Radioimmunotherapy of Fibroblast Activation Protein Positive Tumors by Rapidly Internalizing Antibodies

Fibroblast activation protein (FAP) has emerged as a key player in cancer physiology with multiple biologic functions. Most prominently, FAP is highly expressed on carcinoma-associated fibroblasts where it is thought to contribute to cancer initiation, progression, and metastasis (1–3). In addition, FAP is also present in some cancer cells, including bone and soft tissue sarcoma (4) and some melanoma (5) cells. FAP is a type II transmembrane serine protease that gains its enzymatic activity upon homodimerization of 2 identical 97 kDa subunits (6, 7). The natural substrates of FAP are still not completely identified, but the protease is capable of cleaving N-terminal dipeptides from polypeptides with proline or alanine in the penultimate position (8) and also has collagenase activity (9–11). The dipeptidyl-peptidase activity of FAP is one of the mediators of tumor progression (12), extracellular matrix remodeling (13–18), and metastasis formation (10, 19), but recent studies strongly suggest additional effects of FAP in the absence of its enzymatic activity (20) by hitherto unclear mechanisms. Very recently, FAP-expressing stromal cells have been shown to suppress antitumor immunity (21), adding yet another aspect to the manifold contributions of FAP to tumor growth.

Because of this multifaceted influence on tumor physiology, its restricted expression pattern within diseased areas and correlation with poor clinical outcome in patients with cancer (22, 23), FAP has emerged as a promising target for cancer therapy. A humanized version of mAb F19 (sibrotuzumab), a first generation anti-FAP antibody, however, failed to show measurable therapeutic activity (24) in phase I/II clinical trials despite excellent tumor stroma targeting properties (25). In addition, 8 of 26 sibrotuzumab-treated patients developed human–anti-human antibodies (HAHA) with a change in pharmacokinetics and reduced tumor uptake in 4 of them. Therefore, further clinical development of sibrotuzumab was halted.

Strategies to target cancer with modified FAP-specific antibodies have recently gained interest and new technologies for the conjugation of toxins or cytokines to FAP-specific antibodies have been established. For example, an antibody–maytansinoid conjugate (26) and antibody–TNF fusion proteins (27, 28) have shown promise at the preclinical stage and are pending clinical development.

Radioimmunotherapy, the targeted delivery of therapeutic radioisotopes to the tumor site by monoclonal antibodies, is an attractive strategy for cancer treatment. Notably, antigen expression is usually spatially heterogeneous within the tumor. With FAP being restricted to carcinoma-associated fibroblasts, the therapeutic efficacy can be augmented by the “cross-fire” effect of β-emitting radionuclides on neighboring FAP-negative tumor cells. Clinical trials with ¹³¹I-labeled sibrotuzumab have been conducted to investigate targeting properties and pharmacokinetics but were discontinued because of the above-mentioned issues associated with HAHA (25). While radioiodination with ¹³¹I is relatively easy to conduct, the radionuclide is readily released from the tumor upon internalization into cells. In contrast, high doses of residualizing radioisotopes may accumulate at the tumor site with internalizing antibodies and lead to a persistent irradiation of the tumor tissue. The radiolanthanide ¹⁷⁷Lu has a comparable range of the β-particles in tissue as ¹³¹I (2 and 3 mm, respectively). In contrast to the high energy γ-photons of ¹³¹I, which are a disadvantage from a radiation safety perspective, ¹⁷⁷Lu emits 208 keV γ-photons at much lower abundance (11%).

For preclinical evaluation of radioimmunoconjugates, human–mouse cross-reactive antibodies are preferentially used to detect possible accumulation in tissue expressing homologous FAP. This allows more accurate estimation of side effects that may arise from radioactive doses delivered to healthy organs. In addition, for targeting of the tumor stroma in preclinical models, crossreactivity with the homologous antigen of the model species is a prerequisite.

In this report, we describe the successful selection of 2 human monoclonal antibodies, ESC11 and ESC14, from a phage-display library. In contrast with F19 (sibrotuzumab), we found that these novel antibodies were rapidly internalized by FAP-positive cells. Using ESC11 and ESC14 labeled with the β-emitting radiolanthanide ¹⁷⁷Lu, we found that both antibodies specifically accumulated in xenografted FAP-positive human melanoma and delayed tumor growth in vivo.

