Modular Medical Imaging Agents Based on Azide–Alkyne Huisgen Cycloadditions: Synthesis and Pre‐Clinical Evaluation of 18F‐Labeled PSMA‐Tracers for Prostate Cancer Imaging

Abstract Since the seminal contribution of Rolf Huisgen to develop the [3+2] cycloaddition of 1,3‐dipolar compounds, its azide–alkyne variant has established itself as the key step in numerous organic syntheses and bioorthogonal processes in materials science and chemical biology. In the present study, the copper(I)‐catalyzed azide–alkyne cycloaddition was applied for the development of a modular molecular platform for medical imaging of the prostate‐specific membrane antigen (PSMA), using positron emission tomography. This process is shown from molecular design, through synthesis automation and in vitro studies, all the way to pre‐clinical in vivo evaluation of fluorine‐18‐ labeled PSMA‐targeting ‘F‐PSMA‐MIC’ radiotracers (t1/2=109.7 min). Pre‐clinical data indicate that the modular PSMA‐scaffold has similar binding affinity and imaging properties to the clinically used [68Ga]PSMA‐11. Furthermore, we demonstrated that targeting the arene‐binding in PSMA, facilitated through the [3+2]cycloaddition, can improve binding affinity, which was rationalized by molecular modeling. The here presented PSMA‐binding scaffold potentially facilitates easy coupling to other medical imaging moieties, enabling future developments of new modular imaging agents.

The aqueous residue was dried by freeze drying. The product 7 was isolated as a white solid

Radiochemistry
All executed syntheses and experiments were performed in agreement with the local radiation safety regulations by well-trained / licensed radiochemists. This includes that all actions were performed in lead-shielded fumehoods and HPLC systems, reaction vials were kept in lead containers as much as possible and the radiochemists were working with long tweezers to increase the distance between the extremities of the radiochemist and radiation source. The radiation burden of the radiochemists were checked every month by the radiation safety manager. The FlowSafe synthesizer module was kept in a closed, leadshielded HotCell to avoid any radiation burden for the radiochemists. times to remove residues of water.

Automation with FlowSafe Click Synthesis Module.
After successful manual synthesis, the 18 F-radiolabeling was automated for scaling-up

Radiotracer stability of [ 18 F]PSMA-MIC01 and [ 18 F]PSMA-MIC02
The stabilities of [ 18 F]PSMA-MIC01 and [ 18 F]PSMA-MIC02 were tested, to ensure the integrity of the tracer in solution. The reference compounds for both, F-PSMA-MIC01 and F-PSMA-MIC02, gave a retention time of 20 min. Since HPLC was used for purification, the first step is to collect the radioactive peak eluting at 20 min (A). After purification and formulation into an injectable solution of 10 % EtOH in PBS, the radiotracer was analysed by HPLC again (B),which was repeated after 2 h (C) and 4 h (D).

Cell binding studies.
For the determination of the binding affinity, a competitive binding radioassay was performed.
Two 24 well plates were incubated with 50.000 cells 3 to 4 days prior to the cell experiments.

Organ distribution and Metabolite analysis.
After sacrificing the animals, the organs were dissected and measured in a -counter. The obtained CPM's were normalized to %ID/g (Table shown in Table 1  In order to check for the significant differences, we also checked for the Cohen's d, a measurement to determine the effect size. It is calculated on the following formula: ) . Figure S4. Representative metabolite analysis of two different animals bearing a PC3-or a LNCaP-xenograft. While in Plasma, no metabolite were formed, the urine samples were not conclusive.

Molecular docking
The proteins were prepared through the Protein Preparation Wizard in Maestro, performing the assignment of bond orders, hydrogens addition, hydrogen bonds definition and optimization, waters removal and restrained minimization with the OPLS3 force field. [12] The grid was created through the Receptor Grid Generation, picking the ligand to define the centroid of the receptor box, and rotation of the hydroxyl groups of Ser501, Ser513, Tyr552, Tyr700 were allowed. LigPrep was used to prepare the ligands and to generate possible states at pH 7.0 ± 2.0 with Epik. The ligands were docked with Glide XP, [13]  Two protein-ligand complexes were considered for this docking study, PSMA complexed with MeO-P4 [14] (PDB: 2XEJ) and ARM-P2 [14] (PDB: 2XEI), so that two distinct conformation of Trp541 were included (see main text). In those PDB complexes, electron density was absent for the PEG chain (due to its flexibility and lack of specific interactions) and, in 2XEI, for the nitro groups (because the ring is in more different conformations).
Redocking of the co-crystallized ligands were carried out on 2XEJ and 2XEI ( Figures Figure S13). However, they have a suboptimal orientation for π-π interactions (Table S7) compared to the cutoffs of the Ligand Interaction Diagram.
In Maestro's User Manual, a π-π interaction is defined as an interaction between two aromatic rings in which either (a) the angle between the ring planes is less than 30° and the distance between the ring centroids is less than 4.4 Å (face-to-face), or (b) the angle between the ring planes is between 60° and 120° and the distance between the ring centroids is less than 5.5 Å (edge-to-face). These criteria are the adaptation of literature cutoffs. [15] Figure S13. Detail of the π-π interactions for the co-crystallized ARM-P2 (green), the docked F-PSMA-MIC02 (violet) and F-PSMA-MIC04 (pink) into 2XEI. The π-π interactions behavior of these three compounds was further evaluated with an MD study.

