Design of PSMA ligands with modifications at the inhibitor part: an approach to reduce the salivary gland uptake of radiolabeled PSMA inhibitors?

Aim To investigate whether modifications of prostate-specific membrane antigen (PSMA)-targeted radiolabeled urea-based inhibitors could reduce salivary gland uptake and thus improve tumor-to-salivary gland ratios, several analogs of a high affinity PSMA ligand were synthesized and evaluated in in vitro and in vivo studies. Methods Binding motifs were synthesized ‘on-resin’ or, when not practicable, in solution. Peptide chain elongations were performed according to optimized standard protocols via solid-phase peptide synthesis. In vitro experiments were performed using PSMA+ LNCaP cells. In vivo studies as well as μSPECT/CT scans were conducted with male LNCaP tumor xenograft-bearing CB17-SCID mice. Results PSMA ligands with A) modifications within the central Zn2+-binding unit, B) proinhibitor motifs and C) substituents & bioisosteres of the P1′-γ-carboxylic acid were synthesized and evaluated. Modifications within the central Zn2+-binding unit of PSMA-10 (Glu-urea-Glu) provided three compounds. Thereof, only natLu-carbamate I (natLu-3) exhibited high affinity (IC50 = 7.1 ± 0.7 nM), but low tumor uptake (5.31 ± 0.94% ID/g, 1 h p.i. and 1.20 ± 0.55% ID/g, 24 h p.i.). All proinhibitor motif-based ligands (three in total) exhibited low binding affinities (> 1 μM), no notable internalization and very low tumor uptake (< 0.50% ID/g). In addition, four compounds with P1′-ɣ-carboxylate substituents were developed and evaluated. Thereof, only tetrazole derivative natLu-11 revealed high affinity (IC50 = 16.4 ± 3.8 nM), but also this inhibitor showed low tumor uptake (3.40 ± 0.63% ID/g, 1 h p.i. and 0.68 ± 0.16% ID/g, 24 h p.i.). Salivary gland uptake in mice remained at an equally low level for all compounds (between 0.02 ± 0.00% ID/g and 0.09 ± 0.03% ID/g), wherefore apparent tumor-to-submandibular gland and tumor-to-parotid gland ratios for the modified peptides were distinctly lower (factor 8–45) than for [177Lu]Lu-PSMA-10 at 24 h p.i. Conclusions The investigated compounds could not compete with the in vivo characteristics of the EuE-based PSMA inhibitor [177Lu]Lu-PSMA-10. Although two derivatives (3 and 11) were found to exhibit high affinities towards LNCaP cells, tumor uptake at 24 h p.i. was considerably low, while uptake in salivary glands remained unaffected. Optimization of the established animal model should be envisaged to enable a clear identification of PSMA-targeting radioligands with improved tumor-to-salivary gland ratios in future studies. Supplementary Information The online version contains supplementary material available at 10.1186/s41181-021-00124-1.

