PET Imaging of Self‐Assembled 18F‐Labelled Pd2L4 Metallacages for Anticancer Drug Delivery

Abstract To advance the design of self‐assembled metallosupramolecular architectures as new generation theranostic agents, the synthesis of 18F‐labelled [Pd2L4]4+ metallacages is reported. Different spectroscopic and bio‐analytical methods support the formation of the host‐guest cage‐cisplatin complex. The biodistribution profiles of one of the cages, alone or encapsulating cisplatin have been studied by PET/CT imaging in healthy mice in vivo, in combination to ICP‐MS ex vivo.


Introduction
Supramolecular systems are attracting increasing attention in the development of nanomaterials for different applications. [1] Self-assembled porous metallacages are particularly attractive supramolecular coordination complexes featuring discrete mo-lecular 3-dimensional (3D)-architectures with various appealing applications, including storage, separation, catalysis, recognition, as well as light emitting materials amongst others. [2] In medicine, the biological properties of these well-defined molecular vessels have recently been gaining momentum for drug delivery of therapeutics [3] and imaging agents, [4] as well as for the development of novel theranostic platforms. [5] Among the different metal assemblies, some of us have focused on the advantages of a particular type of palladium-based metallacage scaffold [Pd 2 L 4 ] 4 + (L = 3,5-bis(3-ethynylpyridine)phenyl) as potential delivery system for the well-established anticancer drug cisplatin. [6] Thus, we developed the exo-functionalization of the ligands to add different bioactive components, including fluorescent tags [7,8,9] and peptidic domains. [10] It was also demonstrated that encapsulation of cisplatin in integrin targeted metallacages leads to reduced nephrotoxicity with respect to free cisplatin. [11] Encapsulated cisplatin showed also higher in vitro cytotoxicity against cancer cells expressing the integrin receptors. [11] Noteworthy, [Pd 2 L 4 ] 4 + cages tethered to a blood brain barrier (BBB)-translocating peptide and encapsulating radioactive pertechnetate were recently studied for their biodistribution in mouse models, and demonstrated the stability of the host-guest (cage-pertechnetate) complex and its brain penetration capability. [12] While all the prominent proof-of-concept reports mentioned above can give a glimpse of a bright future for the use of metallacages as drug delivery systems, in vivo imaging studies of these supramolecular entities are still scarce. [13,4] Most importantly, their design as novel theranostic platforms featuring both therapeutic and imaging modalities is still in its infancy. Besides, the ultimate question of the structural integrity of the metallacages upon in vivo injection remains unresolved. In this context, we considered that 18 F-labelling followed by in vivo Positron Emission Tomography (PET) imaging might offer interesting insights to that end. Additionally, and in contrast to optical imaging techniques, PET is fully translational into the clinical setting to assess whole body biodistribution. We considered the positron emitter fluorine-18 ( 18 F) as an appropriate radionuclide due to its wide availability and

Results and Discussion
Among the possible 18 F-labelling strategies, our attention was drawn away from the common 18 F-carbon bond-forming processes, focusing instead on the [ 19 F]-to-[ 18 F]-boron isotopic exchange using ammonium trifluoroborate functionalities (AMBF 3 ). These types of prosthetic groups, introduced by the Perrin laboratory, allow for efficient, single-step, aqueous 18 Flabelling using as the labelling agent aqueous [ 18 F]fluoride directly produced in the cyclotron, thus avoiding time-consuming evaporation steps for solvent exchange. [14a,b] As an efficient route to append the AMBF 3 fragment to the metallacage, we used a classical "click" approach, via a propargyl-bearing AMBF 3 reagent (PPG-AMBF 3 , [14c] Scheme 1) that could be conjugated to the azide-modified cage ligand by Cu-catalyzed alkyne-azide cycloaddition reaction (CuAAC). Hence, two azide-functionalized bis(pyridyl)ethynyl ligand precursors were initially considered, namely the derivative 1 [9a] with the azide attached directly to the central phenylene unit, and compound 2 in which a spacer is introduced between the ligand portion and the azide (see Figure S1a -b). The latter was prepared in an 85 % yield by Steglich-type esterification of the benzylic alcohol exo-functionalized ligand 2' [15] with 2-azido acetic acid. Next, the AMBF 3modified ligands L1 and L2, were synthesized through a CuAAC reaction of azides 1 and 2 with PPG-AMBF 3 using the CuBr-PMDETA (N,N,N',N'',N''-pentamethyldiethylenetriamine) as catalyst (Scheme 1). 16 Data related to the NMR characterization of the ligands are reported in Figure S3-S4 in the Supporting Information.
