biochemical characterization of amg 102: a neutralizing ... · abstract amg 102 is a fully human...

400

Published OnlineFirst February 2, 2010; DOI: 10.1158/1535-7163.MCT-09-0824

Research Article Molecular
Cancer
Therapeutics
Biochemical Characterization of AMG 102: A Neutralizing,Fully Human Monoclonal Antibody to Human andNonhuman Primate Hepatocyte Growth Factor
Teresa L. Burgess1, Jan Sun1, Susanne Meyer5, Trace S. Tsuruda2, Jilin Sun3, Gary Elliott4,Qing Chen2, Mitsuru Haniu2, Will F. Barron2, Todd Juan2, Ke Zhang6,Angela Coxon1, and Richard L. Kendall1

Abstract

Authors'3Clinical ImPharmac5DepartmeUniversityof CancerBethesda,

Note: SupCancer The

CorresponCenter Dri805-447-24

doi: 10.115

©2010 Am

Mol Canc

D

AMG 102 is a fully human monoclonal antibody that selectively targets and neutralizes hepatocytegrowth factor/scatter factor (HGF/SF). A detailed biochemical and functional characterization of AMG102 was done to support its clinical development for the treatment of cancers dependent on signalingthrough the HGF/SF:c-Met pathway. In competitive equilibrium binding experiments, AMG 102 boundto human and cynomolgus monkey HGF with affinities of approximately 19 pmol/L and 41 pmol/L,respectively. However, AMG 102 did not detect mouse or rabbit HGF on immunoblots. Immunoprecipita-tion experiments showed that AMG 102 preferentially bound to the mature, active form of HGF, and in-cubation of AMG 102/HGF complexes with kallikrein protease indicated that AMG 102 had no apparenteffect on proteolytic processing of the inactive HGF precursor. AMG 102 inhibited human and cynomolgusmonkey HGF-induced c-Met autophosphorylation in PC3 cells with IC50 values of 0.12 nmol/L and0.24 nmol/L, respectively. AMG 102 also inhibited cynomolgus monkey HGF-induced migration of humanMDA-MB-435 cells but not rat HGF-induced migration of mouse 4T1 cells. Epitope-mapping studies ofrecombinant HGF molecules comprising human/mouse chimeras and human-to-mouse amino acid substi-tutions showed that amino acid residues near the NH2-terminus of the β-chain are critical for AMG 102binding. Bound AMG 102 protected one trypsin protease cleavage site near the NH2-terminus of theβ-chain of human HGF, further substantiating the importance of this region for AMG 102 binding.Currently, AMG 102 is in phase II clinical trials in a variety of solid tumor indications. Mol Cancer Ther;

9(2); 400–9. ©2010 AACR.

Introduction

Hepatocyte growth factor [HGF; also known as scatterfactor (SF)] is the only known ligand for c-Met, a receptortyrosine kinase expressed in epithelial tissues that playsan essential role in the growth and maintenance of cells(1). Signaling through HGF/SF:c-Met mediates a numberof normal cellular functions, including proliferation, sur-vival, motility, migration, angiogenesis, and morphogen-esis (1–3). Abnormal signaling via the HGF/SF:c-Metpathway has been associated with many types of cancers

Affiliations: 1Oncology Research, 2Protein Science,munology, and 4High Throughput Screening-Molecular

o logy, Amgen Inc. , Thousand Oaks, Ca l i forn ia ;nt of Molecular, Cellular and Developmental Biology,of California, Santa Barbara, California; and 6LaboratoryBiology and Genetics, National Cancer Institute, NIH,Maryland

plementary material for this article is available at Molecularrapeutics Online (http://mct.aacrjournals.org/).

ding Author: Teresa L. Burgess, Amgen Inc., One Amgenve, M/S 14-1-B, Thousand Oaks, CA 91320-1799. Phone:93; Fax: 1-805-375-8368. E-mail: [email protected]

8/1535-7163.MCT-09-0824

erican Association for Cancer Research.

er Ther; 9(2) February 2010

Researcon September 2mct.aacrjournals.org ownloaded from

(1–4), with evidence for both paracrine and autocrine ac-tivation of c-Met by HGF; expression of these moleculesin tumors and elevated concentrations of circulating HGFhave been associated with an increased risk of tumor me-tastasis and poor prognosis (5–9).Mutations in c-Met resulting in constitutive activation

have been described in hereditary and sporadic humancancers (10), and expression of some of these mutant al-leles in mice enhances their tumorigenicity (11, 12). Someof these c-Met mutants remain sensitive to HGF activa-tion, and can be blocked by HGF antagonists (10, 13).HGF is normally secreted from cells of mesenchymal

origin as a monomeric, inactive precursor (pro-HGF)with molecular mass of ∼90 kDa. Extracellular proteol-ysis at a site COOH-terminal to the fourth kringle do-main generates the fully active, disulphide-linkeddimeric form of HGF capable of activating signalingthrough binding to c-Met (1, 14–16). The NH2-terminalportion of HGF, the α-chain, contains the NH2-terminalhairpin domain and four kringle domains. Protein frag-ments derived from NH2-terminal regions of HGF (e.g.,NK1, NK2, and NK4) contain a high-affinity c-Met re-ceptor binding domain and can act as HGF agonistsor as antagonists, showing that HGF binds to c-Met

h. 5, 2020. © 2010 American Association for Cancer

http://mct.aacrjournals.org/

AMG 102 Neutralizes Hepatocyte Growth Factor


and can function through determinants in the α-chain(17, 18). The COOH-terminal β-chain of HGF is struc-turally homologous to serine proteases, but lacks pro-teolytic activity (1). The β-chain also interacts directlywith c-Met and seems to be critical for mature HGFsignaling through its receptor (19, 20).The importance of HGF/SF:c-Met signaling in cancer