In particular, radiolabeled small molecule PSMA inhibitors, based on the L-Glu-urea-X binding motif have been shown to exhibit noticeable uptake in kidneys (proximal tubules), lacrimal and salivary glands (Zechmann et al. 2014;Kratochwil et al. 2016b;Klein Nulent et al. 2018;Oh et al. 2019). Elevated activity accumulation in these organs can be partially assigned to endogenous PSMA expression, but also to a non-PSMArelated uptake mechanism, as PSMA expression was assumed to be present to a considerable lower extent than on the surface of PCa (Ghosh and Heston 2004;Rupp et al. 2019). Thus, beside kidneys, the lacrimal and salivary glands have been identified as critical organs (Tönnesmann et al. 2019;Kratochwil et al. 2017). In this context it has become apparent, that 177 Lu-labeled low molecular weight PSMA inhibitors can cause xerostomia Kratochwil et al. 2018), which can affect patients' quality of life and was found to be partially irreversible when α-emitter-labeled ( 225 Ac) compounds were administered (Rupp et al. 2019;Kratochwil et al. 2018;van Kalmthout et al. 2019). Hence, for the future optimization of PSMA-targeted radioligand therapies, the development of strategies to design improved small molecule-based PSMA ligands with reduced salivary gland uptake is of high priority.
This working hypothesis presupposes a specific uptake of the aforementioned PSMA inhibitors in salivary glands, although not necessarily PSMA-specific. Whereas PSMA protein and mRNA expression in salivary glands was confirmed by Western blot and genetic analyses (Troyer et al. 1995;Israeli et al. 1994;O'Keefe et al. 2004), the high accumulation of radiolabeled small molecule-based ligands (e.g. [ 177 Lu]Lu-PSMA-617 and [ 68 Ga]Ga-PSMA-11) did not correlate with the rather low PSMA expression density detected by immunohistochemistry and in patients, treated with the radiolabeled antibody [ 177 Lu]Lu-J591 (Rupp et al. 2019;Horoszewicz et al. 1987;Lopes et al. 1990;Bander et al. 2005). Indeed, unwanted non-target tissue uptake in salivary glands is markedly reduced using radiolabeled antibodies. Anyhow, therapeutic concepts using PSMA targeting antibodies are affected by slow diffusion into solid lesions and myelotoxicity due to longer blood circulation (Bander et al. 2005;Maurer et al. 2016;Tagawa et al. 2013).
Based on the fact that administration of monosodium glutamate in mice prior to [ 68 Ga]Ga-PSMA-11 markedly reduced activity uptake in salivary glands whilst maintaining high tumor uptake, non-PSMA specific interactions such as small molecule/ anion/glutamate transporter mechanisms may be conceivable (Rousseau et al. 2018). However, the detailed mechanism of uptake into salivary and lacrimal glands and the respective target structure has not been elucidated so far (Rupp et al. 2019;Rousseau et al. 2018).
This lack of knowledge led to various different approaches towards the reduction of radioactivity uptake into the salivary glands and thus corresponding unwanted side effects during radioligand therapy. After reduction of stimulus conduction by injection of botulinum toxin, a significant decrease of the SUVmean (up to 64%) in the right parotid gland compared to the left (control) was observed (Baum et al. 2018). In contrast, external cooling with ice packs to decrease the overall blood perfusion showed no effect on ligand uptake and xerostomia when 177 Lu-labeled peptides were used (van Kalmthout et al. 2018;Rathke et al. 2019;Yilmaz et al. 2019). Furthermore, excretion stimulus via vitamin C was investigated, but did not lead to any measurable uptake reduction of [ 68 Ga]Ga- PSMA-11 (Afshar-Oromieh et al. 2015). Sialendoscopy with dilatation, saline irrigation and steroid injection (prednisolone) after targeted α-therapy ([ 225 Ac]Ac-PSMA-617) showed beneficial effects on salivary gland function preservation and for patients' quality of life (Rathke et al. 2019). However, not all radiation induced damages on the parenchyma could be avoided, as macroscopic findings during sialendoscopy revealed endothelial avascularity with the presence of stenosis (Rathke et al. 2019). All in all, those efforts did not show the desired efficacy.
With the aim to investigate whether modifications at the urea-based inhibitor unit could reduce the non-PSMA-specific or perhaps 'transporter-mediated' uptake Rousseau et al. 2018) of PSMA-binding ligands into salivary glands, three structural alterations were investigated and evaluated ( Fig. 1): A) Modifications within the central Zn 2+ -binding unit B) Proinhibitor motifs C) PSMA-binding motifs with substituents & bioisosteres of the P1'-γ-carboxylic acid

Modifications within the central Zn 2+ -binding unit
The C-terminal glutamate represents a common feature of both phosphoramidateand urea-based PSMA ligands. However, the absorbed dose in salivary glands of PCa patients could be decreased by a factor of~4 via administering a phosphoramidate-based PET agent compared to the urea-based radiotracer [ 68 Ga]Ga-PSMA-I&T (Behr et al. 2019;Herrmann et al. 2015). Therefore, inhibitors containing modifications within the Zn 2+ -binding unit were synthesized, assuming that fragments of the Zn 2+ -binding group might serve as potential recognition sites for small molecule/anion/glutamate transporter systems or other transporterindependent uptake mechanisms.
Based on the results of Rousseau et al. (Rousseau et al. 2018), the C-terminal glutamate of PSMA ligands was presumed to cause for unwanted uptake into salivary glands. In currently used EuE-and KuE-based PSMA inhibitors, this structural feature is set as an essential part for proper ligand binding. However, at the same time it represents the fragment which is most similar to monosodium glutamate, by which activity accumulation could be markedly reduced in salivary glands in mice if administered 15 min prior to [ 68 Ga]Ga-PSMA-11. Therefore, masking of this ligand-inherent glutamate was tried to realize by proinhibitors (reversible masking) and PSMA ligands with substituents & bioisosteres of the P1'-γ-carboxylic acid (irreversible masking).

Proinhibitors
Prodrug approaches for targeted delivery of cytotoxic agents to PCa lesions have been recently published and were further evaluated in an open label dose-escalation trial (NCT01056029) (Denmeade et al. 2012;Ristau et al. 2014). Hence, this prodrug principle was considered as a reasonable option for reducing non-target tissue uptake of PSMA ligands. Thereby, PSMA as hydrolytical active enzyme could act as instrument for on-site liberation of the respective inhibitor and yet serve as the target for RLT of PCa. Non-target tissue accumulation would thus be reduced to a remarkably low level, as binding of the masked inhibitor is only possible to enzymatically active sites. Variations from the original PSMA substrate N-acetyl-L-aspartyl-L-glutamate (NAAG), were recently published by Barinka et al. and Plechanovová et al. (Barinka et al. 2002;Plechanovova et al. 2011) Thereof, Ac-L-Glu-L-Met and Ac-L-Asp-L-2-Aoc showed turnover numbers (k cat ) and cleavage efficiencies (k cat /K m ) closest to the respective values indicated for NAAG. For this reason, both glutamate surrogates, L-Met as well as L-2-Aoc were implemented into the existing structure of PSMA-10 (1). They were linked to the αor γ-carboxylate of the C-terminal glutamate, as the preferred cleavage site could not be clearly specified upfront. Hydrolysis might occur between α-linkages (NAALADase activity of GCP II in the central and peripheral nervous system) or between γ-linkages (FOLH1 activity in the gastrointestinal tract) (Barinka et al. 2012). In addition, prodrug approaches focusing just on γ-linkages were adversely affected by metabolic instability in human plasma (Mhaka et al. 2004).

