Targeting fibroblast activation protein (FAP): next generation PET radiotracers using squaramide coupled bifunctional DOTA and DATA5m chelators

Background Fibroblast activation protein (FAP) is a proline selective serine protease that is overexpressed in tumor stroma and in lesions of many other diseases that are characterized by tissue remodeling. In 2014, a most potent FAP-inhibitor (referred to as UAMC1110) with low nanomolar FAP-affinity and high selectivity toward related enzymes such as prolyl oligopeptidase (PREP) and the dipeptidyl-peptidases (DPPs): DPP4, DPP8/9 and DPP2 were developed. This inhibitor has been adopted recently by other groups to create radiopharmaceuticals by coupling bifunctional chelator-linker systems. Here, we report squaric acid (SA) containing bifunctional DATA5m and DOTA chelators based on UAMC1110 as pharmacophor. The novel radiopharmaceuticals DOTA.SA.FAPi and DATA5m.SA.FAPi with their non-radioactive derivatives were characterized for in vitro inhibitory efficiency to FAP and PREP, respectively and radiochemical investigated with gallium-68. Further, first proof-of-concept in vivo animal study followed by ex vivo biodistribution were determined with [68Ga]Ga-DOTA.SA.FAPi. Results [68Ga]Ga-DOTA.SA.FAPi and [68Ga]Ga-DATA5m.SA.FAPi showed high complexation > 97% radiochemical yields after already 10 min and high stability over a period of 2 h. Affinity to FAP of DOTA.SA.FAPi and DATA5m.SA.FAPi and its natGa and natLu-labeled derivatives were excellent resulting in low nanomolar IC50 values of 0.7–1.4 nM. Additionally, all five compounds showed low affinity for the related protease PREP (high IC50 with 1.7–8.7 μM). First proof-of-principle in vivo PET-imaging animal studies of the [68Ga]Ga-DOTA.SA.FAPi precursor in a HT-29 human colorectal cancer xenograft mouse model indicated promising results with high accumulation in tumor (SUVmean of 0.75) and low background signal. Ex vivo biodistribution showed highest uptake in tumor (5.2%ID/g) at 60 min post injection with overall low uptake in healthy tissues. Conclusion In this work, novel PET radiotracers targeting fibroblast activation protein were synthesized and biochemically investigated. Critical substructures of the novel compounds are a squaramide linker unit derived from the basic motif of squaric acid, DOTA and DATA5m bifunctional chelators and a FAP-targeting moiety. In conclusion, these new FAP-ligands appear promising, both for further research and development as well as for first human application.

UAMC1110 is currently still under evaluation as a potential therapeutic in diseases characterized by FAP expression. At the same time, the molecule is being used as a FAP-targeting moiety in so-called activity-based probes that can be used to visualize and quantify FAP activity in tissues and organisms (De Decker et al. 2019). Highly relevant examples have also been published that rely on radionuclide-based reporter systems, such as XY-FAP-02 developed by Yang et al. (Yang et al. 2019). They used a DOTAGA chelator combined with an alkyl chain as linker system bound to the FAPinhibitor.
