Therapeutic Mechanism and Efficacy of the Antibody–Drug Conjugate

Published OnlineFirst December 6, 2011; DOI: 10.1158/1535-7163.MCT-11-0523

Molecular Cancer Therapeutic Discovery Therapeutics

Therapeutic Mechanism and Efﬁcacy of the Antibody–Drug Conjugate BAY 79-4620 Targeting Human Carbonic Anhydrase 9

Heike M. Petrul, Christoph A. Schatz, Charlotte C. Kopitz, Lila Adnane, Timothy J. McCabe, Pamela Trail, Sha Ha, Yong S. Chang, Andrei Voznesensky, Gerald Ranges, and Paul P. Tamburini

Abstract Carbonic anhydrase IX (CAIX) is a cell surface glycoprotein that is expressed in many different tumors and yet restricted in normal tissues to the gastrointestinal tract. It is upregulated by hypoxia and correlates with tumor grade and poor survival in several tumor indications. Monoclonal antibodies (mAb) with single digit nanomolar binding affinity for CAIX were derived by panning with the recombinant ectodomain of CAIX against the MorphoSys HUCAL Gold library of human Fabs. Highest affinity Fabs were converted to full-length IgGs and subjected to further characterization based upon their avidity and selectivity for CAIX, their capacity to undergo internalization in CAIX-expressing cell lines, and their selective localization to CAIX-positive human xenografted tumors when administered to mice as fluorescent conjugates. Through this selection process, the 3ee9 mAb was identified, which upon conjugation to monomethyl auristatin E through a self-immolative enzyme-cleavable linker yielded the potent and selective CAIX antibody–drug conjugate CAIX-ADC (BAY 79-4620). In preclinical human xenograft models in mice representing several tumor indications, BAY 79-4620 showed potent antitumor efficacy and in some models showed partial and complete tumor shrinkage even following a single dose. The mechanism of action was shown by histology to involve the sequelae of events typical of antitubulin agents. Efficacy in murine preclinical models correlated semiquantitatively, with CAIX expression levels as determined by immunohistochemistry and ELISA. These preclinical data collectively support the development of BAY 79-4620 for the treatment of cancer patients with CAIX overexpressing tumors. Mol Cancer Ther; 11(2); 340–9. 2011 AACR.

Introduction to the tumor marker G250 (named after the antibody with which it was identified) published by Oosterwijk Carbonic anhydrase IX (CAIX) was first identified and colleagues (3). by Pastorekova as an endogenous HeLa cell antigen CAIX is a cell surface glycoprotein that is expressed in that was recognized by the antibody M75 raised carcinomas of several histologic types, including a strik- against this cervix carcinoma cell line (1). Although ingly high proportion of renal cell carcinomas (4, 5), initially referred to as MN antigen, it was shortly carcinomas of the esophagus (4, 6), cervical carcinomas thereafter identified as a new carbonic anhydrase by (4, 7), malignant colon carcinomas (4, 8), non–small cell thesamegroup(2)andwasconfirmedtobeidentical lung carcinomas (NSCLC; refs. 4, 9), and, to a lesser degree, breast carcinomas (4, 10). By contrast, the expression of CAIX on normal tissues is largely restricted to the Authors' Affiliation: Bayer HealthCare AG, Berlin, Germany apical surface of cells of the stomach, bile duct mucosa Note: Supplementary data for this article are available at Molecular Cancer (4, 11), and small intestine (4, 12). The extracellular Therapeutics Online (http://mct.aacrjournals.org/). domain of this type I transmembrane protein comprises both a proteoglycan domain implicated in cell adhesion Current address for T.J. McCabe: Johnson & Johnson, PA; current address for P. Trail: Regeneron Pharmaceuticals, Inc., NY; current address for S. Ha: through homotypic interaction (13) and the carbonic Merck&Co., Inc. PA; current address for Y.S. Chang: Aileron Therapeutics, anhydrase domain that catalyzes the reversible hydration MA; current address for A. Voznesensky: Novartis Institutes for BioMedical Research, MA; current address for G. Ranges: Bayer HealthCare AG, of carbon dioxide to bicarbonate and protons (14) and is retired; and current address for P.P. Tamburini: Alexion Pharmaceuticals. involved in the regulation of the pH within the tumor Inc, CT. environment (15). Corresponding Author: Heike M. Petrul, Bayer HealthCare AG, Berlin CAIX gene expression is under the direct control 13342, Germany. Fax: 49-30-18079; E-mail: [email protected] of the transcription factor hypoxia-inducible factor-1 doi: 10.1158/1535-7163.MCT-11-0523 and is significantly upregulated by tumor hypoxia (16). 2011 American Association for Cancer Research. CAIX expression was found to correlate with (i) a high

