Novel Generation of FAP Inhibitor-Based Homodimers for Improved Application in Radiotheranostics

Simple Summary Radiopharmaceuticals targeting the fibroblast activation protein alpha (FAP) can be used in many different cancer types since FAP is highly expressed in the tumor microenvironment of almost all epithelial cancers. Monomeric radiotracers have shown great potential in molecular imaging (diagnosis), but the tumor retention time is relatively short (few hours). For effective radioligand therapy (RLT), the biological half-life of the radiotracer should ideally match the physical half-life of the important therapeutic radionuclides 177Lu and 225Ac (6.7 and 9.9 days). The tumor retention was improved with the FAPi homodimer DOTAGA.(SA.FAPi)2. In terms of optimizing, the new FAPi homodimers DO3A.Glu.(FAPi)2 and DOTAGA.Glu.(FAPi)2.were synthesized. DOTAGA.Glu.(FAPi)2 showed superior radiolabeling properties (including successful 225Ac-labeling, higher hydrophilicity), in vitro affinity and selectivity compared to DOTAGA.(SA.FAPi)2. In addition, significantly reduced uptake in the critical organs (liver, colon) compared to [177Lu]Lu-DOTAGA.(SA.FAPi)2 was observed with [177Lu]Lu-DOTAGA.Glu.(FAPi)2 in a first patient study (medullary thyroid cancer) while maintaining high and prolonged tumor uptake. Abstract Radiopharmaceuticals based on the highly potent FAP inhibitor (FAPi) UAMC-1110 have shown great potential in molecular imaging, but the short tumor retention time of the monomers do not match the physical half-lives of the important therapeutic radionuclides 177Lu and 225Ac. This was improved with the dimer DOTAGA.(SA.FAPi)2, but pharmacological and radiolabeling properties still need optimization. Therefore, the novel FAPi homodimers DO3A.Glu.(FAPi)2 and DOTAGA.Glu.(FAPi)2. were synthesized and quantitatively radiolabeled with 68Ga, 90Y, 177Lu and 225Ac. The radiolabeled complexes showed high hydrophilicity and were generally stable in human serum (HS) and phosphate-buffered saline (PBS) at 37 °C over two half-lives, except for [225Ac]Ac-DOTAGA.Glu.(FAPi)2 in PBS. In vitro affinity studies resulted in subnanomolar IC50 values for FAP and high selectivity for FAP over the related proteases PREP and DPP4 for both compounds as well as for [natLu]Lu-DOTAGA.Glu.(FAPi)2. In a first proof-of-principle patient study (medullary thyroid cancer), [177Lu]Lu-DOTAGA.Glu.(FAPi)2 was compared to [177Lu]Lu-DOTAGA.(SA.FAPi)2. High uptake and long tumor retention was observed in both cases, but [177Lu]Lu-DOTAGA.Glu.(FAPi)2 significantly reduces uptake in non-target and critical organs (liver, colon). Overall, the novel FAPi homodimer DOTAGA.Glu.(FAPi)2 showed improved radiolabeling in vitro and pharmacological properties in vivo compared to DOTAGA.(SA.FAPi)2. [177Lu]Lu-DOTAGA.Glu.(FAPi)2 and [225Ac]Ac-DOTAGA.Glu.(FAPi)2 appear promising for translational application in patients.


Introduction
Fibroblast activation protein alpha (FAP), also called Seprase, is a S9B family serine protease and an integral type-II transmembrane glycoprotein [1]. It is mainly expressed on activated fibroblasts such as cancer-associated fibroblasts (CAFs), fibroblasts present during wound healing [2], at sites of inflammation (e.g., arthritis) [3] and fibrosis (e.g., liver cirrhosis) [4]. However, FAP is absent in resting fibroblasts of normal healthy tissue. CAFs behave as activated stromal fibroblasts and occupy a major part of the tumor microenvironment (TME) that further consists of vascular, inflammatory and immune cells as well as extracellular matrix components (ECM) [5]. The TME plays an essential role in tumorigenesis, tumor growth, metastasis and angiogenesis and can account for up to 90% of the total tumor mass [5][6][7][8]. Each epithelial tumor more than 1 mm in diameter consists of an extensive TME. More than 90% of these epithelial tumors strongly express FAP, including prostate, breast, lung, ovarian and colorectal cancer, among others [6].
Compared to other tumor targets that are vastly used in nuclear medicine such as prostate-specific membrane antigen (PSMA) or somatostatin receptor (SSTR), radiopharmaceuticals targeting FAP do not address the cancer cells directly but indirectly via its TME. PSMA inhibitor-based tracers can basically only be applied in prostate cancer and octreotide-based SSTR-targeting tracers for neuroendocrine tumors (NETs), while targeting FAP allows for a broader application in many different types of cancers, which makes it a more universal pan-tumor target. Jeremie Calais discussed FAP as potentially being "the next billion dollar nuclear theranostics target" following [ 177 Lu]Lu-PSMA-617 and [ 177 Lu]Lu-DOTA-TATE that has had a great impact in nuclear medicine in the last years [23].
