Fibroblast Activation Protein Triggers Release of Drug Payload from Non-internalizing Small Molecule Drug Conjugates in Solid Tumors

Modern chemotherapy of cancer is based on the systemic administration of cytotoxic agents that interfere with the growth of rapidly dividing cells, including tumor cells (1). Accumulation of antiproliferative drugs in healthy organs causes severe side effects and narrows their therapeutic window (2–4). The use of tumor-targeting antibodies as active delivery vehicles of potent cytotoxic drugs has been proposed and deeply investigated in clinical settings in which conventional therapeutic modalities are not effective (1). Twelve antibody–drug conjugates (ADC) have recently gained marketing authorization for the treatment of various solid and liquid malignancies (5, 6). While the clinical use of approved ADCs is now established (5, 6) and their benefit in some specific clinical indications is becoming evident, various molecular strategies to improve their efficacy and safety are being investigated (e.g., site-specific ADCs, new drug payloads, high drug–antibody ratios; refs. 7–11). The performance of ADCs in solid tumors may be limited by their large molecular size, which impairs their diffusion into the neoplastic mass (12–15). The replacement of antibody-targeting agents with small organic ligands with exquisite selectivity for tumor-associated antigens represents a promising approach to overcome limitations of macromolecule-based products. Indeed, small molecule drug conjugates (SMDC) benefit from a rapid extravasation, a deep penetration into solid tumor lesions and from the possibility to be manufactured with low cost-of-goods (12, 15–17).

In the past, scientists at Endocyte Inc. have developed SMDCs directed against internalizing cellular antigens such as prostate-specific membrane antigen and folate receptor (Fol1), using disulfide reducible bonds as cleavable linkers and highly potent vinblastine and tubulysin payloads. Those products failed to show clinical benefit in patients with ovarian (18, 19) and prostate (20, 21) cancer in large clinical trials. More recently, bicyclic peptides have been considered as delivery vehicles for the generation of drug conjugates based on Vedotin (Maleimidocaproyl-Valine-Citrulline-PABC-MMAE), a linker-payload module frequently used in approved (i.e., Adcetris, Polivy, and Padcev) and advanced ADC products (5, 6, 22–24). Monomethyl auristatin E (MMAE) is a potent cytotoxic agent that inhibits tubulin polymerization in cells undergoing rapid replication and growth (25, 26). The Valine-Citrulline linker is designed to be cleaved by intracellular proteases [e.g., cathepsin B (CathB)], but recent preclinical studies have revealed a possible instability of this linker in circulation due to the presence of other extracellular proteases such as carboxylesterase 1C (Ces1c; refs. 27, 28). Initial signs of clinical activity have recently been reported for BT8009 and BT5528, two drug conjugates based on bicyclic peptides targeting Nectin-4 and EphA2, respectively (29).

Fibroblast activation protein (FAP) is a type II integral membrane serine protease mainly localized and overexpressed in the stroma of different tumors (30). FAP is expressed on the cell surface of cancer-associated fibroblasts in the tumor microenvironment of more than 90% of epithelial human malignancies, while negligible amounts of the target are found in healthy tissues (31–34). In the recent years, Haberkorn and co-workers described a variety of small molecules targeting FAP with excellent biodistribution in human patients with cancer. Those compounds present extremely low uptake in normal organs and selective accumulation in target lesions (32, 34–37). We have reported the development of OncoFAP, a highly potent small organic FAP ligand with a dissociation constant of 680 pmol/L to the cognate target (38). The tumor homing performance of OncoFAP has been validated by PET imaging studies with the [⁶⁸Ga]Ga-OncoFAP-DOTAGA derivative (39). The highly portable chemical structure of OncoFAP enables conjugation not only to radiometal chelators, but also to linker-cytotoxic drug modules for the development of SMDCs (38). We have previously described the generation of a SMDC based on OncoFAP and on the Vedotin linker-payload module (38). The product was safe and efficacious in therapy studies conducted in tumor-bearing mice. Modest tumor growth retardation and no signs of acute toxicity were observed at the dose of 500 nmol/kg (∼1 mg/kg), while the product induced complete responses in all treated animals when combined with L19-IL2, a clinical stage targeted IL2 product (38).

The design of efficacious anticancer drug conjugates crucially relies on an efficient localization at the site of disease and on the exploitation of suitable chemical processes for the selective release of potent cytotoxic payloads. It would be desirable to exploit methodologies which provide a quantitative assessment of the tumor targeting and drug release process. In this article, we describe the development of novel OncoFAP-MMAE SMDCs designed to be selectively cleaved by FAP in the tumor microenvironment. We used a novel mass spectrometry methodology for the quantitative evaluation of the amount of SMDC delivered to tumors and to normal tissues, as well as of the efficiency of MMAE release. Enzymatically cleavable linkers (Val-Cit and Phe-Lys) and disulfide bridges, widely applied in clinical-stage drug conjugates (16, 40, 41), were used as controls. We found that linkers based on the FAP-sensitive Gly-Pro dipeptide enabled the most efficient delivery of MMAE to FAP-positive tumors curing the majority of tumor-bearing mice at doses that were very well tolerated (i.e., 250 nmol/kg).

[⁶⁸Ga]Ga-OncoFAP-DOTAGA (compound 1) was synthesized as reported previously (39). Radiolabeling was performed on the basis of the German Pharmaceuticals Act [AMG §13 (2b)], that is, magistral preparation. Briefly, radiogallium (T_½ = 68 minutes, β⁺ = 89%, and EC = 11%) was automatically eluted with 0.1 mol/L HCl (0.36%) from a 50 mCi (1.85 GBq) ⁶⁸Ge/⁶⁸Ga radionuclide generator (EZAG) without prepurification of the eluate and transferred into the reaction vessel of the fully automated ⁶⁸Ga-radiosynthesis module (miniAllinOne, Trasis), containing a preheated, buffered OncoFAP-DOTAGA (compound 18) solution. After incubation for a few minutes at approximately 100°C, the reaction mixture was loaded onto a SPE cartridge, eluted with EtOH in the product vial, and formulated with additional 0.9% NaCl. A full quality control (QC) was performed for each preparation of [⁶⁸Ga]Ga-OncoFAP-DOTAGA. All QC parameters were in accordance with the Ph. Eur. standards for ⁶⁸Ga-DOTA-TOC (monograph 2482).