340 Mol Cancer Ther; 11(2) February 2012

Efﬁcacy of Anti-CA9 Antibody–Drug Conjugate BAY 79-4620

mean vessel density, cancer stage, and degree of necrosis Cell binding by flow cytometry in head and neck carcinoma (17); (ii) poor survival in Adherent CAIX-expressing PC-3 mm2 cells, non-CAIX nasopharyngeal carcinoma (18); (iii) tumor grade, nega- expressing MiaPaCa-2 cells, and MiaPaCa-2 cells trans- tive estrogen receptor status, higher relapse rate, and poor fected with full-length CAIX were detached from their survival for invasive breast carcinoma (10). This associa- flasks with 1:10 trypsin/Versene in PBS solution for 5 to 10 tion with tumor grade and overall survival, together with minutes. Cells were spun down (1,000 rpm, 5 minutes), its cell surface distribution and restricted expression in washed once with ice-cold RPMI 10% FBS, and resus- þ þ normal tissues, implicates CAIX as an important thera- pended in ice-cold staining buffer [Ca Mg -free PBS, 2% peutic target for monoclonal antibody (mAb)-based ther- bovine serum albumin (BSA), and 0.05% sodium azide] to apy. In this article, we report the identification of a potent achieve 6 106 cells/mL. Anti-CAIX IgG1 or control and selective internalizing human antibody directed human IgG1 antibodies at 25 mg/mL were incubated with against CAIX that, when conjugated to monomethylaur- 6 105 cells on ice for 1 hour. Unbound antibody was istatin E (MMAE), yielded a highly efficacious antibody– removed by washing the cells with the ice-cold staining drug conjugate BAY 79-4620, with activity against a vari- buffer. The cells were fixed with 2% formaldehyde in PBS ety of solid tumor types. for 10 minutes, then washed twice with staining buffer. The cell pellet was resuspended in 100 mL ice-cold staining Materials and Methods buffer containing a 1:200 dilution of Alexa Fluor 488– labeled secondary antibody (Molecular Probes/Invitro- Reference antibodies gen) and incubated on ice for 1 hour. The unbound The hybridoma producing CAIX mAb M75 (1, 13) was antibody was washed from the cells 2 times with flow obtained from American Type Culture Collection (ATCC) buffer (PBS containing 2% BSA), and the cells were resus- and used to express and purify mAb M75 using standard pended in 1 mL flow buffer. Fluorescence-activated cell protocols. sorting (FACS) analysis of the resuspended cells was done on a Beckman FACS Calibur instrument. All cell lines Antibody discovery using the HuCAL Gold were obtained from ATCC; MiaPaCa-2-CAIX was gener- Fab-phage library ated using standard transfection methods. Cell lines are The HuCAL Gold Fab-phage library was obtained from regularly checked for authenticity by DNA fingerprinting MorphoSys AG and was prepared as described elsewhere at the DSMZ (German Collection of Microorganisms and (19), comprising a highly diverse library of 1010 different Cell Cultures), Braunschweig, Germany. monovalent phage encoded within phagemid vector pMORPH23 and allowing for monovalent CysDisplay of Assessment of cellular internalization of mAbs Fab fragments. Solid phase panning was carried out as Antibody internalization was assessed using a Cello- described in the Supplementary Data. mics Array Scan automated confocal microscope system (see Supplementary Data). Fab expression Soluble Fab fragments were produced from the isolated Generation of immunoconjugates with phage clones as described in the Supplementary Data. monomethylauristatin E mAb conjugation with MMAE was done essentially as Identification of CAIX-binding Fabs described elsewhere for conjugation with an anti-CD30 Isolated Fabs were tested for binding to the purified mAb (21). ectodomain of CAIX in an ELISA, positive Fabs were recloned in the IgG format and expressed in Chinese Cell cytotoxicity assays hamster ovary cells (see Supplementary Data). Anti- Antigen-positive and antigen-negative cells were plat- CAIX IgG antibodies were purified using protein A ed at 5,000 cells per well in 96-well plates in 100 mL media sepharose. per well overnight at 37 C, in 5% CO2 to adhere. The media was replaced with fresh media containing antibody Antibody binding kinetics using surface plasmon or antibody-vcMMAE conjugate and the plates further resonance incubated at 37 C, in 5% CO2 for 72 hours. Cytotoxicity Surface plasmon resonance (SPR) was carried out on a was assessed using the Alamar Blue cell viability assay. BIAcore 3000 instrument (BIAcore; see Supplementary Alamar Blue was added at a final concentration of 10% for Data). the last 4 hours of the incubation, and its transformation to fluorescent product by viable cells determined spectro- Immunoprecipitation fluorometrically by plate reader (544 nm excitation–590 The complexes formed between the antibodies and nm emission). biotin-labeled cell proteins were immunoprecipitated and visualized by immunoblots developed with enzyme- In vivo antibody distribution in mice linked streptavidin using an adaptation of methods Noninvasive in vivo imaging studies were conducted described elsewhere (20). using the CRI Maestro in vivo imaging system with

www.aacrjournals.org Mol Cancer Ther; 11(2) February 2012 341

Petrul et al.