PSMA-binding motifs with substituents & bioisosteres of the P1'-γ-carboxylic acid
Modifications at the P1'-γ-carboxylic acid moiety were preferred, as previous studies conducted by Kozikowski et al. revealed substitutions at the P1' glutamate to be more tolerated at the γ-carboxylate than at the α-carboxylate (Kozikowski et al. 2004). Moreover, the P1'-γ-carboxylic acid possibly acts as the relevant recognition site for small molecule/anion/glutamate transporter systems (Rousseau et al. 2018). Besides Kozikowski et al., Wang et al. and Plechanovová et al. investigated the effect of carboxylic acid bioisosteres and aliphatic substituents on the affinity of urea-based inhibitors towards PSMA (Plechanovova et al. 2011;Wang et al. 2010). On the basis of these studies, incorporation of the most auspicious carboxylic acid substituents was pursued. All these modifications were investigated based on one of our recently developed PSMA ligands, PSMA-10 (1) (Wurzer et al. 2020). In particular, a clear distinction between the salivary gland uptake values of EuE-and non-EuE-based radioligands and higher tumorto-salivary gland ratios at 24 h p.i. in comparison to EuE-based PSMA-10 (1) were of primary interest. The latter might indicate for a possible transferability to humans, despite certain species-dependent differences, i.e. a general lower uptake in mouse salivary glands (Knedlik et al. 2017;Roy et al. 2020).

Methods
For detailed information on all methods for synthesis and analysis as well as on the used instruments, see the 'GENERAL INFORMATION', 'MATERIALS' and 'METHODS' section in the supporting information (supplemental materials are available on https://ejnmmipharmchem.springeropen.com). Radiolabeling 177

Lu-labeling
Previously published procedures were applied with minor modifications for 177 Lu-labeling of DOTA-conjugated peptides Sosabowski and Mather 2006 were added and the reaction mixture was heated for 25 min at 70 to 95°C. The radiochemical purity (RCP) was determined using radio-RP-HPLC and radio-TLC. Occasionally, purification by cartridge (HLB Plus Light, 30 mg) and radio-RP-HPLC was necessary. For detailed procedures for each peptide, see the 'METHODS' section in the supporting information.

Lipophilicity
The log D values were determined, using the shake-flask method as previously described (Wurzer et al. 2020). The radiolabeled tracer (0.5-1 MBq, 20 μL) was dissolved in 1 mL of a 1/1 mixture (v/v) of PBS (pH 7.4) and n-octanol in a reaction vial (n = 6). After vigorous mixing of the suspension for 3 min at r.t., the vial was centrifuged at 9000 rpm (ca. 7700×g) for 5 min at room temperature (Heraeus Biofuge 15, Thermo Fisher Scientific, Darmstadt, Germany) and 150 μL aliquots of both layers were measured in a ɣ-counter.

I-labeling
The radioiodinated reference ligand ([ 125 I]I-BA)KuE was synthesized in analogy to a previously described method ).

Affinity determinations (IC 50 ) and internalization studies
Detailed information on affinity and internalization experiments is provided in the supporting information. In brief, competitive binding studies were determined on LNCaP cells (1.50 × 10 5 cells in 1 mL/well) after incubation at 4°C for one hour, using ([ 125 I]I-BA)KuE (0.20 nM/well) as the reference radioligand (n = 3). IC 50 values of ( nat I-BA)KuE were determined on the same day for monitoring assay performance and to identify abnormal deviations in uptake and/or affinity. Only values for which the affinity of ( nat I-BA)KuE was within a range of 3.95 ± 1.35 nM (n = 6) were considered. Internalization studies of the radiolabeled ligands (1.0 nM/well) were performed on LNCaP cells (1.25 × 10 5 cells in 1 mL/well) at 37°C for one hour and accompanied by ([ 125 I]I-BA)KuE (0.20 nM/well), as reference ligand. Data were corrected for non-specific binding and normalized to the specific internalization observed for the radioiodinated reference compound (n = 3).

In vivo experiments
All animal experiments were conducted in accordance with general animal welfare regulations in Germany (German animal protection act, as amended on 18.05.2018, Art. 141 G v. 29.3.2017 I 626, approval no. 55.2-1-54-2532-71-13) and the institutional guidelines for the care and use of animals. To establish tumor xenografts, LNCaP cells (approx. 10 7 cells) were suspended in 200 μL of a 1/1 mixture (v/v) of DMEM F-12 and Matrigel (BD Biosciences, Heidelberg, Germany) and inoculated subcutaneously onto the right shoulder of 6-8 weeks old CB17-SCID mice (Charles River Laboratories, Sulzfeld, Germany). Mice were used for experiments when tumor size reached 5-10 mm in diameter (3-6 weeks after inoculation).