Scans from four different patients 36–76 years old are shown in Fig. 1. Patients underwent whole-body PET/CT (mCT, Siemens Healthineers) or PET/MRI (mMR, Siemens Healthineers) imaging approximately 1 hour after injection of 125–160 MBq of [⁶⁸Ga]Ga-OncoFAP-DOTAGA as part of clinical workup for cancer. Of one patient, additional [¹⁸F]FDG PET/CT 9 days prior to OncoFAP-PET is shown. Maximal standard uptake value (SUV_max) measurements were acquired in Syngovia (Siemens Healthineers) with circular or spherical volumes of interests.

After collection, samples were snap frozen in optimal cutting temperature medium (Thermo Fisher Scientific) and stored at −80°C. Ice-cold acetone-fixed 10 μm cryostat sections were blocked with 20% FBS in 3% BSA/PBS for 1 hour and washed twice with PBS. Cell nuclei were stained with 150 μL/slide of DAPI (Invitrogen, D1306; 1 μg/mL in PBS), depicted in blue. FAP (in green) was stained with 150 μL/slide of OncoFAP-Fluorescein (250 nmol/L in PBS) or Neg-Fluorescein (250 nmol/L in PBS), washed twice with PBS and detected with 150 μL/slide of rabbit anti-FITC (Bio-Rad, 4510-7804; 1:700 in 3% BSA/PBS) and anti-rabbit AlexaFluor488 (Invitrogen, A11008; 1:500 in 3% BSA/PBS). For vascular staining (in red) rat anti-murine CD31 (BD Biosciences, 550274; 1:500 in 3% BSA/PBS) and anti-rat AlexaFluor594 (Invitrogen, A21209; 1:500 in 3% BSA/PBS) antibodies were used on SK-RC-52.hFAP sample, while on human-derived samples mouse anti-human CD31 (Invitrogen, 13-0311-82; 1:500 in 3% BSA/PBS) and goat anti-mouse Alexa 594 (Invitrogen, A11032; 1:500 in 3% BSA/PBS) were used. After washing three times with PBS in the dark, slides were mounted with fluorescent mounting medium (Dako) and analyzed with a Leica DMI6000B microscope equipped with 20×0.7 NA HC PlanApo objective (20× magnification) using 405, 488, 561 nm laser lines. Pictures were collected using Andor iXon EM CCD camera and elaborated using Leica LAS AF software and Fiji. Results are shown in Supplementary Fig. S1.

Detailed chemical procedures and compound characterization are reported in the Supplementary Data.

IC₅₀ values were determined by fitting data to the four-parameter logistic equation, using a Prism 6 software (GraphPad Software) for data analysis.

Mouse serum (19 μL, Sigma-Aldrich) or human serum (19 μL, Sigma-Aldrich) were preincubated at 37°C for 5 minutes. DMSO solutions of compounds 2–9 (1 mmol/L, 1 μL) were added at the final concentration of 50 μmol/L. Serum enzymatic activity was assessed in parallel using procaine as positive control and following its enzymatic hydrolysis 15 minutes after incubation. The solution was incubated at 37°C and the reaction was stopped by adding ACN (9 volumes) at different timepoints (time 0, 30 minutes, 2 hours, 4 hours). Samples were processed and analyzed by LC/MS as reported below. Each experiment was performed in triplicate.

DMSO solutions of test compounds 10–17 (1 mmol/L, 1 μL) were added to 18 μL of HEPES buffer (HEPES 100 mmol/L, NaCl 50 mmol/L, pH = 7.4) to reach a final compound concentration of 50 μmol/L. Human FAP (hFAP), produced as reported previously (38), was added to the compound solutions to a final enzyme concentration of 50 nmol/L. The solution was incubated at 37°C and the reaction was stopped by adding ACN (9 volumes) at different timepoints (time 0, 30 minutes, 2 hours, 4 hours). Samples were processed and analyzed by LC/MS as reported below. Each experiment was performed in triplicate. Before starting the experiment, catalytic activity of recombinant hFAP was determined as described in the Supplementary Data (Supplementary Fig. S2).

The assays were carried out following protocols already described in the literature (42). Shortly, CathB (Merck Millipore, 0.444 mg/mL) was diluted 1:1 with a 2.2 mol/L AcONa solution (pH = 5.5). The enzyme was diluted and activated by adding a 30 mmol/L DTT, 15 mmol/L EDTA aqueous solution (3.43 mL) to afford a final enzyme concentration of 120 nmol/L. After 15 minutes, activated CathB solution (19 μL, 120 nmol/L) was incubated with test compounds 2–9 (1 mmol/L, 1 μL) at a final compound concentration of 50 μmol/L. The solution was incubated at 37°C and the reaction was stopped by adding ACN (9 volumes) at different timepoints (time 0, 30 minutes, 2 hours, 4 hours). Samples were processed and analyzed by LC/MS as reported below. Each experiment was performed in triplicate.

To 19 μL of a 500 μmol/L solution of glutathione (GSH) reduced form (Sigma-Aldrich) in 1 mmol/L PBS pH = 7.4, 1 μL of a 1 mmol/L DMSO solution of compound 2–9 was added to afford a final compound concentration of 50 μmol/L. The solution was incubated at 37°C and the reaction was stopped by adding 1 μL of HCOOH at different timepoints (time 0, 2 hours, 6 hours, 24 hours). Samples were processed and analyzed by LC/MS as reported below. Each experiment was performed in triplicate.