With ligands L1 and L2 in hand, we proceeded to the selfassembly of the homoleptic cages C1 and C2 with palladium dinitrate precursor which reached completion within 10 min after mixing (Scheme 1). The assembly of both [Pd 2 L 4 ] 4 + cages was unequivocally confirmed by 1 H, DOSY NMR experiments, as well as by high-resolution electrospray mass spectrometry (HR-ESI-MS) ( Figure S5-S9 in the Supporting Information). [6] Next, we verified that the newly assembled cages retained their ability to encapsulate cisplatin, an important feature for their potential as drug delivery systems. X-ray diffraction analysis has shown that Pd 2 L 4 cages can encapsulate up to 2 cisplatin molecules. [3c,6a] It should be noted that the cage's 3,5bis(3-ethynylpyridine)phenyl) scaffold creates a hydrophobic cavity whereby cisplatin encapsulation should be favoured over occupancy of the cavity by water molecules in solution. [17] Thus, the host-guest properties of C1 and C2 were studied by 1 H and 195 Pt NMR spectroscopy, as well as by HR-ESI-MS (Figure S10-S14). Metallacages C1 or C2 (1 equiv.) were dissolved in DMF-d 7 and then up to 3 equiv. of cisplatin were sequentially added stepwise, with each addition followed by a 10-min sonication step. The 1 H NMR spectrum recorded after the last addition revealed an identifiable downfield chemical shift of the exofacing proton H b (Δδ = 0.02 ppm) for both cages, previously observed for similar cage systems upon cisplatin encapsulation. [3c,6b] In addition, an observable upfield chemical shift (Δδ = 0.02 ppm) was found for the endohedral cavityfacing H e , a plausible sign of the guest presence ( Figure 1 and Figure S10). Afterwards, 195 Pt NMR spectroscopy was also utilized to gain further insights into the cisplatin encapsulation in these cavities. Compared with free cisplatin, an upfield Scheme 1. Structure of the clickable zwitterionic ammonium trifluoroborate (AMBF 3 ) tag PPG-AMBF 3 , and synthesis of ligands L1 and L2 and of the AMBF 3modified cages C1 and C2. chemical shift of about À 2 ppm was observed upon addition of 2 equiv. of cisplatin to a DMF-d 7 solution of complex C1 ( Figure S12), in line with previous studies, [9] and corroborating the idea of cisplatin encapsulation.
HR-ESI-MS analysis of [C1(NO 3 ) 4 �cisplatin] and [C2(NO 3 ) 4 �cisplatin], respectively, provided further evidence of the cagecisplatin encapsulation properties. For example, when a 1 : 2 mixture of C1 and cisplatin in DMF was analysed, two clear peaks at m/z = 839.8452 and m/z = 1290.7616 appeared that could be unambiguously assigned to [Pd 2 (L1) 4 (NO 3 ) 1 �cisplatin] 3 + and [Pd 2 (L1) 4 (NO 3 ) 2 �cisplatin] 2 + host-guest complexes, respectively (see Figure S13a -d). Moreover, a peak at m/ z = 1440.7548 was attributed to [C1(NO 3 ) 2 �(cisplatin) 2 ] 2 + adducts. These species were also detected in 4 % DMSO in water. Analogous behaviour was observed for cage C2 (see . The encapsulation of cisplatin in C1 was also studied by 1 H DOSY NMR in DMF-d 7 /D 2 O (90 : 10). Cisplatin alone in DMF shows a broad signal at ca. 4.2 ppm (H from NH 3 ), with a diffusion coefficient ca. 6 × 10 À 6 m 2 /s. 11 Upon addition of 2 equiv. cisplatin to 1 equiv. metallacage, the typical signal of free cisplatin disappeared in the DOSY plot (although marginally present in the 1 H spectrum likely due to the fast exchange of free vs. encapsulated cisplatin species), while significant broadening of the cage signals were observed, accompanied by small shifts in their diffusion coefficients ( Figure S15, spectrum c), suggesting that the cage cavity has been saturated to form the [C1�cisplatin] host-guest complex. 11 The cisplatin peak reappears only upon addition of a third equivalent of cisplatin to the sample ( Figure S15, spectrum d), which should not undergo encapsulation. [11] These data agree with the aforementioned HR-ESI-MS studies.