HEK293huFAP, HEK293muFAP, and HEK293huCD26 were generated by cloning huFAP, CD26, or muFAP cDNA into the pEAK8 vector (Edge Biosystems) and stable transfection of HEK293 c-18 cells (ATCC CRL-10852) followed by monoclonalization by limited dilution. The cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% FBS, penicillin–streptomycin, and puromycin (3 μg/mL). HT1080 FAP⁺, huFAP-expressing HT1080 [American Type Culture Collection (ATCC) CCL-121], were maintained in RPMI supplemented with 10% FBS (Gibco, Invitrogen) and 200 μg/mL G418. Human melanoma cell lines SK-Mel-187 and SK-Mel-16 were kindly provided by Ludwig Institute for Cancer Research (New York) and maintained in RPMI 10% FBS. Murine antibody F19 and human anti-A33 antibody were obtained from Ludwig Institute for Cancer Research (Melbourne, Australia). Humanized vF19 was generated by veneering of F19 and was supplied by the Ludwig Institute for Cancer Research.

For selection of huFAP-muFAP–specific antibodies, a nonimmunized phage library–expressing human antibody Fab fragments was used (29). Details of the selection procedure are outlined in the Supplementary Data.

For screening of FAP-binding phage by ELISA, 96-well microtiter plates (MaxiSorp Nunc) were coated with cell extracts from HEK293 huFAP, HEK293 huCD26, HEK293 muFAP, or mock-transfected HEK293, blocked with 5% milk powder in PBS, incubated with phage-containing supernatants for 1 hour at room temperature, and developed using an anti-M13-HRP–conjugated antibody and tetramethylbenzidin as substrate.

Fab fragments were produced in E. coli TG-1 by induction with 1 mmol/L isopropyl-l-thio-B-[scap)d[r]-galactopyranoside (IPTG) for 4 hours at 30°C. Soluble Fab was released from the periplasmic fraction by incubation in PBS, pH 8 at 4°C overnight. Crude fractions were incubated with Talon resin (CLONTECH Laboratories) for 1 hour at 4°C, washed with wash buffer (20 mmol/L Tris, 100 mmol/L NaCl, 0.1% Tween 20), eluted with 100 mmol/L imidazole, and dialyzed against PBS at 4°C overnight.

Binding analysis of Fabs was conducted on a BIAcore T100 instrument. Recombinant huFAP (R&D systems) was immobilized at low density on a CM5 sensor chip (Biacore AB) using amine coupling chemistry according to the manufacturer's instructions. ESC11 and ESC14 Fabs were injected in 10 mmol/L HEPES, pH 7.4, 3.4 mmol/L EDTA, 0.15 mmol/L NaCl, and 0.005% Tween 20 at a flow rate of 80 μL/min at 25 °C, and concentrations ranging from 0.5 to 16 nmol/L and 6.25 to 200 nmol/L, respectively. The sensor chip was regenerated between injections with 0.1 mol/L sodium carbonate and rate constants were calculated using BIAevaluation data analysis program.

The variable sequences of heavy and light chain were cloned via DraIII and RsrII sites, respectively, into a modified pEE12.4 vector (LONZA Biologics) encoding human constant IgG1 domains. Stable antibody-producing cell lines were developed using NS0 cells and the glutamine synthetase selection system according to the manufacturer's instruction (LONZA Biologics). Human IgG were purified from the culture supernatant by binding to CaptureSelect Fab kappa Affinity Matrix (CaptureSelect) followed by elution with 0.1 mmol/L glycine pH 2.5 and dialysis against PBS at 4°C overnight.

Cells were harvested at 50% confluence and stained with primary and secondary antibodies for 60 and 30 minutes on ice, respectively. Each staining step was followed by 2 washing steps (1 mL PBS and 2% FBS). Cells were fixed with 2% paraformaldehyde for 10 minutes on ice before conducting flow cytometry. At least 10,000 cells were measured on a FACScan (BD) cytometer and analyzed using FlowJo software (Treestar). For determination of the apparent K_D, the mean fluorescence intensities of a serial dilution were plotted and a dose–response curve was fitted using the equation f = Bmax*abs(x)/(K_Dapp+abs(x))+Ns*x with SigmaPlot software (Systat Software Inc.). Human primary antibodies were detected with FITC/DyLight488 conjugated goat-anti-human antibody, (Jackson Immunoresearch, cat. no. 109-485-003; 1:200 dilution), whereas murine antibodies were stained with anti-mouse DyLight 488-labeled antibody (Jackson Immunoresearch, cat. no.115-486-146; 1:200 dilution). Fab antibodies were detected with anti-myc tag antibody 9E10, followed by an anti-mouse immunoglobulin-PE conjugate, (Jackson Immunoresearch, cat. no. 115-165-003; 1:200 dilution). Quantification of cell-surface FAP in internalization experiments was conducted after overnight incubation with human IgGs at 37°C. Cells were then stained with 10 μg/mL F19 followed by anti-mouse FITC-conjugated antibody and % FAP expression was calculated relative to the mean fluorescence intensity of cells that have not been incubated with human IgG. In case of vF19, directly labeled ESC11 (DyLight 549) was used to detect cell-surface FAP.

