Synthesis and in vitro and in vivo evaluation of urea-based PSMA inhibitors with increased lipophilicity

The PSMA inhibitors 1–9 (Fig. 1) were synthesized in a combined solution and solid-phase synthesis strategy. The acetylated three-amino acid peptides (amino acid spacer in Fig. 1) were assembled using Fmoc-protocol solid-phase peptide synthesis starting from approximately 100 mg TCP resin. Acidic cleavage from the resin and precipitation in diethyl ether yielded the peptides in medium to high yields and purity. Acetyl ester formation at the unprotected side chain hydroxyl group of 8 and 9 necessitated an additional “deprotection step,” which was achieved by alkaline hydrolysis using methanol/water/saturated NaHCO₃ (2/1/1) in almost quantitative yields. The reaction of the crude peptides with NHS-Sub-(OtBu)KuE(OtBu)₂ was performed according to a literature procedure [22]. After removal of the solvent and acidic deprotection, the PSMA inhibitors were purified using RP-HPLC.

The DOTAGA-conjugated inhibitor 10 was synthesized from the DOTAGA-y-nal-k peptide and PfpO-Sub-(OtBu)KuE(OtBu)₂ yielding the PSMA inhibitor after tBu deprotection and HPLC purification. The NHS ester of suberic acid is commercially available; however, the PfpO ester (Sub(OPfp)₂) was synthesized in 68% yield from affordable reagents resulting in a less hydrolysis-prone building block.

Hydrolytic stability was especially important for the synthesis of PSMA inhibitor 11. To address plasma proteins, a PSMA inhibitor with a halogenated aromatic moiety (4-iodo-d-phenylalanine; I-f) was developed on the basis of 10. Substitution of the suberic acid spacer with glutaric acid-(I-f) resulted in DOTAGA-y-nal-k(Glut-(I-f)-KuE) (11). For the synthesis of 11, PfpO-Glut-(I-f)-(OtBu)KuE(OtBu)₂ was synthesized in solution from (OtBu)KuE(OtBu)₂, Fmoc-d-iodo-phenylalanine, and the bis-pentafluorophenylester of glutaric acid (Glut(OPfp)₂) and was purified using RP-HPLC. Glut(OPfp)₂ was prepared from glutaric acid using pyridine and pentafluorophenol in 87% yield after flash chromatography. Synthesis yields of PSMA inhibitors 1–11 are summarized in Additional file 1: Table S1.

The effect of the lipophilic (aromatic) modifications in the spacer unit of the acetylated PSMA inhibitors 1–9 on the IC₅₀ to PSMA was determined in a competitive binding assay on PSMA-expressing LNCaP cells. (((S)-1-carboxy-5-(4-([13]iodo-benzamido)pentyl)carbamoyl)-l-glutamic acid (¹²⁵I-IBA)) in a concentration of 0.2 nM was used as radioligand [22]. The means of three independent measures are summarized in Table 1. The acetylated L-derivative, Ac-YFK(Sub-KuE) (1), was included to ensure comparability with the DOTAGA-conjugated compound PSMA I&T [17, 22] and revealed a lower affinity compared to it (15.0 ± 1.3 nM vs. 10.2 ± 3.5 nM, 1 vs. PSMA I&T, respectively).

Increased affinity of PSMA inhibitors was reported by interaction with the 20-Å-deep amphipathic funnel-shaped tunnel leading from the enzyme surface to the S₁ pocket in the active center of PSMA [32]. The bulky substituent biphenylalanine in 4 showed the lowest affinity in this series, most likely due to steric repulsion in this narrow tunnel and suboptimal fit into the arene-binding site [13, 33]. Yet, in a recent investigation, a PSMA inhibitor with a similar biphenyl residue demonstrated the highest K_i among the tested set of ligands [34].

Exhibiting similar lipophilicity and steric demand with PSMA I&T [17], iodo-tyrosine-like substituents were investigated with inhibitors 7 (4-nitro-phenylalanine) and 9 (methyl-tyrosine). The affinity of 7 and 9, as well as the tryptophan-containing inhibitor 2, and the benzothienylalanine-containing inhibitor 3 was comparable to PSMA I&T. The diiodo-tyrosine-containing inhibitor 8, as well as 5 (1-naphtylalanine) and 6 (2-naphtylalanine), revealed two- to threefold higher affinities compared to PSMA I&T. Due to synthetic problems caused by the unprotected diiodo-tyrosine side chain of 8 and the good availability of the naphthylalanine derivatives, inhibitor 6 was selected for further PSMA inhibitor development.

Interestingly, the lutetium complexes of PSMA I&T, 10, and 11 revealed higher affinity to PSMA compared to the respective gallium complexes.

