Preclinical Efficacy of the Auristatin-Based Antibody–Drug Conjugate BAY 1187982 for the Treatment of FGFR2-Positive Solid Tumors

Antibody–drug conjugates (ADC) represent a promising therapeutic approach for cancer treatment (1, 2). These agents combine the specificity of a monoclonal antibody directed against a cell surface antigen with the targeted delivery of a highly potent cytotoxic drug. Stable conjugation and specific targeting result in the accumulation of high local concentrations of cytotoxic agents that would not be tolerable if administered systemically. Currently, there are more than 40 ADCs in phase I/II of clinical development. Most of these contain a microtubule-destabilizing payload, either of the auristatin or the maytansine class (3). In addition, other payload classes, including DNA targeting pyrrolobenzodiazepines, indolinobenzodiazepine pseudodimers, duocarmycins, calicheamicins, as well as others derived from traditional chemotherapeutics, such as camptothecin and doxorubicin (4, 5), are currently under preclinical and clinical evaluation.

FGFR2 (fibroblast growth factor receptor 2) is a receptor tyrosine kinase with an important role in both embryonic development and tissue repair (6). Alternative gene splicing of the C-terminal half of the Ig-like domain III yields FGFR2 isoforms IIIb and IIIc exhibiting different ligand-binding specificities as well as distinctive expression profiles (6). FGFR2 aberrations have been implicated in multiple cancer types, associated with poor prognosis and resistance to cancer treatments. Oncogenic FGFR2 functions, promoted by FGFR2 overexpression, gene amplification, gene fusions, and autoactivating mutations of the receptor, have been described in several cancers, including gastric, breast, and ovarian cancer (7–14). FGFR2 gene amplification is found in 4% of triple-negative breast cancers (TNBC) and appears to promote breast tumorigenicity by maintaining breast tumor–initiating cells (11, 14). In gastric cancer, FGFR2 is amplified in 5% to 10% of tumors, and FGFR2 mRNA overexpression is associated with poor overall survival (9). Both the prominent expression of FGFR2 in several cancers and the low cell surface expression in normal tissues render FGFR2 a particularly promising target for an ADC.

Herein, we introduce a novel ADC, BAY 1187982, consisting of a fully human FGFR2-specific monoclonal antibody BAY 1179470 conjugated via lysine side chains and a noncleavable linker to an innovative, highly potent microtubule-disrupting auristatin W derivative (15, 16). We show that FGFR2-ADC BAY 1187982 is highly potent and selective in vitro, is stable in circulation, and exhibits strong tumor enrichment in vivo. Furthermore, it demonstrates remarkable antitumor activity in breast, gastric, and ovarian cancer models including patient-derived xenograft (PDX) models.

SNU-16, KATO III, 4T1, MDA-MB-231, and NCI-H716 cells were obtained from ATCC; KYSE-180 and CACO-2 from DSMZ; MFM-223 from ECACC; SUM-52PE from Asterand; and MDR1-LLC1 from Prof. A. H. Schinkel from the Netherlands Cancer Institute (Amsterdam, The Netherlands). Cancer cell lines were obtained between 2002 and 2012 and they were authenticated using short tandem repeat (STR) DNA fingerprinting by the DSMZ before using them in the experiments. Cells were maintained in an incubator at 5% CO₂, 90% humidity, and 37°C in standard cell culture media as indicated by provider.

The discovery of the FGFR2-specific antibody (FGFR2-Ab BAY 1179470) using the n-CoDeR Fab phage library (BioInvent International AB; refs. 17, 18), the expression, purification, and characterization of the Fab precursor and the FGFR2-Ab are described in Supplementary Methods. The FGFR2-Ab (BAY 1179470) bound specifically to FGFR2-positive cells in fluorescence-activated cell-sorting (FACS) experiments, induced internalization into FGFR2-positive cells, and resulted in FGFR2 degradation as observed in an FGFR2 ELISA. Colocalization of the internalized FGFR2-Ab was analyzed with immunocytochemistry and also by a quantitative detection performed with amnis FlowSight imaging flow cytometer. Details of the internalization and localization experiments are described in Supplementary Methods and below.