Metabolite analysis
Detailed information on organ preparation and extraction procedures is provided in the supporting information. In brief, [ 177 Lu]Lu-11 (9.64 MBq) was injected into the tail vein of a LNCaP tumor xenograft-bearing CB17-SCID mouse. The animal was sacrificed 1 h p.i. and subjected to the standard procedure for biodistribution studies. In addition, urine was taken from all mice that were investigated in this experiment (8.4-9.0 MBq injected activity) and pooled (n = 5). Liver, tumor and kidneys were homogenized with a MM-400 ball mill (Retsch GmbH, Haan, Germany) at 30 Hz for 20 min and extracted with 1 mL radioimmunoprecipitation assay (RIPA) buffer containing 2 μmol of 2-(phosphonomethyl)pentane-1,5-dioic acid (2-PMPA). The suspension was centrifuged (15,200 rpm, 10 min, 21°C) and the supernatants were loaded onto a preconditioned Strata-X cartridge (200 mg). Elution of the activity was conducted by 750 μL of MeCN/H 2 O (6/4, 0.1% TFA) and the extracts were analyzed via radio-RP-HPLC. Prior to solid phase extraction, blood samples were centrifuged twice (13,000 rpm, 5 min) to separate the plasma from the blood cells. Urine samples were also centrifuged (13,000 rpm, 5 min) and the supernatant was directly analyzed via radio-RP-HPLC.
μSPECT/CT imaging Imaging experiments were conducted using a MILabs VECTor 4 small-animal SPECT/ PET/OI/CT. The resulting data were analyzed by the associated PMOD (version 4.0) software. Mice were anaesthetized with isoflurane and the 177 Lu-labeled PSMA compounds were injected via the tail vein. Mice were euthanized 1 h or 24 h p.i. by CO 2 asphyxiation and cervical dislocation and blood samples for later biodistribution studies or metabolite analysis were taken by cardiac puncture before image acquisition. Static images were acquired with 45 min acquisition time using the HE-GP-RM collimator and a step-wise multi-planar bed movement. All images were reconstructed using the MILabs-Rec software (version 10.02) and a pixel-based Similarity-Regulated Ordered Subsets Expectation Maximization (SROSEM) algorithm with a window-based scatter correction (20% below and 20% above the photopeak, respectively). Voxel size CT: 80 μm, voxel size SPECT: 0.8 mm, 1.6 mm (FWHM) Gaussian blurring post processing filter, with calibration factor in kBq/mL and decay correction, no attenuation correction.

Synthesis
PSMA ligands containing modified binding motifs (Fig. 2) were synthesized according to known or modified organic chemical synthesis procedures. On-resin synthesis of binding motifs was established and adjusted in individual cases (Schemes 1, 2 and 3). Peptide chain elongation was performed according to standard solid-phase peptide synthesis protocols for PSMA derivatives and optimizations concerning (radio)metal complexation reactions were performed if necessary (Tables 1 and 2 in the supporting information). The following sections cover the syntheses of compounds 2 to 11 (Fig. 2), highlighting special synthetical aspects, improvements to already described procedures as well as methods for preservation of the mandatory L-configuration of the PSMAbinding motif during inhibitor modification.
Synthesis of PSMA derivatives containing modifications within the central Zn 2+ -binding unit: thioureate 2, carbamate I (3) and carbamate II (4) In order to accomplish substitution of the urea by a thiourea moiety, 1,1'-thiocarbonyldiimidazole was used instead of 1,1'carbonyldiimidazole (CDI) for the synthesis of Scheme 1 General, simplified synthetic routes for the preparation of thioureate 2, carbamate I (3) and carbamate II (4). Synthesis of the binding motif of thioureate 2 was conducted by a solid phase procedure, whereas binding motifs 15 (carbamate I) and 17 (carbamate II) were obtained by solution phase synthesis prior to coupling to compound 18. Detailed synthesis procedures are given in the supporting information . Synthesis of the binding motif of proinhibitor I (5) was conducted by a solid phase procedure, whereas binding motif 21 was obtained by a mixed solid/solution phase synthesis prior to coupling to compound 18. Compound 22 (proinhibitor III) could only be obtained by solution phase synthesis. Detailed synthesis procedures are given in the supporting information Scheme 3 General, simplified synthetic routes for the preparation of L-2-aha (8), furyl (9), alkyne (10) and tetrazole (11) derivatives. Syntheses of the binding motifs of compounds 8 (L-2-aha), 9 (furyl) and 10 (alkyne) were conducted by a solid phase procedure, whereas binding motif 26 (tetrazole) was obtained by solution phase synthesis prior to coupling to compound 18. Detailed synthesis procedures are given in the supporting information  (13). This solid phase approach was successfully implemented for the Glu-thiourea-Glu moiety and transferred to further synthesis procedures of binding motifs in this study. Purified thioureate 2 was directly complexed with nat Ga 3+ for in vitro studies. In contrast to thioureate 2, syntheses of the PSMA-binding moieties of carbamate I (15) and carbamate II (17) were performed completely by a solution phase strategy. All synthesis steps towards compound 15 and 17, starting from enantiopure (S)-5-oxotetrahydrofuran-2-carboxylic acid, focused on conditions to ensure and maintain Lconfiguration of the entire PSMA-binding motif (Zhang et al. 2009;Shin et al. 2000;Weineisen et al. 2014;Yang et al. 2016). By introduction of different protective groups (tert-butyl or benzyl) at the γ-carboxylate of compounds 14 and 16, accommodation of the carbamate oxygen either within the S1' or S1 pocket of PSMA is already defined. Carbamate I (3) was complexed with nat Ga 3+ , nat Lu 3+ and [ 177 Lu]Lu 3+ for in vitro and in vivo studies, whereas carbamate II (4) was complexed with nat Ga 3+ and evaluated solely in in vitro studies.