Further development of radiotracers by Lindner and Loktev et al. based on the FAP inhibitors from Antwerp have shown promising results in preclinical and first clinical patient studies. Applications of these molecules cover both diagnosis and therapy (Giesel et al. 2019b;Giesel et al. 2019a;Kratochwil et al. 2019;Lindner et al. 2018;Loktev et al. 2019;Loktev et al. 2018). First, a DOTA-FAPI conjugate using piperazine as linker (referred to as FAPI-02 in the original reference) was synthesized and characterized with respect to binding, internalization, and efflux in cells expressing human and murine FAP as well as CD26. PET-imaging studies of HT-1080 tumor xenografts showed low [ 68 Ga]Ga-FAPI-02 accumulation in normal tissues and a rapid clearance from the blood via kidneys and bladder. In addition, a high tumor uptake resulting in high tumor-to-normal organ-ratio was determined. By structural variation, especially in the linker region, more analogous gallium-68 labeled compounds were obtained. Several of these had improved imaging parameters, with FAPI-04,  being relevant examples Loktev et al. 2019). These compounds also had low nanomolar FAP-affinities, higher tumor uptakes in vivo and longer tumor retention times. First PET/CT imaging studies of patients diagnosed with different tumor entities were performed with the gallium-68 compounds indicating high tumor uptake and low background in healthy organs. As an example of a first therapeutic application, patients diagnosed with metastatic breast cancer were treated with [ 90 Y]Y-FAPI-04. The 68 Ga/ 90 Y-DOTA-derivatives represent promising tracers for both diagnostic imaging and, possibly, targeted therapy of malignant tumors with high accumulated activated fibroblasts. In this work, novel FAP-targeting radiotracers were evaluated using bifunctional DOTA and DATA 5m chelators coupled by squaramide as linker moiety. The basic motif squaric acid (SA) is a cyclic aromatic diacid (Ian Storer et al. 2011;Wurm and Klok 2013). One advantage of SA is the simple chemistry regarding coupling to chelator and target vector including that no protecting groups are necessary due to its selectivity for primary amines. Especially reactions with biomolecules are attractive and no side reactions are observed. The coupling with SA-diester is a highly selective, pH controlled asymmetric amidation under mild conditions (Tietze et al. 1991). In a neutral pH, only one ester of the SA-diester reacts with an amine and by increasing the pH to basic conditions, amidation of the second ester takes place. The use of SA as a linker unit between a chelator-biomolecule conjugate as a radiopharmaceutical was demonstrated using DFO and conjugation on a peptide to complex iron and using DFOsquaric acid coupled to antibodies for complexing zirconium-89 (Rudd et al. 2016;Yoganathan et al. 2011). Recently, our group published the usage of SA as a linker forming a radiotracer with the bifunctional hybrid chelator AAZTA 5 coupled to a PSMA inhibitor unit (KuE) and evaluated those AAZTA 5 .SA.PSMA conjugate with various radionuclides such as scandium-44, copper-64, gallium-68 and lutetium-177 (Greifenstein et al. 2019 (Greifenstein et al. 2020).
Here, the preparative synthesis of DOTA.SA.FAPi and DATA 5m .SA.FAPi and the metal-analogs [ nat Ga]Ga-DOTA.SA.FAPi, [ nat Ga]Ga-DATA 5m .SA.FAPi and [ nat Lu]Lu-DOTA.SA.FAPi are described. The macrocyclic chelator DOTA was used to allow labeling with both gallium-68 and lutetium-177. However, one disadvantage of these chelator types are the requirement of high temperatures for complexation (Price and Orvig 2014). DATA 5m , a bifunctional version of the hybrid chelator DATA, was used to allow instant gallium-68 labeling at room temperature (Seemann et al. 2017;Seemann et al. 2015;Sinnes et al. 2019). Radiochemical evaluation with regard to labeling and in vitro stability studies were performed with gallium-68 for DOTA.SA.FAPi and DATA 5m .SA.-FAPi. For all the five cold compounds, inhibition assays were carried out and IC 50 values obtained for FAP and PREP. In a first proof-of-principle PET-study, [ 68 Ga]Ga-DOTA.SA.FAPi was tested in vivo using a FAP-expressing HT-29 human colorectal adenocarcinoma xenograft model. Figure 2 shows the squaric acid containing tracers DOTA.SA.FAPi and DATA 5m .SA.FAPi based on UAMC1110.
Synthesis of DATA 5m .SA.FAPi DATA 5m -3 t Bu 5 was synthesized as described by Seemann et al. (Fig. 4) (Seemann et al. 2017). DATA 5m -3 t Bu provides a bifunctional carbonyl group for further coupling with spacer molecules or target vectors. Terminal primary amines are required for binding to SA-diethylester. Therefore N-boc-ethylenediamine was attached to the carboxylic acid group of DATA 5m via common coupling reagents HATU in DIPEA and acetonitrile receiving 6. Amidation of SA-monoester 7 with the terminal amine of NH 2 -UAMC1110 was executed analogously to DOTA.SA.-FAPi to receive DATA 5m .SA.FAPi 8. In the inhibition assays, DOTA.SA.FAPi and DATA 5m .SA.FAPi, along with their nonradioactive, metal complexed analogues were characterized for inhibitory potency towards FAP and PREP. Earlier work had shown that the lack of a basic amine function in UAMC1110-based molecules, precludes DPP-affinity in this series (De Decker et al. 2019;Jansen et al. 2014a). Nonetheless, the FAP/PREP selectivity was shown to be a particularly important parameter to check. Obtained results are summarized in Table 1. Parent compound UAMC1110 was used as a reference in this assay. All the evaluated molecules displayed highly satisfactory, low nanomolar FAP potencies, in the same range as the parent inhibitor UAMC1110. This implies that introduction of a linker, a chelator and a metal ion at the selected position of the quinoline ring are tolerated by FAP and have no negative influence on target affinity. Likewise, equally satisfactory compound selectivities with respect to PREP were measured, again comparable with UAMC1110.