has led to the development of specific inhibitors of thispathway that may provide novel therapeutic strategiesfor intervention in human disease (21–23). In previousstudies (24, 25), we reported on the biological activityof novel, fully human anti-HGF antibodies in both in vitroand in vivo models of cancer. The five antibodies ana-lyzed were all competitive inhibitors of HGF bindingto its receptor c-Met. Based on its in vitro and in vivoactivities, AMG 102 was selected as the clinical candi-date. Phase I clinical studies with AMG 102 have beencompleted (26), and phase II clinical trials in a varietyof solid tumor indications are underway. Here weprovide a detailed biochemical and functional character-ization of this molecule in support of its clinical develop-ment for treatment of cancers dependent on HGFsignaling. We show that AMG 102: (a) binds and neutra-lizes human and cynomolgus monkey HGF but notmouse, rat, or rabbit HGF; (b) preferentially binds tothe mature, active, heterodimeric form of HGF; (c) hasno apparent effect on the proteolytic processing leadingto HGF activation; and (d) interacts functionallywith regions of human HGF including the NH2-terminalportion of the β-chain.

Materials and Methods

Reagents and Cell LinesAMG 102 was prepared as previously described (24).

cDNA cloning and DNA sequencing of cynomolgusmonkey HGF (submitted to GenBank; full-length HGF =GQ477910; d5HGF =GQ477911) and rabbit HGF (submit-ted to GenBank; full length HGF = GQ477909) were doneusing standardmethods and an automated sequencer. Theamino acid sequences aligned with those of human andmouse full-length HGF are shown in SupplementaryFig. S1. The calculated percent similarities are shown inSupplementary Table S1. MAB294 (a mouse anti-humanHGF monoclonal antibody) and AF-294-NA (a goat anti-HGF polyclonal antibody) were purchased from R&DSystems. 93R1B9 (an anti-HGFmonoclonal antibody iden-tified by phage display that detects human and rabbitHGF equally well) and the rabbit anti-mouse HGF poly-clonal antibody were generated using standard immuni-zation procedures; both were prepared at Amgen Inc.Recombinant human HGF, human d5-HGF (a naturallyoccurring splice-variant of HGF with a five–amino aciddeletion of residues 163 to 167 in the α-chain; ref. 27),cynomolgus monkey d5-HGF, and mouse HGF wereexpressed by Chinese hamster ovary (CHO) cells andpurified with heparin sulfate affinity chromatographyand gel filtration. Rabbit HGF and chimeric and mutated

www.aacrjournals.org

Researcon September 2mct.aacrjournals.org Downloaded from

HGF constructs were expressed in human embryonickidney (HEK293) cells and collected from serum-freeconditioned medium without further purification. RatHGFwasobtained fromthe Institute of Immunology, Japan.

Equilibrium BindingTo evaluate the physical interaction between HGF and

AMG 102, a competition equilibrium binding methodwas developed. Analysis of AMG 102 was done on a Bia-core 3000. AMG 102 was immobilized on a research-gradeCM5 sensor chip via primary amine groups using theAmine Coupling Kit (Biacore Life Sciences), according tothe manufacturer's suggested protocol. Two different con-centrations (1 nmol/L and 3 nmol/L) of purified humand5-HGF or cynomolgus monkey d5-HGF were incubatedwith a dose range of AMG 102 (1 pmol/L to 25 nmol/L)in PBSwith 0.005% surfactant P-20 and 0.1mg/mL bovineserum albumin (BSA). Protein samples were incubated atroom temperature for at least 6 h to allow binding to reachequilibrium. The binding signal obtained after the sampleswere injected over the immobilized AMG 102 surface wasconsidered to be proportional to the freeHGF in solution atequilibrium. The equilibrium dissociation constants (KD)were obtained from nonlinear regression analysis of thecompetition curves using a dual-curve, one-site homoge-neous binding model (KinExA software, Sapidyne Instru-ments Inc.).

AMG102 Binding toHuman,Mouse, and Rabbit HGFProtein samples were resolved by reducing SDS-PAGE,

transferred to Hybond ECL membranes (Amersham), andblocked with TBS, 0.25% Tween 20, and 5% BSA. The pri-mary blotting antibody was either AMG 102 or 93R1B9 asindicated. The secondary antibodywas horseradish perox-idase–conjugated antigoat polyclonal antibody (1:5,000 di-lution). The membranes were washed and the complexeswere detected using electrochemiluminescence.

Immunoprecipitation StudiesHuman d5-HGF (1 μg) was mixed in an ∼1:1 molar ra-

tio with 2 μg of either AMG 102 or MAB294 in 1 mL of0.1% BSA-PBS buffer and incubated for 2 h at room tem-perature. The antibody-antigen complexes were capturedon Protein G beads (Pierce), washed with PBS, andpelleted. The immunoprecipitated proteins were elutedwith 2× SDS-PAGE loading buffer containing 5%β-mercaptoethanol. The mixture was boiled, and the elut-ed, immunoprecipitated proteins were separated and im-munoblotted as above. The primary antibody was goatanti-HGF polyclonal antibody AF-294-NA (100 ng/mL),and the secondary antibody was horseradish peroxi-dase–conjugated antigoat polyclonal antibody (1:5,000).The membrane was washed and the immune-precipitatedcomplexeswere detected using electrochemiluminescence.

HGF Proteolysis/ProcessingA preliminary experiment determined that AMG 102

was resistant to kallikrein proteolysis (data not shown).

Mol Cancer Ther; 9(2) February 2010 401



Burgess et al.