Preparation of the FGFR2-ADC BAY 1187982 is explained in Supplementary Methods. The microtubule-depolymerizing activity of the main toxophore metabolite (BAY 1168650) of FGFR2-ADC BAY 1187982 and the induction of apoptosis were analyzed in vitro in microtubule polymerization and caspase-3/7 assays, respectively, as explained in Supplementary Methods. CellTiter-Glo (CTG) Luminescent Cell Viability Assay (Promega) was used to determine the half maximal inhibitory concentration of cell viability (IC₅₀) in a panel of cancer cell lines. Cells (7,000 of SNU-16, KATO III, SUM-52PE, NCI-H716, and MFM-223 cells; 4,000 of MDA-MB-231 cells; 3,000 of KYSE-180 cells; and 750 of 4T1 cells per well) were seeded in 96-well plates and incubated for 24 hours. FGFR2-ADC BAY 1187982 or a control ADC (BAY 1160535), consisting of the isotype control Ab BAY 1138806 and the same linker toxophore as the FGFR2-ADC BAY 1187982, were added and the plates were incubated for 72 hours. Cell proliferation was quantified using the CTG assay according to manufacturer's instructions. The raw data were analyzed with a dose–response curve Analysis Spreadsheet developed by Bayer Pharma AG and Bayer Business Services on the IDBS (ID Business Solutions Ltd.) E-WorkBook Suite platform.

To determine the FGFR2 receptor density, the number of antibodies bound per cell (ABC) was measured by flow cytometry in a panel of cancer cell lines as explained in Supplementary Methods. FGFR2 mRNA expression was determined by RNAscope 2.0 following the kit instructions from Advanced Cell Diagnostics using FGFR2 and PPIB (peptidyl prolyl isomerase B) target probes (19). RNA expression was scored according to kit instructions. RNAscope H score 0–400 was calculated as predominant intensity score multiplied by percentage of tumor cells with RNA signal. To analyze FGFR2 amplification, formalin-fixed, paraffin-embedded (FFPE) slides (3 μm) were pretreated and incubated with FISH probes FGFR2-20-RE and CHR10-10-GR from Empire Genomics. For analysis, red (FGFR2) and green (CEN10) dots were counted in 40 nuclei per slide. Samples with large clusters of FGFR2 signal were scored as FGFR2:CEN10 > 5.

The concentration of the toxophore metabolite of FGFR2-ADC BAY 1168650 was analyzed by HPLC coupled to ionization/tandem mass spectrometer detection. Mouse plasma, tumor, or organ homogenates were spiked with 0.5 to 500 μg/L of BAY 1168650 and used for calibration.

All animal experiments were conducted in accordance with the German animal welfare law or Guide for the Care and Use of Laboratory Animals (NIH, Bethesda, MD) and approved by local authorities or with the Singapore animal welfare law and by SingHealth Animal Care Committee authorities. For the SNU-16 xenograft model, tumor cells (2 × 10⁶) were injected subcutaneously to female NOD/SCID mice (Taconic M&B). For MFM-223 and NCI-H716 xenograft models, tumor cells (1 × 10⁷ or 2 × 10⁶, respectively) were injected subcutaneously to female NMRI nu/nu mice (Taconic M&B). Cell inoculation was supplemented with 50% Matrigel (Basement Membrane Matrix, BD Biosciences). For MFM-223 xenografts, mice were implanted with 17β-estradiol pellets (0.37 mg, 90-day release, Innovative Research of America) subcutaneously 1 day prior to the tumor cell inoculation.

For the PDX models BR1115, GA0114, GA0033, PA0787, and ES0199, tumor fragments were inoculated subcutaneously to female BALB/c nude mice (Beijing HFK Bioscience Co. Ltd.) at CrownBio. For BR1115 xenografts, mice received subcutaneous implantation of 17β-estradiol pellets (0.05 mg, 60-day release, Innovative Research of America) on the day of tumor inoculation. For OV30-0511A, GC10-0608, and GC12-0811 xenografts, tumors were inoculated subcutaneously to female (OV30-0511A) or male (GC10-0608 and GC12-0811) SCID mice (Animal Resources Centre) at the laboratory of Prof. Huynh Hung (NCC Singapore). MAXF 857 tumor fragments were inoculated subcutaneously to female NMRI nu/nu mice (Harlan Laboratories, BV, Venray 5804 AB) at Oncotest GmbH. MRL 2003100375 tumor cells (5 × 10⁵) were suspended in 50% Cultrex ECM (Trevigen) and injected subcutaneously to female NOD/SCID mice (Harlan Laboratories) at Molecular Response Laboratories.