Stability studies were performed in ethanol (EtOH), human serum (HS) and saline 0.9% (NaCl) over a period of 2 h at 37°C. In all three media, [ 68 Ga]Ga-DOTA.SA.FAPi showed high stabilities over 98% intact conjugate (SI, Fig. S4). In addition, stability against transmetallation and transchelation were carried out (SI, Fig. S5, S6). Against DTPA and EDTA the stability values were > 98% and against Cu, Mg and Ca the stabilities were > 95% after 2 h. Stabilities against Fe showed > 95% after 90 min and a slightly lower value however still over 92% after 2 h.

PET/CT-imaging and ex vivo biodistribution data of [ 68 Ga]Ga-DOTA.SA.FAPi
For investigation of the tumor uptake by [ 68 Ga]Ga-DOTA.SA.FAPi HT-29 tumorbearing mice (n = 3) were sacrificed after PET/CT scans and an ex vivo biodistribution study was executed. In the PET images, the tumor accumulation is clearly visible (SUV mean of 0.75 ± 0.09) and the ratio to nonspecific organs and tissues is very high (SUV mean : 0.15 ± 0.01 in the heart, 0.18 ± 0.07 in the muscle, 0.37 ± 0.14 in the small intestine, 0.27 ± 0.11 in kidneys and 0.22 ± 0.08 in the liver)). Figure 7 shows the maximum intensity projection (MIP) images of three mice. Ex vivo biodistribution is shown in Fig. 8a. The accumulation in the tumor at 60 min post injection (p.i.) as found in both the PET images and biodistribution with high with an overall uptake of 5.2%ID/g. In general, the tumor-to-organ ratios are high after 1 h p.i. which is shown by, e.g., tumor-to-blood (9.2 ± 1.1), tumor-to-large intestine (24.9 ± 1.7) and tumor-to-muscle (11.5 ± 2.2) ratios ( Fig. 8b). Uptake in other organs are also low such as in heart, lungs, liver, spleen, pancreas, stomach, fat and skin. In addition to the tumor accumulation, a slightly higher accumulation at the bones and small intestine were found, which cannot yet be fully explained. One suggestion could be that FAP is also expressed in these tissues. However, it is important that the main accumulation is located in the tumor and although the unexpected radiotracer uptake in bone and small intestine the tumor-to-bone (1.5 ± 0.2) and tumor-to-small intestine (2.9 ± 0.8) ratios are still high enough to provide high contrast PET images.

Discussion
Two novel bifunctional chelator-linker conjugates based on the FAP inhibitor UAMC1110 were developed. As bifunctional chelators, macrocyclic DOTA as well as the hybrid chelator DATA 5m were used. DATA is well known for fast and stable complexation of gallium-68 and to show high stabilities over a long period of time. DOTA is interesting because of its ability to complex other radiometals preferring higher coordination numbers, e.g., the long-lived therapy radionuclide lutetium-177 but also shorter-lived therapeutic radionuclides such as bismuth-213, lead-210 and yttrium-90. DOTA.SA.FAPi allows to use the same precursor for both diagnosis with gallium-68 and therapy with lutetium-177 in nuclear medicine. SA is the main component of the linker system forming a squaramide unit and accordingly substitutes the heterocyclic nitrogen moieties in the structures reported by the Heidelberg group Loktev et al. 2019;Loktev et al. 2018). In recent works from our group, SA has shown good results, both in chemistry and physiologically, as a linker unit coupled with PSMA inhibitors (Greifenstein et al. 2020;Greifenstein et al. 2019). The asymmetrically substituted squaramide unit in the target compounds was efficiently installed with SADE, relying on the latter's elegant pH-dependent chemistry and selectivity for primary amines. Due to this selectivity, no protecting groups are required. In addition, the pH dependent reactivity of SADE is explained by changes of the aromatic stabilization energy in the ring system over the course of the sequential amidation steps (Tietze et al. 1991).