multispectral acquisition and analysis (CRI; see Supple- conjunction with Animal Research Kit Peroxidase (Dako), mentary Data). mouse IgG block, peroxidase block and biotin block. Epitope demasking for FFPE tissue was done for 17 In vivo tumor growth inhibition and minutes in a vegetable steamer in Tris-buffered solution pharmacodynamic studies (pH 9). All animal experiments were conducted in accordance IHC slides were analyzed by a pathologist with an with either the United Kingdom Coordinating Committee intensity scale of 0, 1þ,2þ,or3þ. on Cancer Research regulations for the welfare of animals, Correlation of CAIX level with antitumor efficacy of the German animal welfare law, or the Institutional Ani- CAIX-ADC across a variety of human tumor xeno- mal Care and Use Committee, in addition to approval by grafts. IHC slides were quantified using Histoscore local authorities. For xenograft experiments based on (H-score). H-score considers staining intensity per cell human tumor cell lines, immunocompromised mice (8– as well as the distribution of the staining in the tumor 12 weeks of age) were subcutaneously implanted on day 0 tissue and is determined as percentage of tumor cells with the respective tumor cells in 0.1 mL volume as showing 3þ staining intensity 3 þ (% cells staining detailed in Supplementary Table ST2. For xenograft 2þ) 2 þ 1 (% cells staining 1þ); range 0–300. Percent experiments based on patient-derived NSCLC tumors, tumor growth inhibition (TGI) at 1 mg/kg BAY 79-4620 tumor fragments were subcutaneously passaged on naive was plotted against the average H-Score for each model. NMRI nu/nu mice (Lu7506, Lu7298, Lu7913, Lu7406, CAIX content of several tumors was also determined Lu7747; refs. 22, 23). Treatment was initiated when tumors by Western blot (Supplementary Data). reached an average size of 60 to 160 mm3 depending on the model used (Supplementary Table ST2). Dosing of test Statistical analysis compounds was carried out according to the dosing To compare TGI between treatment groups and their levels, schedules, and routes of administration described respective vehicle group, 2-sided Dunnett tests (24) were in Supplementary Table ST2. Carboplatin was obtained carried out. The Dunnett test is a one-step test procedure from Hexal, taxol (Paclitaxel) from Bristol Myers Squibb, that is very powerful in the given situation of comparing cisplatin from Sigma Aldrich, and gemcitabine from Eli several groups in a common control. Dunnett’s procedure Lilly. A dose volume of 0.1 mL/10 g body weight was keeps the multiple significance level. In each study, a used throughout and intravenous administration was by significance level of 0.05 was chosen. The log-transformed tail vein injection. The health status of animals was mon- ratios of tumor volume at the end of experiment to itored daily. The length and width of each tumor was baseline tumor volume at day 0 were defined as the measured by electronic calipers 2 to 3 times per week and response variable of interest. tumor volumes (mm3) were calculated as [length (mm) All calculations were carried out using SAS 9.1.3. width (mm)2]/2. Percentage increases in tumor size during the study were calculated by the formula [100 tumor Results volume at treatment end/tumor volume at treatment start] 100 for every single animal. Final tumor volume Human antibody fragments (Fabs) directed at the was defined by the last time point in which the vehicle ectodomain of CAIX were selected by in vitro "panning" treated control remained within the experiment. Antitu- of the MorphoSys HuCal Gold nonimmune biased Fab- mor efficacy was also assessed from the incidence of phage display library with a purified preparation of the regression in which a tumor regression is defined as a extracellular domain of human CAIX. More than 50 reduction in tumor size of more than 30% at study end CAIX-binding Fabs with unique VH and/or VL relative to the initial size. sequences were identified that bound CAIX by ELISA In pharmacodynamic experiments, tumors were col- with signal to noise ratios of more than 10. The top 10 lected at 4 hours, 1, 3, and 5 days following single admin- Fabs disclosed in Tamburini and colleagues (25) with istration of BAY 79-4620, formalin-fixed, paraffin-embed- respect to selective binding and internalization by ded (FFPE), sectioned at 5 mm, deparaffinized, and stained CAIX-expressing cancer cell lines were converted to using standard protocol for fluorescent immunohis- full-length human IgG1 antibodies, expressed in Chi- tochemistry (IHC) of human tissues using mouse anti- nese hamster ovary cells, and purified by affinity chro- bodies against -a/b-tubulin, phospho-Histone H3, and matography. Seven of these antibodies with acceptable DNA. Slides were observed by fluorescent microscopy. expression and solubility were further characterized and found to exhibit high affinity binding to the CAIX Immunohistochemistry ectodomain in vitro, with several mAbs exhibiting sin- In parallel with efficacy studies, a group of animals gle-digit nmol/L Kd values by SPR (Supplementary carrying each tumor type was sacrificed, the tumors were Table ST1). Next, the capacity of 6 of these antibodies preserved as FFPE and cryo (frozen in Tissue-Tek) tissues. to selectively immunoprecipitate antigen from whole- Tumor tissue was cut in 3-mm slices and stained with the cell lysates derived from biotin-labeled PC-3mm2 whole þ anti-CAIX mouse mAb M75 at 0.5 mg/mL (cryo sections) cells that express CAIX (CAIX ) or DLD-1 cells that do and 1 mg/mL (FFPE sections) final concentrations in not (CAIX ) was examined by immunoblots developed

342 Mol Cancer Ther; 11(2) February 2012 Molecular Cancer Therapeutics

Efﬁcacy of Anti-CA9 Antibody–Drug Conjugate BAY 79-4620

A Control IgG 3 ee 9 1 aa 3 ef 2 5 aa 3 3a4 5A6 Control IgG 3 ee 9 1 aa 3 ef 2 5 aa 3 3a4 5A6 None; biotinylated ctrl

191

PC3mm2 DLD1 B PC-3MM2 MiaPaCa-2 MiaPaCa-2-CAIX

Untreated Sec. antibody Anti-CAIX 3ee9

Figure 1. In vitro selectivity of fully human mAbs directed against human carbonic anhydrase IX. A, immunoprecipitation of biotinylated cell surface proteins þ from CAIX PC-3mm2 cells (left) and CAIX DLD-1 cells (right). Lanes are labeled according to the particular anti-CAIX mAb used. Far right lane contains biotinylated CAIX ectodomain directly applied to the lane. Arrows locate positions of migration of Mr markers of the indicated sizes (kDa). B, binding of mAb 3ee9 to CAIX-positive PC-3mm2 and CAIX-transfected MiaPaca-2-CAIX cells (left and right) but not to CAIX-negative parental MiaPaCa-2 cell line (middle) by flow cytometry. with streptavidin–horseradish peroxidase (Fig. 1A). Fig. S1). A similar selectivity was observed with the ADC mAbs 3ee9 and 1aa1 selectively coimmunoprecipitated comprising mAb1aa1 (data not shown), and both CAIX- þ þ from biotin-labeled CAIX cells—but not from the specific mAb conjugates selectively killed CAIX cells in CAIX cells—a single protein that comigrated with culture. For example, BAY 79-4620 killed CAIX-trans- recombinant biotinylated CAIX (Fig. 1A). mAbs 3ef2, fected MiaPaCa-2-CAIX cells, with an IC90 of 10 5aa3, and 5a6 were not selective and coprecipitated nmol/L, whereas CAIX-negative MiaPaca2 wild-type many additional cell surface proteins in addition to cells were quite resistant (Fig. 2), with at least a 10-fold CAIX.Importantly,theconjugateofmAb3ee9with higher IC90. The mechanism of cell killing by BAY 79-4620 MMAE (BAY 79-4620; CAIX-ADC) retained the affinity involved targeted tubulin disruption. Thus although þ of the parent antibody for CAIX (Supplementary Table treatment of CAIX PC-3mm2 cells in culture resulted in ST1), exhibiting a kDa of 3.6 nmol/L by SPR. The fragmented fibers and prevention of normal spindle for- conjugate did not bind to 13 other carbonic anhydrases mation, as visualized by tubulin staining, CAIX H460 tested (data not shown). Binding to mitochondria-associ- cells retained normal spindle formation in the presence of ated CA5 was observed, but this isozyme would be inac- the conjugate (Supplementary Fig. S2). cessible to the antibody in vivo. BAY 79-4620 selectively The high selectivity of the 3ee9 targeting antibody bound (Fig. 1B) and underwent internalization by (Sup- component of BAY 79-4620 was further shown using þ plementary Fig. S1) CAIX PC-3mm2 cells in culture but established tumors derived from the CAIX-negative MIA- did not bind CAIX MiaPaca2 cells (Fig. 1B) or undergo PaCa-2 pancreatic carcinoma cell line, and its CAIX-pos- internalization by CAIX DLD-1 cells (Supplementary itive transfectant. Administration of an 3ee9-Alexa Fluor

www.aacrjournals.org Mol Cancer Ther; 11(2) February 2012 343

Petrul et al.

