Small peptide-based GLP-1R ligands: an approach to reduce the kidney uptake of radiolabeled GLP-1R-targeting agents?

Aim Elevated kidney uptake in insulinoma patients remains a major limitation of radiometallated exendin-derived ligands of the glucagon-like peptide 1 receptor (GLP-1R). Based on the previously published potent GLP-1R-activating undecapeptide 1, short-chained GLP-1R ligands were developed to investigate whether kidney uptake can be reduced by means of direct 18F-labeling (nuclide-based accelerated renal excretion) or the reduction of the overall ligand charge (ligand-based reduced kidney uptake). Materials & methods GLP-1R ligands were prepared according to optimized standard protocols via solid-phase peptide synthesis (SPPS) or, when not practicable, via fragment coupling in solution. Synthesis of (2‘-Et, 4‘-OMe)4, 4’-L-biphenylalanine ((2′-Et, 4′-OMe)BIP), required for the preparation of 1, was accomplished by Suzuki-Miyaura cross-coupling. In vitro experiments were performed using stably transfected GLP-1R+ HEK293-hGLP-1R cells. Results In contrast to the three reference ligands glucagon-like peptide 1 (GLP-1, IC50 = 23.2 ± 12.2 nM), [Nle14, Tyr(3-I)40]exendin-4 (IC50 = 7.63 ± 2.78 nM) and [Nle14, Tyr40]exendin-4 (IC50 = 9.87 ± 1.82 nM), the investigated GLP-1R-targeting small peptides (9–15 amino acids), including lead peptide 1, exhibited only medium to low affinities (IC50 > 189 nM). Only SiFA-tagged undecapeptide 5 (IC50 = 189 ± 35 nM) revealed a higher affinity than 1 (IC50 = 669 ± 242 nM). Conclusion The investigated small peptides, including lead peptide 1, could not compete with favorable in vitro characteristics of glucagon-like peptide 1 (GLP-1), [Nle14, Tyr(3-I)40]exendin-4 and [Nle14, Tyr40]exendin-4. The auspicious EC50 values of 1 provided by the literature could not be transferred to competitive binding experiments. Therefore, the use of 1 as a basic scaffold for the design of further GLP-1R-targeting radioligands cannot be recommended. Further investigations might include the scaffold of 5, although substantial optimizations concerning affinity and lipophilicity would be required. In sum, GLP-1R-targeting radioligands with reduced kidney uptake could not be obtained in this work, which emphasizes the need for further ligands addressing this particular issue. Supplementary Information The online version contains supplementary material available at 10.1186/s41181-021-00136-x.