After the first amidation, which is carried out at neutral pH, the obtained uncharged monoamide is characterized by a higher aromatic stabilization than the starting material, squaric acid diester. This stabilization prevents addition of a second amine molecule. By increasing the pH, the monoamide is deprotonated, loses aromatic stabilization, and this allows for addition of a second amine molecule to provide the diamide (Quiñonero et al. 2000;Wurm and Klok 2013). Correspondingly, C 2 -symmetric diamide derivatives of SA can be obtained, if the reaction is performed at higher pH: in that case, two equivalents of amine will directly substitute the ethoxy groups in SADE. Both DOTA.SA.FAPi and DATA 5m .SA.FAPi could be well used for radiolabeling with gallium-68. DATA has already demonstrated good complexation with gallium-68 and has the advantage of labeling even under mild conditions such as room temperature. Quantitative gallium-68 labeling results were observed for both FAPi-conjugates with gallium-68. The DOTA complex showed quantitative radiochemical yields at temperatures of 95°C with precursor amounts of > 15 nmol. For the DATA 5m conjugate quantitative yields could be achieved at room temperature for amounts > 15 nmol. In addition, it could be seen that with lower competing cations of gallium-68, less precursor amount (≥ 10 nmol) is required to achieve quantitative complexation, whereas with higher gallium activity (> 200 MBq) more substance > 15 nmol is needed for quantitative yields.
The stability for both derivatives against different media was high with > 95% intact conjugates. Preclinical in vivo animal studies were performed with HT-29 xenograft mice. The HT-29 cancer cell line is a human colorectal adenocarcinoma cell line with epithelial morphology (Cheng et al. 2002;Henry et al. 2007). When inoculated into nude mice, they produce undifferentiated tumors with modest stroma. Significant FAP expression is present in this stroma, that typically trabeculates between nests of HT-29 cells. Notably, and similar to the situation in most tumor types, FAP staining is distinctly absent from the actual HT-29 cancer cells within the tumors (Cheng et al. 2002).
The highest accumulation in the tumor was found in both the PET images with SUV mean of 0.75 ± 0.09 60 min post injection and biodistribution with an overall uptake of 5.2%ID/g. In addition, the tumor-to-organ ratios in the biodistribution data were quite high (tumor-to-blood (9.2 ± 1.1), tumor-to-large intestine (24.9 ± 1.7) and tumor-to-muscle (11.5 ± 2.2)), which is also reflected by the high contrast in the images. Besides the high tumor uptake, accumulation in the bladder could also be observed, suggesting renal clearance to be the predominant excretion.
The reference FAPI-04 could not be included in the experiments because it was not available yet. While head-to-head comparison with 68 Ga-FAPI-04 was not possible, we may compare the SUV mean value data from reference publications. It should be noted, however, that different tumor models in different test series were used. Therefore, a direct comparison of the results is not advisable, but tendencies could still be observed.

Conclusion
In this work, two potential theranostic radiopharmaceuticals were successfully synthesized, based on the selective FAP-inhibitor UAMC1110. Key elements of these compounds are a squaramide motif (introduced via amidation of SADE) and a DOTA or DATA 5m -type chelator. Due to the unique chemistry of SADE, it was possible to avoid complex synthesis routes and protective group strategies. DOTA.SA.FAPi and DATA 5m .SA.FAPi showed very good in vitro complexations of gallium-68 and a very high stability in different media of more than 95% intact conjugate. In general, the hybrid chelator DATA 5m shows a quantitative complexation under mild conditions and is therefore very well suited to label temperature sensitive target molecules with radiometals. Both FAPi-precursors as well as their gallium and lutetium versions showed excellent affinity and selectivity to FAP, in the low nanomolar range, with IC 50 -values between 0.7 and 1.4 nM. Conversely, PREP IC 50 -values were found to be in the μM-range, implying excellent FAP/PREP selectivity indices.