After addition of organic solvent (ACN or HCOOH), samples were centrifuged at 15,000 × g for 10 minutes. Supernatant (120 μL) was dried under vacuum at 37°C and resuspended in 120 μL of an aqueous solution containing 10% ACN and 0.1% HCOOH. Samples were analyzed by LC/MS on an Agilent 1200 Series LC System, using as column an InfinityLab Poroshell 120 EC-C18 (4.6×56 mm, particle size: 2.7 μm, pore size: 120 Å) at a flow rate of 0.8 mL/minute with linear gradients of solvents A and B (A = Millipore water with 0.1% HCOOH, B = ACN with 0.1% HCOOH) from 40% to 100% of B in 3 minutes. Eluents were analyzed in full mass scan in positive ion mode with an Agilent 6100 Series Single Quadrupole MS 5 System.

SK-RC-52 and HT-1080 cells were transduced to express hFAP on the cell membrane (38). All cell lines were tested for Mycoplasma every 12 months by PCR. Cells were expanded for six passages and preserved in cryotubes (5 × 10⁶ cells/mL). After further expansion (8–10 passages in total), cells were used for in vivo and in vitro experiments.

All animal experiments were conducted in accordance with Swiss animal welfare laws and regulations under the license number ZH006/2021 granted by the Veterinäramt des Kantons Zürich.

Tumor cells were grown to 80% confluence using RPMI1640 (Gibco) for SK-RC-52 cells or DMEM (Gibco) for HT-1080 cells supplemented with 10% FBS (Gibco), 1% Antibiotic-Antimytotic (Gibco), and detached with Trypsin-EDTA 0.05% (Gibco). Tumor cells were then resuspended in Hanks’ Balanced Salt Solution medium (Gibco) at a concentration of 50 × 10⁶ cells/mL. Aliquots of 5 × 10⁶ cells (100 μL of suspension) were injected subcutaneously in the right and/or left flanks of female athymic Balb/c AnNRj-Foxn1 mice (6 to 8 weeks of age, Janvier).

SK-RC-52.hFAP and SK-RC-52 wild-type (.wt) tumors were implanted into the right and left flanks of female athymic Balb/c AnNRj-Foxn1 mice (Janvier, 6 to 8 weeks old) as described above and allowed to grow to an average volume of approximately 200 mm³. Tumor-bearing mice were injected intravenously with test compounds 3–8 (5 nmol dissolved in sterile saline solution 0.9% NaCl containing 2% of DMSO) and sacrificed at different timepoints after the intravenous injection (1, 6, and 24 hours). Fresh blood was collected in lithium heparin tubes (BD Microcontainer LH Tubes), vortexed, and centrifuged (15,000 × g, 15 minutes). Plasma was frozen and stored at −80°C. Healthy organs and tumors were subsequently excised, frozen with dry ice, and stored at −80°C. Frozen plasma and mouse tissues (∼50 mg) were thawed and 600 μL of a solution containing 95% ACN and 0.1% HCOOH were added to induce protein precipitation. A solution of internal standard (d₈-MMAE, 50 μL, 50 nmol/L and ¹³C₄-OncoFAP-Val-Cit-MMAE, 50 μL, 1.2 μmol/L) was added. Samples were then homogenized with a tissue lyser (TissueLyser II, QIAGEN) for 15 minutes at 30 Hz. After homogenization, samples were centrifuged (15,000 × g for 10 minutes). Supernatant was collected and dried at room temperature with a vacuum centrifuge. Pellets were then resuspended in 1 mL of an aqueous solution containing 3% ACN and 0.1% of trifluoroacetic acid (TFA) and subsequently purified on Oasis HLB SPE columns (Waters) following instructions indicated by the manufacturer. Eluates were dried under vacuum at room temperature. Dry pellets were resuspended in 0.4 mL of an aqueous solution with 3% ACN and 0.1% of TFA and further purified using Macro-Spin SPE columns (Harvard Bioscience). Eluates were dried under reduced pressure at room temperature. Dry samples were finally resuspended in 30 μL of an aqueous solution containing 3% of ACN and 0.1% of HCOOH. Each sample (3 μL, 10% of the total) was then injected in the nanoLC-HR-MS system. Chromatographic separation was carried out on a Acclaim PepMap RSLC column (50 μm × 15 cm, particle size 2 μm, pore size 100 Å, Thermo Fisher Scientific) with a gradient program from 95% A (0.1% HCOOH), 5% B (ACN, 0.1% HCOOH) to 5% A, 95% B in 45 minutes on an Easy nanoLC 1000 (Thermo Fisher Scientific). Sample clean-up and concentration was carried out on an Acclaim PepMAP 100 precolumn (75 μm × 2 cm, particle size 3 μm, pore size 100 Å; Thermo Fisher Scientific). The LC system was coupled to a Q-Exactive mass spectrometer (Thermo Fisher Scientific) via a Nano Flex ion source (Thermo Fisher Scientific). Ionization was carried out with 2 kV of spray voltage, 250°C of capillary temperature, 60 S-lens RF level. Mass spectrometry analysis was carried out in positive-ion mode with the following parameters: resolution 70,000 FWHM (full width at half maximum) at 200 m/z, AGC target 5 × 10⁴, and maximum injection time 200 ms. The detector was working in single-ion monitoring mode following the transition reported in Supplementary Table S1. Data analysis was carried out with Thermo Xcalibur Qual Broswer v2.2 (Thermo Fisher Scientific). MMAE ionization and recovery from biological matrix has been determined as described in the Supplementary Data (Supplementary Figs. S3 and S4).