Most of them are conjugated with bifunctional chelators such as DOTA to enable PET imaging with 68 Ga. The rapid renal clearance of monomers such as the squaramide (SA)-based [ 68 Ga]Ga-DOTA.SA.FAPi or [ 68 Ga]Ga-FAPI-04 resulted in high-contrast PET images in various cancers even at early time points (10 min p.i.) [25,40]. However, the short tumor retention time and as such the biological half-life does not match the physical half-life of therapeutic radionuclides such as 177 Lu (t 1/2 = 6.7 d) or 225 Ac (t 1/2 = 9.9 d), which limits the use of these compounds for radioligand therapy (RLT) with the beta minus particle emitter 177 Lu or targeted alpha-particle therapy (TAT) with the alpha emitter 225 Ac.
Recently, peptides were introduced as FAP-targeting vectors in compounds such as FAP-2286, which showed prolonged retention in CAFs-expressing tissue and which represent a new approach regarding FAP radiotherapeutics with a similar approach to TOC or TATE used for peptide receptor radionuclide therapy by targeting SSTRs of NETs [41,42]. Most of them are conjugated with bifunctional chelators such as DOTA to enable PET imaging with 68 Ga. The rapid renal clearance of monomers such as the squaramide (SA)based [ 68 Ga]Ga-DOTA.SA.FAPi or [ 68 Ga]Ga-FAPI-04 resulted in high-contrast PET images in various cancers even at early time points (10 min p.i,) [25,40]. However, the short tumor retention time and as such the biological half-life does not match the physical half-life of therapeutic radionuclides such as 177 Lu (t1/2 = 6.7 d) or 225 Ac (t1/2 = 9.9 d), which limits the use of these compounds for radioligand therapy (RLT) with the beta minus particle emitter 177 Lu or targeted alpha-particle therapy (TAT) with the alpha emitter 225 Ac.
Recently, peptides were introduced as FAP-targeting vectors in compounds such as FAP-2286, which showed prolonged retention in CAFs-expressing tissue and which represent a new approach regarding FAP radiotherapeutics with a similar approach to TOC or TATE used for peptide receptor radionuclide therapy by targeting SSTRs of NETs [41,42].
Another approach is to apply the dimeric concept, which has already shown, to improve accumulation and to prolong tumor retention time in the context of different targets (e.g., PSMA) beforehand [43][44][45][46][47]. This has led to the development of the first homodimeric structure DOTAGA.(SA.FAPi)2 (see Figure 1c). Other groups have adapted this approach, and several other dimeric FAPi-based radiopharmaceuticals were published recently [48][49][50][51]. Zhao et al. developed the dimer DOTA-2P(FAPI)2 based on the FAPI-46 structure, and the 68 Ga-labeled derivative was investigated in vivo [48,49]. Galbiati et al. developed the dimeric BiOncoFAP-DOTAGA [50]. The publication mainly focused on the 177 Lu-labeled therapeutic agent that showed promising results in preclinical studies. Both studies Another approach is to apply the dimeric concept, which has already shown, to improve accumulation and to prolong tumor retention time in the context of different targets (e.g., PSMA) beforehand [43][44][45][46][47]. This has led to the development of the first homodimeric structure DOTAGA.(SA.FAPi) 2 (see Figure 1c). Other groups have adapted this approach, and several other dimeric FAPi-based radiopharmaceuticals were published recently [48][49][50][51]. Zhao et al. developed the dimer DOTA-2P(FAPI) 2 based on the FAPI-46 structure, and the 68 Ga-labeled derivative was investigated in vivo [48,49]. Galbiati et al. developed the dimeric BiOncoFAP-DOTAGA [50]. The publication mainly focused on the 177 Lu-labeled therapeutic agent that showed promising results in preclinical studies. Both studies confirmed that the homodimeric concept is a successful strategy to target FAP and to prolong tumor retention as first demonstrated with DOTAGA.(SA.FAPi) 2 .
The 177 Lu-labeled derivative [ 177 Lu]Lu-DOTAGA.(SA.FAPi) 2 has already been investigated extensively in RNT of radioiodine-refractory differentiated thyroid cancer (RR-DTC) patients, among others (e.g., breast cancer) [39,52]. Much longer tumor residence times of up to one week could be measured via planar scintigraphy and single-photon emission computed tomography (SPECT) imaging [39]. The colon was identified as the critical organ regarding the radiation dose due to slow biliary excretion [39,52]. Furthermore, the linear molecular design with the central chelator may have resulted in steric hindrance during complexation, which led to slower reaction kinetics and/or lower radiochemical yields, in particular with 225 Ac.