In the HT-29 colon cancer xenograft model, first proof-of-concept animal studies with [ 68 Ga]Ga-DOTA.SA.FAPi showed good tumoral accumulation with high uptake (SUV mean 0.75 ± 0.09) at 60 min p.i. Ex vivo biodistribution revealed 5.2 ± 0.2% ID/g on average and low background activity, i.e. an overall good tumor-toorgan ratio. Comparison of the different tumor models with the reference compound FAPI-04 has shown that DOTA.SA.FAPi offers comparable results to FAPI-04. The values should not be compared in direct relation, as there are different test series as well as different tumor models, but nevertheless a tendency can be exhibited.
The potential of the novel compound family to target FAP could be clearly demonstrated. The introduction of squaric acid as linker forming a squaramide bond between bifunctional chelator and pharmacophore firstly simplified the preparative work and secondly showed pharmacological improvements due to the excellent in vitro binding affinities and the great in vivo/ ex vivo data. Further preclinical characterizations for both precursors are planned for publication at a later stage. In meantime, a first clinical trial was carried out in cooperation with the University Medical Center Bonn showing specific uptake in focal nodular hyperplasia (Kreppel et al. 2020). Further patient investigations are ongoing and we expect that our FAPi based radiotracers could be of importance characterize various malignant and benign tumor types in nuclear medicine.

Organic synthesis
Synthesis of DOTA.SA.FAPi

Inhibitory potency determination
Enzymes: A gateway-entry clone for human FAP was purchased from Dharmacon (Accession number DQ891423) and the human secretion signal was replaced with the HoneyBee mellitin secretion signal. For transfection and expression of FAP in Sf9 insect cells, the C-terminal BaculoDirect kit from LifeTechnologies was used. The enzyme was purified from the supernatant of the insect cells using immobilized Nichelating chromatography (GE healthcare, Diegem, Belgium), followed by anionexchange chromatography using a 1 mL HiTrap Q (GE healthcare, Diegem, Belgium). Human recombinant PREP was expressed in BL21(DE3) cells and purified using immobilized Co-chelating chromatography (GE healthcare) followed by anion-exchange chromatography on a 1 ml Mono Q column (GE healthcare).
FAP: IC 50 measurements of the inhibitors were carried out using Z-Gly-Pro-7-amino-4methylcoumarine (AMC) (Bachem, Switzerland) as the substrate at a concentration of 50 μM at pH 8 (0.05 M Tris-HCl buffer with 0.1% glycerol, 1 mg/mL BSA and 140 mM NaCl). Eight concentrations of inhibitors were tested. The final DMSO concentration was kept constant during the experiment to exclude any solvent effects. Inhibitors were preincubated with the enzyme for 15 min at 37°C, afterwards the substrate was added and the velocities of AMC release were measured kinetically at λ ex = 380 nm, λ em = 465 nm for at least 10 min at 37°C. The Infinite 200 (Tecan Group Ltd.) micro-titer plate reader and the Magellan software were used for measurement and data processing respectively.
Note: a slightly different protocol, involving a different FAP substrate (Ala-Pro-pNA), was used to determine the originally published FAP IC 50 -value for reference UAMC1110 (3.2 +/− 0.4 nM). This accounts for the non-identical value published here.
PREP: IC 50 measurements of the inhibitors were carried out using N-succinyl-Gly-Pro-AMC (Bachem, Switzerland) as the substrate at a concentration of 250 μM at pH 7.4 (0.1 M K-phosphate, 1 mM EDTA, 1 mM DTT). Eight concentrations of inhibitors were tested. The final DMSO concentration is kept constant during the experiment to exclude any effects. Inhibitors were pre-incubated with the enzyme for 15 min at 37°C, afterwards the substrate was added and the velocities of AMC release were measured kinetically at λ ex = 380 nm, λ em = 465 nm for at least 10 min at 37°C. The Infinite 200 (Tecan Group Ltd.) micro-titer plate reader and the Magellan software were used for measurement and data processing, respectively.
The data were fitted using a non-linear fit model in GraFit 7 software, according to the following equation: where y is the value of the residual enzymatic activity compared to a non-inhibited sample, x is the final inhibitor concentration in the assay, s is the slope factor and the IC 50 is the half maximal inhibitory concentration.