402


Purified human d5-HGF was preincubated for 20 min atroom temperature with or without AMG 102 (1:1 and5:1 molar ratios of antibody:HGF were tested) to allowcomplex formation prior to the addition of kallikrein.When included together, kallikrein and HGF were addedin a 1:10 (weight:weight) ratio, and reactions were fur-ther incubated at 37°C for 0, 4, or 12 h. Results were an-alyzed by SDS-PAGE and Coomassie Blue staining.

Autophosphorylation StudiesElectrochemiluminescence assays to determine the

phosphorylation status of c-Met were done as described(24) using human PC3 cells (a human prostate cancer linethat expresses c-Met but not HGF) treated with serial di-lutions of AMG 102 that had been preincubated with200 ng/mL human d5-HGF or cynomolgus monkey d5-

Mol Cancer Ther; 9(2) February 2010


HGF. Percent inhibition was calculated relative to the re-sults for cells stimulated with HGF in the absence of AMG102. The IC50 values were determined using the 4-param-eter fit equation in XLfit (IDBS).

Cell Invasion AssaysCell invasion assays were done as previously de-

scribed (24). Cynomolgus monkey d5-HGF (50 ng/mL)was used as the attractant for human breast carcinomaMDA-MB-435 cells. In one experiment (Fig. 2C), ratHGF (50 ng/mL) was used as the attractant for themigration of mouse mammary tumor 4T1 cells. In thisexperiment, rabbit anti-mouse HGF polyclonal anti-body was employed as an inhibitory control. Datarepresent mean ± SE, n = 3. Data are representative oftwo experiments.

Figure 1. Binding specificity of AMG 102. AMG 102 (1 pmol/L to 25 nmol/L) was incubated with two concentrations (▪, 1 nmol/L;⋄, 3 nmol/L) of human(A) or cynomolgus monkey (B) d5-HGF. The fraction of free HGF in solution at equilibrium was measured by Biacore, and the values were used tocalculate affinity constants (KD). C, recombinant human, rabbit, or mouse HGF was separated by nonreducing SDS-PAGE, transferred to membranes,and probed with either 93R1B9 (anti-pan HGF, top) or AMG 102 (bottom). Top, lanes 1 to 5, 25, 50, 100, 250, and 500 ng purified human d5-HGF;lanes 6 to 8, 5, 15, and 30 μL conditioned medium from CHO cells stably expressing rabbit HGF; lane 9, 30 μL conditioned medium from human HEK293cells transiently expressing rabbit HGF. Bottom, lanes 1 to 6, 10, 25, 50, 100, 100, and 250 ng purified human d5-HGF; lanes 7 to 10, 10, 100,250, and 500 ng purified mouse HGF; lane 11, 20 μL conditioned medium from CHO cells expressing rabbit HGF; lane 12, culture medium alone.D, the preparation of purified human HGF used in this experiment contained approximately equivalent amounts of mature HGF and pro-HGF. Afterimmunoprecipitation of HGF, the samples were separated by reducing SDS-PAGE and analyzed with immunoblotting using goat anti-human HGFpolyclonal (AF-294-NA), which detects pro-HGF as well as the larger, α-chain of mature-HGF. Lane 1, purified recombinant human HGF withoutimmunoprecipitation; lane 2, recombinant human HGF immunoprecipitated with a 1:1 molar ratio of AMG 102; lane 3, recombinant human HGFimmunoprecipitated with a 1:1molar ratio of MAB 294. E, humanHGFwas incubatedwith or without AMG102 at a 1:5molar ratio, andwith or without kallikreinprotease. After incubation, the samples were analyzed with reducing SDS PAGE, followed by Coomassie Blue staining. Lanes 7 and 10 (arrowheads)contain Seablue molecular weight standards. Pro-HGF and mature HGF as well as the heavy and light chains of AMG 102 are as indicated.

Molecular Cancer Therapeutics





Epitope MappingHuman-mouse chimeric HGF-avidin fusion proteins

were constructed using standard recombinant DNA andPCR methodologies as described by Burgess et al. in 2006(24). As shown in Table 1 and Supplementary Table S2,chimeras 1, 2, and 3 comprised a portion of the NH2-terminus of mouse HGF linked to the COOH-terminusof human HGF. Chimera 7 comprised the NH2- andCOOH-terminus of human HGF, with the internal aminoacid residues from the mouse β-chain (because of thestrong sequence identity between human and mouseHGF, there are only nine murine residues in chimera 7;Fig. 3B). Chimera 8 mirrored chimera 7, with the mouseand human sequences interchanged, thus the only human



HGF residues present in chimera 8 were from the NH2-terminal portion of the β-chain (only seven unique humanresidues and a two–amino acid deletion relative to themouse sequence are present in chimera 8). We also gener-ated five humanHGF-avidin constructs, each containing asingle, nonconservative amino acid substitution fromthe mouse HGF sequence: a human HGF construct con-taining a two–amino acid insertion from mouse HGF,and amouse HGFwith the same two–amino acid deletion(Table 1, Supplementary Table S2, and Fig. 3).Each avidin-HGF fusion-chimeric and point-mutated

protein was transiently expressed in HEK293 cells.Conditioned media were incubated with BSA-blocked,biotin-coated beads and stained with FITC-labeled

Figure 2. Functional specificity of AMG 102. Human PC3 cells were stimulated with human (A) or cynomolgus monkey (B) d5-HGF to induceautophosphorylation of endogenous c-Met receptors. AMG 102 (0.01–200 nmol/L) was preincubated with HGF (200 ng/mL) and then added to the cellsfor 10 min. Phosphorylation was quantified using a plate-based electrochemiluminescence assay. The percent inhibition was calculated relative toreactions lacking AMG 102. C, MDA-MB-435 human breast carcinoma cells were stimulated with cynomolgus monkey d5-HGF and migration throughFluoroBlok Matrigel inserts was measured. There were three reactions: no HGF (negative control), 50 ng/mL HGF, and HGF plus 10 μg/mL AMG 102.After incubation for 18 h, the inserts were stained with ER-Tracker Green, washed, and read. Insets, representative fluorescent images of inserts containingcells stimulated with cynomolgus HGF with and without AMG 102. D, the same methods were used to evaluate the migration of mouse 4T1 cells stimulatedwith rat HGF. NS, not significantly different from HGF alone. *, P = 0.003 compared with HGF alone. Data were analyzed with ANOVA using StatViewsoftware v5.0.1.