On the one hand, the new compounds DO3A.Glu.(FAPi) 2 and DOTAGA.Glu.(FAPi) 2 (Figure 1d,e) were designed, synthesized and investigated to improve radiolabeling, espe-  225 Ac and therefore allowing for potential application in TAT. On the other hand, the pharmacokinetic and pharmacodynamic properties should be optimized while maintaining or even further improving the desired long tumor retention time. In contrast to the original linear dimer DOTAGA.(SA.FAPi) 2 , these new FAPi dimers are designed in a branched structure where the two FAPi targeting vectors and the terminal chelator are coupled via a central glutamic acid (Glu) linker. The further use of squaramide as a linker or spacer was renounced with the intention to synthesize more hydrophilic compounds instead. This could accelerate clearance from non-target tissues as well as excretion to reduce the radiation dose for healthy organs such as colon and liver.
Following the organic synthesis of the precursors, the radiolabeling properties with different radiometals ( 68 Ga,177 Lu, 225 Ac and 90 Y) as well as the complex stabilities were investigated. In addition, the lipophilicity of the 68   Lu-labeled derivatives was evaluated. The new FAPi dimers were tested for their in vitro affinity for FAP as well as PREP, DPP4, DPP8, and DPP9. Beyond that, first proof-of-concept patient investigations were carried out with [ 177 Lu]Lu-DOTAGA.Glu.(FAPi) 2 , which demonstrated faster excretion kinetics and therefore reduced the radiation dose for healthy organs such as the colon and liver. All these studies were performed in comparison to DOTAGA.(SA.FAPi) 2 .
Thin-layer chromatography (TLC) plates coated with silica gel 60 F 254 from Merck were used. For analysis, they were investigated via UV lamp (λ = 254 and 366 nm) as well as staining with potassium permanganate. Silica 60 (0.040-0.063 mm particle size) was used for column chromatography.
Nuclear magnetic resonance (NMR) spectra were recorded in deuterated solvents on an Avance II 400 spectrometer (400 MHz, 5 mm BBFO sample head with z-gradient and ATM, SampleXPress 60 sample changer) or an Avance III 300 spectrometer (300 MHz, 5 mm BBFO sample head with z-gradient and ATM, BACS 60 sample changer) from Bruker (Rheinstätten, Germany). MestReNova 14.2.0 software from Mestrelab Research (Santiago de Compostela, Spain) was used to analyze the spectra.
The reaction solutions were heated at 95 • C, and reaction controls were taken at different time points to investigate reaction kinetics analyzed via radio-TLC. Additional AmOAc/MeOH (1:1) radio-TLCs and radio-HPLC were carried out for the last reaction control to double-check radiochemical conversion (RCC) and purity (RCP). Small scale labeling was carried out as follows: ca. 500 kBq [ 225 Ac]AcCl 3 (in 50 µL 0.04 M HCl) was added to 500 µL 0.1 M sodium ascorbate (pH = 7.0) buffer solution, and then, the respective quantity of precursor 11 (10 or 20 nmol, 10 or 20 µL of 1 mM stock solution) was added and heated at 95 • C for 60 min.
Reaction controls were taken at different time points to investigate reaction kinetics. The radio-TLCs were dried with a heating gun and imaged at different time points (1 h, 1 d) after the TLC had been developed. Additional AmOAc/MeOH (1:1) radio-TLCs were developed for the last reaction control.
Subsequent purification using SepPak ® Light C18 cartridge yielded the product. The cartridge was conditioned with 2 mL EtOH and 5 mL H 2 O. The reaction solution was pushed through the cartridge. The reaction vial was rinsed with 0.1 M sodium ascorbate, and the step was repeated once. The product was washed with 1 mL H 2 O and eluted with 2 mL EtOH/saline (1:1). It was diluted with 8 mL of saline containing 100 mg sodium ascorbate to give the final formulation. The radiochemical yield (>98%) was determined via radio-TLC and high-resolution gamma spectroscopy with HPGe detector. For the gamma spectroscopy, a TLC was developed in standard 0.1 M citrate buffer, cut into two pieces at R f = 0.2-0.3 and measured separately after one hour for 15 min each. The integral of the 221 Fr line (E γ = 218 keV) of each of the two pieces was considered for the RCC determination.

Complex Stability Measurements
After radiolabeling was carried out (RCP > 95%), the stability in human serum (HS) and phosphate-buffered saline (PBS) was investigated (n = 3) by incubating ca. 10 MBq of the labeled tracer solution in 0.5 mL HS and PBS at 37 • C for 120 min ( 68 Ga), for 6 d ( 90 Y) and 14 d ( 177 Lu). The complex stability was determined via radio-TLC with 0.1 M citrate buffer (pH = 4.0) as the mobile phase.
For 225 Ac, 350-400 kBq was incubated in 0.5 mL HS and PBS as well as 200 kBq in 1 mL final formulation solution ("pure"). The complex stability was investigated over 20 days, and the citrate radio TLCs were imaged 1 d after they were developed.