ing animal experiments the dosing schedule of 3 doses 100 every 4 days was used. At this dose scheduling, BAY 79- 80 4620 was highly active against established tumors derived from the HT-29 cell line (Fig. 3A, right panel). 60 Whereas tumors from untreated, PBS vehicle treated, and MiaPaCa-2-CAIX 40 10 mg/kg unconjugated CAIX mAb-treated control MiaPaCa-2 groups grew progressively, with mean doubling times of 20 around 6.4 days, treatment with BAY 79-4620 on a 3 doses % Inhibition every 4 days schedule, intravenous schedule produced 0 0 20 40 60 80 100 120 robust antitumor efficacy at all doses examined, with –20 0.625, 1.25, 2.5, 5, and 10 mg/kg doses yielding 54, 72, c (nmol/L) 97, 100, and 100% TGI, respectively. The percentages of the tumors that exhibited regression were 20, 90, 100, 100, respectively, at the 1.25, 2.5, 5, and 10 mg/kg doses, and Figure 2. Cytotoxicity of BAY 79-4620 toward CAIX-transfected BAY 79-4620 was well tolerated with no weight loss at all MiaPaCa-2 (diamonds) but not CAIX-negative MiaPaCa-2 cells (circles). doses. Paclitaxel (Taxol) at a dose of 15 mg/kg showed efficacy very similar to 2.5 mg/kg of BAY 79-4620. Free 750 conjugate, but not a control hIgG Alexa Fluor 750 MMAE toxophore administered at a dose of 0.2 mg/kg, conjugate, to mice with bilaterally implanted CAIX-neg- equivalent to the amount of MMAE administered in ative and CAIX-positive MiaPaca tumors resulted in 10 mg/kg of BAY 79-4620 was much less active than the exclusive retention of the Alexa-750 conjugate in the antibody conjugate, producing 60% TGI and no tumor CAIX-positive tumor following elimination of the bulk regression. Although this dose of free MMAE was well of the unbound fluorescence via the liver and bladder on tolerated, a 1 mg/kg dose produced 50% lethality with or around day 4 (Supplementary Fig. S3). Furthermore, severe weight loss across all survivors (not shown). BAY 79-4620 exhibited minimal TGI (50%) when admin- BAY 79-4620 was highly active against a variety of other istered to mice implanted with established CAIX-negative established tumor types (Fig. 4), including those estab- þ MiaPaca2 tumors, even at the higher dose of 10 mg/kg, lished using the following CAIX human cancer cell lines: but exhibited significant antitumor activity when admin- cervix carcinoma HeLa-MaTu, colorectal Colo205, pros- istered to mice implanted with established CAIX-positive tate PC3mm2, gastric NCI-N87, lung A549, gastric SNU16, tumors derived with MiaPaca2 cells transfected with and gastric MKN45. Tumors derived using HeLa-MaTu CAIX (data not shown), in which doses of 2.5, 5, and 10 were the most sensitive wherein a minimal effective dose mg/kg yielded TGI of 63%, 82%, and 94%, respectively, of 0.625 mg/kg was observed. The 60 mg/kg dose of and 100% shrinkage at the latter dose. BAY 79-4620 produced 10% lethality and 20% body BAY 79-4620 was highly active against colorectal cancer weight loss establishing this dose of BAY 79-4620 as the HT-29 tumors using a range of dosing schedules, includ- MTD, whereas all other treatment levels were well toler- ing, quite remarkably, single intravenous doses (Fig. 3A, ated. Eighty percent of the HeLa-MaTu tumors exhibited left panel) in the range 0.625 to 10 mg/kg. To explore the in tumor shrinkage at the 1.25 mg/kg dose, whereas at the vivo mechanism of action, tumors were harvested follow- 2.5 mg/kg dose and higher, 100% of the tumors showed ing single BAY 79-4620 doses of 1.25 and 5 mg/kg and regression or complete regression. BAY 79-4620 was also evaluated by IHC (Fig. 3B). Whereas little effect was active against established tumor xenografts derived from observed at 4 hours, increased numbers of cells in the human prostate PC-3mm2 cell line (Fig. 4). Thus, G2/M arrest, the appearance of multipolar spindles, and whereas PBS vehicle or unconjugated mAb (not shown) a decreased level of tubulin was clearly seen by day 1. This at 30 mg/kg exhibited no antitumor efficacy, BAY 79-4620 pattern intensified through to 5 days postdosing when produced a dose-dependent TGI ranging from 45% at almost all cells in the 5 mg/kg dose group were severely 1 mg/kg to 100% at doses of 10 mg/kg and higher and affected as judged by the presence of defective spindles or was well tolerated at all doses. apoptosis. These data indicate a mechanism of tumor BAY 79-4620 was also active in the gastric carcinoma killing by BAY 79-4620 that is both consistent with the model NCI-N87 (Fig. 4). This aggressive tumor model effects of the compound on CAIX-expressing cells (Sup- responded well to the higher dose of 60 mg/kg on a plementary Fig. S2) and in line with the known properties 3 doses every 4 days schedule, although neither cisplatin of MMAE as a tubulin polymerization inhibitor that nor paclitaxel treatment reached statistical significance induces G2/M arrest through prevention of normal spin- compared with the vehicle group on day 37, when the dle formation (26). vehicle group was sacrificed. Despite the impressive results obtained in this model, Efficacy of BAY 79-4620 in heterogeneous tumors the tumor response to single dose (Fig. 3A, left panel) and derived from patient tumor explants was shown in several the 1 dose every 7 days (not shown) schedules, was, unlike lung carcinoma models (Lu series). As these tumors the effects of the 3 doses every 4 days schedule (Fig. 3A, consist of both CAIX-positive and CAIX-negative cells right panel), transient in nature. Therefore, for all follow- (Fig. 5), the observed efficacy in these models suggested