In all in vivo experiments, mice (n = 5–30 per group, see figure legends) were randomized according to primary tumor size prior to treatment. Unless otherwise indicated, FGFR2-ADC BAY 1187982 and the control ADC were administered once weekly intravenously and unconjugated FGFR2-Ab twice weekly (intravenously). Doxorubicin (10 mg/kg; Sigma) was given every 2 weeks (intravenously). Vinorelbine (10 mg/kg; Actavis) and paclitaxel (24 mg/kg; Lapharm GmbH) were administered once weekly (intravenously). PBS was used as the vehicle control. Tumor volume [(length × width²)/2] was measured by caliper at least twice weekly, and the treatment response was defined using the RECIST criteria (20). Progressive disease (PD) was defined as greater than 20% increase in tumor size. Partial response (PR) was defined as greater than 30% reduction in tumor size. Complete response (CR) was defined as an absence of any palpable tumor mass. No tumor growth or a slight reduction (<30%) or small increase (<20%) in tumor size was defined as stable disease (SD). Treatment-to-control ratios (T/C) were calculated on the basis of mean tumor volumes.

To determine total (t-FGFR2) and phosphorylated (P-FGFR2) FGFR2 in tumor lysates, tumors were cut into 5 × 5 mm² fragments, lysed and analyzed by Prometheus’ collaborative enzyme enhanced reaction (CEER) assay (21). FGFR2 immunoblot analysis was performed on a panel of ovarian cancer PDX models. Tumor samples were lysed and separated on SDS-PAGE followed by blotting and incubation with the anti-FGFR2 rabbit polyclonal antibody Bek C-17 (sc-122) from Santa Cruz. To analyze the mode of action of BAY 1187982 in vivo, FFPE samples of the PDX model BR1115 were stained for phospho-histone H3 (pHH3, Ser10) as marker of mitotic cells with chromosomes in G₂–M phase, cleaved PARP1 as marker of apoptosis, and for α-tubulin to indicate tubulin structures as explained in Supplementary Methods.

For the comparison of t-FGFR2 and P-FGFR2, statistical significance was determined using Kruskal–Wallis test, followed by Mann–Whitney U test with Holm–Bonferroni correction. The comparison of final tumor weights was performed by one-way ANOVA, followed by a Dunnett test. For the final tumor volume, the log-transformed data were analyzed using one-way ANOVA and Tukey HSD test or Kruskal–Wallis test, followed by Mann–Whitney U test with Holm–Bonferroni correction. All analyses were compared to vehicle group and performed using statistical software R (version 3.1.2). P < 0.05 was considered statistically significant.

The monoclonal antibody BAY 1179470 (FGFR2-Ab) was selected because of its high binding affinity to the extreme N-terminus of human FGFR2 (K_d of 75 nmol/L) as detected by surface plasmon resonance, the critical residues being Pro-2, Leu-6, and Glu-8. The N-terminal epitope recognized by this FGFR2-Ab is present in all described FGFR2 splice variants and is 100% identical in human, rat, rhesus monkey, and murine FGFR2. Consequently, the selected FGFR2-Ab binds specifically to FGFR2 (but not FGFR1, FGFR3, or FGFR4) and shows a wide cross-species reactivity (EC₅₀ of 0.25–0.35 nmol/L for mouse, rat, dog, pig, and rhesus monkey FGFR2 protein) allowing informative preclinical safety studies. Fluorescence microscopy analyses revealed that the FGFR2-Ab induced rapid internalization of FGFR2 in FGFR2-positive cancer cells (SUM-52PE, SNU-16), which was not observed in FGFR2-negative cancer cells (MDA-MB-231) or with an isotype control Ab (Fig. 1A). Antibody-mediated internalization was confirmed by FACS analysis and measurement of FGFR2 degradation by ELISA in SNU-16 cells (Fig. 1B and C, respectively). Internalized FGFR2-Ab colocalized with lysosomal Rab7 (early endosome marker) and LAMP1 (lysosomal-associated membrane protein 1) but not with recycling endosome marker Rab11 or late endosome marker M6PR (mannose-6-phosphate receptor), indicating that FGFR2-Ab induced FGFR2 intracellular trafficking will result in routing of FGFR2-Ab to the lysosome as shown also with immunocytochemistry (Fig. 1D) and lead to degradation of the antibody (Fig. 1C). The colocalization of FGFR2-Ab with the lysosomal marker LAMP1 or with the early endosome marker Rab7 was observed in approximately 60% of SUM-52PE cells as determined using imaging flow cytometer.