A conventional method used for lowering the kidney uptake of radiotracers makes use of pre-administration of amino acids (Rolleman et al., 2003). Gotthardt et al. applied this method for [ 111 In]In-DTPA-exendin-4 in rats and revealed that both gelofusine (18.7% decrease) and poly-L-glutamic acid (29.4% decrease) as well as the combination of both (47.9% decrease) had a significant impact on kidney uptake, whereas the administration of L-lysine did not show any effect (Gotthardt et al., 2007). Besides, coinfusion of albumin-derived peptide fragments reduced the kidney uptake of [ 111 In]In-DTPA-exendin-3 by 26% in rats (Vegt et al., 2010).
In order to circumvent any co−/pre-administration steps, modifications of the exendin-4 scaffold itself were pursued to improve the tumor-to-kidney ratio. As a possible strategy, radiotracers with kidney-cleavable linkers have been introduced (Yim et al., 2013;Jodal et al., 2015), and resulted in a 50% reduction of radiation dose in mice, although a general proof-of-concept in humans is still pending (Zhang et al., 2019). For diagnostic purposes, studies with 18 F-labeled exendin-4 revealed high tracer uptake in INS-1 tumor cells and xenograft models and a rapid clearance from the kidneys compared to radiometallated/[ 18 F]AlF analogs (Kiesewetter et al., 2012a;Mikkola et al., 2016;Gao et al., 2011;Wu et al., 2013). Although these results suggested high potential of 18 F-labeled exendin-4 derivatives, none of the investigated compounds were transferred into clinical application (Jansen et al., 2019;https://clinicaltrials.gov/ct2/ results?cond=Insulinoma&term=glp-1R&cntry=&state=&city=&dist=&Search=Search, 2020). Occasionally, difficulties in separating the precursor from the respective radiolabeled agent resulted in low molar activities (A m ) (Wu et al., 2013;Yue et al., 2014). As a consequence, optimal tumor uptake could not be attained, since a low peptide dose (~1 μg) and hence, a high molar activity (~200 GBq/μmol (Boerman & Gotthardt, Felber and Wester EJNMMI Radiopharmacy and Chemistry (2021) 6:29 Page 2 of 16 2012)) is mandatory. Moreover, dependent on the labeling methodology, a relatively high nonspecific accumulation in liver and intestines occurred in preclinical studies (Kiesewetter et al., 2012a;Gao et al., 2011). Direct 18 F-labeling via a SiFA-modified precursor provided only low 18 F-incorporation (6%), low radiochemical yields (RCY, 1.0-1.5%) and a slow blood clearance (3.3 ± 0.8%ID/g, 2 h p.i.) (Dialer et al., 2018). Moreover, PRRT of GLP-1R-overexpressing tumors has not been envisaged yet. Potential nephrotoxicity disallows the use of radiometallated αor β − -emitting agents and for 131 I-labeled derivatives, an ingenious treatment protocol with Irenat® would be indispensable to avoid severe 131 I-induced damage of the thyroid (Lappchen et al., 2017). In consequence, the use of the exendin-3/4 scaffold currently allows for a very limited number of peptide-radionuclide conjugates that can be only applied for insulinoma imaging in patients.
Since a low amount of GLP-1R expression was detected in the kidneys and high activity accumulation by this organ could not be specifically blocked by an excess of unlabeled analog, a non-saturable, GLP-1R independent mechanism was presumed for tubular reabsorption (Gotthardt et al., 2006;Brom et al., 2010;Brom et al., 2012). Accordingly, further studies revealed that the megalin transporter system of the renal proximal tubules is crucial for uptake and retention of 111 In-labeled exendin-4 and potential metabolites (Gotthardt et al., 2007;Jodal et al., 2015;Vegt et al., 2011). Kidney extracts of mice that received [ 18 F]AlF-NOTA-exendin-4 showed a single very polar radioactive metabolite at 1 h p.i. and no detectable parent peptide. The identity of this metabolite could not be determined, but the general observation that exendin-4 peptides with C-terminal radiometal chelates exhibit very high uptake suggests that some metabolite from the C-terminal end might be responsible for the slow egress from the kidneys (Kiesewetter et al., 2012b). By comparing the renal accumulation of the 111 Inlabeled G protein-coupled receptor (GPCR) ligands octreotide, minigastrin, bombesin and exendin-4, Gotthardt et al. observed that the number of charged amino acids in these peptides correlates with their kidney uptake. Therefore, beside radiometal chelate-induced renal retention, the high number of charged amino acids of exendin-4based radioligands is supposed to play a critical role for tubular reabsorbtion, although the exact mechanism still remains unknown (Gotthardt et al., 2007).
Based on the aforementioned study, we hypothesized that downsized GLP-1Rbinding structures might have the advantage to exhibit a reduced kidney uptake and might provide accelerated urinary excretion, due to their inherent reduced number of charged amino acids. However, at present, GLP-1R-targeting radioligands, that structurally clearly differ from exendin-4 are not available. Although low-molecular-weight organic molecules were developed for addressing the GLP-1R (orally available antidiabetics) (Donnelly, 2012;Graaf et al., 2016), none of them was able to show favorable properties comparable or even superior to exendin-4/GLP-1-based polypeptide ligands (Graaf et al., 2016;Willard et al., 2012;Knudsen et al., 2007). As a result, these structures were not considered for further use as GLP-1R-targeted radiopharmaceuticals.

Felber and Wester EJNMMI Radiopharmacy and Chemistry
(2021) 6:29 Page 4 of 16 In addition, cumbersome labeling procedures as currently required for radiofluorination might become dispensable if the respective ligands would allow for derivatization with a silicon-based fluoride acceptor moiety. Improved labeling conditions in combination with the 'Munich method' (Wessmann et al., 2012) for 18 F-drying might enable direct radiofluorination with higher RCYs, exploiting the favorable imaging characteristics of 18 F (low β + energy of max. 0.635 MeV, β + decay ratio of 97%, half-life 109.8 min) and the possibility of generating GLP-1R-targeting radiohybrid ligands (Jacobson et al., 2015;Wurzer et al., 2020).
Derivatives of 1, suitable for direct radiolabeling by isotopic exchange on a SiFAbearing prosthetic group, that could also be valuable precursors for future studies on the insertion of other modifications (e.g. attachment of a chelator), were planned to be preselected by affinity determinations. As a result, the main objective of this study was to generate precursor molecules with low nanomolar IC 50 values that preferably lie in the range of endogenous GLP-1.