Although crystal structure-based characterization of the active center of PSMA revealed an additional lipophilic binding pocket (S₁ accessory lipophilic pocket) near the S₁ site [32, 35], the affinity of ^natGa- and ^natLu-11 was comparable to metallated 10 being in the low nanomolar range. Thus, substitution of suberic acid in inhibitor 10 by glutaric acid-(iodo-phenylalanine) in inhibitor 11 does not increase the affinity to PSMA as reported previously for PSMA inhibitors with aromatic moieties conjugated to the KuE motif [28].

To determine an effect on the cell binding and internalization of radiolabeled 10 and 11 (Additional file 1: Figure S1), LNCaP cells (125,000/well) were incubated with 0.2 nM ⁶⁸Ga or 0.5 nM ¹⁷⁷Lu-10 or ¹⁷⁷Lu-11, respectively, for up to 1 h at 37 °C as described in the literature [17, 22]. A 2-(phosphonomethyl)pentane-1,5-dioic acid (PMPA) wash step (10 μm, 10 min, 4 °C) was conducted to differentiate between specifically membrane-bound (< 3% of applied dose) and internalized activity. Nonspecific binding, determined by co-incubation with 10 μm PMPA, was below 1% of the applied dose. To compensate for differences in cell count and viability, the external reference ¹²⁵I-IBA was always assayed in parallel, and the binding data in Table 2 are given as percentage of the external reference. As expected from the affinity data, the internalization of ¹⁷⁷Lu-10 was comparable to ¹⁷⁷Lu-PSMA I&T. Interestingly, the cell binding and internalization kinetics revealed a significantly increased cellular uptake of ⁶⁸Ga-10 in LNCaP cells. An increased internalization was also observed for ⁶⁸Ga- and ¹⁷⁷Lu-11 compared to radiolabeled 10 and PSMA I&T. Unlike that reported for radiolabeled DOTAGA-ffk(Sub-KuE) and PSMA I&T [17, 22], the internalization experiments do not correlate with the affinity to PSMA. Thus, the affinity and cell binding data indicate that the interaction of PSMA inhibitors 10 and 11 with PSMA differs from PSMA I&T [17, 22] and was therefore further investigated with regards to lipophilicity and in vivo behavior.

For the novel PSMA inhibitors, the partition coefficient between n-octanol and PBS (pH 7.4) was determined using the shake-flask method. The logP values (n = 6 for each) of the ⁶⁸Ga- and ¹⁷⁷Lu-labeled inhibitors 10 (− 3.8 ± 0.1 and − 4.1 ± 0.1 for ⁶⁸Ga- and ¹⁷⁷Lu-10, respectively) and 11 (− 3.5 ± 0.1 and − 3.1 ± 0.1 for ⁶⁸Ga- and ¹⁷⁷Lu-11, respectively), designed for optimized lipophilic interaction with lipophilic PSMA pockets, was correspondingly higher compared to the radiolabeled derivative PSMA I&T (− 4.3 ± 0.3 and − 4.1 ± 0.1 for ⁶⁸Ga- and ¹⁷⁷Lu-PSMA I&T, respectively). However, having the hydrophilic KuE motif and the chelator (DOTAGA) at both ends of the molecule, radiolabeled PSMA inhibitors 10 and 11 were still highly hydrophilic compared to other peptides, such as peptides binding the CXCR-4 receptor (logP (⁶⁸Ga-CPCR4–2) = − 2.90 ± 0.08) [36].

High in vivo plasma protein binding increases the plasma half-life of the radiopharmaceutical and therefore might offer beneficiary effects on the tracer distribution (higher uptake into target tissue) but can also lead to increased background activity especially at early time points [27]. In general, drugs binding to plasma proteins with high affinity feature moderate to high lipophilicity, in many cases due to halogenated aromatic groups. To estimate the bioavailability of ¹⁷⁷Lu-PSMA I&T, ¹⁷⁷Lu-10, and ¹⁷⁷Lu-11 in the blood circulation, the extent of plasma protein binding was determined by in vitro incubation in human plasma and subsequent ultracentrifugation. Human albumin binding was determined, applying a modified HPLC method [37]. In accordance with an almost similar lipophilicity of ¹⁷⁷Lu-PSMA I&T and ¹⁷⁷Lu-10, the plasma protein binding of these PSMA inhibitors was 82% and 81%, respectively. These high values might be explained by the multiple negative charges (carboxylates of KuE and DOTAGA) at both ends of the molecules, being connected over a lipophilic peptide spacer, another structural motif reported to bind plasma proteins [31]. In addition, the intercalation of an additional iodo-phenylalanine residue increased the lipophilicity of ¹⁷⁷Lu-11 compared to ¹⁷⁷Lu-10. In consistency with the increased lipophilicity, the iodo-phenyl group insertion resulted in almost quantitative plasma protein binding of 97% for ¹⁷⁷Lu-11. Similar results were obtained for the HSA binding. While ^natLu-PSMA I&T and ^natLu-10 showed values of 79% and 83% bound to HSA, the para-iodo-phenyl-substituted derivative ^natLu-11 exhibited 97%. The results indicate that the modification with the halogenated aromatic residue increases in first line the albumin binding, which accounts almost completely for the almost quantitative plasma protein binding of ^natLu-11 in vitro.