To generate the FGFR2-ADC, FGFR2-Ab was coupled to a novel, highly potent N-methyl auristatin W derivative. Cysteine-linked ADCs with thiosuccinimide linkages can undergo deconjugation via retro-Michael reaction resulting in a partial loss of drug load in vivo (22). To achieve maximum stability and to avoid FGFR2-ADC deconjugation in vivo, the auristatin W was attached to lysine side chains of the Ab via an N-(5-carboxypentyl) linker (16). The N-(5-carboxypentyl)–modified auristatin W derivative was transformed into an activated N-hydroxy succinimide ester and was subsequently coupled to the FGFR2-Ab to generate the FGFR2-ADC BAY 1187982 (Fig. 2A) with an average drug-to-antibody ratio (DAR) of 4 (range, 1.7–5.1). Both the unconjugated antibody FGFR2-Ab and the FGFR2-ADC BAY 1187982 exhibited antigen-binding affinities of 0.29 nmol/L as determined by ELISA, indicating that linker payload conjugation does not affect the antigen recognition of the FGFR2-Ab in vitro.

The microtubule-depolymerizing activity of the auristatin toxophore metabolite (BAY 1168650) of BAY 1187982 was first confirmed in an in vitro microtubule polymerization assay with toxophore linker metabolite (Supplementary Fig. S1A). In vitro cytotoxicity of BAY 1187982 was subsequently analyzed in a broad panel of cancer cell lines in comparison to the nonbinding control ADC containing an identical linker payload. To correlate in vitro potency with FGFR2 receptor density on the cell surface, the number of FGFR2 ABC count was determined for a panel of cancer cell lines by quantitative FACS analysis. High FGFR2 levels were observed in KATO III, SUM-52PE, NCI-H716, MFM-223, and SNU-16 cell lines (Supplementary Table S1). The ABC count, an indirect measure for cell surface FGFR2 expression, correlated well with the in vitro potency of FGFR2-ADC BAY 1187982 with IC₅₀ values ranging from 0.097 to 0.83 nmol/L in cell lines harboring at least 10,000 antibody-binding sites. Importantly, BAY 1187982 was more than 100-fold selective compared with the nonbinding control ADC. Both caspase-3/7 activation assay and FACS-based cell-cycle analysis demonstrated that treatment with FGFR2-ADC, but not with the control ADC, activated caspase-3/7 and increased the number of cells with a sub-G₁ content in FGFR2-positive SNU-16, respectively. This was not observed in MDA-MB-231, which express very low levels of FGFR2 (Supplementary Fig. S1B and S1C), strongly indicating that FGFR2-ADC induces apoptosis in vitro specifically in FGFR2-positive cells.

Because the therapeutic effects of ADCs are dependent on the tumor-specific delivery of the toxophore, we next studied the concentration of the main toxophore metabolite BAY 1168650 in tumors, liver, kidneys, and spleen in NMRI nu/nu mice bearing NCI-H716 tumors following intravenous administration of 5 mg/kg FGFR2-ADC BAY 1187982 with a DAR of 1.7 or 4.5 (Fig. 2B). The plasma concentration values for BAY 1187982 with a DAR of 4.5 were C_max of 165 mg/L, AUC of 13,614 mg h/L, and t_1/2 of 196 hours. The respective values for BAY 1187982 with a DAR of 1.7 were in a comparable range (C_max = 131 mg/L, AUC = 13,345 mg *h/L, and t_1/2 = 182 hours). The active toxophore metabolite BAY 1168650 was found to be enriched more than 30-fold in tumors versus normal tissues (liver, spleen, and kidneys) as expected because of the high FGFR2 expression level on NCI-H716 tumor cells. BAY 1187982 with DAR of 4.5 yielded an approximately 2-fold higher concentration of the toxophore metabolite in tumors as compared with the ADC with a DAR of 1.7. However, the tumor/organ ratio of the metabolite was independent of the DAR value of the BAY 1187982 with AUC(0–t_last) ratios 27.7 versus 26.6 in liver, 32.3 versus 41.0 in spleen, and 103.2 versus 125.3 for DAR 1.7 and 4.5, respectively. Importantly, there is no evidence that BAY 1187982 would exert a bystander effect, as the toxophore metabolite has a low membrane permeability in CACO-2 cells as well as in the P-glycoprotein–expressing MDR1-LLC1 cell line (data not shown).