Molecular Dynamics
The protocol was adapted from a previous MD study on the same system. [13] The crystal structure of ARM-P2 in complex with PSMA (PDB: 2XEI), the second-best docking pose of F-PSMA-MIC02 into 2XEI and the top-ranked docking pose of F-PSMA-MIC04 into 2XEI were used to setup the MD calculations. The structures were embedded in a orthorhombic box of circa 20600 TIP3P [16] water molecules, the dimension of the box was circa 106x86x83 Å. The net charge of the system was neutralized by addition of five sodium ions to the solvent box. The total number of atoms was circa 73,000 atoms. The simulations were performed with the Desmond molecular dynamics package, [17] with default settings for bondconstrains, Van der Waals and electrostatic interactions cutoffs, PME method for long range electrostatic interactions.
Each system was subjected to the following relaxation and equilibration protocol: 100 ps of Brownian dynamics at 10 K in the NVT ensemble with harmonic restraints (50 kcal/mol/A) [14] on the solutes heavy atoms, followed by 12 ps in the NVT ensemble (Berendsen thermostat) [20] at 10 K and retaining harmonic restraints on the solutes heavy atoms, followed by 12 ps in the NPT ensemble (Berendsen thermostat and barostat) at 10 K and retaining harmonic restraints on the solutes heavy atoms, followed by 24 ps in the NPT ensemble (Berendsen thermostat and barostat) at 300 K and retaining harmonic restraints on the solutes heavy atoms, followed by 24 ps in the NPT ensemble (Berendsen thermostat and barostat) at 300 K without harmonic restraints on the solutes heavy atoms. The production simulations were run for 100 ns in the NPT (300 K, 1 bar, Martyna-Tobias-Klein barostat and Nose-Hoover thermostat), [21,22] in three replicas. Coordinates were saved every 100 ps and analyzed in Maestro.
Ring distances and ring angles between the aromatic ring of the ligands and the sixmembered ring of Trp541 were measured in Maestro, through the Plot>Measurements tool.
Following Maestro's User Manual, π-π interaction cutoffs were defined as follows: (a) the angle between the ring planes is less than 30° and the distance between the ring centroids is less than 4.4 Å (face-to-face), or (b) the angle between the ring planes is between 60° and 120° and the distance between the ring centroids is less than 5.5 Å (edge-to-face). These criteria were the adaptation of literature cutoffs [15] . The choice of the six-membered ring in the indole of Trp541 as the ring for the distances and angles measurements was supported by QM calculations (see next section), because the negative electron potential was localized on top of the six-membered ring.
The following Figures (12 -17) depict the time traces of ring distances (left) and ring angles (right) between the aromatic ring of the ligands and the six-membered ring of Trp54, over the course of the three MD replicas for each of the protein-ligand complex. Replica 1 is red, Replica 2 is green and Replica 3 is green. The areas that correspond to the geometry cutoffs for face-to-face and edge-to-face π-π interactions are highlighted in pink and yellow, respectively. Hydrophobic (π-cation, π-π, non-specific interactions), Ionic and Water bridges.

Ab initio calculations
The models were constructed using the software Maestro. [1] Geometries were initially optimized with MacroModel (Force Field: OPLS3, [14] vacuum, Method: PRCG). Afterwards, the geometries were further optimized at the M06-2X-D3/6-311G**++ level, in vacuo, with Ultrafine accuracy, 100 max iterations, tight convergence criteria for SCF (1e-6 energy change, 1e-7 density matrix change), tight convergence criteria (iaccg=5 in the input file) and the option "Switch to analytic integrals near convergence" on. Single point energies were calculated at the same level of theory and with the same options. Frequency analysis showed zero imaginary frequencies for all the optimized structures. Electrostatic potential surfaces of the fragments were generated by mapping the electrostatic potentials onto surfaces of molecular electron density (0.001 electron/Å) and rainbow color-coding, using the software Maestro. [1] The potential energy values range from +25 kcal/mol to -25 kcal/mol, where red signifies the maximum in negative potential and violet signifies the maximum in positive potential.