344 Mol Cancer Ther; 11(2) February 2012 Molecular Cancer Therapeutics

Efﬁcacy of Anti-CA9 Antibody–Drug Conjugate BAY 79-4620

A PBS Vehicle 2.5 mg/kg BAY 794620 0.625 mg/kg BAY 794620 10 mg/kg 3ee9 5 mg/kg BAY 794620 1.25 mg/kg BAY 794620 10 mg/kg 3ee9 + 0.2 mg/kg MMAE 10 mg/kg BAY 794620 2.5 mg/kg BAY 794620 0.625 mg/kg BAY 794620 15 mg/kg Taxol 5 mg/kg BAY 794620 1.25 mg/kg BAY 794620 0.2 mg/kg MMAE 10 mg/kg BAY 794620 2,500 2,000

] ] Mean ± SD

3 3 2,000 Mean ± SD 1,500 1,500

1,000 1,000

500 Tumor volume [mm Tumor volume [mm 500

0 0 6050403020100 6050403020100 Time after tumor cell inoculation [d] Time after tumor cell inoculation [d] B

Day 1

Day 5

0 mg/kg 1.25 mg/kg 5.0 mg/kg

Figure 3. Antitumor efficacy of CAIX-ADC against HT-29 tumors. A, dose-dependent TGI resulting from single dose (left) and 3 doses every 4 days (right) treatments. Mean values SD are plotted. Red arrows indicate time points of treatment. B, immunohistochemistry of HT-29 tumors at 1 or 5 days after a single dose of 0, 1.25, or 5 mg/kg BAY 79-4620. Sections were stained for a/b tubulin (green fluorescence), phospho-histone-H3 (red fluorescence), and DNA (blue fluorescence). Highlighted features are normal M-phase cells with chromosomes and spindles (arrows); M-phase cells with defective or absent spindles (arrowheads); normal tubulin in the cytosol (asterisks); apoptotic cells with condensed or fragmented DNA, aggregation or absence of tubulin (stars). that the release of toxophore from the CAIX-positive cells sacrificed at the time when treatment was started and kills the CAIX-negative cells through a bystander effect. tested for CAIX expression by IHC. CAIX expression In contrast to the effects observed in the above tumor was semiquantified from the average pixel intensities models, BAY 79-4620 was inactive against the MDR line percell,thepercentageof the tumor cells staining HCT-15 that overexpresses P-glycoprotein (P-gp). The P- positive for CAIX and the location of CAIX within the gp substrate paclitaxel was similarly inactive, whereas tumor tissue (Fig. 5). Most of the tumors, including gemcitabine, a non-Pgp substrate, exhibited TGI in this A431, A549, Lu7913, Lu7298, and Lu7506, expressed tumor model (Fig. 4). CAIX in the area around necrotic cores. In most cases, Anti-CAIX–antibody drug conjugates are expected to CAIX staining was absent from the immediate vicin- release the attached toxophore specifically in antigen- ity of blood vessels consistent with its upregulation expressing tumors. To test whether the activity of BAY by hypoxia. The intensity of CAIX expression across 79-4620 correlated with CAIX expression levels in the various tumor types varied strongly. HeLa-MaTu tumors, an immunohistochemical staining protocol for showed a strong homogenous CAIX expression pat- CAIX was developed. In parallel to the above studies of tern, whereas Hs746T showed almost no staining. The the antitumor activity of BAY 79-4620 in the various efficacy of BAY 79-4620 in tumor models showed a xenograft models, the levels of CAIX expression within strong positive association with tumor CAIX expression thesetumorswasdetermined.Groupsofanimalswere level and/or H-score (Fig. 6). HeLa-MaTu, HT29, and

www.aacrjournals.org Mol Cancer Ther; 11(2) February 2012 345

Petrul et al.

1,400 1,200 1,000 800 600 Figure 4. CAIX-ADC exhibits potent 400 * antitumor activity against a variety 200 * * * * ** of human tumor types. Waterfall 0 plots for change in tumor size are –50 shown for tumors derived from –100 * * * *** ** * * ** * * various human cancer cell lines [% size at treatment of start] [% –150 ***** HT-29 Colo205 HCT-15 NCI-N87 MKN45 SNU16 A549 PC3mm2 HeLa- (top row) and patient-derived Increase of tumor size during study Increase of tumor size during MaTu tumor fragments (bottom row). Colored bars identify the particular treatment depicted in the inset. Vehicle 2,000 Mean þ SD is plotted. Asterisks 1,800 0.625 mg/kg BAY 794620 1 mg/kg BAY 794620 indicate statistical signiﬁcance 1,600 1.25 mg/kg BAY 794620 versus respective vehicle controls 1,400 2.5 mg/kg BAY 794620 (detailed statistical information are 1,200 3 mg/kg BAY 794620 presented in Supplementary Table 1,000 5 mg/kg BAY 794620 800 10 mg/kg BAY 794620 ST3). 600 30 mg/kg BAY 794620 400 * * 60 mg/kg BAY 794620 * * Gemcitabine 200 * * 0 Taxol –50 Cisplatin –100 * ** * 75 mg/kg Carboplatin

[% of size at treatment start] treatment at size of [% –150 * ** ** Lu7913 Lu7298 Lu7406 Lu7506 Lu7747 Increase of tumor size during study Increase of tumor size during