Burgess et al.

404


anti-avidin antibodies (Vector Lab) to evaluate proteinexpression level. The fluorescent probe allowed normali-zation of the amount of protein expression among differ-ent samples (24).To evaluate AMG 102 binding to the chimeric and

mutated HGF proteins, a second fluorescent probe, phyco-erythrin (PE)-labeled goat anti-human F(ab')2, was used.The presence of both FITC (avidin/HGF) and PE (AMG102) labeling, detected on a Becton Dickinson BioscienceFACScan (BD), indicated that AMG 102 bound to the HGFportion of the fusion construct. Single-color fluorescence-activated cell sorting (FACS) histograms without nor-malizationwere generated from the dual-labeled data sets.Biotin-coated beads complexed with avidin alone wereused as a negative control, enabling detection of shifts influorescence associated with AMG 102 binding. Becausethe spectral properties of FITC and PE overlap, controlsamples labeled with FITC only or PE only were analyzedand fluorescence signals were adjusted accordingly to en-hance sensitivity and to facilitate measurement of bindingwith constructs that were expressed at low levels.

Protease Protection StudiesHuman HGF was preincubated with and without

AMG 102 to allow complex formation (∼1:1 molar ratioof antibody:HGF was tested). Digestion with trypsin wascarried out at 37°C for 1 h (weight:weight trypsin:HGF =1:30). The digested material from each sample was sub-jected to reverse-phase high-performance liquid chroma-



tography (HPLC) for peptide separation. The HPLCmaps were compared to identify trypsin cleavage sitesprotected by AMG 102 binding. NH2-terminal protein se-quencing was done to identify the protected peptides.A complementary experiment was designed to isolate

and identify the AMG 102-binding peptides of HGF. TheHGF-AMG 102 complex was digested as above and thensubjected to Microcon centrifugal filtration (Millipore)with a cutoff at a molecular weight of 10 kDa to captureAMG 102 with the bound peptides, while the unboundpeptides flowed through the filter. AMG 102 was re-leased from the bound peptides captured on the filterby addition of intact human HGF (HGF elution). Themixture of HGF-bound AMG 102 and free peptideswas incubated overnight at 4°C and was again filteredto collect the AMG 102–binding peptides. These peptidesand those eluted at the first filtration step were analyzedwith reverse-phase HPLC. Bound peptides were isolatedand subjected to NH2-terminal sequencing.

Results

Binding Specificity of AMG 102A competition bindingmethodwas used to evaluate the

affinity of AMG 102 for human and cynomolgus monkeyd5-HGF. The equilibrium dissociation constants (KD) wereobtained from nonlinear regression analysis of the compe-tition curves (Fig. 1A and B) using a dual-curve, one-site

Table 1. AMG 102 binding results for chimeric and mutant HGF molecules

HGF chimera or mutantdesignation

(Domains or amino acid numbers)

Molecular Can

h. 5, 2020. © 2010 American Association

AMG 102binding

Human HGF
Mouse HGF
Human full length*
N, K1, K2, K3, K4, β None Yes Mouse full length† None N, K1, K2, K3, K4, β No Ch 1 HGF K1, K2, K3, K4, β N-domain Yes Ch 2 HGF K2, K3, K4, β N, K1 Yes Ch 3 HGF K3, K4, β N, K1, K2 Yes Ch 7 HGF N , K1, K2, K3, K4, β (586 to 728) β 507 to 585 No Ch 8 HGF β (507 to 585) N, K1, K2, K3, K4, β (586 to 728) Yes G555E HGF N, K1, K2, K3, K4, β‡ 555‡ No C561R HGF N, K1, K2, K3, K4, β‡ 561‡ No D592N HGF N, K1, K2, K3, K4, β‡ 592‡ Yes N601S HGF N, K1, K2, K3, K4, β‡ 601‡ Yes R647Q HGF N, K1, K2, K3, K4, β‡ 647‡ Yes Human NK (insertion) HGF N, K1, K2, K3, K4, β‡ 2 amino acid insertion of mouse 540 to
541‡
No
Mouse NK (deletion) HGF
None N, K1, K2, K3, K4, β (540, 541 deleted)‡ No
Abbreviations: N, amino terminus; K1, K2, K3, and K4, the four conserved kringle domains of HGF; NK, insertion or deletion of Asn-Lys; Ch, chimera.*Human HGF sequence, GenBank NM_000601.†Mouse HGF sequence, GenBank NM_010427.‡Mutationally altered domain or amino acid substitution.

cer Therapeutics

for Cancer




Figure 3. Epitope mapping of AMG 102 binding to HGF. A, single-color FACS histograms for binding of human/mouse chimeric and point-mutatedHGF to AMG 102. PE fluorescence is a measure of AMG 102 binding to the various HGF proteins captured on avidin-coated beads (red) or beads withavidin alone as negative controls (black). Each human/mouse hybrid HGF construct is described in Table 1 and Supplementary Table S2. B, HGFwas subjected to partial trypsin digestion without AMG 102 (first panel) or with bound AMG 102 (second panel), and peptides were analyzed byreverse-phase HPLC. AMG 102-bound HGF was partially digested with trypsin and then subjected to filtration to separate the unbound (third panel) fromthe bound peptides (fourth panel). Prior to analysis, AMG 102 was removed from the bound peptides by incubation with intact HGF. C, summary ofAMG 102 epitope mapping. The amino acid sequences of the mouse and human HGF β-chain are aligned. Red boxes, both conservative andnonconservative differences between the two sequences; green box, a region of human HGF implicated in AMG 102 binding. Red stars (both closedand open), residues that were investigated using the point-mutation strategy; closed stars, residue changes that disrupted AMG 102 binding tohuman HGF.