SK-RC-52.hFAP tumors were implanted into the right flank of female athymic Balb/c AnNRj-Foxn1 mice (Janvier, 6 to 8 weeks old) as described above and allowed to grow to an average volume of approximately 50–100 mm³. Mice were randomly assigned into therapy groups of four animals. Compounds 2 and 4–7 (250 nmol/kg) were systemically administered by intravenous injection in the lateral tail vein (following the schedule presented in Fig. 5). Test compounds 2 and 4–7 were injected as sterile saline solution (0.9% NaCl) containing 2% of DMSO. Tumors were measured with an electronic caliper, and the animals were weighed daily. Tumor volume (mm³) was calculated with the formula (long side, mm) × (short side, mm) × (short side, mm) × 0.5. Animals were euthanized when one or more termination criteria indicated by the experimental license were reached (e.g., weight loss > 15%). Prism 6 software (GraphPad Software) was used for data analysis (regular two-way ANOVA followed by Bonferroni test). Results of statistical analysis for all therapy experiments are reported in the Supplementary Data.

Analysis has been approved by the Ethics Committee of the Medical Association of Westphalia-Lippe and the Medical Faculty of the University of Münster (Az. 2021-408-f-S) and studies were conducted in accordance with the Declaration of Helsinki. Patients gave written informed consent for [¹⁸F]FDG and [⁶⁸Ga]Ga-OncoFAP-PET/CT imaging and retrospective scientific analysis. Consent for publication is not necessary, as no potential identifying details and/or images are published.

Processed data are available in the main text or the Supplementary Data. Access to raw data can be requested to [email protected].

OncoFAP can be conjugated to radionuclide chelators such as DOTAGA, as reported in Fig. 1A, to obtain precursors for FAP-specific radiotracers for imaging. OncoFAP-DOTAGA (compound 18) and [⁶⁸Ga]Ga-OncoFAP-DOTAGA (compound 1) were synthesized following well-established procedures described in the literature (38, 39) as reported in the Supplementary Data (39). Figure 1 depicts PET scans performed on a patient with esophageal cancer (75 years old, female). The same patient was imaged both with [¹⁸F]FDG (Fig. 1B) and [⁶⁸Ga]Ga-OncoFAP-DOTAGA (Fig. 1C). [⁶⁸Ga]Ga-OncoFAP-DOTAGA efficiently and selectively localized to primary tumor and metastases, thus defining the rims of cancer lesions with favorable tumor-to-background ratio. In contrast, FDG-PET only visualized primary tumor with worse contrast against background compared with that obtained by OncoFAP-PET procedure. Figure 1D–F display PET/CT scans of three additional patients with different cancer types. [⁶⁸Ga]Ga-OncoFAP-DOTAGA efficiently localized to primary tumors and metastatic lesions in patients with breast (Fig. 1D), hepatocellular (Fig. 1E), and pancreatic (Fig. 1F) cancer.

OncoFAP-Fluorescein (compound 19) was afforded following established literature protocols (38). Similarly, Neg-Fluorescein (compound 20) was synthesized as described in the Supplementary Data. Expression of FAP was assayed by immunofluorescence studies with OncoFAP-Fluorescein on human healthy colon and human colorectal cancer specimens. Results of these experiments confirmed high and selective expression of FAP in the stroma of tumor samples (Supplementary Fig. S1). No expression in healthy colon was observed by immunofluorescence staining.

OncoFAP-Val-Cit-MMAE (OncoFAP-Vedotin, compound 2) was synthesized starting from OncoFAP-Asp-Lys-Asp-Cys precursor (compound 22) and the commercially available reactive Maleimidocaproyl-Val-Cit-PABC-MMAE module (Levena Biopharma) as described previously (38). To study in vivo biodistribution of OncoFAP-Vedotin (prodrug) and the efficiency of drug release (MMAE) in tumors, plasma, and healthy organs, the compound was systemically administered in mice bearing FAP-positive and wild-type SK-RC-52 tumors in the two flanks (Fig. 2A). OncoFAP-Vedotin was administered at the dose of 5 nmol/mouse (corresponding to ∼250 nmol/kg), a dose which is compatible with tumor targeting and which stands below saturation of tumors as previously demonstrated by quantitative biodistribution with OncoFAP-radionuclide derivatives in the same tumor model (38). Tumors, plasma samples, and organs were collected at different timepoints after single or multiple administrations. MMAE and OncoFAP-Vedotin were extracted by sample preparation to enable quantification by LC/MS. Figure 2A shows schedule of the experiment and depicts the LC/MS profile of free MMAE and d₈-MMAE (used as internal standard) after extraction from biological matrix. OncoFAP-Val-Cit-MMAE efficiently targeted FAP-positive tumors with a growing accumulation from approximately 15% to approximately 35% ID/g after one, four, or seven daily consecutive injections, at the 6-hour timepoint (Fig. 2B). No or little prodrug uptake was observed in wild-type tumors and healthy organs both at the 6- and 24-hour timepoints. Absolute quantification of the precursor was achieved by using ¹³C₄-OncoFAP-Val-Cit-MMAE (compound 43) as internal standard. MMAE in vivo biodistribution results did not show selective uptake of the active free drug in SK-RC-52.hFAP tumors as compared with SK-RC-52.wt lesions and healthy organs. After 6 hours from a single administration of OncoFAP-Vedotin between 1% and 4% ID/g of free MMAE was found in FAP-positive cancer lesions, wild-type tumors, spleen, kidney, lung, and stomach (Fig. 2C). Poor biodistribution of free MMAE potentially reflects lack of efficient Val-Cit linker cleavage at the tumor site. Lower uptake (∼1%–2% ID/g) of MMAE in both SK-RC-52.hFAP and SK-RC-52.wt tumors was observed at the 24-hour timepoint. Supplementary Figure S5 presents results of the in vivo biodistribution of prodrug and MMAE released by the untargeted control compound Neg-Val-Cit-MMAE (compound 10), a product in which the 2-cyano-4,4′-difluoropyrrolidine nucleus has been replaced by a simple hydroxyl group (-OH). No selective uptake of Neg-Val-Cit-MMAE (compound 10) and free MMAE in tumors was observed at the 6-hour timepoint. As an additional negative control, we administered free MMAE in SK-RC-52.hFAP tumor-bearing mice (5 nmol/animal) and assessed its biodistribution. MMAE did not show selective accumulation in FAP-positive tumors, but presented high uptake in the kidney (excretion organ; Supplementary Fig. S6).