PC3mm2 models characterized by the highest expres- select fully human antibodies that are highly selective sion of CAIX showed 56% to 92% TGI, when treated for binding to, and internalization by, CAIX-expressing with 1 mg/kg BAY 79-4620. This dose did not show TGI cancer cell lines. These antibodies exhibited undetect- in the Hs746T, A549, and A431 models—characterized able binding to other proteins within complex target by the lowest CAIX levels observed–-but these models cell proteomes by immunoprecipitation, did not bind did respond to doses higher than 1 mg/kg. to other extracellular carbonic anhydrases, and in the case of the CAIX-targeting antibody 3ee9 (BAY 79- Discussion 4682), showed speciﬁc retention in CAIX-positive tumors in vivo by whole animal imaging using a ﬂuores- Although some recently published small molecule cently labeled version. The binding selectivity of 3ee9 inhibitors (27, 28) successfully target the enzymatic was fully retained in the corresponding antibody–drug activity of CAIX, the work described here uses a dif- conjugate BAY 79-4620, which proved to be both potent ferent therapeutic approach in exploiting CAIX as an and highly selective in the killing of cultured CAIX- anchor for the selective delivery of a toxic payload to expressing cell lines through a mechanism that involved tumors via an antibody-based therapeutic. A rigorous tubulin disruption. Consistent with the in vitro selectiv- discovery cascade was implemented to generate and ity of the conjugates, conjugate BAY 79-4620 exhibited

HeLa-MaTu Lu7913 HT29 PC3mm2 Lu7298 Lu7506 HCT116 A549 A431 Hs746T

Figure 5. Distribution of CAIX expression in xenograft models. Overview Staining intensity of CAIX-positive tumor cells was done as 0, 1þ,2þ, or 3þ. Representative images of 20x the tumor models are shown. The 3 panels show an overview. Magniﬁcations, 20 and 60. 60x

Cervix cancer Lung cancer Colon cancer Prostate cancer Lung cancer Lung cancer Colon cancer Lung cancer Skin cancer Gastric cancer

346 Mol Cancer Ther; 11(2) February 2012 Molecular Cancer Therapeutics

Efﬁcacy of Anti-CA9 Antibody–Drug Conjugate BAY 79-4620

driven mainly by targeted delivery of the MMAE

100 toxophore to CAIX-expressing tumor cells. Targeting of a CAIX-expressing tumor by the anti-CAIX antibody 80 was shown in in vivo imaging studies, in which the

HeLa-MaTu antibody was selectively retained in the antigen-pos- 60 Hs746T 50% TGI itive, but not the antigen-negative, tumor implanted on HT29 A549 thesameanimal.Thetherapeuticefﬁcacyoftargeted 40 HCT116 delivery of the toxophore was also shown in the A431

at 1 mg/kg CA9-ADC at 20 PC3mm2 patient-derived lung tumor model Lu7406, in which Lu 7506 both carboplatin and paclitaxel show little efficacy, %Tumor growth inhibition (TGI) %Tumor growth Lu 7913 0 Lu7298 whereas BAY 79-4620—which delivers a much lower total dose of cytotoxic compound to the tumor— 0 100 200 300 showed high efficacy. The efficacy against this and H-Score [0-300] 3 x (% cells staining 3+) + 2 x (% cells staining 2+) + 1 x (% cells staining 1+) other patient-derived models with heterogeneous expression of CAIX is consistent with BAY 79-4620– dependent tumor cell killing both by direct and fi Figure 6. Correlation of CAIX level with antitumor ef cacy of CAIX-ADC bystander mechanisms. The mechanistic IHC studies across a variety of human tumor xenografts. Tumor-bearing animals were treated with vehicle or 1 mg/kg BAY 79-4620 3 doses every 4 days, of HT-29 tumors showed that the response to a single intravenously. Tumor size was determined using caliper. Percent TGI was dose of CAIX-ADC was rapid, being detectable at determined at the end of the experiment. IHC slides were quantified using day 1, and seemed to affect the majority of tumor cells Histoscore (H-score), which was determined by 3 (% of tumor cells by day 5. The efficacy of BAY 79-4620 seems to be staining 3þ) þ 2 (% of tumor cells staining 2þ) þ 1 (% of tumor cells staining 1þ); range 0 to 300. Percent TGI at 1 mg/kg BAY 79-4620 is limited to tumors that are susceptible to tubulin inhi- plotted against the average H-Score in that model. bition, as was shown by the lack of response to the ADC in the highly chemoresistant HCT-15 colon carcinoma model. little or no antitumor activity against a human tumor The safety and toxicologic evaluation of the ADCs xenograft with low or no CAIX expression but was described presently has been facilitated by cross-reactiv- highly potent against xenografts comprising the same ity with primate CAIX and BAY 79-4620 is currently in cell line that had been stably transfected with CAIX. The phase I of clinical development. selectivity of the conjugate BAY 79-4620 was further We anticipate that the high potency and selectivity of confirmed in human and primate normal tissue binding the ADCs described presently will translate into the clinic studies, wherein the binding pattern mimicked exactly and that these compounds may possess single-agent the expected tissue distribution of CAIX with significant efficacy, possibly rising to become first-line therapeutics. staining only in the gastrointestinal tract and where no We note that these therapeutics are combinable pre- off-target binding was observed (data not shown). clinically with standard of care therapeutics such as cape- There remains a high unmet need for safe and ef- citabine (Xeloda; data not shown), and it is conceivable fective antitumor agents targeting different tumor that therapeutic use could be further augmented in indications, yet despite the large amount of informa- combination. tion about the association of CAIX with various solid The mechanism of action of BAY 79-4620 suggests tumors, there is no approved therapeutic drug target- clinical development in tumor indications characterized ing this antigen. Our approach uses an anti-CAIX by a high expression of CAIX and a reported sensitivity to antibody with a linker toxophore system based on the spindle poisons. NSCLC and gastric cancer are 2 of several tubulin inhibitor MMAE thatofferspotencyatthe tumor indications showing CAIX expression as well as site of action and a high level of stability in the peri- sensitivity to taxanes (29–31). The expression of CAIX phery. Accordingly, the conjugate BAY 79-4620 proved varies between tumor indications as well as between to be highly effective and well tolerated when used individual patients with a tumor of the same histologic against CAIX-expressing human xenografts repre- origin. Clinical development of BAY 79-4620 should, senting cervical, prostate, colorectal, gastric, and lung therefore, be supported by biomarker measurements of tumors. In each case, a 100% complete response rate CAIX expression. Clinical studies should explore cut-offs couldbeshownatthehigherdoseswithina3doses for CAIX expression associated with response to BAY 79- every 4 days schedule. At lower doses of BAY 79-4620 4620. on or around 1 mg/kg, differences in the antitumor Noninvasive assays to identify patients likely to activity across the various models was observed with respond to BAY 79-4620 treatment and assays monitoring higher activity associated with tumor types exhibiting early response to treatment would aid clinical develop- high CAIX expression, such as HeLa-MaTu, and lower ment of this drug. Potential approaches currently under activity against xenograft models with less CAIX evaluation include the quantification of circulating levels expression such as PC-3mm2. The data suggests that of CAIX ectodomain and imaging-based methods such as the antitumor activity of BAY 79-4620 at 1 mg/kg is positron emission tomography.