Mol Cancer Ther; 9(2) February 2010www.aacrjournals.org 405

Research. on September 25, 2020. © 2010 American Association for Cancermct.aacrjournals.org Downloaded from


Burgess et al.

406


homogeneous binding model. AMG 102 showed KD

values of 19 pmol/L for human d5-HGF and 41 pmol/Lfor cynomolgus monkey d5-HGF.To investigate the interaction of AMG 102 with HGF

of other species, recombinant HGF molecules wereseparated, blotted, and probed with 93R1B9, an anti-HGF monoclonal antibody that detects human andrabbit HGF equally well (Fig. 1C, top), or withAMG 102 (Fig. 1C, bottom). Antibody 93R1B9 de-tected both human HGF (top, lanes 1 to 5) and rabbitHGF (top, lanes 6 to 8). AMG 102 detected as little as10 ng of human HGF (bottom, lanes 1 to 6) but didnot detect up to 500 ng of either mouse HGF (bottom,lanes 7 to 10) or rabbit HGF (bottom, lane 11).To investigate the specificity of AMG 102 for pro-HGF

and mature HGF, we did immunoprecipitation reactionscontaining ∼1:1 molar ratio of AMG 102 to total HGF.The CHO-derived preparation of HGF contained a mix-ture of both the mature and pro forms of HGF, whichseparate on a reducing gel as shown (Fig. 1D, lane 1). Ma-ture HGF was the predominant form immunoprecipi-tated by AMG 102 under these conditions (lane 2),whereas both forms were immunoprecipitated using acommercial monoclonal anti-HGF antibody MAB294(lane 3). At higher molar ratios (e.g., 5:1) of AMG 102:HGF, both mature HGF and pro-HGF were immunopre-cipitated (data not shown).We next investigated whether AMG 102 alters proteo-

lytic processing of pro-HGF to mature HGF under con-ditions of excess AMG 102 (at a 5:1 molar ratio; a 1:1molar ratio was also tested; data not shown). Pro- andmature forms of HGF were stable throughout the 12-hour incubation period in the absence (Fig. 1E, lanes 1to 3) and the presence of AMG 102 alone (Fig. 1E, lanes4 to 6; note the prominent antibody heavy and lightchains of AMG 102 in lanes 4 to 5 and 12 to 14). Incu-bation with kallikrein alone led to a time-dependentconversion of pro-HGF to the mature form that wascomplete by 12 hours (Fig. 1E, lanes 8, 9, and 11). Incu-bation with AMG 102 had no apparent impact on therate or extent of pro-HGF processing catalyzed by kalli-krein (Fig. 1E, lanes 12 to 14). Thus, AMG 102 bindingto pro-HGF did not seem to enhance or inhibit proteo-lytic processing of human HGF.

Functional Specificity of AMG 102The ability of AMG 102 to inhibit c-Met autophosphor-

ylation in response to human and cynomolgus monkeyHGF was tested in PC3 cells. AMG 102 treatment led topotent and complete inhibition of c-Met phosphoryla-tion induced by either human or cynomolgus monkeyHGF, with calculated IC50 values of 0.12 nmol/L or0.24 nmol/L (Fig. 2A and B).The functional activity of AMG 102 was further stud-

ied using Matrigel cell-invasion assays. Treatment ofhuman MDA-MB-435 cells with cynomolgus monkeyd5-HGF (50 ng/mL for 18 hours) stimulated theirmigration through a Matrigel matrix-coated membrane.



Incubation with 10 μg/mL AMG 102 completelyinhibited monkey HGF–induced migration (Fig. 2C).Consistent with the binding studies showing lack of in-teraction between AMG 102 and nonprimate HGF, amigration assay employing mouse 4T1 cells stimulatedwith rat HGF revealed no inhibition by AMG 102 atconcentrations up to 10 μg/mL (Fig. 2D, left). In con-trast, migration of these cells was significantly inhibited(P = 0.003) by a polyclonal rabbit anti-mouse HGFantibody (Fig. 2D, right).

Epitope Mapping of AMG 102 Binding SitesAlthough the lack of cross-reactivity of AMG 102 for

rodent HGF precluded the use of rodents in safety stud-ies, we took advantage of the species specificity of AMG102 to probe its binding sites on human HGF. As in ourprevious studies (24), we employed engineered human/mouse chimeric HGF fusion proteins (Table 1, Supple-mentary Table S2) and a novel two-color FACS methodfor these experiments. FACS results are shown inFig. 3A and are summarized in Table 1. As was shownin our previous experiments, AMG 102 bound to fullyhuman HGF but not to fully mouse HGF (24). AMG102 also bound to chimeras 1, 2, and 3, which are humanHGF molecules containing increasing lengths of the NH2-terminal α-chain of mouse HGF including the N, K1, andK2 domains. These data suggested that AMG 102 wasbinding predominantly to the β-chain of human HGF.Most significantly, AMG 102 bound to chimera 8, amouse HGF containing only seven unique human aminoacids and a two–amino acid deletion relative to themouse HGF sequence (Fig. 3A). In contrast, AMG 102did not bind to chimera 7, a human HGF containing onlynine mouse HGF-specific amino acids from an NH2-terminal region of the β-chain. These data suggested thatthe key binding epitope for AMG 102 comprised se-quences including one or more of the indicated residues.To further refine our analysis, we identified the non-