Cells were seeded on sterile 20 mm fibronectin-coated (Sigma F1141-1MG, at 10 μg/mL) coverslips (Deckglaser; Carolina Biological) in 12-well plates and grown to 20 to 30% confluency. Cells were fixed with 2% paraformaldehyde on ice and permeabilized using Cytofix/Cytoperm solution (51-2090KZ; BD Biosciences). Primary antibodies were diluted in 1× Perm/Wash (51-2091KZ; BD Biosciences) at following dilutions: EEA1, 1:100 (2411; Cell Signaling Technologies), LAMP1, 1:1,000 (ab24170; Abcam), F19, 1:100. All secondary antibodies were used at a 1:200 dilution and were obtained from Jackson Immunoresearch. Dynasore hydrate (D7693; Sigma) and 5-(N-ethyl-N-iso-propyl)amiloride (EIPA; A3085; Sigma) were added 30 minutes before the experiment at a final concentration ranging from 0 to 100 μmol/L and 0 to 10 μmol/L, respectively. Cells were fixed with 4′, 6-diamidino-2-phenylindole anti-fade (Invitrogen). All images were acquired on Leica SP5 UV/Vis confocal microscope (Leica Microsystems) at the Center for Microscopy and Image Analysis (University of Zurich, Zurich, Switzerland) using LAS AF software. Images were kept in 1024 × 1024 formats and acquired with a zoom factor of 4 at 700 Hz frequency with HCxAPO Lambda blue 63.0× oil UV objective lens at a numerical aperture of 1.40. Images were stored in Tagged Image File Format (TIFF) and further processed using Image J software.

For internalization experiments, cells were incubated with 30 μg/mL human IgGs for the indicated time. FAP was then detected with F19 and DyLight 549 anti-mouse antibody, (Jackson Immunoresearch, cat. no. 115-506-062; 1:200 dilution). After incubation with vF19, FAP was detected with directly conjugated ESC11-DyLight 549. Alternatively, internalization was directly induced and monitored by DyLight 549-conjugated human IgG1 (30 μg/mL).

Antibodies were labeled with ¹⁷⁷Lu using the chelator CHX-A″-DTPA (N-[(R)-2-Amino-3-(p-isothiocyanato-phenyl) propyl]-trans-(S,S)- cyclohexane-1,2-diamine-N,N,N′,N″,N″-pentaacetic acid, B355; Macrocyclics). Conjugation was conducted in borate buffer (0.07 mol/L, pH 9.15) under metal-free conditions and using a 5-fold molar excess of CHX-A″-DTPA. The reaction mixture was incubated for 1 hour at 37°C. For all antibodies, 0.5 to 1 moles CHX-A″-DTPA, were coupled per one mole IgG1, as determined by incubation of a fixed amount of conjugated antibody with increasing amounts of ¹¹¹In (Mallinckrodt). The conjugated antibody was separated from free CHX-A″-DTPA using a NAP-5 column equilibrated with ammonium acetate buffer (0.25 mol/L, pH 5.5). CHX-A″-DTPA-antibody conjugates were labeled with ¹⁷⁷Lu ([¹⁷⁷Lu]Cl₃ in 0.04 mol/L HCl, ITG Isotope Technologies Garching-GmbH) for 1 hour at 37°C.

The reaction was quenched with 5 mmol/L EDTA and after 5 minutes of incubation at 37°C, the radiolabeled antibody was purified on a Superose-12 FPLC column (Amersham Biosciences). Elution was carried out in PBS at a flow rate of 0.5 mL/min. Radiochemical purity was analyzed by fast protein liquid chromatography (FPLC). The immunoreactive fraction was determined on HT1080 FAP cells using the Lindmo method as previously described (30).