Figure 3A presents chemical structures of novel CathB-cleavable, FAP-cleavable, and reducible OncoFAP-MMAE conjugates, together with their corresponding untargeted controls. A first set of novel CathB-cleavable (compounds 2 and 3) and FAP-cleavable (compounds 4–7) OncoFAP-MMAE SMDCs were designed on the basis of OncoFAP-Vedotin structure, already described in the literature (38).

For the design of novel FAP-cleavable SMDCs, a small set of AA₂-Pro₁ dipeptide linkers (i.e., linkers that include Proline) was carefully selected on the basis of the amino acid sequences that are recognized and cleaved by FAP. In particular, the Glycine-Proline (Gly-Pro), l-Alanine-Proline (l-Ala-Pro), d-Alanine-Proline (d-Ala-Pro), and Valine-Proline (Val-Pro) motifs were included in our study as linker structures in OncoFAP-MMAE conjugates (14, 44, 45). In this set of compounds, the Val-Cit dipeptide structure was replaced by Phe-Lys or AA₂-Pro₁. To afford maleimidocaproyl-dipeptide-PABC-MMAE derivatives, 6-maleimidocaproic acid was sequentially coupled with the two desired amino acids, followed by coupling with p-aminobenzyl alcohol, activation with p-nitrophenyl chloroformate and reaction of the resulting PNP carbonates with free MMAE. Final SMDC products were obtained in excellent yields (Supplementary Data) and purity [≥97%, high-performance liquid chromatography (HPLC)] by Michael addition with OncoFAP-Asp-Lys-Asp-Cys (compound 22). A second set of OncoFAP-MMAE SMDCs (compounds 8 and 9) containing disulfide reducible linkers were designed on the basis of previous work from Endocyte Inc. and Bicycle Therapeutics (46, 47). In particular, two different versions of mixed disulfides between the MMAE-mercaptoethanol carbamate and OncoFAP-Asp-Lys-Asp-Cys or OncoFAP-Asp-Lys-Asp-Pen were produced following standard synthetic protocols (Supplementary Data). Untargeted analogs (compounds 10–17) devoid of the 2-cyano-4,4′-difluoropyrrolidine FAP-targeting nucleus were synthesized as reported in the Supplementary Data.

Cytotoxicity of MMAE, OncoFAP-Val-Cit-MMAE (compound 2), OncoFAP-Gly-Pro-MMAE (compound 4), and Neg-Gly-Pro-MMAE (compound 12) was assessed in cultures of both SK-RC-52.wt and SK-RC-52.hFAP cells with a MTS assay. MMAE exhibited a low nanomolar potency on both cellular models, in accordance to literature values (48), while the targeted prodrugs 2 and 4 showed an approximately 1,000-fold higher IC₅₀. Neg-Gly-Pro-MMAE (compound 12) inhibited cell proliferation similarly to the targeted prodrugs on SK-RC-52.wt cells (IC₅₀ = 1.47 μmol/L), while displayed an enhanced activity on SK-RC-52.hFAP cells (IC₅₀ = 7.84 nmol/L) due to the efficient cleavage of the Gly-Pro linker by hFAP. Experimental details are reported in the Supplementary Data (Supplementary Fig. S7).

In vitro stability and cleavage efficiency of the linkers was assessed by incubating the OncoFAP-MMAE conjugates or their untargeted analogs in presence of hFAP, CathB, and GSH (Fig. 3B). Results of the assessment of in vitro release of free MMAE are further presented in Supplementary Fig. S8.

Untargeted analogs of OncoFAP-MMAE SMDCs (compounds 10–17) were incubated with hFAP to evaluate kinetics of prodrug degradation by LC/MS. As expected, Gly-Pro and d-Ala-Pro dipeptides are rapidly hydrolyzed in presence of hFAP (∼3% residual compounds left 30 minutes after incubation with the enzyme), while all other test SMDCs are stable in the assay conditions (Fig. 3B).

OncoFAP-MMAE derivatives (compounds 2–9) were incubated with recombinant CathB and kinetics of prodrug degradation was followed by LC/MS. All test compounds were gradually hydrolyzed at the level of Lys-Asp bond of the peptide spacer sequence, with almost full degradation 4 hours after incubation with the enzyme. OncoFAP-Val-Cit-MMAE (compound 2) and OncoFAP-Phe-Lys-MMAE (compound 3) efficiently released the MMAE payload at early timepoints as a consequence of the cleavage of Val-Cit or Phe-Lys moiety (Supplementary Fig. S8).

OncoFAP-MMAE derivatives (compounds 2–9) were incubated with a 10-fold molar excess of GSH to study in vitro stability of the compounds in the presence of reducing agents. A significant release of MMAE was measured only for the unsubstituted disulfide derivative OncoFAP-SS-MMAE (compound 8; Supplementary Fig. S8). All other compounds, including methylated OncoFAP-SS(Me)₂-MMAE analogue (compound 9), were stable in reducing conditions with a partial degradation (i.e., less than ∼30% in 24 hours).