conservative amino acid differences between the humanand mouse HGF sequences within the NH2-terminal por-tion of the HGF β-chain and generated five constructs,each of which contained a point mutation that changeda human residue to a mouse residue (Table 1, Supple-mentary Table S2, Fig. 3C). Substitutions at position 555(G555E HGF) or 561 (C561R HGF) completely inhibitedbinding (Fig. 3A). Substitutions at 592 (D592N HGF), 601(N601S HGF), or 647 (R647Q HGF) had little effect onbinding between HGF and AMG 102. Insertion of twomouse-specific residues, 540N and 541K (Fig. 3C), whichare not present in the human HGF sequence, completelydisrupted binding [human NK (insertion) HGF; Table 1,Fig. 3A]. However, deletion of these same two residuesfrom the mouse sequence did not restore AMG 102 bind-ing [mouse NK (deletion) HGF; Table 1, Fig. 3A]. Thesedata identified key residues within the β-chain of humanHGF that seemed to be critical for AMG 102 binding.Finally, to physically map the interaction between

AMG 102 and human HGF, we identified trypsin






cleavage sites in HGF that were protected by AMG 102binding. As the comparison of the two upper HPLCtraces in Fig. 3B indicate, AMG 102 binding to humanHGF protected two peptide-containing peaks, designat-ed T33 and T38.6. Peak T38.6 contained two peptideswith NH2-terminal sequences of VVNGIPTR andGIPTRT corresponding to sequences beginning at resi-due 495 or 498, respectively, at the NH2-terminus ofthe β-chain of mature HGF (Fig. 3C). Based on an ob-served mass of peptides 7165 and 6878 (determined bymass spec), the likely protected trypsin cleavage site isR556 (calculated masses for the two predicted peptideswere 7152 and 6840, respectively). The NH2-terminalsequence of T33 (VNTADQ) corresponded to a regionwithin the α-subunit of mature HGF (beginning at res-idue 65 of the NH2 domain of HGF; SupplementaryFig. S1). The interaction of AMG 102 with the α-sub-unit of HGF was not confirmed by other methods.However, with the use of a complementary approachto directly capture AMG 102-binding HGF peptides,the same two NH2-terminal β-chain sequences wereidentified (contained in peak T48; Fig. 3B, bottom).AMG 102-bound HGF was trypsin digested and un-bound peptides were separated from bound peptidesby filtration. Bound peptides were eluted from AMG102 with intact HGF and separated by HPLC (Fig. 3B,bottom). NH2-terminal protein sequencing showed thatthe VVNGIPTR and GIPTRT peptides were containedin peak T48. Three other peaks (*, Fig. 3B, bottom)contained either no peptide or a peptide of unknownorigin, unrelated to HGF or AMG 102. Taken together,the epitope mapping data strongly indicated thatAMG 102 bound to the NH2-terminus of the β-chainof HGF, specifically including residues from Valine495 or Glycine 498 to Arginine 556 of human HGF.

Discussion

In the present study, we extended the characterizationof fully human anti-HGF–neutralizing monoclonal anti-bodies described by Burgess et al. (24) by focusing onthe most promising candidate for clinical investigation,now called AMG 102 (25). AMG 102 bound to humanand cynomolgus monkey d5-HGF with high affinity(KD = 19 pmol/L and 41 pmol/L, respectively) but didnot bind to rodent or rabbit HGF. AMG 102 preferentiallybound to the mature, heterodimeric form of human HGF,and binding had no apparent effect on the proteolyticprocessing that leads to pro-HGF activation. AMG 102completely inhibited in vitro c-Met autophosphorylationstimulated by human and by cynomolgus monkeyHGF in cells with IC50 values of 0.12 nmol/L and 0.24nmol/L, respectively. AMG 102 inhibited cynomolgusmonkey d5-HGF–stimulated migration of humanMDA-MB-435 cells through Matrigel but had no such ef-fect on mouse 4T1 cells stimulated with rat HGF. Epitopemapping studies revealed that AMG 102 bound directly



to amino acid residues at the NH2-terminus of the humanHGF β-chain of human HGF.The species specificity of AMG 102 binding and activ-

ity is reflected in the extent of similarity in amino acidsequence among the HGF molecules. Although the muta-tional epitope mapping done in this study is not exhaus-tive, binding of AMG 102 was disrupted by substitutionof only one or two amino acids from the human sequencewith those from mouse HGF. Furthermore, the changesthat disrupted binding of AMG 102 were all within theregion of the β-chain defined by the studies employingchimeric molecules. Substitutions outside of this region(592, 601, and 647) did not disrupt AMG 102 binding. Fi-nally, protection from trypsin cleavage and identificationof AMG 102-bound human HGF peptides point to a dis-crete region at the NH2-terminus of mature human HGFβ-chain that is both necessary and sufficient for bindingof AMG 102 to human HGF.The functional interaction of HGF and c-Met involves

multiple points of physical contact, including high-affinitybinding between c-Met and the α-chain of HGF (28, 29)and themore recently described low-affinity, but function-ally critical, binding between the β-chain of HGF andthe Sema-domain of c-Met (19, 20). AMG 102 seemed tobe unique in its ability as a single agent to completelyneutralize human HGF binding to c-Met and to fully in-hibit in vitro and in vivo functional assays (24); it had beensuggested that at least three antibodies with overlappingepitopes are required for full activity (30). More recently,other neutralizing anti-HGF mouse monoclonal antibo-dies have been described (31); however, characterizationof the binding epitope has not been reported. Our studiesprovide evidence that the fully human monoclonal anti-body AMG 102 bound predominantly near the NH2-ter-minal portion of the β-chain of human HGF. Thebinding of AMG 102 to this epitope on HGF seems toconfer excellent functional neutralizing properties to thisclinical candidate.The lack of binding and functional activity of AMG

