting that maximal efficacy in this model could be reached at relatively low doses of h10H5. Time to the first tumor doubling was also significantly longer (P = 0.03-0.0009) in all h10H5 groups with a range of 4.5 to 6.4 days compared with 2.7 days for the vehicle group. Thus, h10H5 shows strong single-agent activity in this model. Furthermore, a Fc-mutated (D256A) version of h10H5 that is no longer capable of binding to Fcγ receptors (data not shown) resulted in identical TGI (Supplementary Fig. S3),¹ showing for the first time in vivo that antibody-dependent cell-mediated cytotoxicity is not part of the mechanism of action of h10H5.

In a separate xenograft experiment, we examined whether h10H5 treatment in vivo affected tumor-associated IGF-IR signaling. SK-N-AS tumor samples treated at 5 or 20 mg/kg h10H5 were collected at 6, 24, and 48 h after dosing and analyzed by Western blotting. IGF-IR down-regulation was observed as early as 6 h and maintained at least up to 48 h (Fig. 4B). Furthermore, phosphorylated AKT levels decreased significantly, whereas total AKT levels remained unchanged in these tumors (Fig. 4B), in agreement with prior observations that the small-molecule inhibitor of IGF-IR kinase NVP-ADW742 and anti-IGF-IR antibody A12 can inhibit AKT activation in the xenograft tumors (16, 27). Thus, IGF-IR-mediated signaling in the SK-N-AS tumors was effectively inhibited by h10H5.

Although it is feasible to collect xenograft tumors to measure drug activity in preclinical studies, obtaining tumor biopsies from metastatic cancer patients poses a significant challenge. Noninvasive imaging techniques, such as measurement of glucose uptake in tumors, would be a highly desirable alternative for monitoring drug responses in clinical trials. We therefore first examined any possible effect of h10H5 on glucose uptake in SK-N-AS cells in vitro. Indeed, [5,6-³H]FDG uptake into SK-N-AS cells in the presence of h10H5 was inhibited with an IC₅₀ of 3 μg/mL (Fig. 4C), independent of serum, encouraging us to assess the utility of FDG-PET in an additional SK-N-AS xenograft study. As expected, a single injection of h10H5 at 10 mg/kg resulted in significant inhibition of SK-N-AS tumor growth compared with vehicle treatment (data not shown). TGI on days 3, 7, 10, and 14 was 42.2% (P = 0.004), 63.2% (P < 0.0001), 72.6% (P < 0.0001), and 64.1% (P < 0.0001), respectively. FDG-PET imaging of the SK-N-AS tumors was done in parallel and revealed decreases of 37%, 28%, and 37% in the rate of FDG uptake on days 3, 7, and 14, respectively (Fig. 4D), relative to the baseline rate on day 0 (P < 0.05 for all time points of h10H5 treatment). These results show that FDG-PET changes accompany h10H5 treatment and correlate with the inhibition of SK-N-AS xenograft tumor growth. Therefore, FDG-PET, as an indicator of tumor cell density/function (28), is useful for monitoring the h10H5 antitumor activities in vivo, similar to its applicability to other cancer therapeutics (29). Although FDG-PET provides a viable alternative to routine imaging techniques, its utility is limited to FDG-avid tumors, which varies in frequency depending on tumor types.

In addition to the h10H5 effect on IGF-IR signaling and glucose uptake, we also investigated global regulation of gene expression induced by the antibody. Previously, differential gene expression, particularly those involved in cell cycle regulation and apoptosis, was noted by analyzing LuCap 35V human prostate xenograft tumors treated with the anti-IGF-IR antibody A12 (30). However, those tumors were harvested from a single time point after treatment. To compare and better characterize the major processes affected by blockade of IGF-IR signaling during h10H5 treatment, we followed the kinetic alterations of gene expression as early as 24 h as well as at a later time point of 2 weeks in SK-N-AS tumors receiving weekly 5 mg/kg h10H5 treatment. Interestingly, a total of 316 genes exhibited differential expression with a false discovery rate of <5% comparing vehicle and h10H5 treatment at 24 h (Supplementary Table S2),¹ with 200 genes being down-regulated (classes I and II) and 116 being up-regulated (classes III and IV). DNA metabolism and cell cycle were identified as the top ranked functional categories (Supplementary Table S3),¹ with highly significant ESAE scores (2.6E-19 to 3.8E-23). This gene expression signature emerged 24 h after treatment before obvious TGI and correlated with significant efficacy observed several days later (Fig. 4A). Interestingly, the majority of the genes that were affected after 24 h were no longer affected after 2 weeks: classes I and IV showed recovery to vehicle levels by this time point (Fig. 5A), However, this recovery was incomplete, because a much smaller set of genes remained down-regulated (class II) or up-regulated (class III) at 2 weeks (Fig. 5A). These data suggest that parallel and/or alternative signaling pathways are used for tumor adaptation to growth in the absence of IGF-IR signaling, consistent with the observed cytostatic but not cytotoxic effect of h10H5 on tumor growth. We further validated the expression patterns of selected genes using quantitative real-time PCR analysis (Fig. 5B). PCNA, E2F1, and CDC6 recapitulated the pattern of the class I genes expression, with statistically significantly reduced expression at 24 h and full recovery at 2 weeks. In contrast, the p55 γ-regulatory subunit of phosphoinositide 3-kinase (PIK3R3) and Aurora kinase A (AURKA) are examples of the class II genes that remain significantly down-regulated after 2 weeks. PIK3R3 is frequently up-regulated in ovarian, prostate, breast, and liver cancers (31), and its down-regulation of PIK3R3 has been associated with decreased IGF-II-dependent growth of glioblastoma-derived neurospheres (32) and increased apoptosis in ovarian cancer cells (31). Thus, these microarray data implicate intermediate molecular links between IGF-IR pathway inhibition and decreased tumor growth and highlight the dynamic and complex effects induced by a therapeutic antibody. Such a gene expression signature could also be useful for monitoring the transient and/or persistent responses to h10H5 therapy. However, independent verification with other tumor types would be necessary to further establish the significance and broader applicability of this gene signature.