In the view of further in vivo investigation, serum stability of novel OncoFAP-MMAE SMDCs (compounds 2–9) was assessed (Fig. 3C). All test compounds were stable in human serum (half-lives >4 hours). Compounds 4, 5, and 7–9 showed to be stable over the time frame of the experiments both in mouse and human serum (less than ∼20% loss of intact product quantified by LC/MS over 4 hours at 37°C). OncoFAP-Val-Cit-MMAE, OncoFAP-Phe-Lys-MMAE, and OncoFAP-d-Ala-Pro-MMAE displayed a reduced stability in mouse serum (half-lives of around ∼4 hours), which might be determined by the presence of certain enzymes only in mouse serum (e.g., Ces1c; ref. 27).

OncoFAP binds to murine recombinant FAP (mFAP) with subnanomolar affinity (k_D = 513 pmol/L), as determined by fluorescence polarization (Supplementary Fig. S9A).

Neg-Gly-Pro-MMAE (compound 12) cleavage efficacy in presence of mFAP was assessed by determining the percentage of intact prodrug and the release of MMAE over time by LC/MS. mFAP rapidly hydrolyzes the Gly-Pro linker and enables MMAE release. Neg-Val-Cit-MMAE (compound 10), used as a negative control in this experiment, is not cleaved by mFAP (no MMAE release observed in the presence of the enzyme; Supplementary Fig. S9B).

Figure 4 presents in vivo quantitative biodistribution results of free MMAE after systemic administration in tumor-bearing mice of OncoFAP-Phe-Lys-MMAE (compound 3), OncoFAP-Gly-Pro-MMAE (compound 4), OncoFAP-l-Ala-Pro-MMAE (compound 5), OncoFAP-d-Ala-Pro-MMAE (compound 6), OncoFAP-Val-Pro-MMAE (compound 7), and OncoFAP-SS-MMAE (compound 8).

The release pattern of free MMAE from OncoFAP-Phe-Lys-MMAE (compound 3) was comparable with the MMAE biodistribution profile observed after administration of OncoFAP-Val-Cit-MMAE at the equivalent molar dose (5 nmol/mouse), thus providing a second example of inefficient drug release in this model from cathepsin-cleavable linkers on noninternalizing SMDCs.

The FAP-cleavable OncoFAP-Gly-Pro-MMAE (compound 4) selectively delivers high and stable amounts of MMAE at the site of disease (∼8% and 10% ID/g in SK-RC-52.hFAP tumors at the 6- and 24-hour timepoints, respectively), with a tumor-to-kidney ratio of 16-to-1, 24 hours post-injection (p.i.). Low MMAE uptake was determined in all other healthy organs, including lung, liver, stomach, spleen, heart, and plasma at all timepoints (1, 6, and 24 hours after administration of OncoFAP-Gly-Pro-MMAE). When directly compared with a single injection of OncoFAP-Val-Cit-MMAE, the Gly-Pro analogue delivers over a 10-fold higher amount of MMAE in FAP-positive tumors at 24 hours p.i.

OncoFAP-l-Ala-Pro-MMAE (compound 5) and OncoFAP-Val-Pro-MMAE (compound 7) exhibit a significant and sustained payload accumulation in FAP-positive tumors (∼5%–6% ID/g at 24 hours p.i.). Notably, the biodistribution profile of free MMAE payload released by OncoFAP-Val-Pro-MMAE is characterized by higher tumor-to-organ ratios compared with OncoFAP-l-Ala-Pro-MMAE. Both OncoFAP-Gly-Pro-MMAE and OncoFAP-Val-Pro-MMAE showed selective accumulation of MMAE at the FAP-positive tumor site over time (from 1 to 24 hours p.i.).

In vivo biodistribution of MMAE released by OncoFAP-d-Ala-Pro-MMAE (compound 6) did not show selective accumulation of drug payload in SK-RC-52.hFAP tumors as compared with wild-type tumors, plasma, and healthy organs. Remarkably, no in vivo release of MMAE in tumors and healthy structures was observed in animals administered with OncoFAP-SS-MMAE (compound 8), thus indicating full in vivo stability of the disulfide bridge both in healthy organs and in the tumor.

To investigate the sole contribution of FAP-cleavable linkers to tumor targeting of MMAE, we generated and characterized a nontargeted Gly-Pro–based prodrug (compound 21). In compound 21, the OncoFAP targeting moiety has been entirely replaced by a capping quinoline group. Compound 21 is a prodrug substrate of FAP that does not possess FAP-inhibitory activity. In vitro MMAE-release assays show that the molecule is efficiently cleaved by human recombinant FAP, while the stability in mouse serum is poor (i.e., half-life of ∼30 minutes). In vivo biodistribution results in SK-RC-52.hFAP tumor-bearing mice present no preferential accumulation of MMAE in FAP-positive tumors over healthy tissues. Our results are in line with the lack of the OncoFAP targeting function in compound 21 and with its poor serum stability. Direct comparison with the MMAE biodistribution data obtained with OncoFAP-targeted SMDCs clearly indicates the synergistic contribution of OncoFAP and FAP-cleavable linkers to a favorable in vivo performance of this class of products (Supplementary Fig. S10).