Post mortem SPECT scans (2 hours 180 seconds/view) were conducted with a 4-head multiplexing multipinhole camera (NanoSPECT/CT; Bioscan Inc.) 72 hours after injection of ¹⁷⁷Lu-CHX-A″-DTPA-IgG1 (8 MBq, 15 μg) into the tail vein of nude mice bearing an SK-MEL-187 tumor on the hind leg and an SK-MEL-16 tumor on the shoulder. SPECT data were reconstructed iteratively with HiSPECT software (Scivis GmBH). The fused SPECT and CT datasets were analyzed using InVivoScope post processing software (Bioscan Inc.).

Female nu/nu mice (Charles River) were injected subcutaneously in the right hind leg with 5 × 10⁶ SK-Mel-187 cells in RPMI: BD Matrigel (cat. 354248; BD Biosciences) = 1:1 in a final volume of 200 μL. Tumors were allowed to grow to approximately 200 mm³ before animals were randomly divided into groups of 6 mice for biodistribution studies (4 mice for time-dependent biodistribution of ¹⁷⁷Lu-ESC11). Four micrograms (3 MBq) of each radiolabeled antibody was injected intravenously in 100 μL PBS. Animals were sacrificed by cervical dislocation. Radioactivity for each organ was measured using a gamma counter. The experiments were approved by the local ethics committee for animal research.

Mice were injected subcutaneously with 5 × 10⁶ FAP-positive SK-Mel-187 cells as described above and randomized into groups of 7 to 8 mice. 8 MBq radiolabeled antibody (100 μL in PBS, 15–17 μg) or 100 μL PBS was injected into the tail vein 5 days after tumor cell implantation. Tumor growth was measured every 2 to 3 days with a caliper and tumor volumes were calculated according to the formula: volume = length × width² × 0.52. Animals were sacrificed when tumor volumes exceeded 1,000 mm³. Additional mice were sacrificed 3 and 7 days after treatment with 8 MBq ¹⁷⁷Lu-ESC11 or ¹⁷⁷Lu-A33 and tumors were processed for immunohistochemistry.

Whole sections (3 μm thickness) were cut from formalin-fixed, paraffin-embedded blocks. Tissues were mounted on glass slides and deparaffinized in xylene. Rehydration was carried out in decreasing concentrations of ethanol and stained with hematoxylin–eosin using standard histologic techniques. Ki-67 immunohistochemistry was conducted using the automated Bond Max platform (Leica Microsystems) with a rabbit monoclonal MIB-1 antibody (1:200; clone SP6, Neomarkers, LabVision). Heat-induced epitope retrieval pretreatment was conducted using H2-Buffer (Leica Microsystems) and boiling at 95°C for 30 minutes. Bound antibody was detected with a corresponding secondary antibody included in the Refine-DAB detection kit (Leica Microsystems).

Data were analyzed using Prism (version 5.01) software (Graph Pad). Survival times were analyzed by Kaplan–Meier analysis, followed by the log-rank test. To correct for multiple comparisons of survival times, the Bonferroni-corrected threshold was applied to determine significance.

Monoclonal antibody Fab-fragments were selected from a large human Fab antibody library by phage display. The first 4 panning rounds were conducted on human FAP (huFAP; Supplementary Fig. S1). Enrichment for Fab-fragments that crossreacted with mouse FAP (muFAP) was conducted in the fifth panning round (Supplementary Table S1). Screening of 300 selected clones by ELISA followed by PCR fingerprinting and sequencing of FAP-specific binders (data not shown) led to the identification of Fabs ESC11 and ESC14. By ELISA, both Fab fragments specifically bound to huFAP and muFAP and did not crossreact with the homologous protein CD26 (Fig. 1A). Furthermore, Fab ESC11 and ESC14 bound to native FAP on the cell membrane of muFAP or huFAP-expressing cells (Fig. 1B). The binding affinity to recombinant huFAP was determined by surface plasmon resonance measurement using chips coated with a low density of huFAP. K_D values of 10 ± 5 nmol/L and 210 ± 35 nmol/L were calculated for Fabs ESC11 and ESC14, respectively (Fig. 1C). Similar affinities were determined by flow cytometry with calculated K_Dapps on huFAP-expressing cells of 4.7 ± 1.6 nmol/L for ESC11 and 200 nmol/L for ESC14, respectively (Table 1). On muFAP-expressing cells, K_Dapps were 51 ± 11 nmol/L for ESC11 and 251 ± 42 nmol/L for ESC14 (Table 1).