To investigate an influence of the increased internalization of ⁶⁸Ga-10 and the almost quantitative plasma protein binding of radiolabeled 11 on the in vivo behavior, the biodistribution of ⁶⁸Ga-10, ⁶⁸Ga-11, and PSMA I&T was determined 1 h after injection in LNCaP tumor-bearing CD-1 nu/nu mice (Fig. 3a). As expected from the highly hydrophilic tracers, their clearance was fast and exclusively via the kidneys. After 1 h, the uptake of ⁶⁸Ga-labeled 11 into the tumor xenograft, the kidneys, and the spleen (all of which are organs with documented PSMA expression [38]) was comparable to ⁶⁸Ga-labeled 10 and PSMA I&T. In consistency with the 97% in vitro plasma protein binding for ⁶⁸Ga-11, 1.3 ± 0.1% ID/g was found in the blood after 1 h compared to 0.4 ± 0.2% ID/g for ⁶⁸Ga-PSMA I&T. Interestingly, increased blood retention was also observed for ⁶⁸Ga-10, although the HPLC retention, the logP, and the plasma protein binding were comparable to ⁶⁸Ga-PSMA I&T. Thus, the increased internalization and elevated blood level of ⁶⁸Ga-10 and ⁶⁸Ga-11 compared to ⁶⁸Ga-PSMA I&T merits further investigation at later time points.

With respect to a potential endoradiotherapeutic benefit of the higher blood levels of radiolabeled 10 and 11, the biodistribution of ¹⁷⁷Lu-10 and ¹⁷⁷Lu-11 was determined at 24 h p.i. in LNCaP tumor-bearing SCID mice (Fig. 3b). The 24-h p.i. biodistribution of ¹⁷⁷Lu-10 revealed washout from all organs and a tumor retention (4.5 ± 1.1% ID/g), which is comparable to ¹⁷⁷Lu-PSMA I&T (4.1 ± 1.1% ID/g). A fourfold increase of the activity in the LNCaP tumor xenograft at 24 h p.i. was observed for ¹⁷⁷Lu-11 with 16.1 ± 2.5% ID/g. The high ¹⁷⁷Lu-11 retention in the tumor xenograft is explained by 97% plasma protein binding, which decelerates the excretion and thus increases the uptake in PSMA-specific tissues with time. This hypothesis is supported by the higher blood activity of radiolabeled 11 compared to PSMA I&T at 1 h p.i., which might allow subsequent delivery of the PSMA inhibitor to the tumor over time. However, although the initial blood activity at 1 h p.i. of ⁶⁸Ga-10 and ⁶⁸Ga-11 was similar, ¹⁷⁷Lu-10 did not show the same effect of increased tumor uptake at 24 h p.i. The data suggest that the initial distribution of both ligands is not significantly influenced by the HSA binding; however, its effect becomes more apparent at later time points. The initial α-phase of the blood clearance is dominated by especially the distribution process for strong and weak plasma protein binding small molecules and leads to a rapid decline of blood pool activity after i.v. injection. However, the difference in the clearance kinetics becomes more visible in the terminal elimination phase. The terminal phase is stronger dominated by elimination processes and influenced by the plasma protein binding properties of the molecule. The ¹⁷⁷Lu-11 activity in the spleen (6.6 ± 3.3% ID/g) and the kidneys (100.9 ± 45.4% ID/g) was also higher for ¹⁷⁷Lu-11 compared to ¹⁷⁷Lu-10, confirming the effect of higher plasma protein binding on PSMA-expressing tissue uptake, PSMA-specific uptake in these organs. Similar results were observed by other groups, in which the strong albumin-binding ligands exhibited increased tumor accumulation and stronger retention [26, 27, 39]. Kelly et al. suggested that differences in kidney uptake are mediated by albumin binding and that strong binding ligands should display lower renal accumulation. This is, however, in contrary to our observation. Kidney uptake of ¹⁷⁷Lu-11 was the highest after 24 h among the tested ligands and resembled more the results of Choy et al. [27].