Synthesis of proinhibitors I, II & III (5, 6 & 7)
Dipeptide H-L-Glu(L-Met-OtBu)-OtBu (19) was generated in solution and further coupling to resin-bound compound 20 provided the tris-tBu-protected proinhibitor I binding motif, which afforded product 5 after on-resin elongation. Functionalization of resin-bound dipeptide 13 (H-L-Glu[D-Orn(Dde)-2-CT]-OtBu) to yield its carbonylimidazole derivative 20 was required, as functionalization of dipeptide 19 with CDI was not successful. Dimerization as a competing reaction was more favored and hence, only the urea-conjugated dimer of 19 could be isolated after this attempt (unpublished observations). Notably, carbonylimidazole functionalization of resin-bound peptide 13 worked properly in this approach, as a certain distance to the resin anchor (2-chlorotrityl group) was preserved. In contrast, this functionalization did not work at the first 2-CT resin-bound amino acid, provided that it was coupled to the resin by its αcarboxylate (unpublished observations). Proinhibitor I (5) was complexed with nat Lu 3+ and [ 177 Lu]Lu 3+ for in vitro and in vivo studies.
In analogy to proinhibitor I (5), on-resin synthesis of the binding motif was also pursued for proinhibitor III (7). However, in contrast to the procedure for proinhibitor I, the starting dipeptide for proinhibitor III (H-L-Glu(OtBu)-L-2-Aoc-OtBu) did not react and only on-resin urea-conjugated dimerization of peptide 20 was detected (unpublished observations). Therefore, the strategy was changed to solution phase synthesis to obtain tris-tBu-protected binding motif 22. Coupling to 18 and subsequent on-resin elongation afforded proinhibitor III (7), which was complexed with nat Lu 3+ and [ 177 Lu]Lu 3+ for in vitro and in vivo studies.
Synthesis of PSMA-binding motifs with substituents & bioisosteres of the P1'-γ-carboxylic acid: L-2-aminoheptanoic acid (2-aha) (8), furyl (9), alkyne (10) and tetrazole (11) derivatives Reactants 23 and 24 were generated in solution, using the respective amino acid and tert-butyl acetate, according to a previously published procedure by Hyun et al. with some minor modifications (Hyun et al. 2010). This strategy was used to assure esterification of the carboxylic acid moiety only, with no unwanted alkylation/carbamation at the free primary amine. O-tert-butyl-N,N′-diisopropylisourea could not be used in this case, as its usage would have led to simultaneous alkylation of the free primary amine (Mathias 1979). Coupling of 23/24 with resin-bound 20 provided an alkane-/furyl-functionalized binding motif, which then afforded product 8/9 after on-resin elongation. Both, L-2-aminoheptanoic acid derivative 8 as well as was furyl derivative 9 were complexed with nat Lu 3+ for in vitro studies.
Reactant 25 was commercially available and used directly for on-resin coupling with compound 20. Subsequent peptide chain elongation afforded alkyne derivative 10, which was complexed with nat Lu 3+ and [ 177 Lu]Lu 3+ for in vitro and in vivo studies.
Synthesis of the tetrazole bioisostere was successfully accomplished following a multiple step procedure described by Kozikowski et al. (Kozikowski et al. 2004) Binding motif 26 was completed by a solution phase strategy. All synthesis steps, starting from enantiopure N-Cbz-L-Glu-OtBu, focused on conditions to ensure and maintain Lconfiguration of the entire PSMA-binding motif (Kozikowski et al. 2004;Weineisen et al. 2014). After coupling to resin-bound 18 and further peptide chain elongation, tetrazole derivative 11 was obtained and complexed with nat Lu 3+ and [ 177 Lu]Lu 3+ for in vitro and in vivo studies.

Cold metal complexation and radiolabeling
Cold metal complexation with molar excess of Ga(NO 3 ) 3 *6 H 2 O (3.50-fold molar excess) or LuCl 3 (6-fold molar excess) led to formation of the respective nat Ga/ nat Lu-PSMA ligands. Purification of the crude nat Ga-ligand was performed by RP-HPLC. nat Lu-complexation mixtures (0.5-1.0 mM in DMSO/H 2 O = 1/1) were directly used as stock solutions for affinity determination, as chemical purity was always > 92%. Reaction conditions, chemical purities and yields of the investigated nat Ga-and nat Lu-PSMA ligands are given in the supporting information (Table 1).
Radiolabeling ([ 177 Lu]Lu 3+ ) was performed using manual procedures and radiochemical purity (RCP), as determined by radio-RP-HPLC and radio-thin-layer chromatography was 96.4 ± 2.2% (n = 16) with exception of proinhibitor I (5) (89.3 ± 1.9% RCP, n = 4). In case of incomplete complexation, removal of free [ 177 Lu]Lu 3+ via HLB cartridge (30 mg) was performed. Moreover, in case of proinhibitor I, radioactive by-products were tried to remove by preparative radio-RP-HPLC. Reaction conditions, isolated radiochemical yields (RCY), apparent molar activities (A m ) at the end of synthesis and purification, the range of used activities as well as radiochemical purities (RCP) are given in the supporting information (Table 2). Detailed synthesis procedures for cold metal complexation and radiolabeling are described in the 'METHODS' section in the supporting information. In vitro characterization