Determination of logD 7.4 (Lipophilicity Measurement)
After radiolabeling was carried out (RCP > 95%), the logD 7.4 value was determined by diluting 10 MBq of the labeled tracer solution to 700 µL with PBS (n = 4). To each, 700 µL of 1-octanol was added and shaken vigorously for 1-2 min (1500 rpm) followed by centrifugation for 1-2 min. Then, 400 µL of the organic and aqueous phase were separated into new Eppendorf tubes. Samples of 3 µL (PBS) or 6 µL (1-octanol) were pipetted onto a TLC plate. Since most of the activity was in the aqueous phase, this was diluted to 700 µL with PBS, and again, 700 µL of 1-octanol was added. The procedure was repeated twice. The TLC plate was imaged (exposure time: 5-10 min), and the integral of each spot (octanol phase: I O , aqueous PBS phase: I PBS ) was determined. The logD 7.4 value was calculated using Equation (1) where the different volumes of V O = 6 µL und V PBS = 3 µL were taken into account.
Only the values of the second and third extraction were considered in the calculation.
All enzymatic activities were determined kinetically for 15 min at 37 • C by measuring initial velocities of AMC release (λex = 380 nm and λem = 465 nm) or pNA release (405 nm) from the substrates mentioned above with an Infinite 200 plate reader from Tecan Group Ltd. (Männedorf, Switzerland). The Magellan software was used to process the data, and data fitting was performed using a non-linear fitting model in Grafit 7.4. All IC 50 measurements were carried out in triplicate with at least eight different inhibitor concentrations tested. The clinical study was approved by the Institute Ethics committee at All India Institute of Medical Sciences (IECPG-22/27.02.2020). All patients gave their written informed consent.

Organic Synthesis
FAPi-NH 2 1 was synthesized according to the procedure in the supporting information (cf. Figure S1).
First, 1 was coupled twice to obtain Boc-Glu.(FAPi) 2 3. Here, HOBt and EDC*HCl with DIPEA in DMF were found to be the best coupling conditions, and 1 needs to be used in slight excess (2.5-3.0 eq). The cleavage of the Boc-protective group can be carried out under very mild conditions with 4 M HCl in 1,4-dioxane (8.6 eq) and acetonitrile as solvent, alternatively using trifluoracetic acid (TFA). The following coupling of the FAPi dimer Glu.(FAPi) 2 4 with DOTA-tris(tert-butyl ester) 8 and DOTAGA( t Bu) 4 5 were both carried out with HATU and DIPEA in DMF. For DO3A, only small amounts of product could be found in LC-MS measurements, and no product could be isolated after column chromatography. Therefore, an alternative route was tried out. Compound 8 was first converted into DOTA( t Bu) 3 -NHS 9. After coupling with 4 with DIPEA in DMF, the tertbutyl ester groups of the chelator were cleaved and following RP-HPLC purification yielded DO3A.Glu.(FAPi) 2 11 with 20% over four steps. DOTAGA.Glu.(FAPi) 2 7 could be obtained with similar yields.
The nat Lu-complexes were obtained by adding 0.1 M LuCl 3 solution (2.0 eq) to 1 M HEPES buffer (pH = 5.5) and the corresponding precursors 7 and 11, respectively. Subsequent RP-HPLC purification gave the Lu-complexed derivatives nat Lu-7 and nat Lu-11.

[ 177
The complex stability of 177 Lu-7 was determined similarly to the 68 Ga-complex 68 Ga-7, except that the incubation at 37 °C was carried out for 14 days. 177 Lu-7 showed no degradation, and the complex was stable over two half-lives ( Figure S6).
The radio-TLCs were always imaged after one hour and one day. Waiting at least one hour is necessary since the equilibrium between actinium-225 and its daughters is disrupted while running the TLC, and the 225 Ac/ 221 Fr-equilibrium has to be reestablished. At this time point, bismuth-213 also contributed to the TLC imaging, which is why the 225 Ac-RCY/RCC is systematically underestimated at this time point. This was broadly investigated with a statistic model by Kelly et al. [59] where they found that a RCP > 90% after 2 h results in a "true" RCP > 97% after 1 day. After one day, bismuth-213 and all the other daughter nuclides decayed. Imaging at this time point gives the distribution of actinium-225. In gamma spectroscopy the gamma-ray lines of francium-221 and bismuth-213 can be distinguished. The amount of actinium-225 can be calculated via the 221 Fr peak; thus, waiting one hour is sufficient in this case.  In a proportional upscale, this would lead to 10 nmol per GBq lutetium-177. The second radio-TLC system (RCC > 99%) and radio-HPLC (cf. Figure S14: RCP = 99.8%) also showed quantitative labeling without any subsequent purification step to separate free noncomplexed lutetium-177.