The feasibility of the 18 F-isotopic labelling of exo-functionalized cages was then assessed. Initially, direct 19 F-to-[ 18 F] exchange ( 18 F-EX) from the preassembled C1 and C2 was attempted as a more straightforward approach. To this end, a wet no carrier added solution of 18 F-fluoride ion was generated and subsequently trapped on an anion exchange resin (QMA Cartridge). The 18 F-fluoride was next eluted with isotonic saline and then added to a vial containing unlabelled metallacages C1 and C2 in a DMF/water-pyridazine·HCl buffer (pH = 2) following previously established procedures. [14a] The mixture was heated at 90°C for 30-60 min. Unfortunately, the radio-HPLC analysis of the 18 F-EX reaction evidenced major disassembly of the Pd 2 + cages to the corresponding isotopically marked 18 F-L1 and 18 F-L2 ligands. Nevertheless, this experiment proved the efficient 18 F-EX reaction of the ligand framework prior to the selfassembling process.
Given the fast kinetics of the assembly (few minutes), the final generation of the 18 F-labelled-C1 was straightforwardly achieved within the radioisotope half-life time. Thus, 18 F-L1 was prepared as above with slight modification of the experimental conditions (T = 85°C; reaction time = 45 min, see Supporting Information for details). After confirmation of the quantitative formation of 18 F-L1 by radio-HPLC ( Figure S21, chromatogram a), the labelled 18 F-C1 cage was assembled by mixing 18 F-L1 (2 equiv.) and 1 equiv. of Pd(NO 3 ) 2 · 2H 2 O in DMSO for 30 min in 95 % chromatographic yield (see Figure S21, chromatogram b). The same conditions were applied to achieve 18 F-L2 and cage 18 F-C2. However, we decided to pursue the in vivo study only with L1 since chromatographic yield of ligand 18 F-L2 proved lower (40 %) than the corresponding 18 F-L1. Afterwards, we proceeded with encapsulation of cisplatin in the radio-labelled cage. It should be noted that the radio-HPLC retention time of the parent cage 18 F-C1 and the same cage incubated for 5 min with 2 equiv. of cisplatin were virtually indistinguishable (Figure S21, chromatogram c). However, inductively coupled plasma optical emission spectrometry (ICP-MS) analysis of the manually collected fractions of the latter sample did reveal presence of Pt for cisplatin loaded 18 F-C1 (ca. 0.5 ng) but not in samples of free 18 F-C1 and 18 F-L1 (ca. 0.03 and 0.05 ng, respectively) used as controls; an observation that constitutes another indirect evidence of cisplatin encapsulation ( Figure S22).
In parallel, the stability of cage C1 over several hours was assessed through a series of 1 H NMR experiments in different conditions that included 100 % DMSO-d 6 and 4 % DMSO-d 6 in D 2 O (Figure S16-S17). The stability of the [C1(NO 3 ) 4 �cisplatin] complex was also monitored by 1 H NMR over time in 4 % DMSO-d 6 in D 2 O, and the results showed that the signals of the cage remain prominent only during the first hour ( Figure S18). The origin of this enhanced cage instability in the presence of cisplatin is presently under investigation. Moreover, the stability of the free cage C1 (0.15 mM injection concentration, 4 % DMSO in H 2 O) was also gauged by HR-ESI-MS. The resulting spectra showed the prominent presence of intact [C1(NO 3 ) n ] z + species ( Figure S19). Importantly, the presence of similar C1related species was also observed in 4 % DMSO in saline solution (up to 0.09 % NaCl) ( Figure S20).