Although an almost fourfold higher uptake of ¹⁷⁷Lu-11 compared to ¹⁷⁷Lu-10 and ¹⁷⁷Lu-PSMA I&T was found at 24 h p.i., the tumor-to-organ ratios (Fig. 3c) of ¹⁷⁷Lu-PSMA I&T were comparable or higher than that for ¹⁷⁷Lu-11. The high and persistent tumor uptake of ¹⁷⁷Lu-11 was caused by a decelerated blood clearance and probably higher metabolic stability due to stronger interactions with albumin and the resulting decreased steric accessibility of metabolizing enzymes towards the PSMA ligand. The increased tumor uptake led to high tumor-to-organ ratios at 24 h p.i., e.g., 326 ± 0.3 (tumor-to-blood) or 143 ± 0.3 (tumor-to-bone). Thus, endoradiotherapeutic application of ¹⁷⁷Lu-11 might deliver higher radiation doses to the target tissue compared to ¹⁷⁷Lu-PSMA I&T and ¹⁷⁷Lu-10. However, the likewise increased blood level and especially the kidney and spleen uptake of ¹⁷⁷Lu-11 after 24 h p.i. (Fig. 3b) most likely limits the maximal dose and has to be considered in terms of potential nephro- or hematotoxicity [40]. First-in-man studies are necessary to address the issue if the renal uptake in preclinical studies is a predictive parameter to assess the nephrotoxicity in men. The difference of preclinical results regarding the renal uptake of PSMA I&T (high) and PSMA-617 (low) and the almost identical uptake in clinical studies suggest that the kidney uptake in mice has only limited predictive value [21].

Figure 4a shows PET images of LNCaP tumor-bearing mice 1 h after injection of ⁶⁸Ga-10 and ⁶⁸Ga-11, respectively. In accordance with the biodistribution data, both tracers were primarily taken up into the tumor and the kidneys, with excretion into the bladder. The PSMA specificity of in vivo binding was confirmed by co-injection of the structurally independent PSMA inhibitor PMPA. In the ⁶⁸Ga-11 time-activity curves of a 1.5-h observation period (Fig. 4c), the plasma protein bound activity seems to be stronger delivered to PSMA-specific organs compared to ⁶⁸Ga-10 (Fig. 4b) and therefore leads to an increase in tumor uptake over time, which is consistent with the biodistribution data of ¹⁷⁷Lu-11 at 24 h p.i.

The PSMA inhibitors 1–9 were synthesized according to previously published methods [22, 41] with minor modifications as described in Additional file 1. For amino acid nomenclature, the one-letter code according to IUPAC-IUB was used.

For lipophilicity determination, samples of 1–2 MBq ⁶⁸Ga/¹⁷⁷Lu-10 or ⁶⁸Ga/¹⁷⁷Lu-11 in 500 μL PBS, respectively, were added to 500 μL n-octanol (n = 6), mixed gently for 3 min, and centrifuged for 3 min. A 100-μL octanol and a 50-μL PBS sample were quantified using a γ-counter, respectively.

Plasma protein binding was determined in fresh blood samples collected in heparinized tubes, which were centrifuged at 6000 rpm (Biofuge 15, Heraeus Sepatech, Osterode, Germany) to separate plasma from the blood cells. Subsequently, 1–2 MBq ¹⁷⁷Lu-10 or ¹⁷⁷Lu-11 was added to the fresh plasma, respectively, and was incubated at 37 °C for 15 min. The sample was transferred into a VWR 30K (low protein binding) modified PES ultrafiltration vial and centrifuged at 16000 rpm. Quantification of the activity concentration before and after ultrafiltration was performed in a γ-counter. The data was corrected for non-specific association to the membrane (ultrafiltration of tracer sample in PBS).

HSA binding experiments were performed according to a previously reported method using a binary gradient HPLC system connected to a Chiralpak HSA (5 μm, 50 × 3 mm) analytical column connected to a Chiralpak HSA (5 μm, 10 × 3 mm) guard cartridge (Daicel Chemical Industries) purchased from Chiral Technologies Europe (Illkirch, France) with minor modifications [37]. Mobile phase A was an ammonium acetate buffer (50 mM, pH 6.9), mobile phase B was 2-propanol, the flow rate was 0.5 mL/min, and equilibration with 0% B for 3 min was followed by a gradient of 0% B to 20% B to the end of each run within 45 min. To check the column performance and to conduct the non-linear regression, the HSA column was calibrated each day with nine reference substances. Afterward, the PSMA inhibitors with unknown HSA binding were measured. The retention times and factors together with correlation curve are found in Additional file 1.

Experimental details for the determination of the affinity to PSMA (IC₅₀), the cell binding and internalization kinetics, and the specificity of binding in vitro were described previously [17, 22] and are summarized in Additional file 1.

All animal experiments were conducted in accordance with the German Animal Welfare Act (Deutsches Tierschutzgesetz, approval #55.2-1-54-2532-71-13). Tumor induction and experimental details were described previously [17, 22] and are summarized in Additional file 1.