The complex stability of 177 Lu-7 was determined similarly to the 68 Ga-complex 68 Ga-7, except that the incubation at 37 • C was carried out for 14 days. 177 Lu-7 showed no degradation, and the complex was stable over two half-lives ( Figure S6).
The radio-TLCs were always imaged after one hour and one day. Waiting at least one hour is necessary since the equilibrium between actinium-225 and its daughters is disrupted while running the TLC, and the 225 Ac/ 221 Fr-equilibrium has to be reestablished. At this time point, bismuth-213 also contributed to the TLC imaging, which is why the 225 Ac-RCY/RCC is systematically underestimated at this time point. This was broadly investigated with a statistic model by Kelly et al. [59] where they found that a RCP > 90% after 2 h results in a "true" RCP > 97% after 1 day. After one day, bismuth-213 and all the other daughter nuclides decayed. Imaging at this time point gives the distribution of actinium-225. In gamma spectroscopy the gamma-ray lines of francium-221 and bismuth- At first, test labeling experiments were carried out. Therefore, 10 or 20 nmol of 7 were added to a solution of ca. 500 kBq [ 225 Ac]AcCl 3 and 0.1 M sodium ascorbate (pH = 7.0), which was heated at 95 • C for 60 min (n = 2). For 10 nmol, the RCC was 86.7 ± 3.6% (60 min, 1 d) when imaging after one day. As expected, the RCCs are slightly lower when imaging after one hour, in this case 84.3 ± 4.0% (60 min, 1 h). For 20 nmol, the RCCs were > 90% (15 min) and 91.8 ± 4.7% (60 min, 1 h) and 96.5 ± 1.9% (60 min, 1 d). Exemplary reaction kinetics are shown in Figure S5. It can be seen that at early imaging time points (e.g., 1 h), four spots can be seen, with R f ≈ 0.5 being free actinium-225 and R f = 0.0 being the 225 Ac-complex [ 225 Ac]Ac-DOTAGA.Glu.(FAPi) 2 225 Ac-7. They can also be seen when the exact same TLC is measured again the next day. The other two spots disappear at the radio-TLc after one day. When measuring at additional time points (e.g., 2 h, 3 h etc.), gradually decreasing intensity is observed. This most likely fits the physical half-life of 213 Bi and the ability of Bi to form complexes with DOTA conjugates as a trivalent metal. Therefore, R f > 0.7 is free 213 Bi, whereas the spot at R f = 0.1-0.2 is considered to be [ 213 Bi]Bi-DOTAGA.Glu.(FAPi) 2 , also due its gradual increasing intensity during the reaction.
Higher amounts were used for labeling in similar conditions: 1.6-3.2 MBq [ 225 Ac]AcCl 3 and 30-40 nmol/MBq were heated in 1 mL 0.1 M sodium ascorbate (pH = 7.0) at 95 • C for 60 min (n = 3). Using 40 instead of 30 nmol/MBq did not significantly influence the RCC; hence, the data were merged. The reaction kinetics were analyzed via radio-TLCs ( Figure 5) that were imaged one hour after the TLC had been developed, and the exact same TLC was measured again the next day. . Using 40 instead of 30 nmol/MBq did not significantly influence the RCC; hence, the data were merged. The reaction kinetics were analyzed via radio-TLCs (Figure 5) that were imaged one hour after the TLC had been developed, and the exact same TLC was measured again the next day.  Figure 5 shows the underestimation of the imaging after one hour compared to imaging after one day. The maximum RCC of 87.7 ± 3.2% (1 h) and 94.3 ± 2.1% (1 d) is reached after 15 min of heating. The RCC is not quantitative, but increasing to 40 nmol/MBq did not increase RCC. Higher activity concentrations per volume (MBq/mL) might improve the yield, for example by using more activity, less volume or both.
With longer heating times, the RCC decreases again, which might indicate radiolysis. The effect is stronger in the data after one hour. Kelly et al. [59] also found out that the underestimation is not linear and varies depending on the value of RCP/RCC, which might play a role here. In conclusion, the heating should be shortened (15-30 min) in future 225 Ac-labelings of 7. Lower reaction temperatures should also be investigated in future labeling experiments since the solution became slightly yellow after long heating times, which is probably due to the oxidation of ascorbate.
The RCCs were determined with AmOAc (pH = 4.0)/MeOH (1:1) radio-TLC and gamma spectroscopy measurements of a citrate TLC strip match and confirm the results (Table 1).    Figure 5 shows the underestimation of the imaging after one hour compared to imaging after one day. The maximum RCC of 87.7 ± 3.2% (1 h) and 94.3 ± 2.1% (1 d) is reached after 15 min of heating. The RCC is not quantitative, but increasing to 40 nmol/MBq did not increase RCC. Higher activity concentrations per volume (MBq/mL) might improve the yield, for example by using more activity, less volume or both.