The variable heavy chain and light chain domains of ESC11 and ESC14 were cloned and expressed as fully human IgG1. IgG1 was purified by affinity chromatography from cell culture supernatant to more than 95% purity (Supplementary Fig. S2). As expected, the binding affinity of bivalent IgG1 for huFAP was higher than that of monovalent Fab due to avidity effects with apparent K_D values around 1.1 nmol/L on huFAP for both ESC11 and ESC14 IgGs (Table 1). These binding affinities are about 4-fold higher than for vF19 IgG. Interestingly, affinity constants of both ESC11 and ESC14 for muFAP were similar (around 2 nmol/L) after conversion to IgG1.

To investigate whether ESC11 and ESC14 recognize different epitopes of FAP, we used ESC11 to compete with ESC14 or F19 for binding to native surface human FAP and vice versa followed by flow cytometry. Binding of ESC11 Fab could be blocked by ESC14 IgG and vice versa, whereas F19 (28) did not compete with either of the ESC antibodies for FAP binding (Supplementary Fig. S3). These results suggest that antibodies ESC11 and ESC14 recognize the same epitope or overlapping/spatially close epitopes that are different from the epitope recognized by the F19 antibody.

Incubation of SK-MEL-187, a FAP-expressing melanoma cell line, with F19, ESC11, or ESC14 on ice resulted in detectable surface staining. However, incubation at 37°C for 3 hours with ESC11 or ESC14, but not with F19, induced internalization of FAP-antibody complexes (Fig. 2A and B). This was not observed with monovalent Fabs indicating that cross-linking of FAP molecules by bivalent IgG1 was essential (data not shown). Internalization of surface FAP occurred very rapidly. Already after 20 minutes at 37°C, an almost complete colocalization of internalized antibody with early endosomes could be observed, whereas after 40 minutes, colocalization with LAMP-1, a marker for late endosomes and lysosomes (Fig. 2C) was seen. We next tested pharmacologic inhibitors of both dynamin-dependent endocytosis (dynasore) and macropinocytosis (EIPA) to investigate the mechanism of endocytosis. Dynasore inhibited antibody-mediated FAP internalization (Supplementary Fig. S4A), whereas EIPA had no effect (Supplementary Fig. S4B), showing that endocytosis is mediated by a dynamin-dependent mechanism.

After overnight incubation with ESC11 and ESC14 antibodies, only very little FAP was detectable by flow cytometry on the surface of SK-MEL-187 cells. In contrast, still around 80% of cell-surface FAP could be detected after prolonged incubation with vF19 (Fig. 3A). We next investigated the time and concentration dependency of the observed FAP downmodulation for ESC11. Overnight internalization was concentration dependent and the half-maximal effective concentration (EC₅₀) was 3.3 ± 0.5 nmol/L (Fig. 3B). At this concentration, most of the FAP epitopes are occupied by ESC11 antibody, which presumably results in rafting of FAP molecules and concomitant internalization. To assess time dependency, we then chose a concentration of 200 nmol/L at which FAP downmodulation is expected to be more than 90%. FAP was rapidly internalized, and the majority of FAP molecules disappeared from the cell surface within 1 hour (Fig. 3C). When excess antibody was removed after overnight incubation, full FAP expression was restored not before 3 days after antibody removal (Fig. 3D). In contrast, FAP expression remained low when ESC11 was present over the entire observation period.

ESC11, ESC14, vF19, and A33 were conjugated with CHX-A″-DTPA and subsequently labeled with ¹⁷⁷Lu. Radiochemical purity was more than 95% by FPLC for all 4 antibodies, and the specific activities were 0.4 to 0.45 MBq/μg. Immunoreactivity of the radiolabeled antibodies was between 50% and 70% as determined by Lindmo analysis. All 3 ¹⁷⁷Lu-labeled FAP-binding antibodies bound specifically to SK-MEL-187 cells and could be displaced with increasing amounts of unlabeled antibody, whereas only a low amount of nonspecific binding could be observed on the FAP-negative SK-MEL-16 cells (data not shown). Accordingly, SPECT/CT imaging 72 hours after injection showed high and specific uptake of ¹⁷⁷Lu-ESC11 in the SK-MEL-187 tumor (Fig. 4A), but low uptake in a SK-MEL-16 xenograft. Specific uptake was also observed with ¹⁷⁷Lu-vF19 in the FAP-expressing tumor (Fig. 4B), whereas the control antibody ¹⁷⁷Lu-A33 did not accumulate in either tumor (Fig. 4C).