To analyze the in vivo efficacy of BAY 1187982, the FGFR2-ADC was first tested in the human SNU-16 gastric cancer xenograft model using two different DARs (1.8 and 4.5) and two treatment schedules (Q4D×3 and Q10D×3). In contrast to the control ADC, treatment with FGFR2-ADC BAY 1187982 at 5 mg/kg resulted in partial tumor regression in at least 90% of the animals irrespective of the DAR and treatment schedule used (Fig. 3A and B). Lower FGFR2-ADC doses of 0.5 or 1 mg/kg did not significantly inhibit tumor growth compared with the vehicle control. To further assess the duration of tumor growth inhibition induced by BAY 1187982 at 5 mg/kg, animals were maintained for additional 30 days without treatment. On day 62, tumor regrowth was evident in all BAY 1187982-treated animals (Fig. 3A). The antitumor activity of BAY 1187982 was selective, as the nontargeting control ADC had no apparent effect on tumor growth.

In a separate experiment using an intermediate dosing schedule of Q7D×3, the lowest dose sufficient to induce PR was determined as 1.25 mg/kg. Weekly administration of BAY 1187982 (Q7D×3) at 2.5, 5, and 10 mg/kg resulted in a significant inhibition of tumor growth including the induction of PRs (P < 0.05, Fig. 3C). Treatment with BAY 1187982 was well tolerated. The highest dose of 10 mg/kg resulted in 10.5% transient reduction of the mean body weight 4 days after start of therapy. Animals regained normal body weights within 1 week.

When tested in the MFM-223 TNBC model, BAY 1187982 treatment at 1 and 5 mg/kg resulted in a marked decrease in tumor volume (P < 0.001), similarly to the standard-of-care doxorubicin (P = 0.001, Fig. 3D and E). Moreover, BAY 1187982 at 5 mg/kg resulted in PRs in all mice, whereas the lower dose of 1 mg/kg was sufficient to achieve PRs in 6 of 10 mice (Fig. 3F). Similarly to our experiment with the SNU-16 cell line, neither the control ADC (Fig. 3D and E) nor the unconjugated FGFR2-Ab (antitumor activity with T/C ratio of 1.22) had an effect on tumor growth. Finally, the impact of different treatment schedules was also assessed in the MFM-223 model in which administration of BAY 1187982 at all doses resulted in strong antitumor effects as indicated by a high number of mice with CR (P < 0.001, Supplementary Fig. S2A and S2B).

Interestingly, the significant antitumor efficacy of BAY 1187982 correlated with a substantial decrease in t-FGFR2 and P-FGFR2 protein levels in MFM-223 tumors at the end of the experiment (P < 0.001, Fig. 3G). In contrast, the nontargeted control ADC showed variable effects on FGFR2 protein, with 5 mg/kg suppressing the t-FGFR2 level and 1 mg/kg leading to a minor increase in P-FGFR2 protein content in the tumor.

In addition to being highly effective in MFM-223 and SNU-16 models, FGFR2-ADC BAY 1187982 at 7.5 mg/kg resulted in notable inhibition of tumor growth also in another cell line–derived xenograft model overexpressing FGFR2, the NCI-H716 human colorectal cancer model (P < 0.001, Table 1). In this model, neither paclitaxel nor vinorelbine inhibited tumor growth (T/C ratios, 0.91 and 1.02, respectively).

To further substantiate the antitumor efficacy of BAY 1187982, the FGFR2-ADC was evaluated in several PDX mouse models representing various cancer types with different FGFR2 expression and amplification level (Table 1).