Radiolabeling and stability studies with gallium-68
Gallium-68 was obtained manually utilizing ethanol-based post-processing from a 68 Ge/ 68 Ga-generator (ITG Garching, Germany). Elution process was performed following the protocol established by Eppard et al. (Eppard et al. 2014). After elution of gallium through the generator with a 0.05 M HCl (5 mL) solution gallium-68 was distributed on the microchromatography CEX column AG 50 W-X4. The column was washed with 1 mL 80% EtOH/ 0.15 M HCl and the Ga (III) was eluted from the column with 400 μL 90% EtOH/ 0.9 M HCl. The washing step ensures that unwanted chemical and radiochemical impurities are separated and only 68 Ga 3+ remains on the column. Reaction controls for radiochemical purity were executed using radio-TLC (TLC Silica gel 60 F 254 Merck) with citrate buffer pH 4 and radio-HPLC using an analytical HPLC 7000 series Hitachi LaChrom with a Phenomenex (Aschaffenburg, Germany) Luna C18 column (250 × 4.6 mm, 5 μ), linear gradient of 5-95% MeCN (+ 0,1% TFA)/ 95-5% Water (+ 0,1% TFA) in 10 min). TLC's were measured in TLC imager CR-35 Bio Test-Imager from Duerr-ndt (Bietigheim-Bissingen, Germany) with the analysis software AIDA Elysia-Raytest (Straubenhardt, Germany). The citrate TLCs show free radio metal with a R f value of 0.8-0.9. The labeled complexes are observed at a R f value of 0.1-0.2.

In vivo animal studies and ex vivo biodistributions
After quantitative radiolabeling of [ 68 Ga]Ga-DOTA.SA.FAPi with a tracer amount of 20 nmol at 95°C in 20 min. The solution was purified via C-18 column (Sep-Pak Light C18, Waters Corporation, Massachusetts, USA). Conditioning of the SPE was performed using 5 ml abs. Ethanol and 5 mL water. Crude reaction mixture was pressed over the SPE and then washed with 5 ml water. Afterwards, the gallium-68 labeled product was eluted with 1 mL of 50 vol% ethanol. Finally, the ethanol was evaporated and the tracer was reformulated in 5% ethanol in saline solution (500 μl total volume). The radiochemical purity was > 99% and no traces of free gallium could be detected by radio-TLC analysis (mobile phase: citrate buffer pH 4.0) and RP-HPLC (5-95% MeCN (+ 0,1% TFA)/95-5% Water (+ 0,1% TFA) in 10 min). The activity after purification process was 200 ± 10 MBq (10 GBq/ μmol) with a total RCY of 56%.
In vivo tumor model: HT-29 (human colon adenocarcinoma, ATCC, Rockville, Maryland) cells were routinely cultured in Dulbecco's Modified Eagle Medium supplemented with 10% heat inactivated foetal bovine serum (FBS), 2 mM glutamine, 1% sodium pyruvate and 1% penicillin/streptomycin (Gibco, Life technologies). After detaching the cells, the number of viable cells was counted with the automated Muse™ Cell Analyzer (Merck Millipore). For the HT-29 subcutaneous model, 10.10 6 viable cells, suspended in 100 μl PBS, were inoculated in the right hind leg of female 6-week-old CD1 −/− Foxn1nu mice (n = 3), obtained from Charles River Laboratories (L'Arbresle, France). The animals were kept under environmentally controlled conditions (12 h light/dark cycle, 20-24°C and 40-70% relative humidity) with food and water ad libitum. When tumors reached an approximate volume of 400 mm 3 , 3 mice underwent μPET imaging. All experimental procedures and protocols were performed in accordance with European Directive 86/609/EEC Welfare and Treatment of Animals and were approved by the local ethical commission (2017-070, University of Antwerp, Belgium).
Micro-PET imaging: Micro-PET scans were carried out using an Inveon small-animal PET/CT scanner (Siemens), after i.v. injection of 4 nmol of [ 68 Ga]-DOTA.SA.FAPi (8.6 MBq, molar activity of 2.1 GBq/ μmol) into tumor bearing mice (n = 3), under isoflurane anesthesia (5% for induction, 2% for maintenance). Static whole-body PET images were acquired 60 min after injection of the radiotracer. Following each PET acquisition, a wholebody CT scan was acquired to obtain the animal's anatomical information individually.
Ex vivo biodistribution: Immediately after the CT scans, the animals were sacrificed, the blood, tissues and organs were collected, weighed and the radioactivity was measured using an automatic γ-counter (Wizard 2 2480, PerkinElmer). Values were expressed as percentage of the injected dose per gram (%ID/g).