Materials & methods
For detailed information on all methods for synthesis and analysis as well as on the used instruments, see the supporting information (available on https:// ejnmmipharmchem.springeropen.com).

Synthesis
Small peptide GLP-1R ligands 2 to 14 (except 3 and Tyr(3-I) 40 -3) were synthesized completely on Rink amide ChemMatrix® (RACM) resin. Synthesis of lead peptide 1 had to be accomplished via fragment coupling in solution (Scheme 1). Despite several attempts to realize on-resin synthesis as described by Mapelli et al. (Mapelli et al., 2009), 1 could not be obtained by a linear reaction protocol. On-resin Suzuki-Miyaura cross-coupling using compounds 15 and 16 provided the (2′-Et, 4′-OMe)BIP fragment 17 still functionalized with a cross-coupling resistant Boc-protective group at the N-terminus. Cleavage of 17 from the resin by HFIP/DCM (1/4) was not mandatory, since a major part was already cleaved from the 2-CT resin by the cross-coupling conditions. Boc-protected 17 was coupled to RACM resin-bound H-L-homoPhe and subsequent cleavage with TFA/TIPS/H 2 O (95/2.5/2.5) revealed dipeptide 18. Fragment coupling of purified 18 and 19 provided lead peptide 1*2TFA as colorless powder (2.05 mg, 16.1% referred to 18) after removal of all acid-labile protective groups and RP-HPLC purification (Fig. 2). This strategy involved three RP-HPLC purifications steps in total, after synthesis of 17, 18 and 1.

In vitro characterization
For direct comparison, IC 50 data of all new derived agents as well as of reference compounds 1, 2, 3 and Tyr(3-I) 40 -3 are displayed in Table 1. Due to different in vitro displacement behaviors initially observed for 4 and 5, SiFA-tagged peptides were investigated in assays supplemented with (w/) and without (w/o) BSA to determine optimum binding conditions.

SiFA-tagged undecapeptides
Based on the assumption, that bulky hydrophobic residues are tolerated by the GLP-1R at position 10 and 11 (structure of 1, 'Introduction'), (2′-Et, 4′-OMe)BIP at position 10 was substituted by a SiFA moiety with concomitant variation of the α-quaternary amino acid at position 6 ( Table 1). Both variants were synthesized to evaluate the effect of SiFA at position 10 and hence, their suitability as small peptide agents directly usable for 18 F-labeling. However, initial competitive binding experiments with bovine serum albumin (BSA) in the incubation buffer gave no detectable affinity neither for 4 nor for 5. Due to the presence of the lipophilic SiFA moiety (Bernard-Gauthier et al., 2014), experiments were repeated in assay buffer (HBSS) without bovine serum albumin (BSA) as additive. Thereby, lipophilicity-induced nonspecific adhesion and retention by BSA should be avoided (Lexa et al., 2014). The curves of displacement indicated IC 50 values of 2.40 μM (n = 1) for 4 and 189 ± 35 nM (n = 2) for 5, though still high unspecific binding could be observed (maximal radioligand displacement 61-69% of maximal radioligand binding, Fig. 3).
In general, a positive trend towards 5 could be determined indicating more favorable properties for undecapeptides with (α-Me)Phe(2-F) instead of (α-Me)Phe as αquaternary amino acid at position 6. A value of~13 was obtained for the IC 50 ratio 4/ 5 and thus confirmed the results obtained by Mapelli et al. (factor~3.2) (Mapelli et al., 2009) to some extent, despite the higher factor determined within this study.