The therapeutic efficacy of OncoFAP-MMAE SMDCs (compounds 2 and 4–8) was evaluated in immunodeficient Balb/c nude mice subcutaneously xenografted with SK-RC-52.hFAP tumors (∼50 mm³ average tumor volume). Compounds were administered by intravenous injection at 250 nmol/kg (equivalent molar doses) following the schedule indicated in Fig. 5. Compounds 4–7 significantly inhibited tumor growth, compared with saline treatment (P < 0.0001 at day 18 after tumor implantation) and to treatment with OncoFAP-Val-Cit-MMAE (compound 2; P < 0.0001 at day 22 after tumor implantation; Fig. 5A). Potent in vivo antitumor efficacy of OncoFAP-Gly-Pro-MMAE (compound 4), OncoFAP-Val-Pro-MMAE (compound 7), and OncoFAP-l-Ala-Pro-MMAE (compound 5) strongly correlates with high and stable release of active drug in FAP-positive SK-RC-52 tumors, as assessed by LC/MS quantification (Fig. 4). The most active compound in our study was OncoFAP-Gly-Pro-MMAE (compound 4), with the longest retardation of tumor growth. Administration of OncoFAP-Val-Pro-MMAE (compound 7) and OncoFAP-l-Ala-Pro-MMAE (compound 5) resulted in complete tumor regression (2/4 and 1/4 complete tumor regression, respectively). As anticipated by results of in vivo quantification of free active MMAE (Fig. 4), OncoFAP-SS-MMAE (compound 8) did not show any anticancer activity in this model (no statistical difference compared with control mice treated with saline). All compounds were well tolerated, with no body weight loss observed over the duration of the therapy experiment (Fig. 5B).

In a second experiment, we further characterized the in vivo therapeutic activity of OncoFAP-Gly-Pro-MMAE (compound 4) in mice bearing large SK-RC-52.hFAP tumors (∼250 mm³ average tumor volume; Fig. 5C). A second group of animals bearing small well-established SK-RC-52.hFAP tumors (∼70 mm³ average tumor volume) was included in the study to confirm the results obtained with the same compound as the first therapy studies. OncoFAP-Gly-Pro-MMAE showed a very potent anticancer activity both on large and small subcutaneous SK-RC-52.hFAP tumors (with 1/4 complete responses in both groups), with no body weight loss (Fig. 5D). No statistical differences are measured comparing tumor volumes of the two groups of animals at the end of the in vivo phase of the experiment (day 30 after tumor implantation).

The Gly-Pro linker is efficiently and selectively cleaved by FAP, in antigen-positive tumors. To better understand the kinetics of the prodrug uptake and release of MMAE at the tumor site, we ran a quantitative biodistribution in mice following the concentration of OncoFAP-Gly-Pro-MMAE and of free MMAE by mass spectrometry. In addition, with the aim to evaluate whether the delivery of MMAE at the tumor site solely derives from the linker structure, we compared the in vivo MMAE release from OncoFAP-Gly-Pro-MMAE (compound 4) and its untargeted analogue Neg-Gly-Pro-MMAE (compound 12). SK-RC-52.hFAP tumor-bearing mice were injected with the two prodrugs and the biodistribution of both prodrugs and MMAE was assessed at different timepoints (1, 6, 24, and 32 hours). ¹³C₄-OncoFAP-Gly-Pro-MMAE (47) and ¹³C₄-Neg-Gly-Pro-MMAE (49) were used as internal standards for the quantification of intact prodrugs. The SMDC OncoFAP-Gly-Pro-MMAE (compound 4) rapidly accumulated in the tumor (∼25% ID/g at 1 hour p.i.) and is depleted over time while it is hydrolyzed and releases the MMAE payload (Fig. 6A). The concentration of untargeted prodrug Neg-Gly-Pro-MMAE (compound 12) at the FAP-positive tumor was below the limit of quantification at all timepoints. The untargeted prodrug produces a transient increase of MMAE concentration at the tumor site at early timepoints, which rapidly decreases over time (∼5% ID/g at 6 hours p.i.). In contrast, tumor-targeted OncoFAP-Gly-Pro-MMAE produces a high and prolonged release of MMAE in FAP-positive lesions, with MMAE concentrations ranging from approximately 5% ID/g to approximately 17% ID/g from 1 to 32 hours after intravenous administration (Fig. 6B). The contribution of the targeting agent is therefore essential for an efficient and sustained release of the cytotoxic payload at the tumor site.

In a subsequent therapy experiment, we compared the efficacy of OncoFAP-Gly-Pro-MMAE (compound 4) when administered following three different dosing schedules. SK-RC-52.hFAP tumor-bearing mice were treated daily, every 3 days, or weekly with the same dose of OncoFAP-Gly-Pro-MMAE (compound 4; 250 nmol/kg). Compound 4 potently mediated anticancer effect in mice treated according to all schedules proposed in the therapy experiment (Fig. 6C) with no effect on body weight (Fig. 6D). OncoFAP-Gly-Pro-MMAE induced a longer tumor growth retardation when administered once a week (with 2/4 cancer cures). The therapeutic activity observed when OncoFAP-Gly-Pro-MMAE (compound 4) is administered weekly supports the importance of the prolonged release of the drug payload at the tumor site. Notably, the first administration of compound 4 induced significant anticancer activity and tumor shrinkage in the SK-RC-52.hFAP model. For this reason and to add preclinical evidence of the OncoFAP-Gly-Pro-MMAE therapeutic potential, we performed an additional therapy study with single administration of compound (compound 4; 250 nmol/Kg) in mice bearing HT-1080.hFAP fibrosarcoma lesions. We included unconjugated cytotoxic MMAE (systemic administration at the equimolar dose of 250 nmol/kg) as negative control in addition to a group of mice treated with saline. Compound 4 promoted rapid and complete tumor regression in all treated animals (4/4 CR; Fig. 6E). Single administration of untargeted MMAE displayed only modest and negligible anticancer activity. These results well correlate with the biodistribution data at 6 hours p.i. (Supplementary Fig. S6) which show poor and nonselective tumor uptake of the untargeted cytotoxic agent. No sign of acute toxicity (body weight loss) was observed in animals treated with free MMAE or with OncoFAP-Gly-Pro-MMAE (Fig. 6F).

In the field of ADCs and SMDCs, the amount of drug payload that is effectively released in tumors and healthy tissues directly impacts on the efficacy and selectivity of treatment (25, 46, 50). Insufficient drug release at the site of disease, which may result from an inefficient or premature linker cleavage, often constitutes the main reason behind clinical failures of ADC (51) and SMDC products (18–21). Mass spectrometric methodologies enable a quantitative measurement of how much drug conjugate is delivered and released within the tumor mass, thus facilitating pharmaceutical development (43, 46, 50, 52. 53).