With longer heating times, the RCC decreases again, which might indicate radiolysis. The effect is stronger in the data after one hour. Kelly et al. [59] also found out that the underestimation is not linear and varies depending on the value of RCP/RCC, which might play a role here. In conclusion, the heating should be shortened (15-30 min) in future 225 Ac-labelings of 7. Lower reaction temperatures should also be investigated in future labeling experiments since the solution became slightly yellow after long heating times, which is probably due to the oxidation of ascorbate.
The RCCs were determined with AmOAc (pH = 4.0)/MeOH (1:1) radio-TLC and gamma spectroscopy measurements of a citrate TLC strip match and confirm the results (Table 1). Table 1. Comparison of the determined radiochemical conversion rates (RCCs) with the different methods at different time points of the reaction (15 and 60 min: n = 3; after C18 cartridge purification: n = 1). The imaging/measurement was carried out 1 hour (1 h) or 1 day (1 d) after the radio TLC had been developed. For the complex stability measurements of 225 Ac-7 in HS and PBS, 350-400 kBq of the labeling solution was directly added to HS and PBS (0.5 mL, n = 3) and incubated for 20 days at 37 • C ( Figure S9). The complex [ 225 Ac]Ac-DOTAGA.Glu.(FAPi) 2 remains stable in PBS for a couple of days (89.9 ± 0.7% after 2 d), but then, gradual radiolysis occurs (45.7 ± 0.2% after 20 d). In contrast, the stability in HS is much higher despite having the same activity concentration (85.5 ± 1.7% after 20 d). Higher dilution and the addition of sodium ascorbate and ethanol results in a high stability of the pure final formulation (95.4 ± 1.0% after 20 d). Both are known to be scavengers and are able to reduce radiolysis. It seems reasonable to dilute the final product and add scavengers as much as possible, especially for 225 Ac-radiopharmaceuticals due to the high-energy alpha-particles and the high potential for radiolysis.

[ 68 Ga]Ga-DO3A.Glu.(FAPi) 2 ( 68 Ga-11)
Labeling of DO3A.Glu.(FAPi) 2 11 with gallium-68 was performed similarly to labeling of 7. Since the generator had much lower activities, cationic post-processing was carried out before labeling [53]. In addition, 1 M HEPES buffer was tested at two different pH values (4.5 and 5.5). Three different precursor amounts were used to label 100 MBq 68 Ga at each pH value ( Figure 6). For the complex stability measurements of 225 Ac-7 in HS and PBS, 350-400 kBq of the labeling solution was directly added to HS and PBS (0.5 mL, n = 3) and incubated for 20 days at 37 °C ( Figure S9). The complex [ 225 Ac]Ac-DOTAGA.Glu.(FAPi)2 remains stable in PBS for a couple of days (89.9 ± 0.7% after 2 d), but then, gradual radiolysis occurs (45.7 ± 0.2% after 20 d). In contrast, the stability in HS is much higher despite having the same activity concentration (85.5 ± 1.7% after 20 d). Higher dilution and the addition of sodium ascorbate and ethanol results in a high stability of the pure final formulation (95.4 ± 1.0% after 20 d). Both are known to be scavengers and are able to reduce radiolysis. It seems reasonable to dilute the final product and add scavengers as much as possible, especially for 225 Ac-radiopharmaceuticals due to the high-energy alpha-particles and the high potential for radiolysis.
The complex stability measurements of 177 Lu-11 were performed as described before for 177 Lu-7. Slight decomposition could be observed in PBS (94.7 ± 2.0% intact conjugates after 14 days), whereas 177 Lu-11 seems to be stable in HS, with both having 20 MBq/mL in the beginning ( Figure S11). 177 Lu-11 is slightly less stable than 177 Lu-7 since the additional carboxyl group of DOTAGA might be able to take part in the complexation of lutetium-177, which could give it additional stability. Lutetium can form a complex with unconjugated DOTA where all eight donor ligands (4 × N in the macrocyclic + 4 × O − from the acetate arms) are part of the complex [60]. Whereas the RCC is >99% after 1.5 min for 2 nmol 13, it takes 5 min to reach >95% with 177 Lu-11, and even 5 min to obtain RCC > 98% with 5 nmol 177 Lu-11. Nevertheless, 177 Lu-11 can be synthesized in practically quantitative RCCs (>98% with 2 nmol for 30 min; >99% with 5 nmol for 10 min). The second TLC system and radio-HPLC (cf. Figure S16) also gave almost quantitative RCC/RCP.
The complex stability measurements of 177 Lu-11 were performed as described before for 177 Lu-7. Slight decomposition could be observed in PBS (94.7 ± 2.0% intact conjugates after 14 days), whereas 177 Lu-11 seems to be stable in HS, with both having 20 MBq/mL in the beginning ( Figure S11). 177 Lu-11 is slightly less stable than 177 Lu-7 since the additional carboxyl group of DOTAGA might be able to take part in the complexation of lutetium-177, which could give it additional stability. Lutetium can form a complex with unconjugated DOTA where all eight donor ligands (4 × N in the macrocyclic + 4 × O − from the acetate arms) are part of the complex [60].  Figure 8).