www.aacrjournals.org Mol Cancer Ther; 11(2) February 2012 347

Petrul et al.

Disclosure of Potential Conﬂicts of Interest Agostino (protein expression); Carla Pellegrino, Robert Dreyer, Steven Fisk (protein chemistry), Elizabeth Bourret, Peggy Bourguillon, Marina Ichetovkin, Susan Gawlak, Marina Reinelt, Katrin Weidner (cellular All authors are current or former employees of Bayer Healthcare. assays), Elizabeth Bortolon, Arris Henderson, Dahai Xue, Bianka Timp- ner, Karola Henschel, Katrin Jansch€ (preclinical animal studies), Sabine Acknowledgments Jabusch (IHC), Anna Behnke (biochemical assays) and Tina Muller€ (statistics). The costs of publication of this article were defrayed in part by the BAY 79-4620 uses the Seattle Genetics linker-toxophore chemistry; payment of page charges. This article must therefore be hereby marked the antibody 3ee9 is derived from the HuCAL phage display library advertisement in accordance with 18 U.S.C. Section 1734 solely to indi- (Morphosys AG). The authors thank Silvia Pastorekova and Jaromir cate this fact. Pastorek, Institute of Virology, for scientiﬁc consultancy and also the following individuals for their important contributions to this work: Dana Wirak, Richard Altman, Tom Donaldson, Haren Vasavada Received July 20, 2011; revised November 2, 2011; accepted November (molecular biology); Dave Wunderlich, Jennifer Pendleton, Karla D’ 11, 2011; published OnlineFirst December 6, 2011.