Peptides without C-terminal bulky substituents
The N-terminal fragments of 4 and 5, represented by 6 and 7 respectively, were investigated but exhibited no detectable affinity (Table 1). Even introduction of an amidated C-terminus in 7 installed for mimicking resemblance to a true peptide bond, did not result in any benefit. Likewise, substitution of the L-Dap(SiFA) moiety of 4 by a simple L-Phe(4-I) residue (8) eliminated any detectable affinity. Obviously, this hydrophobic residue cannot compensate for the sterically demanding SiFA-moiety as in 4 (IC 50 = 2.40 μM). Consequently, the C-terminal dipeptide of lead peptide 1 or at least a hydrophobic bulky substituent at position 10 seemed to be indispensable for radioligand displacement within the 10 − 11 to 10 − 5 M range. Since a certain tolerance, and especially for undecapeptides a mandatory need for bulky substituents emerged, SiFA-bearing peptides similar to 1 were further investigated, however, with structures closer related to the endogenous ligand GLP-1. Thereby, an increased affinity should be reached since the peptides investigated so far revealed IC 50 values not below 189 ± 35 nM (n = 2) and therefore were not able to compete with high affinity GLP-1R-targeting peptides like GLP-1 (2) and [Nle 14 , Tyr(3-I) 40 ]exendin-4 (Tyr(3-I) 40 -3).

SiFA-tagged chimeras of 1 and GLP-1
Expanding the structural range, peptides with different chain lengths (11-15 amino acids) were synthesized and GLP-1-inherent amino acids were placed at the appropriate positions.
In order to avoid detrimental effects of the SiFA unit on ligand receptor interactions of the crucial N-terminal nonapeptide, repositioning towards the C-terminus was pursued. However, introduction of a SiFA moiety at position 11 was not tolerated, as no affinity was detected for peptide 9. By contrast, compounds with SiFA at position 13 like in 10, 11 and 12 displayed affinities in the micromolar range (2.95 ± 0.31 μM to 348 ± 117 μM, Table 1). Indeed, an aromatic tyrosine residue can be found at this position in endogenous GLP-1, indicating a certain tolerance for this structural modification. It is important to note that further variations from original GLP-1 residues between Asp 9 and Dap(SiFA) 13 seem to be not well tolerated in those peptides, although no crucial role for receptor binding and activation could be ascribed by previous SAR studies (Gallwitz et al., 1994;. Substitution of L-Ser 11 by a L-homoPhe residue as realized in peptide 1, resulted in a~2.4-fold decline in affinity (2.95 ± 0.31 μM for 10 vs. 7.12 ± 2.47 μM for 11). Consequently, L-homoPhe at position 11 might be advantageous for receptorligand-interactions in undecapeptides like 1, but obstructive if incorporated at the same position in extended peptides like 11.
Furthermore, affinity also remarkably decreased by introduction of a short hydrophilic poly (ethylene oxide)-similar linker (O2Oc), which was installed in peptides 12 and 13 (Table 1). First, this linker fragment was supposed to compensate to a certain extent for the overall lipophilic character of the SiFA unit and was therefore introduced adjacent to the SiFA moiety in 12 and 13 (Israelachvili, 1997). Whereas peptide 12 showed low but still detectable affinity (348 ± 117 μM), no radioligand displacement was observed for its D-stereoisomer 13. These results indicate that the natural Lconfiguration at position 13 might be preferred. Moreover, a certain rigidity introduced by the original GLP-1 residues Val 10 , Ser 11 and Ser 12 seems to be indispensable, since pentadecapeptides 10 and 11 still showed higher affinities (2.95 ± 0.31 μM and 7.12 ± 2.47 μM) compared to the more flexible O2Oc-modified analog 12 (348 ± 117 μM). Pentadecapeptides 10, 11, 12 and 13 were synthesized until Glu 15 (Glu 21 in GLP-1 counting method) for positioning a hydrophilic, anionic residue in close spatial proximity to the SiFA unit for partial compensation of the rather hydrophobic character.

C-terminal pentadecapeptide of GLP-1
The importance of the C-terminal fragment of GLP-1 for receptor binding and selectivity was already proven by several studies (Hjorth et al., 1994;Underwood et al., 2010;Wu et al., 2020;Hoare, 2005). Based on these findings and the rather poor results obtained so far with N-terminal fragments of GLP-1 and derivatives thereof, a potential usability of the 15-residue C-terminal fragment of GLP-1 was investigated. Interestingly the isolated, acetylated C-terminal half of GLP-1 (14) exhibited a low but noticeable affinity (2.54 ± 0.70 μM, Table 1). However, expedient optimizations of this pentadecapeptide would be necessary to obtain a basic structure that tolerates modifications like chelator or SiFA attachment whilst keeping high affinity towards GLP-1R. For now, the IC 50 of 14 exceeds the IC 50 of GLP-1 by a factor of 109.