In the ovarian cancer PDX model OV30-0511A, BAY 1187982 at 7.5 and 15 mg/kg led to a remarkable and specific inhibition of tumor growth with T/C values of 0.08 and 0.12, respectively (P < 0.001, Fig. 4A and B). Cisplatin and paclitaxel showed only moderate antitumor efficacy with T/C values of 0.35 and 0.38, respectively. OV30-0511A–bearing mice were sensitive to vinorelbine, which showed comparable efficacy as BAY 1187982 with a T/C value of 0.14. These tumors exhibited strong FGFR2 amplification and high FGFR2 protein expression (Fig. 4C and D).

The potent antitumor efficacy of BAY 1187982 at 7.5 mg/kg was apparent in GA0033 gastric cancer PDX model with high FGFR2 expression (P = 0.0267, Fig. 5A). Tumor eradication was achieved in all mice (5 of 5) treated with BAY 1187982, whereas no tumor growth inhibition was observed in mice treated with the control ADC (P = 0.1667).

Furthermore, BAY 1187982 at 7.5 mg/kg resulted in strong inhibition of tumor growth in an FGFR2-positive BR1115 breast cancer model with a T/C ratio of 0.37 (P < 0.001, Fig. 5B). In contrast, no effect on tumor growth was observed for the control ADC indicating high selectivity of BAY 1187982.

Importantly, the unconjugated FGFR2-Ab was not efficacious in any of the PDX models with T/C ratios 0.97, 1.38, and 1.03 in OV30-0511A, GA0033, and BR1115 models, respectively.

To analyze the mode of action of BAY 1187982 in vivo, FFPE samples of the BR1115 model were stained for phospho-histone H3 (pHH3, Ser10) as marker of cells in G₂–M phase, cleaved PARP1 as marker of apoptosis and for α-tubulin to indicate tubulin structures. In BAY 1187982–treated BR1115 tumors, a 2-fold increase of cells with colocalization of α-tubulin–stained mitotic spindles with pHH3 as marker for G₂–M phase chromosomes compared with the control ADC and the vehicle-treated groups was observed, indicating induction of G₂–M phase arrest after specific uptake of BAY 1187982 into FGFR2-positive cells in vivo. In both BAY 1187982 and vinorelbine-treated BR1115 tumors, a 4-fold increase of cleaved PARP1-positive tumor cells was detected compared to vehicle or control ADC–treated tumors, indicating that BAY 1187982 induces apoptosis in vivo in a comparable range to vinorelbine which has a similar mode of action, that is, microtubule depolymerization.

Taken together, a positive correlation was observed between FGFR2 amplification and/or mRNA expression and antitumor activity in vivo regardless of the cancer type (Table 1). In general, no response in vivo is observed with tumor models with no or low FGFR2 expression. Variable responses were observed when models with intermediate FGFR2 amplification and/or mRNA expression were evaluated.

FGFR2 aberrations, such as gene amplifications as well as protein and RNA overexpression, have been implicated in the development and progression of multiple cancer types and are commonly associated with poor prognosis and resistance to cancer treatments (7, 9–11, 13). FGFR2 is highly expressed in several cancers and exhibits only restricted expression in normal tissues and organs, making it a valuable cancer target and an ideal candidate for the development of an ADC. Here, we have generated an FGFR2-specific human monoclonal Ab (BAY 1179470), which binds to FGFR2-IIIb and FGFR2-IIIc. We demonstrate that the unconjugated control Ab can be used to efficiently target FGFR2-expressing tumor cells, resulting in receptor internalization and FGFR2 degradation. However, use of the unconjugated antibody alone is not sufficient to inhibit growth in most FGFR2-positive tumor models tested indicating an FGFR2-independent mode of cell survival in some tumors. Therefore, we utilized the FGFR2-Ab (BAY 1179470) as a targeting moiety and conjugated it via lysine side chains to a novel microtubule-disrupting auristatin W derivative via a noncleavable linker. Here, we provide the first preclinical evidence of the successful use of a highly potent FGFR2-ADC (BAY 1187982) for the treatment of FGFR2-positive tumors.

BAY 1187982 demonstrates antitumor activity in SNU-16 gastric cancer and, importantly, also in MFM-223 TNBC model previously shown to be unresponsive to the unconjugated FGFR2-Ab. This suggests that the specificity and efficacy of BAY 1187982 in eliminating FGFR2-expressing tumor cells is largely due to the action of the conjugated payload.