PSMA affinity
Binding to PSMA was determined using LNCaP human prostate cancer cells in a competitive binding assay with nat Ga or nat Lu complexes of compounds 2 to 11. For comparison, IC 50 values of nat Lu-PSMA-10 ( nat Lu-1) (2.8 ± 0.5 nM) were also determined.
The results show loss of affinity to a varying extent for all modifications differing from glutamate at P1' position (Table 1). Only nat Lu-3 (carbamate I) and nat Lu-11 (tetrazole derivative) still exhibited high affinity (7.1 ± 0.7 nM and 16.4 ± 3.8 nM, respectively) towards PSMA-expressing LNCaP cells.

Internalization
LNCaP cells were used to investigate internalization of the 177 Lu-labeled compounds 1, 3, 5, 6, 7, 10 and 11. No internalization studies were performed on thioureate 2, carbamate II (4), L-2-aminoheptanoic acid derivative 8 and furyl derivative 9, as these candidates were excluded due to poor IC 50 data. Normalized to the uptake of ([ 125 I]I-BA)KuE, the results in Table 1  at the P1' position, only weak internalization was detected (1.2 ± 0.4% and 9.9 ± 3.2% compared to the reference, respectively). For all proinhibitor motifs internalization studies were conducted prior to any affinity determination, in order to investigate possible substrate cleavage kinetics. Internalization at 37°C was determined at several time points of 0.5 h, 1 h, 2 h and 4 h. All in all, no internalization at any time point could be detected.

Metabolite analysis
Besides biodistribution and μSPECT/CT studies, [ 177 Lu]Lu-11 (tetrazole) was analyzed according to its metabolic stability 1 h p.i. Relevant dissected organs (tumor, kidney, liver) and body fluids (blood and urine) were collected, homogenized if necessary and subjected to extraction of the incorporated activity. Tissue extracts were analyzed by radio-RP-HPLC at a predefined gradient (25-40% MeCN (0.1% TFA) in 20 min), for which the retention time (18.3 min) of the intact cold standard ( nat Lu-11) was previously determined and hence, served as a reference. Only one metabolite with higher hydrophilicity was detected in tumor (7.1%), blood (8.5%) and kidney (11.7%), with the maximum percentage of metabolite amount found in the urine (18.7%). This metabolite was not detectable in the liver homogenate, instead a more lipophilic metabolite was observed at t R = 20.5 min (28.0%). The respective radio-RP-HPLC analyses of extracts from homogenized organs and body fluids are depicted in Figure 1 in the supporting information.

Discussion
With the aim to develop radiolabeled small molecule-based PSMA inhibitors with reduced non-target tissue uptake in patients, especially in salivary and lacrimal glands, several modified PSMA-10 ligands were prepared and further evaluated in vitro and in vivo. Comparison of salivary gland uptake values at 24 h p.i. should give a first indication whether a clear distinction of modified PSMA (pro)inhibitors from classical L-Glu-urea-X binding motifs could be provided by this animal model, established in our group for preclinical evaluation of PCa radiotherapeutics.

Synthesis
For all peptides, the starting material as well as further synthetical procedures were adjusted in order to preserve L-configuration of the inhibitor motif, since high affinity binding to PSMA basically depends on steric orientation (Barinka et al. 2019;Ferraris et al. 2012). The introduction of different protective groups (tert-butyl or benzyl) at the γ-carboxylate of the carbamate-oxygen-bearing moiety was of utmost importance for both carbamates I & II (3 & 4). These steps, already conducted at an early stage of synthesis, were crucial for determination of carbamate orientation. On-resin synthesis of the binding motif was first established for thioureate 2 and simplified the reaction process significantly. In contrast to L-Glu-urea-L-Lys binding motifs (Mosayebnia et al. 2018), PSMA inhibitor motifs based on L-Glu-urea-L-Glu were synthesized exclusively in solution in previously published attempts (Wurzer et al. 2020;  Table 3 in the supporting information Felber et al. EJNMMI Radiopharmacy andChemistry (2021) 6:10 Robu et al. 2018;Lu et al. 2013;Kularatne et al. 2009). Related time-and substanceconsuming steps (purification, removal of an orthogonal protective group at P1 position, etc.) could be avoided by this on-resin method. Hence, this strategy was also used for the synthesis of subsequent derivatives (proinhibitor I (5), L-2-aha derivative 8, furyl derivative 9 and alkyne derivative 10). Moreover, it might serve as a general procedure for the preparation of L-Glu-urea-L-Glu-based PSMA inhibitors. Syntheses of the binding motifs of proinhibitors 5-7 had to be adjusted due to unreactive starting material or unwanted dimerization reactions. In addition, recurring methionine oxidation resulted in low overall yields (0.08%) for proinhibitor I (5) after RP-HPLC purification.