Biodistribution data showed very high uptake of the 3 antibodies in FAP-expressing tumors 72 hours after injection and low activity in nontargeted organs (Fig. 4D). Notably, ¹⁷⁷Lu-vF19 and, to a lesser extent, ¹⁷⁷Lu-ESC14 showed higher accumulation in the spleen than ¹⁷⁷Lu-ESC11, and cleared faster from the blood, which resulted in a reduced uptake in the tumors.

The in vivo biodistribution of ¹⁷⁷Lu- ESC11 was further studied over time in nude mice bearing human SK-MEL-187 tumor xenografts. We found a high (36% of the injected dose per gram of tumor tissue after 48 hours) and prolonged uptake of ¹⁷⁷Lu- ESC11, whereas its presence in blood and other organs was relatively low (Fig. 4E). Radioactivity in melanoma xenografts accumulated until approximately 48 hours after injection. After this time point, uptake values in the tumors remained high, but also became more variable in the individual mice, most likely due to the inhomogeneous tumor growth observed for this tumor model.

Next, we investigated the therapeutic efficacy of a single dose of the ¹⁷⁷Lu radioimmunoconjugates in vivo. We estimated a dose of 8 MBq to be therapeutically active (31) and this dose has been shown to be well below the maximum tolerated dose in this mouse strain with other antibodies displaying similar biodistribution properties. Eight MBq (15 μg) ¹⁷⁷Lu-labeled antibodies were injected intravenously into nude mice bearing established subcutaneous human melanoma tumors (∼150 mm³) 5 days after tumor implantation. Tumor growth was delayed with all 3 anti-FAP antibodies and ¹⁷⁷Lu-ESC11 exhibited the most pronounced growth delay (Fig. 5A). Analysis of survival curves (Fig. 5B) showed an increase of median survival from 18 days (¹⁷⁷Lu-A33) to 32 days (¹⁷⁷Lu-vF19), 43 days (¹⁷⁷Lu-ESC14), and more than 43 days (¹⁷⁷Lu-ESC11) after tumor implantation. Additional mice were sacrificed on day 0, 3, and 7 after injection of radioimmunoconjugates, and tumor specimens were processed for immunohistochemistry. We found a substantial reduction of proliferating (Ki-67⁺) tumor cells in samples taken after 3 and 7 days after injection of ¹⁷⁷Lu- ESC11 but not of the control conjugate (Fig. 5C). In addition, we observed increasing cell-free areas over time in the tumors treated with radiolabeled ESC11 (Fig. 5D).

We here describe the selection of novel human–mouse FAP-specific antibodies which downmodulate FAP expression from the surface of tumor cells, thus converting them into FAP-negative cells. Therapeutic strategies that aim at the complete downmodulation of FAP expression from the cellular surface might be promising approaches per se, as FAP is involved in cell invasion and metastatic processes of many cancers. Thus, antibody-induced internalization of cell surface–expressed FAP could provide major therapeutic impact by disturbing the various interactions on the cellular membrane, for example, the formation of functional invadopodia.

But more importantly, the rapid internalization renders them perfect carrier proteins for the targeted destruction of FAP positive carcinoma-associated fibroblasts or tumor cells. These approaches have the potential to be much more efficient than therapy with naked antibodies (32). While internalization of the antibody is not a prerequisite for successful radioimmunotherapy with a β-particle emitting radionuclide like ¹⁷⁷Lu, it becomes a clear advantage when using α-particle or Auger-electron emitting radionuclides instead. In these cases, the vicinity of the internalized radionuclide to the cellular nucleus augments the therapeutic efficacy by introducing more damage to the radiation-sensitive DNA. Antibody–drug conjugates are a second example for which internalization is an important feature, especially when the drug is coupled to the antibody by an acid-labile linker releasing the cytotoxic drug in the acidic lysosomes.