102 on mouse, rat, or rabbit HGF precluded the use ofthese species for preclinical studies, including efficacyand safety evaluation of the antibody. As we have pre-viously reported (24), the efficacy of AMG 102 wasshown only in preclinical models that were dependenton human HGF. Kakkar et al. (32) did cross-reactivitystudies using fluorescein-conjugated AMG 102 on awide variety of human and cynomolgus monkey tissuesas well as on a smaller panel of mouse, rat, and rabbittissues. The results were consistent with the speciesspecificity of AMG 102 binding/activity seen in ourin vitro experiments. Based on the species specificityshown here and by Kakkar et al., cynomolgus monkeyswere deemed to be a pharmacologically relevant speciesfor further pharmacokinetic and safety studies of AMG102 (32).The therapeutic consequences of preferential binding of

AMG 102 for the active, heterodimeric form of HGF arenot known. However, a preexisting store of pro-HGF




Burgess et al.

408


andmature HGF is expected to be present in many tissuesand the tumors of cancer patients. It is possible that theapparently higher affinity of AMG 102 for processedHGF would enhance its potential anticancer properties.Stimulating the conversion of pro-HGF to mature, activeHGF would be counter to the desired clinical effect, i.e.,inhibition of mature, active HGF; therefore, showing thatAMG 102 binding to HGF does not enhance proteolyticprocessing provides yet another positive characteristicof AMG 102.AMG 102 is a fully human, neutralizing antibody to

human and nonhuman primate HGF. It bound HGF withlow pmol/L affinity and neutralized cell activity at lownanomolar concentrations. We recently reported datafrom the AMG 102 first in human/phase I study (Gordonet al in press, Clinical Cancer Research). The concentra-tion of AMG 102 in plasma was in stoichiometric excesscompared with circulating HGF levels, even in the low-est-dose group (>100-fold molar excess at 0.5 mg/kg).Epitope mapping showed that AMG 102 bound pri-marily to mature human HGF via the β-chain. Cur-rently, AMG 102 is in multiple, controlled phase IIclinical trials in a variety of solid tumor indications. Treat-



ment regimens include AMG 102 in combination withchemotherapy or other targeted agents in gastric, pros-tate, colorectal, small cell lung, and other solid tumors.

Disclosure of Potential Conflicts of Interest

All authors are current or former employees of and own stock inAmgen, Inc.

Acknowledgments

We thank Shu-Yin Ho, Frederick Jacobsen, Helen Kim, RohiniDeshpande, Francis Martin, Hsien Lu, Rashid Syed, and Steve Jordanfrom Amgen for their contributions to this study, and Martin Smith(funded by Amgen Inc.) and Kathryn Boorer (Amgen Inc.) forassistance with writing the manuscript.

Grant Support

Amgen, Inc.The costs of publication of this articlewere defrayed in part by the payment

of page charges. This article must therefore be hereby marked advertisementin accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received 9/2/09; revised 11/6/09; accepted 12/6/09; publishedOnlineFirst 2/2/10.

References
1. Birchmeier C, Birchmeier W, Gherardi E, Vande Woude GF. Met,
metastasis, motility and more. Nat Rev Mol Cell Biol 2003;4:915–25.

2. Thiery JP. Epithelial-mesenchymal transitions in tumour progression.Nat Rev Cancer 2002;2:442–54.

3. Bottaro DP, Liotta LA. Cancer: out of air is not out of action. Nature2003;423:593–5.

4. Trusolino L, Comoglio PM. Scatter-factor and semaphorin recep-tors: cell signalling for invasive growth. Nat Rev Cancer 2002;2:289–300.

5. Yamashita J, Ogawa M, Yamashita S, et al. Immunoreactive hepato-cyte growth factor is a strong and independent predictor of recur-rence and survival in human breast cancer. Cancer Res 1994;54:1630–3.

6. Ghoussoub RA, Dillon DA, D'Aquila T, Rimm EB, Fearon ER, RimmDL. Expression of c-met is a strong independent prognostic factor inbreast carcinoma. Cancer 1998;82:1513–20.

7. Ueki T, Fujimoto J, Suzuki T, Yamamoto H, Okamoto E. Ex-pression of hepatocyte growth factor and its receptor c-metproto-oncogene in hepatocellular carcinoma. Hepatology 1997;25:862–6.

8. Daveau M, Scotte M, Francois A, et al. Hepatocyte growth factor,transforming growth factor alpha, and their receptors as combinedmarkers of prognosis in hepatocellular carcinoma. Mol Carcinog2003;36:130–41.

9. Han SU, Lee JH, Kim WH, Cho YK, Kim MW. Significant correlationbetween serum level of hepatocyte growth factor and progression ofgastric carcinoma. World J Surg 1999;23:1176–80.

10. Dharmawardana PG, Giubellino A, Bottaro DP. Hereditary papillaryrenal carcinoma type I. Curr Mol Med 2004;4:855–68.

11. Graveel C, Su Y, Koeman J, et al. Activating Met mutations produceunique tumor profiles in mice with selective duplication of the mutantallele. Proc Natl Acad Sci U S A 2004;101:17198–203.

12. Jeffers M, Fiscella M, Webb CP, Anver M, Koochekpour S, VandeWoude GF. The mutationally activated Met receptor mediatesmotility and metastasis. Proc Natl Acad Sci U S A 1998;95:14417–22.

13. Michieli P, Basilico C, Pennacchietti S, et al. Mutant Met-mediated

transformation is ligand-dependent and can be inhibited by HGFantagonists. Oncogene 1999;18:5221–31.

14. Naldini L, Tamagnone L, Vigna E, et al. Extracellular proteolyticcleavage by urokinase is required for activation of hepatocyte growthfactor/scatter factor. EMBO J 1992;11:4825–33.