Cold metal complexation and radiolabeling
In general, all derivatives containing modifications within the Zn 2+ -binding unit showed enhanced formation of intramolecular condensation by-products during complexation (unpublished observation). Therefore, all nat/177 Lu-complexation reactions of 3 were performed at 70°C, as formation of unwanted by-products could be almost completely reduced. Oxidation tendency for proinhibitor I (5) remained at 177 Lu-labeling and also during radio-RP-HPLC purification, which impeded the preparation of products with RCP higher than 90%, even at lower temperatures (70-80°C).

In vitro & in vivo characterization
Though IC 50 and lipophilicity data of carbamate I (3) were comparable to [ 177 Lu]Lu-PSMA-10 ([ 177 Lu]Lu-1), internalization was distinctly lower (67.8 ± 0.5% for [ 177 Lu]Lu-3 vs. 177 ± 15% for [ 177 Lu]Lu-1), which might explain decreased tumor accumulation at 1 h and 24 h p.i. However, low internalization may not be the only reason for decreased tumor uptake. As observed for SST 2 antagonists, high tumor uptake can also be reached with a negligible capacity to internalize (Dude et al. 2017). A two-fold lower tumor accumulation compared to [ 177 Lu]Lu-1 already 1 h p.i. (5.31 ± 0.94% ID/g) in combination with a rapid decline to 1.20 ± 0.55% ID/g at 24 h p.i., led to the assumption that in vivo decomposition of the inhibitor motif might have generated a non-PSMAbinding ligand, resulting in fast renal excretion (0.31 ± 0.05% ID/g for [ 177 Lu]Lu-3 vs. 1.97 ± 0.78% ID/g for [ 177 Lu]Lu-1, 24 h p.i.). Applications of carbamate-based prodrugs, liberating the biologically active substance by in vivo hydrolysis, support this theory (Ghosh and Brindisi 2015).
Similar in vitro results as obtained for carbamate I and II were reported by Yang et al. and Barinka et al. (Yang et al. 2016;Barinka et al. 2019) These observations emphasize the necessity of a hydrogen bond donor at the (non-pharmacophore) P1 position and provide a certain flexibility within the pharmacophore S1' subpocket.
Since thioureate derivative 2 revealed sulfur to be less tolerated inside the binding pocket, it was assumed that thiourethane derivatives (= combination of carbamate I or II with thioureate) would also lead to poor results. In consequence, their synthesis was not further pursued.
For all proinhibitors, internalization studies were conducted first in order to investigate possible substrate cleavage kinetics. Since no internalization could be detected at any time point (0.5 h, 1 h, 2 h and 4 h) for [ 177 Lu]Lu-5, -6 or -7 (Table 1), it was assumed that no cleavage occurred under these conditions. As we suggested that cleavage of the proinhibitor motifs might be strongly dependent on the tumor cells' microenvironment, in vivo studies were directly conducted after internalization experiments. With a maximum tumor accumulation of 0.33 ± 0.11% ID/g for [ 177 Lu]Lu-6 (proinhibitor II) and a minimum tumor accumulation of 0.09 ± 0.02% ID/g for [ 177 Lu]Lu-5 (proinhibitor I), all investigated proinhibitors showed very low ability to bind to PSMAexpressing tumors (24 h p.i.), as depicted in Fig. 3. Furthermore, non-target tissue uptake was on the scale of [ 177 Lu]Lu-PSMA-10 ([ 177 Lu]Lu-1), wherefore no tumor-to-tissue ratios were calculated. It was assumed that proinhibitor cleavage probably did not occur in in vitro and in vivo experiments, due to the low (micromolar) affinities of these conjugates determined by additional competitive binding experiments (Table 1). For this reason, synthesis and evaluation of proinhibitor IV (= methionine at the α-carboxylate) was abandoned, as no positive results were expected.
As presumed for the tetrazole moiety (Herr 2002), in vitro studies confirmed a slightly increased lipophilicity for [ 177 Lu]Lu-11 (~7.4-fold increase compared to [ 177 Lu]Lu-1). A noticeable decreased internalization of [ 177 Lu]Lu-11 (9.9 ± 3.2%) with concomitant high affinity (16.4 ± 3.8 nM) did not lead to favorable in vivo results. As the metabolite proportion was rather low in tumor tissue (7.1%) and circulating blood (8.5%), low retention of [ 177 Lu]Lu-11 within the LNCaP tumor xenograft at 1 h p.i. cannot be attributed to severe metabolic instability. Therefore, poor in vivo performance at 1 h p.i. (3.40 ± 0.63% ID/g) as well as at 24 h p.i. (0.68 ± 0.16%ID/g,) was mainly assigned to the overall lower internalization in combination with the decreased hydrophilicity and affinity and of the final ligand.
Low accumulation of alkyne derivative [ 177 Lu]Lu-10 in tumor tissue (0.10 ± 0.03% ID/g, 24 h p.i.) as well as weak internalization (1.2 ± 0.4% compared to the reference) could be attributed to the medium affinity of nat .