We have chosen human melanoma as a model system because the interaction between tumor cells and carcinoma-associated fibroblasts seems to be crucial for tumor growth and the development of drugs resistance (33). As shown by Flach and colleagues, melanoma cells stimulate the recruitment of fibroblasts and activate them by a so far unknown mechanism. In turn, these activated fibroblasts contribute to melanoma progression by providing both structural and chemical support (33). Incomplete abolishment of tumor cells by a short duration of kinase inhibitor therapy led to a transient reduction of tumor cells that was counteracted by an accelerated regrowth of tumor cells supported by the large fibroblast population that remained posttreatment. According to their model, the sole elimination of tumor cells is insufficient to provide tumor control and only a combined approach targeting tumor cells and carcinoma-associated fibroblasts simultaneously will lead to long lasting remissions.

FAP is not ubiquitously expressed in human melanoma at all stages of disease but is transiently present during the development of malignancy, and is routinely confined to the stromal fibroblasts in metastatic melanoma (34). However, high expression levels of FAP have been shown to correlate with a pronounced invasive phenotype in vitro (35) and have recently been postulated to contribute to transmigration of melanoma cells through the blood-brain barrier (36).

In this study, we compared the in vivo targeting properties of the human–mouse cross-reactive antibodies ESC11 and ESC14 to vF19. In a model with relatively high antigen expression on the tumor cells, high accumulation of up to 50% of the injected dose per gram tumor was observed. The only previously reported biodistribution experiment with a FAP-targeting antibody in mice was a study with human skin-grafted SCID mice bearing breast cancer tumors and a ¹³¹I-labeled antibody (37). There, much lower accumulation in the tumor and less favorable tumor-to-organ ratios were achieved, presumably as a result of the tumor model and the choice of the metabolically less stable radiolabel. In phase I clinical trials, ¹³¹I-labeled mAb F19 and its CDR-grafted humanized variant have been used to study pharmacokinetics, biodistribution, and imaging characteristics in patients with metastatic cancer (25, 38, 39). FAP was efficiently targeted by mAb F19 in all patients and its expression was confined to the stroma of metastatic lesions. Given the high expression of FAP in human cancers and the restriction to diseased tissue, FAP is a promising target both for imaging and radioimmunotherapy.

In particular, targeting stromal fibroblasts instead of tumor cells has several appealing advantages: the highly vascularized stroma is thought to be reached more efficiently by bulky molecules such as monoclonal antibodies. Antigen expression is considered to be more stable in carcinoma-associated fibroblasts than in highly proliferating cancer cells.

Furthermore, in many human epithelial cancers, stromal fibroblasts are highly abundant, and the cross-fire effect of low energy β-emitting radionuclides (penetration of 2 mm in tissue) additionally contributes to the eradication of antigen-negative tumor cells and to a homogenous deposition of the dose within the tumor tissue. A recent study showed that antitumor immunity is suppressed by FAP-positive stromal cells (21), which adds to the importance of FAP in tumor physiology. Thus, human–mouse cross-reactive FAP-specific antibodies are excellent tools to address this phenomenon in vivo by specific targeting of FAP-positive stromal fibroblasts and may help to unravel the mechanisms involved in local immunosuppression and tumor promotion.

In conclusion, we have developed novel fully human antibodies with excellent in vivo targeting properties. Because of their efficient internalization and targeting to the lysosomes, they are promising scaffolds for future development of antibody–drug conjugates or radioimmunotherapeutics.

No potential conflicts of interest were disclosed.

Conception and design: E. Fischer, T. Wuest, A.M. Scott, M. van den Broek, S. Bauer, C. Renner

Development of methodology: E. Fischer, K. Chaitanya, T. Wuest, A. Wadle, A.M. Scott, M. van den Broek, S. Bauer, C. Renner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Fischer, K. Chaitanya, T. Wuest, S. Bauer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Fischer, K. Chaitanya, T. Wuest, A. Wadle, S. Bauer, C. Renner

Writing, review, and/or revision of the manuscript: E. Fischer, K. Chaitanya, T. Wuest, A.M. Scott, M. van den Broek, S. Bauer, C. Renner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Wuest, R. Schibli, S. Bauer, C. Renner

Study supervision: E. Fischer, M. van den Broek, R. Schibli, S. Bauer, C. Renner

The authors thank Susan Cohrs and Sascha Kleber for excellent technical assistance.

This study was supported by Ludwig Institute for Cancer Research/Atlantic Philanthropies (to C. Renner), Zürcher Krebsliga (to T. Wuest and C. Renner), Krebsforschung Schweiz (to T. Wuest, E. Fischer, and C. Renner), and the University of Zürich (Forschungskredit to E. Fischer and S. Bauer).

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.