Although the IGF-IR pathway is frequently exploited by cancer cells for dysregulated growth, the microarray data (Fig. 5) indicate that other compensating or parallel pathways may allow tumor cells to evade IGF-IR inhibition. Combination of anti-IGF-IR antibodies with anti-EGFR antibodies, tamoxifen, radiation therapy, or some chemotherapeutic agents have shown improved efficacy over several forms of monotherapy (17, 18, 33). We therefore investigated whether h10H5 can be combined with docetaxel, a commonly used chemotherapeutic agent, in the SW527 human breast cancer xenograft model. h10H5 alone resulted in significant inhibition of SW527 tumor growth compared with vehicle treatment (P < 0.01; Fig. 6), with respective TGI values on days 11 and 14 of 52% and 36% and increased time to the first tumor doubling from 5.6 to 10.2 days (P < 0.002). Docetaxel alone also significantly inhibited tumor growth, with TGI of 54% on day 14 (P < 0.01, relative to the vehicle treatment), whereas docetaxel in combination with h10H5 further increase TGI to 82% (P < 0.01, relative to docetaxel treatment). Time to the first tumor doubling was significantly prolonged in the combination group (docetaxel + h10H5) compared with either h10H5 or docetaxel group (P < 0.01). The combination of h10H5 and docetaxel was therefore significantly more effective at inhibiting tumor growth than either monotherapy.

We also determined whether h10H5 could also cooperate with another class of antitumor agents, the antiangiogenic therapy via VEGF blockade (34). The anti-VEGF antibody B20-4.1, which binds both human and murine VEGF with affinity similar to that of bevacizumab (Avastin; Genentech) to human VEGF (35), was therefore tested in the SW527 breast tumor model alone or in combination with h10H5 as above. At weekly doses of 10 mg/kg, B20-4.1 significantly inhibited tumor growth, with TGI of 60% on day 14 (P < 0.0001, relative to vehicle group). Combining B20-4.1 with h10H5 increased the level of TGI to 68% (P = 0.02, relative to B20-4.1 group) but did not result in a statistically significant increase in time to the first tumor doubling. These results show that although h10H5 or B20-4.1 delays SW527 tumor growth as single agents, the combination of h10H5 with B20-4.1 increases the levels of tumor inhibition. Inhibition of IGF-IR signaling has been shown to reduce VEGF production in tumor cells (36, 37) and to interfere with retinal neovascularization through its effect on endothelial cells (38). Our data provide further support that these two pathways likely also interact in the SW527 tumor model.

In summary, we show that h10H5 is a potent inhibitor of IGF-IR signaling. In an 8-week repeat-dose toxicity study, h10H5 was well tolerated on administration at doses from 5 to 50 mg/kg in cynomolgus monkeys, significantly above the minimum efficacious dose of 2 mg/kg established here. Our data provide support for investigation of this antibody as a single agent or in combination with other treatments in cancer patients. Furthermore, we show that FDG-PET could be a useful approach for monitoring the activities of anti-IGF-IR antibodies in FDG-avid tumors. Multiple anti-IGF-IR antibodies are being pursued in clinical trials. Although some of them are of human origin (14, 18, 19), others (13, 17), including h10H5, are humanized. In addition, they sometimes differ in antibody isotype, binding affinity/epitope, and antibody-dependent cell-mediated cytotoxicity activity (22). Whether or how these combinatorial attributes might contribute to the differences in drug efficacy, safety, and immunogenicity remains to be determined. Critical variables defining the best-in-class anti-IGF-IR antibody would shed light on developing more effective antibody-based cancer therapeutics in the future.

All authors are employees of Genentech.

We thank Jo-Anne Hongo, Kurt Schroeder, and Farzam Farahi for assistance with hybridoma production and purification; Phil Hass, Mark Nagel, and the CMC subteam for producing and purifying h10H5; Yongmei Chen for generating various 10H5 expression vectors; Mikeal Paquet, Stacy Frankovitz, Carol O'Brien, and Heidi Savage for assistance with in vitro cell-based assays; Robert Soriano for microarray analysis; Karissa Peth for assistance with the FDG-PET study; and John Hanson for developing the PET Parser Program.