References 1. Pastorekova S, Zavadova Z, Kostal M, Babusikova O, Zavada J. A 15. NeriD,SupuranCT.InterferingwithpHregulationintumours novel quasi-viral agent, MaTu, Is a two-component system. Virology as a therapeutic strategy. Nature Rev Drug Discov 2011;10: 1992;187:620–6 767–77. 2. Pastorek J, Pastorekova S, Callebaut I, Mornon JP, Zelnk V, Opavsky 16. Wykoff CC, Beasley NJP, Watson H, Turner KJ, Pastorek J, Sibtain A, R, et al. Cloning and characterization of MN, a human tumor-associ- et al. Hypoxia-inducible expression of tumor-associated carbonic ated protein with a domain homologous to carbonic anhydrase and a anhydrases. Cancer Res 2000;60:7075–83. putative helix-loop-helix DNA binding segment. Oncogene 1994;9: 17. Beasley NJP, Wykoff CC, Watson PH, Leek R, Turley H, Gatter K, et al. 2877–88. Carbonic anhydrase IX, an endogenous hypoxia marker, expression in 3. Oosterwijk E, Ruiter DJ, Hoedemaeker PJ, Pauwels EK, Jonas U, head and neck squamous cell carcinoma and its relationship to Zwartendijk J, et al. Monoclonal antibody G 250 recognizes a deter- hypoxia, necrosis and microvessel density. Cancer Res 2001;61: minant present in renal-cell carcinoma and absent from normal kidney. 5262–7. Int J Cancer 1986;38:489–94. 18. Hui EP, Chan ATC, Pezzella F, Turley H, To KF, Poon TCW, et al. 4. Pastorekova S, Zavada J. Carbonic anhydrase IX (CA IX) as a potential Coexpression of hypoxia-inducible factors 1a and 2a, carbonic anhy- target for cancer therapy. Cancer Ther 2004;2:245–62. drase IX, and vascular endothelial growth factor in nasopharyngeal 5. Liao SY, Aurelio ON, Jan K, Zavada J, Stanbridge EJ. Identification of carcinoma and relationship to survival. Clin Cancer Res 2002;8:2595– the MN/CAIX protein as a reliable diagnostic biomarker of clear cell 604. carcinoma of the kidney. Cancer Res 1997;57:2827–31. 19. Rothe C, Urlinger S, Lohning€ C, Prassler J, Stark Y, Jager€ U, et al. The 6. Turner JR, Odze RD, Crum CP, Resnick MB. MN antigen expression in human combinatorial antibody library HuCAL GOLD combines diver- normal, preneoplastic, and neoplastic esophagus: a clinicopatholog- sification of all six CDRs according to the natural immune system with a ical study of a new cancer-associated biomarker. Hum Pathol novel display method for efficient selection of high-affinity antibodies. 1997;28:740–4. J Mol Biol 2008;376:1182–200. 7. Liao SY, Brewer C, Zavada J, Pastorek J, Pastorekova S, Manetta A, 20. Flesch B, Ntambi E, Neppert J. Biotinylation: a nonradioactive method et al. Identification of the MN antigen as a diagnostic biomarker of for the identification of cell surface antigens in immunoprecipitates. cervical intraepithelial squamous and glandular neoplasia and cervical Electrophoresis 1995;16:757–62. carcinomas. Am J Pathol 1994;145:598–609. 21. Francisco JA, Cerveny CG, Meyer DL, Mixan B.J, Klussman K, Chace 8. Saarnio J, Parkkila S, Parkkila A.-K, Haukipuro K, Pastorekova S, DF, et al. cAC10-vcMMAE, an anti-CD30-monomethyl auristatin E Pastorek J, et al. Immunohistochemical study of colorectal tumors conjugate with potent and selective antitumor activity. Blood 2003; for expression of a novel transmembrane carbonic anhydrase, MN/CA 102:1458–65. IX, with potential value as a marker of cell proliferation. Am J Pathol 22. Fichtner I, Rolff J, Soong R, Hoffmann J, Hammer S, Sommer A, et al. 1998;153:279–85. Establishment of patient-derived non-small cell lung cancer xeno- 9. Vermylen P, Roufosse C, Burny A, Verhest A, Bosschaerts T, Pastor- grafts as models for the identification of predictive biomarkers. Clin ekova S, et al. Carbonic anhydrase IX antigen differentiates between Cancer Res 2008;14:6456–68. preneoplastic malignant lesions in non-small cell lung carcinoma. Eur 23. Hammer S, Sommer A, Fichtner I, Becker M, Rolff J, Merk J, et al. Respir J 1999;14:806–11. Comparative profiling of the novel epothilone, sagopilone, in xeno- 10. Chia SK, Wykoff CC, Watson PH, Han C, Leek RD, Pastorek J, et al. grafts derived from primary non–small cell lung cancer. Clin Cancer Prognostic significance of a novel hypoxia-regulated marker, car- Res 2010;16:1452–65. bonic anhydrase IX, in invasive breast cancer. J Clin Oncol 2001; 24. Dunnett CW. A Multiple Comparison Procedure for Comparing Several 19:3660–8. Treatments with a Control. J Amer Statistical Assoc 1955;50:1096– 11. Pastorekova S, Parkkila S, Parkkila AK, Opavsky R, Zelnik V, Saarnio J, 121. et al. Carbonic anhydrase IX, MN/CA IX: Analysis of stomach com- 25. Tamburini P, Ranges G, Adnane L, McCabe T, Trail P. Anti-MN plementary DNA sequence and expression in human and rat alimen- antibodies and methods of using same. WO 2007/070538 A2 2007. tary tracts. Gastroenterology 1997;112:398–408. 26. Li Y, Singh B, Ali N, Sarkar FH. Induction of growth inhibition and 12. Saarnio J, Parkkila S, Parkkila AK, Waheed A, Casey MC, Zhou XY, apoptosis in pancreatic cancer cells by auristatin-PE and gemcitabine. et al. Immunohistochemistry of carbonic anhydrase isozyme IX (MN/ Int J Mol Med 1999;3:647–53. CA IX) in human gut reveals polarized expression in the epithelial cells 27. Lou Y, McDonald PC, Oloumi A, Chia S, Ostlund C, Ahmadi A , et al. with the highest proliferative capacity. J Histochem Cytochem Targeting tumor hypoxia: suppression of breast tumor growth and 1998;46:497–504. metastasis by novel carbonic anhydrase IX inhibitors. Cancer Res 13. Zavada J, Zavadova Z, Pastorek J, Biesova Z, Jezek J, Velek J. Human 2011;71:3364–76. tumour-associated cell adhesion protein MN/CA IX: identification of 28. Pacchiano F, Carta F, McDonald PC, Lou Y, Vullo D, Scozzafava M75 epitope and of the region mediating cell adhesion. Br J Cancer A, et al. Ureido-substituted benzenesulfonamides potently 2000;82:1808–13. inhibit carbonic anhydrase IX and show antimetastatic activity 14. Supuran CT. Carbonic anhydrases: novel therapeutic applications for in a model of breast cancer metastasis. J Med Chem 2011;54: inhibitors and activators. Nat Rev Drug Discov 2008;7:168–81. 1896–902.

348 Mol Cancer Ther; 11(2) February 2012 Molecular Cancer Therapeutics

Efﬁcacy of Anti-CA9 Antibody–Drug Conjugate BAY 79-4620

29. Kon-no H, Ishii G, Nagai K, Yoshida J, Nishimura M, Nara M, et al. esophageal and gastric adenocarcinomas. Ann Surg 2006;243: Carbonic anhydrase IX expression is associated with tumor progres- 334–40. sion and a poor prognosis of lung adenocarcinoma. Lung Cancer 31. Treat J, Edelman MJ, Belani CP, Socinski MA, Monberg MJ, Chen 2006;54:409–18. R, et al. Retrospective analysis of outcomes across histological 30. Driessen A, Landuyt W, Pastorekova S, Moons J, Goethals L, Hauster- subgroups in a three-arm phase III trial of gemcitabine in combi- mans K, et al. Expression of carbonic anhydrase IX (CA IX), a hypoxia- nation with carboplatin or paclitaxel versus paclitaxel plus carbo- related protein, rather than vascular-endothelial growth factor (VEGF), platin for advanced non-small cell lung cancer. Lung Cancer a pro-angiogenic factor, correlates with an extremely poor prognosis in 2010;70:340–6.

www.aacrjournals.org Mol Cancer Ther; 11(2) February 2012 349

Therapeutic Mechanism and Efficacy of the Antibody−Drug Conjugate BAY 79-4620 Targeting Human Carbonic Anhydrase 9

Heike M. Petrul, Christoph A. Schatz, Charlotte C. Kopitz, et al.

Mol Cancer Ther 2012;11:340-349. Published OnlineFirst December 6, 2011.

Updated version Access the most recent version of this article at: doi:10.1158/1535-7163.MCT-11-0523

Supplementary Access the most recent supplemental material at: Material http://mct.aacrjournals.org/content/suppl/2011/12/08/1535-7163.MCT-11-0523.DC1

Cited articles This article cites 30 articles, 10 of which you can access for free at: http://mct.aacrjournals.org/content/11/2/340.full#ref-list-1

Citing articles This article has been cited by 7 HighWire-hosted articles. Access the articles at: http://mct.aacrjournals.org/content/11/2/340.full#related-urls

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

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mct.aacrjournals.org/content/11/2/340. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.