Quantitative RCC > 99% could be achieved with 10 nmol in 15 min. With 5 nmol, the reaction is much slower, but after 45 min, RCC > 99% could also be reached.
Subsequent complex stability studies of 90 Y-11 were carried out the same way as with 68   Lu-cmplexes. The complex can be considered stable (>98%) in both HS and PBS after more than two half-lives (6 days) at 37 • C (see Figure S12). DOTAGA might be able to take part in the complexation of lutetium-177, which could give it additional stability. Lutetium can form a complex with unconjugated DOTA where all eight donor ligands (4 × N in the macrocyclic + 4 × O − from the acetate arms) are part of the complex [60].

Lipophilicity
The lipophilicity was investigated experimentally via the shake-flask method for the 68 Table 2 gives an overview and compares the logD 7.4 values with those of previous works from our group [29,31]. The monomeric radiotracer [ 68 Ga]Ga-DOTA.SA.FAPi showed relatively strong hydrophilic characteristics. This may be one of the reasons that in vivo PET studies indicated a relatively fast renal clearance [58]. In contrast, the dimer [ 177 Lu]Lu-DOTAGA.(SA.FAPi) 2 predominantly showed biliary excretion over renal excretion [39]. The washout from the hepatobiliary system/liver was slower than that of the monomeric radiotracers (the same for the renal excretion). In particular, [ 177 Lu]Lu-DOTAGA.(SA.FAPi) 2 remained in the large intestines/colon region for a longer time during excretion. The colon was identified as the critical organ, and the radiation dose to the kidneys was increased compared to the monomer [39,52]. Therefore, we tried to synthesize compounds that are more hydrophilic than DOTAGA.(SA.FAPi) 2 to reduce biliary excretion and thus the radiation dose to the colon. FAPi) 2 may create radiolabeling limitations due to steric hindrance as a result of the chelator being in the center of the molecule. The DOTA chelator only has six donor ligands because it is attached to the two targeting vectors on two binding sites, which only allow for labeling with gallium-68 [31]. The DOTAGA derivative has seven donor ligands for complexation and can be labeled with lutetium-177, while it was cumbersome in the case of actinium-225. The new FAPi dimer DOTAGA.Glu.(FAPi) 2 7 has eight donor ligands and can be labeled with actinium-225 with high RCC > 90%. 225 Ac-labeling of DO3A.Glu.(FAPi) 2 11 was just as problematic as with DOTAGA.(SA.FAPi) 2 . In general, DOTA-conjugated precursors have been investigated for 225 Ac-labeling quite extensively and allow for efficient labeling when the chelator is in exo-position [61][62][63][64][65][66].
In summary, DOTAGA.Glu.(FAPi) 2 showed superior labeling properties compared to the other dimeric compounds.

In Vitro Inhibition Assays
High affinity for FAP and low affinity for the very similar proteases PREP and DPPs (DPP4, DPP8 and DPP9) are important requirements for FAPi-based radiopharmaceuticals to be promising candidates for subsequent in vivo studies.
The IC 50 values were determined in in vitro studies. Table 3 summarizes the results of the IC 50 measurements. The selectivity index (SI) is the ratio of the IC 50 value of the respective protease (PREP, DPP4, DPP8 and DPP9) to the IC 50 value of FAP. Although selectivity indexes (SIs) are lower compared to UAMC-1110, high selectivity for FAP over PREP and DPP4 was measured, which are most relevant among the related proteases. In contrast to FAP, they are ubiquitously expressed in healthy tissue. A high affinity for PREP or DPP4 would lead to worsened tumor selectivity and a diminished tumor-to-background ratio in PET/SPECT scans and/or RNT. Overall, these results strongly suggest that compounds 7 and 11 might also have superior properties in vivo compared to DOTAGA.(SA.FAPi) 2 and that especially FAPi dimer 7 is suitable for further investigations.
The IC 50 (FAP) values reported for BiOncoFAP-DOTAGA and [ nat Lu]Lu-BiOncoFAP-DOTAGA were 0.17 and 0.19 nM, respectively [50]. These values seem to be equal to 7 and nat Lu-7, although it has to be noted that the assay conditions were different (substrate/enzyme/buffer concentration, temperature, volume etc.). For the dimer [ 68 Ga]Ga-DOTA-2P(FAPI) 2 , the assay setup differed drastically, as a competitive assay of the radiolabeled agent against FAPI-46 was carried out and resulted in an IC 50 = 3.68 ± 1.82 nM [48]. The different assay conditions, especially of Zhang et al. [48], limit the comparability of the different IC 50 values. On top, no IC 50 values for the related proteases and therefore no selectivity indexes were reported by Zhao et al. [48] or by Galbiati et al. [50].