Apparently, tumor-to-submandibular and tumor-to-parotid gland values of [ 177 Lu]Lu-3 both decreased by a factor of 8 when compared to [ 177 Lu]Lu-1. An even higher decrease (16 to 20 times lower) was observed for [ 177 Lu]Lu-11 (tetrazole) and the alkyne analog [ 177 Lu]Lu-10, which exhibited the lowest tumor-tosalivary gland ratios (39-45 times lower than for [ 177 Lu]Lu-1) (Fig. 5). However, reduced tumor-to salivary gland values at 24 h p.i. in comparison to reference compound [ 177 Lu]Lu-1 were found to be mainly induced by the decreased tumor accumulation rather than by an altered salivary gland uptake (Fig. 4 in this manuscript and Table 3 in the supporting information). Salivary gland uptake values of EuEand non-EuE-based ligands ranged between 0.02 ± 0.00% ID/g and 0.09 ± 0.03% ID/ g and hence, revealed no significant difference. At least for EuE-based reference [ 177 Lu]Lu-1 a higher activity accumulation was expected, which in turn accounts for a certain unsuitability of this animal model concerning the development of ligands for RLT with reduced salivary gland uptake (discussed in more detail in 'Critical assessment of the animal model').
High IC 50 values led to cessation of further in vitro and in vivo investigations for thioureate 2, carbamate II (4), L-2-aha derivative 8 and furyl derivative 9. Theoretically, a multitude of other substituents as P1'-γ-carboxylic acid bioisosteres might be conceivable (Lassalas et al. 2016). IC 50 data of the respective PSMA ligands can give a first indication whether further in vitro and in vivo investigations are Felber et al. EJNMMI Radiopharmacy and Chemistry (2021)  Critical assessment of the animal model Activity uptake in the salivary glands of mice was found to be constantly at a very low level (between 0.02 ± 0.00% ID/g and 0.09 ± 0.03% ID/g) irrespective of the use of EuE-(1) or non-EuE-based (3, 5, 6, 7, 10, 11) PSMA ligands ( Fig. 4 in this manuscript and Table 3 in the supporting information). The low overall tracer uptake in salivary glands at 24 h p.i. might indicate for a rather unspecific accumulation and that speciesdependent differences possibly not allow for a similar radioligand uptake as observed in humans (Knedlik et al. 2017;Roy et al. 2020 Critical assessment of the rationale for inhibitor modification PSMA ligands containing modifications within the inhibitor motif (like in compounds 2 to 11) might be overestimated if considered as the sole strategy for reduction of nontarget tissue uptake in PCa patients. Although small molecule/anion/glutamate transporter systems were proposed for non-target tissue accumulation (Rousseau et al. 2018), the exact mechanism of how monosodium glutamate affects non-target binding of radiolabeled PSMA inhibitors remains still unknown. This effect could also originate from differences in physiological pH of healthy salivary gland parenchyma and poorly differentiated tumor tissue and/or tumor-associated stroma (Warburg effect) (Vāvere et al. 2009). Previous Western blot analyses have shown a variation of the PSMA protein within the salivary glands by 20 kDa (in total: 120 kDa; PSMA of LNCaP cell extracts: 100 kDa) (Troyer et al. 1995). Additionally, genetic analyses revealed mRNA expression of PSMA in salivary glands (Israeli et al. 1994;O'Keefe et al. 2004). Posttranslational modified versions of PSMA (e.g. different glycosylation pattern, otherwise than on PCa) on the surface of salivary gland cells might serve as a reasonable concept for integrating all these contradictory findings (Troyer et al. 1995;Barinka et al. 2012).
Binding of antibodies that are highly specific towards epitopes of PCa-related PSMA (intra-or extracellular) would not be possible in this case (Rupp et al. 2019;Horoszewicz et al. 1987;Lopes et al. 1990;Bander et al. 2005). By contrast, small molecule ligands would still be able to bind, as their binding mechanism is probably not affected by posttranslational changes at the surface of the protein. Modifications within the ligand structure would hence result in no benefit. Furthermore, inconsistent in vitro results concerning the binding of PSMA-specific mAbs to salivary gland tissue are still present (Rupp et al. 2019;Wolf et al. 2010). This indicates that not enough fundamental research was performed yet to provide a clear statement of PSMA expression on salivary gland parenchyma. As long as there is no reliable information concerning this issue, further investigations on ligand modifications might be dispensable.

Conclusions
No modified inhibitor structure was found to be able to compete with favorable in vivo characteristics of EuE-based PSMA inhibitor [ 177 Lu]Lu-PSMA-10. In consequence, the development of further strategies to minimize radioligand uptake in salivary glands should be envisaged.
Additional file 1. Supporting Information is provided in addition to data presented in the main manuscript portion, including detailed information on all methods for synthesis and analysis as well as on the used instruments. Furthermore, detailed procedures for ligand synthesis, cold metal complexation and radiolabeling are described. Methods for in vitro and in vivo characterizations are given in more detail, as well as radio-RP-HPLC analyses and extraction efficiencies of the metabolite analysis.