N,N-Dialkyl auristatin W derivatives are highly efficient microtubule-disrupting agents (23). ADCs employing such payloads were optimized for high potency, selectivity, as well as maximum linker stability to avoid ADC deconjugation in vivo. We show that targeted release and intracellular accumulation of BAY 1168650, the non–cell-permeable active metabolite of BAY 1187982, exerts potent cytotoxic effects in several FGFR2-positive tumor models. The mode of action of the toxophore metabolite is shown to be microtubule depolymerization resulting in induction of apoptosis in vivo comparable to vinorelbine. Intriguingly, BAY 1187982 also demonstrates activity in tumor models (e.g., the NCI-H716 colorectal cancer model) that are not sensitive to other microtubule-targeting agents such as paclitaxel and vinorelbine.

BAY 1187982 resulted in dose-dependent tumor regression including PRs and CRs in TNBC and gastric cancer, and tumor stasis in in vivo models of ovarian cancer and was well tolerated. Importantly, we observed a positive correlation between FGFR2 protein levels and BAY 1187982 cytotoxic potency in vitro (Supplementary Table S1). Similarly, a positive correlation was observed between FGFR2 amplification and mRNA expression and antitumor activity in vivo (Table 1). Consequently, FGFR2 mRNA levels and FGFR2 gene amplification present themselves as attractive selection markers for patient stratification during clinical development.

Taken together, we report the identification and characterization of BAY 1187982, a potent and selective FGFR2-ADC for treatment of FGFR2-positive human malignancies. Phase I study (NCT02368951) with BAY 1187982 is currently ongoing.

H. Apeler, M. Braun, F. Dittmer, S. Hammer, R. Jautelat, H.-G. Lerchen, C.F. Nising, F. Reetz, J. Schuhmacher, A. Sommer, and K. Ziegelbauer have ownership interest as shares in Bayer AG. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Sommer, C. Kopitz, C.F. Nising, H.-G. Lerchen, B. Stelte-Ludwig, S. Wittemer-Rump, A. Harrenga, F. Dittmer, F. Reetz, H. Apeler, K. Ziegelbauer, B. Kreft

Development of methodology: C. Kopitz, C. Mahlert, H.-G. Lerchen, B. Stelte-Ludwig, S. Hammer, S. Greven, M. Braun

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Sommer, C. Kopitz, C.A. Schatz, C.F. Nising, C. Mahlert, B. Stelte-Ludwig, S. Hammer, S. Greven, J. Schuhmacher, A. Harrenga, H. Huynh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Sommer, C. Kopitz, C.A. Schatz, C.F. Nising, C. Mahlert, B. Stelte-Ludwig, S. Hammer, S. Greven, J. Schuhmacher, M. Braun, H. Zierz, A. Harrenga, H. Apeler

Writing, review, and/or revision of the manuscript: A. Sommer, C. Kopitz, C.F. Nising, C. Mahlert, B. Stelte-Ludwig, S. Hammer, J. Schuhmacher, M. Braun, S. Wittemer-Rump, A. Harrenga, F. Dittmer, F. Reetz, H. Apeler, K. Ziegelbauer, B. Kreft

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Stelte-Ludwig, S. Hammer

Study supervision: A. Sommer, C. Kopitz, B. Stelte-Ludwig, R. Jautelat, K. Ziegelbauer, B. Kreft

Other [as a chemist in the project, he designed and synthesized the ADC molecule (payload-linker synthesis and conjugation)]: H.-G. Lerchen

We thank Anna Behnke, Susanne Bendix, Sandra Berndt, Tim Brandenburger, Henryk Bubik, Anna DiBetta, Norman Dittmer, Karola Henschel, Sabine Jabusch, Katrin Jänsch, Beate König, Nadja Langner, Stefanie Mai, Bettina Muchow, Jenny Stepan, Rukiye Tamm, Jan Tebbe, Bianka Timpner, Simone Zolchow, Nina Wobst, and Dirk Wolter for excellent technical assistance. We thank Seattle Genetics Inc. for their support and Aurexel Ltd. (www.aurexel.com) for editorial support funded by Bayer Pharma AG.

This work was supported by a grant from the National Medical Research Council of Singapore (NMRC/MOHIAFCat1/0002/2014) to Hung Huynh.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.