Patient Study (Medullary Thyroid Cancer)
A 40-year-old male was diagnosed with medullary thyroid cancer in 2016. The patient underwent total thyroidectomy and bilateral radical neck dissection in prior treatments. 177 Lu-7, which was previously synthesized according to the procedure described in Methods and Materials (Section 2.5.1). Furthermore, 24 h p.i. high uptake in the neck and mediastinal metastases was comparable to the first-generation dimer. In contrast, only negligible radiotracer uptake in the liver and colon was observed. This suggests a different excretion pattern, possibly a higher proportion of renal instead of hepatobiliary excretion (improved pharmacodynamic profile) or a faster washout (improved pharmacokinetic profile), or both, which results in a significantly reduced radiation dose to the critical healthy organs liver and colon.
While this is a first proof-of-concept investigation and not of quantitative dimension, the results support the assumption that the new 177   The patient was administered 5.5 GBq of [ 177 Lu]Lu-DOTAGA.Glu.(FAPi)2 177 Luwhich was previously synthesized according to the procedure described in Methods an Materials (Section 2.5.1). Furthermore, 24 h p.i. high uptake in the neck and mediastina metastases was comparable to the first-generation dimer. In contrast, only negligib radiotracer uptake in the liver and colon was observed. This suggests a different excretio pattern, possibly a higher proportion of renal instead of hepatobiliary excretio (improved pharmacodynamic profile) or a faster washout (improved pharmacokinet profile), or both, which results in a significantly reduced radiation dose to the critica healthy organs liver and colon.
While this is a first proof-of-concept investigation and not of quantitative dimension the results support the assumption that the new 177

Conclusions
DOTAGA.Glu.(FAPi) 2 and DO3A.Glu.(FAPi) 2 represent the second generation of our homodimeric FAPi compounds. They were synthesized with glutamic acid as a central linker unit and successfully labeled with different trivalent radiometals. Both precursors show excellent labeling properties with 68 Ga and 177 Lu. In contrast to DO3A.Glu.(FAPi) 2 and the first-generation dimer DOTAGA.(SA.FAPi) 2 , DOTAGA.Glu.(FAPi) 2 could also be labeled successfully with 225 Ac in high yields. All complexes showed high stability and hydrophilicity. Beyond that, IC 50 values were determined to confirm the very high affinity for fibroblast activation protein alpha (FAP) and high selectivity for FAP over prolyl endopeptidase (PREP) and the dipeptidyl peptidase (DPP4). The in vitro studies showed the best results for DOTAGA.Glu.(FAPi) 2 , but the results for DO3A.Glu.(FAPi) 2 were also better than for DOTAGA.(SA.FAPi) 2 . It is also noteworthy that the respective FAP affinity of the new homodimeric compounds is higher than for the initial FAP inhibitor UAMC-1110 itself, which was measured in the same assay. Similar FAP affinities were reported for BiOncoFAP-DOTAGA and its nat Lu-complex by Galbiati et al. [50], but no selectivity data were reported. The comparison of the IC 50 values is difficult and should be cautiously interpreted, as the inhibition assays are different from each other (especially regarding Zhao et al. [48]).
DOTAGA.Glu.(FAPi) 2 proved to be the most promising candidate for further in vivo studies that are currently ongoing. The compound seems especially suited for therapeu-tic applications with 177 Lu, 90 Y and 225 Ac. Especially, FAP-targeted alpha-particle therapy (FAP-TAT) with 225 Ac-radiopharmaceuticals is a new treatment option that could potentially be more efficacious than with beta-emitting therapeutics since stromal cells, similar to how CAFs are reported to be more resistant to radiation [67][68][69]. Therefore, particles with higher energy and linear energy transfer such as alpha particles might be needed to address the specifics of the tumor stroma. The stroma can build a physical barrier around the tumor that is hard to penetrate for some agents (e.g., chemotherapeutics), which results in poor efficacy [5,70,71]. The destruction of tumor stroma with alpha-emitters in combination with other non-radioactive agents (e.g., chemotherapeutics, targeted drug delivery systems, CAR T-cell therapy etc.) might also be a promising therapy approach, and [ 225 Ac]Ac-DOTAGA.Glu.(FAPi) 2 might be an interesting candidate for FAP-TAT in the future.
A first proof-of-concept patient study with [ 177 Lu]Lu-DOTAGA.Glu.(FAPi) 2 showed promising results. The pharmacological properties are superior to those of [ 177 Lu]Lu-DOTAGA.(SA.FAPi) 2 . with a high radiotracer uptake and long tumor retention in lesions. At the same time, the excretion kinetics are faster, and minimized uptake in non-target tissues, especially the colon as the critical organ for [ 177 Lu]Lu-DOTAGA.(SA.FAPi) 2 , was observed. Improved pharmacologic properties are essential to ensure a safe therapy without adverse side effects. Further studies are currently ongoing and will investigate biodistribution, efficacy and safety in more detail.