Critical data analysis
It has to be noted that during the synthesis of compound 12, racemization was detected after the coupling of Fmoc-L-Dap(Dde)-OH and resulted in an enantiomeric ratio (er) of~3/7 for 12/13. Peptides 12 and 13 were isolated by collecting the respective fractions separately (t R = 6.81 min for 12 and t R = 9.39 min for 13, 80-100% Method B*, 1 mL/min. *Method B: solvent A = water + 0.1% TFA, solvent B = acetonitrile + 5% water + 0.1% TFA) via semi-preparative HPLC. Based on the following affinity determinations, peptide 12 was assumed to represent the stereoisomer comprising L-Dap(SiFA) 13 , since the natural ligand GLP-1 is assembled exclusively of L-amino acids. Further endeavors to determine the exact enantiomeric identity of compounds 12 and 13 were not undertaken, since even the favored enantiomer had a very low affinity (348 ± 117 μM). For all other compounds, no racemization was observed.

Lead peptide 1
In previous studies, GLP-1R binding affinities of 1 and related derivatives either remained undefined, or were not compared to established ligands, such as GLP-1 or exendin-4 (Mapelli et al., 2009;Haque et al., 2010a;Haque et al., 2010b;Hoang et al., 2015;Swedberg et al., 2015). Therefore, the affinity of this GLP-1 mimetic should be determined in cell-based assays and compared to the values of known high affinity ligands GLP-1 and exendin-4. In vitro evaluation of 1 revealed an unfavorable IC 50 value of 669 ± 242 nM (n = 5), which was not expected due to the very close EC 50 of 1 to endogenous GLP-1 (EC 50 (1) = 31 pM, EC 50 (GLP-1) = 34 pM), as determined by Mapelli et al. and Haque et al. (Mapelli et al., 2009;Haque et al., 2010a) Among all GLP-1R-targeting ligands, only 5 provided distinctly lower IC 50 values (189 ± 35 nM, n = 2) than lead peptide 1 (669 ± 242 nM, n = 5). Although this suggests a higher GLP-1R affinity for 5, nonspecific binding of this SiFA-bearing compound was notably high as displayed by the respective dose-response curves (Fig. 3). Introduction of charged amino acid residues (positive and/or negative) or other lipophilicityreducing auxiliaries in close spatial proximity to the SiFA moiety might compensate for lipophilicity-induced unspecific binding of 5 (Bernard- Gauthier et al., 2014;Lexa et al., 2014). Accordingly, competitive binding studies with these ligands may provide more reliable IC 50 values.
All in all, the rather poor results obtained for short-chained GLP-1 receptor ligands unveiled a huge discrepancy between the high potency of these ligands known from the literature and their actual low affinities, determined in HEK293-hGLP-1R cell-based assays. Especially 1 was supposed to possess a high affinity towards GLP-1R positive cells due to its inherent low EC 50 of 31 pM determined by Haque et al. (Haque et al., 2010a). However, 1 lost its potential as a basic structure for novel promising small peptide GLP-1R agonists, since affinity studies revealed an unfavorable IC 50 of 669 ± 242 nM (n = 5). It also remains questionable if small peptide approaches would generally provide high-affinity ligands, since previous SAR studies suggested that non-contiguous residues within the Nand C-terminus of GLP-1 and also in between were essential for high affinity and potency (Parker et al., 1998;Hjorth et al., 1994). As a consequence, yet synthesized SiFA-bearing ligands were not subjected to radiofluorination and no internalization or other in vitro and in vivo experiments were conducted due to the poor IC 50 data generated by these compounds.

Conclusions
In sum, the investigated structures provide no advantage over GLP-1-and exendin-4based ligands. Only SiFA-tagged peptide 5 showed a higher affinity compared to 1, but still not in a range which allows for in vivo evaluation. In consequence, radioligands Felber and Wester EJNMMI Radiopharmacy and Chemistry (2021)  targeting GLP-1R + tumor lesions with reduced kidney uptake and/or enhanced renal excretion could not be generated. In order to reach an improved tumor localization and enable targeted radiotherapy of malignant insulinomas as well as of other GLP-1Roverexpressing malignancies, further investigations on pharmacokinetically optimized peptides should be envisaged.