SMDC products, which have progressed to clinical trials, feature the use of linkers that had originally been designed for internalizing ADC products (54), such as disulfide bridges (14, 47, 55) and the Valine-Citrulline dipeptide (50). These structures are meant to be cleaved intracellularly in the presence of high concentrations of reducing agents (e.g., GSH) or proteolytic enzymes (e.g., cathepsins). However, a growing body of preclinical experimental evidences indicates that CathB-susceptible linkers (e.g., Val-Cit and Phe-Lys) can be efficiently processed in the extracellular space by murine-specific proteases, such as Ces1c (49, 56). Differences in linker cleavage between mouse and man might reflect differences in efficacy and toxicity of drug conjugate constructs. Linkers used for SMDC product development had typically not been optimized for extracellular release and it is not clear whether the suboptimal performance observed in the clinic was due to limitations in the choice of linker or of the cytotoxic payload, or both (18–21). Recent clinical work with bicyclic peptide–drug conjugates suggests that SMDCs can be clinically active if equipped with MMAE-based payloads (50, 57, 58).

In principle, the extracellular release of therapeutic payloads could be implemented using targeting agents specific to stromal antigens (59). Extracellular release mechanisms benefit from a “bystander effect” (59) and may overcome limitations of antigen loss in tumor cells. Indeed, stromal antigens (like FAP) are typically more abundant and genetically stable, compared with markers expressed on the tumor cell membrane (32, 33, 35, 36).

Intracellular and extracellular proteases, which could be exploited for a preferential drug release at the tumor site, vary in concentration and activity not only from tumor to tumor, but also from mouse to man, thus hindering translational activities (56, 60, 61). Therefore, it would be desirable to identify and exploit enzymatic activities which are abundant in the majority of different types of malignancies. FAP is both an excellent target for ligand-based pharmacodelivery, found in many different tumor types, and a protease with a preference for AA₂-Pro₁ substrates. For this reason, we focused our activity on the design of SMDCs that combine a FAP-targeting moiety with FAP-cleavable linkers, thus promoting a preferential homing and processing at the site of disease.

The implementation of mass spectrometry–based analytical methodologies for the quantitative analysis of biodistribution data facilitates the optimization of the drug delivery process. The delivery and release of approximately 10% ID/g of drug payload in the tumor, at the doses used in the therapy experiments described in this article, corresponds to the local deposition of micromolar concentrations of cytotoxic payload within the neoplastic mass. Medium-potency payloads, which are often not considered for ADC applications, may still be suitable for SMDC development activities (12, 62). The use of a relatively large product dose is industrially viable for SMDCs, as those agents are produced by chemical synthesis at acceptably low cost-of-goods (12) and promise to extend the applicability of the methodology to a wide variety of medium to low potency bioactive payloads. Importantly, the mass-spectrometric methodologies used in this article for the study of mouse specimens can also be applied on patient-derived biopsies, thus allowing a quantitative comparison of the drug delivery process in mouse and man.

The nuclear medicine evidence of efficient tumor targeting, gained thanks to PET procedures performed with [⁶⁸Ga]Ga-OncoFAP-DOTAGA (39) and with other FAP-specific PET tracers (31, 32, 34–36, 63, 64), confirms that the vast majority of aggressive human malignancies are strongly positive for FAP and can be efficiently targeted using high-affinity FAP ligands. The results presented in this article suggest that FAP ligands, equipped with judiciously chosen linker payloads, can be used for the selective delivery of potent cytotoxic agents to tumors. In a theranostic approach, PET imaging with ⁶⁸Ga-labeled OncoFAP should enable the identification of those patients which may benefit most from OncoFAP-Gly-Pro-MMAE treatment, on the basis of a preferential accumulation of the FAP ligand at the site of disease.

A. Zana reports grants from Marie Skłodowska-Curie program during the conduct of the study. J. Millul reports a patent for WO2021/160825 pending. T. Sturm reports personal fees from Solvias AG outside the submitted work. L. Stegger reports grants from Philochem outside the submitted work. I. Asmus reports grants from Philochem outside the submitted work. P. Backhaus reports grants from Philochem outside the submitted work. M. Schäfers reports grants from Philochem and Deutsche Forschungsgemeinschaft (DFG) during the conduct of the study. D. Neri reports personal fees from Philogen during the conduct of the study, as well as other support from Philogen outside the submitted work; in addition, D. Neri has a patent for Philogen pending. S. Cazzamalli reports grants from Marie Skłodowska-Curie program during the conduct of the study; in addition, S. Cazzamalli has a patent for WO2021/160825 pending. No disclosures were reported by the other authors.

A. Zana: Data curation, validation, investigation, visualization, methodology, writing–original draft. A. Galbiati: Investigation, writing–review and editing. E. Gilardoni: Investigation, methodology, writing–review and editing. M. Bocci: Investigation, writing–review and editing. J. Millul: Investigation. T. Sturm: Investigation, methodology. R. Stucchi: Data curation, supervision, methodology, writing–review and editing. A. Elsayed: Methodology. L. Nadal: Investigation. M. Cirillo: Investigation. W. Roll: Investigation. L. Stegger: Investigation. I. Asmus: Investigation. P. Backhaus: Investigation, writing–review and editing. M. Schäfers: Resources, supervision, writing–review and editing. D. Neri: Conceptualization, resources, supervision, funding acquisition, writing–original draft, writing–review and editing. S. Cazzamalli: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

The authors wish to thank PD Dr. Jörg Scheuermann and his group for the fruitful and open collaboration.

This project has received funding from the Marie Skłodowska-Curie program (grant agreement 861316).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.