Next, biodistribution of 18 F-L1, 18 F-C1 and cisplatin loaded 18 F-C1 was investigated in healthy mice using PET imaging in combination with computed tomography (CT). PET acquisitions were started immediately after administration of labelled compounds and dynamic scans were acquired for 60 min. Quantification analysis of PET images were performed only in those organs clearly visualized on CT images (brain, heart, lungs, liver, kidneys, and bladder). Visual inspection of PET images obtained over the first 10 min after administration ( Figure 2a) showed a very different profile for 18 F-L1, with presence of radioactivity in heart, lungs, intestines and gall bladder, with respect to cage 18 F-C1. The latter showed major accumulation in the liver and lower accumulation in the kidneys (both statistically significant in the time frame 0-4 min), as confirmed by image quantification (Figure 2b). Images also suggest lower accumulation of 18 F-C1 in the gall bladder, although no quantification was carried out in this vesicle as, due to its small size, results could be subjected to severe partial volume effect. The observed differences in biodistribution provide evidence that the species arising upon injection of the supramolecular structure is distinct from the free ligand. In contrast, the biodistribution profile for 18 F-L1 and 18 F-[C1(NO 3 ) 4 �cisplatin] showed more similarities. This result suggests that a possible disassembly of the cisplatin loaded 18 F-C1 may occur after in vivo injection, in accordance with the above-mentioned NMR studies. Accumulation of radioactivity in the different organs reached similar values at longer time points, irrespectively of the administered compound (see Figure S23 for representative PET-CT images). The lack of uptake in the bone confirms the absence of 18 F-defluorination of our 19 F-to-[ 18 F]trifluoroborate labelling strategy. After 60 min, ex vivo analysis based on dissection and gamma counting was carried out for 18 F-L1 and 18 F-C1 (Figure 2c). In this case, both compounds presented similar distribution pattern with major excretion through intestines (ca. 60 % of injected dose per gram of tissue -%ID/g -in the small intestine), further suggesting a possible non-negligible disassembly of 18 F-C1 to 18 F-L1. Analysis of the palladium levels in selected organs was performed by ICP-MS after 60 min upon injection of 18 F-C1 (Figure 2c) as another way of differentiating between the fates of the cage and the disassembled ligand. Noteworthy, Pd accumulation appears to be mostly uncoupled with respect to the ligand biodistribution. In fact, Pd was detected in spleen, kidney and liver, while being virtually absent in the intestine, where the ligand 18 F-L1 prominently accumulates (Figure 2c). This result is in line with the observed different biodistribution among the 18 F-L1 and 18 F-C1.

Conclusion
In conclusion, we have reported on the straightforward synthesis and purification of metallacages as possible drug delivery systems and studied their biodistribution in vivo by PET imaging. The obtained results show that the species arising upon cages injection accumulates in different organs with respect to their ligands in the early time points. Moreover, cisplatin encapsulation seems to favour cage disassembly in vivo, as suggested by PET. Certainly, further optimization of the ligand system, for example using more electron-donating tripyridyl ligands, [18] is necessary to enable kinetically robust cage complexes. Alternatively, the use of Pt(II) ions instead of Pd(II) could also increase the kinetic stability, although the selfassembly may require different reaction conditions not necessarily compatible with the radiolabelling procedure. 19 The cisplatin encapsulation process also requires in-depth investigation and should be performed on targeted and more hydrophilic cage systems. Overall, owing to their unique physicochemical properties, metal-coordinated supramolecular self-assemblies, including the selected metallacages, can bridge the boundary between traditional inorganic and organic materials, and our work further progresses their design for biomedical applications.