15. Mukai S, Fukushima T, Naka D, Tanaka H, Osada Y, Kataoka H.Activation of hepatocyte growth factor activator zymogen (pro-HGFA) by human kallikrein 1-related peptidases. FEBS J 2008;275:1003–17.

16. Lokker NA, Mark MR, Luis EA, et al. Structure-function analysis ofhepatocyte growth factor: identification of variants that lack mitogen-ic activity yet retain high affinity receptor binding. EMBO J 1992;11:2503–10.

17. Matsumoto K, Nakamura T. NK4 (HGF-antagonist/angiogenesis in-hibitor) in cancer biology and therapeutics. Cancer Sci 2003;94:321–7.

18. Lokker NA, Godowski PJ. Generation and characterization of acompetitive antagonist of human hepatocyte growth factor, HGF/NK1. J Biol Chem 1993;268:17145–50.

19. Stamos J, Lazarus RA, Yao X, Kirchhofer D, Wiesmann C. Crystalstructure of the HGF β-chain in complex with the Sema domain ofthe Met receptor. EMBO J 2004;23:2325–35.

20. Kirchhofer D, Yao X, Peek M, et al. Structural and functional basis ofthe serine protease-like hepatocyte growth factor β-chain in Metbinding and signaling. J Biol Chem 2004;279:39915–24.

21. Knudsen BS, Vande Woude G. Showering c-MET-dependent can-cers with drugs. Curr Opin Genet Dev 2008;18:87–96.

22. Comoglio PM, Giordano S, Trusolino L. Drug development of METinhibitors: targeting oncogene addiction and expedience. Nat RevDrug Discov 2008;7:504–16.

23. Migliore C, Giordano S. Molecular cancer therapy: can our expecta-tion be MET? Eur J Cancer 2008;44:641–51.

24. Burgess T, Coxon A, Meyer S, et al. Fully human monoclonal antibo-dies to hepatocyte growth factor with therapeutic potential againsthepatocyte growth factor/c-Met-dependent human tumors. CancerRes 2006;66:1721–9.

25. Jun HT, Sun J, Rex K, et al. AMG 102, a fully human anti-hepatocytegrowth factor/scatter factor neutralizing antibody, enhances the






efficacy of temozolomide or docetaxel in U-87 MG cells and xeno-grafts. Clin Cancer Res 2007;13:6735–42.

26. Gordon MS, Sweeney CS, Mendelson DS, et al. Safety, pharmacoki-netics, and pharmacodynamics of AMG 102, a fully human hepatocytegrowth factor-neutralizing monoclonal antibody, in a first-in-humanstudy of patients with advanced solid tumors. Clin Cancer Res2010;16:699–710.

27. Rubin JS, Chan AM, Bottaro DP, et al. A broad-spectrum human lungfibroblast-derived mitogen is a variant of hepatocyte growth factor.Proc Natl Acad Sci U S A 1991;88:415–9.

28. Lokker NA, Presta LG, Godowski PJ. Mutational analysis and molec-ular modeling of the N-terminal kringle-containing domain of hepato-cyte growth factor identifies amino acid side chains important forinteraction with the c-Met receptor. Protein Eng 1994;7:895–903.



29. Okigaki M, Komada M, Uehara Y, Miyazawa K, Kitamura N. Func-tional characterization of human hepatocyte growth factor mutantsobtained by deletion of structural domains. Biochemistry 1992;31:9555–61.

30. Cao B, Su Y, Oskarsson M, et al. Neutralizing monoclonal antibodiesto hepatocyte growth factor/scatter factor (HGF/SF) display anti-tumor activity in animal models. Proc Natl Acad Sci U S A 2001;98:7443–8.

31. Kim KJ, Wang L, Su YC, et al. Systemic anti-hepatocyte growth fac-tor monoclonal antibody therapy induces the regression of intracra-nial glioma xenografts. Clin Cancer Res 2006;12:1292–8.

32. Kakkar T, Ma M, Zhuang Y, Patton A, Hu Z, Mounho B. Pharmacoki-netics and safety of a fully human hepatocyte growth factor antibody,AMG 102, in cynomolgus monkeys. Pharm Res 2007;24:1910–8.




2010;9:400-409. Published OnlineFirst February 2, 2010.Mol Cancer Ther Teresa L. Burgess, Jan Sun, Susanne Meyer, et al. Primate Hepatocyte Growth FactorHuman Monoclonal Antibody to Human and Nonhuman Biochemical Characterization of AMG 102: A Neutralizing, Fully

Updated version

10.1158/1535-7163.MCT-09-0824doi:

Access the most recent version of this article at:

Material

Supplementary

http://mct.aacrjournals.org/content/suppl/2010/02/02/1535-7163.MCT-09-0824.DC1

Access the most recent supplemental material at:

Cited articles

http://mct.aacrjournals.org/content/9/2/400.full#ref-list-1

This article cites 32 articles, 12 of which you can access for free at:

Citing articles

http://mct.aacrjournals.org/content/9/2/400.full#related-urls

This article has been cited by 7 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mct.aacrjournals.org/content/9/2/400To request permission to re-use all or part of this article, use this link

Research. on September 25, 2020. © 2010 American Association for Cancermct.aacrjournals.org Downloaded from


http://mct.aacrjournals.org/lookup/doi/10.1158/1535-7163.MCT-09-0824

http://mct.aacrjournals.org/content/suppl/2010/02/02/1535-7163.MCT-09-0824.DC1

http://mct.aacrjournals.org/content/9/2/400.full#ref-list-1

http://mct.aacrjournals.org/content/9/2/400.full#related-urls

http://mct.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://mct.aacrjournals.org/content/9/2/400


biochemical characterization of amg 102: a neutralizing ... · abstract amg 102 is a fully human...

Documents