Experimental Section
Materials and methods: All commercially acquired reagents were used as received unless indicated otherwise. 2-azidoacetic acid, [20] 3,3'-((5-azido-1,3-phenylene)bis(ethyne-2,1-diyl))dipyridine (1), [9a] (3,5-bis(pyridin-3-ylethynyl)phenyl)methanol (2') [6b] and ((dimeth- yl(prop-2-yn-1-yl)ammonio)methyl)trifluoroborate (PPG-AMBF 3 ) [14c] were prepared according to literature procedures or with slight modifications. HPLC grade ethanol, methanol and acetonitrile were purchased from Scharlab (Sentmenat, Barcelona, Spain). Reactions requiring inert atmosphere were conducted under argon atmosphere using standard Schlenk line techniques. Thin layer chromatography (TLC) was performed using Merck plastic-backed plates of TLC Silica gel 60 F254; the plates were revealed using UV light at 254 nm or by staining using potassium permanganate. Standard Flash Column chromatography was accomplished using Merck silica gel (60 Å pore size, 70-230 μm mesh size). Automated Flash Column chromatography was performed by a Teledyne ISCO CombiFlash Rf200 system through pre-packed RediSep Rf silica gel columns. HRMS data were acquired on a X500B SCIEX QTOF highresolution mass spectrometer (ESI mode). Spectroscopic experiments for the characterization of compounds and encapsulation studies were carried out at the Structural Determination facility of IQS on a Varian 400 NMR spectrometer (400 MHz for 1 H, 100.5 MHz for 13 C, 376 MHz for 19 F and 128 MHz for 11 B). 195 Pt and 1 H DOSY experiments were performed at the NMR unit of Universitat de Barcelona on a Bruker Avance III 400 MHz spectrometer and at TUM on a Bruker Avance III 500 MHz spectrometer. Chemical shifts (δ Η ) are quoted in parts per million (ppm) and referenced to the appropriate NMR resonance, which for 1 H measurements would correspond to the residual portion component of the deuterated solvent. The 19 F and 11 B chemical shift are referenced relative to CFCl 3 and BF 3 ·Et 2 O resonance at 0.00 ppm, respectively. The 195 Pt chemical shift was referenced using an external reference of K 2 PtCl 4 in D 2 O (À 1610 ppm). Spin-spin coupling constants (J) are reported in Hertz (Hz). Infrared spectra were recorded on a Thermo Scientific Nicolet iS10 FTIR spectrophotometer equipped with Smart iTR window and are reported in cm À 1 . Mediterranean C18 column (4.6 × 150 mm, 5 μm) as stationary phase and 0.1 % TFA water/acetonitrile (0 min 10 % acetonitrile; 0-2 min 20 % acetonitrile; 2-10 min 70 % acetonitrile; 10-14 min 70 % acetonitrile; 14-16 min 10 % acetonitrile; 16-20 min 10 % acetonitrile) as mobile phase at a flow rate of 1 mL/min and wavelength of 254 nm. ICP-MS measurements were performed on a Thermo iCAP Q ICP-MS instrument.

Synthesis of metallacages C1 and C2
General procedure B: The corresponding AMBF 3 -containing ligand (0.1 mmol, 2 equiv.) and Pd(NO 3 ) 2 · 2H 2 O (12 mg, 0.05 mmol, 1 equiv.) were charged into a 50 mL conical centrifuge tube and dissolved in DMSO (4 mL). The mixture was allowed to stir at room temperature for 1 h. At this point, the product was precipitated by addition of acetone (4 mL) and diethyl ether (40 mL). The mixture was centrifugated, decanted and the obtained solid was washed with diethyl ether (3 × 5 mL). The product was dried under high vacuum.
Formation of 18 F-C1: Prepared following the procedure described above for non-labelled C1, with minor modifications. To a vial containing evaporated 18 F-L1, 40 μL of Pd(NO 3 ) 2 · 2H 2 O in DMSO (1 mg/mL, 0.15 μmol, 1 equiv.) were added and the mixture was stirred for 30 min at room temperature. Quality control was performed via analytical radio-HPLC (tr = 8.6 min, Figure S21, chromatogram b). Then, the resulting solution was diluted with 1 mL of saline for in vivo injections.
Encapsulation of cisplatin in 18 F-C1: Prepared following the procedure described above. To a vial containing evaporated 18 F-L1, 40 μL of Pd(NO 3 ) 2 · 2H 2 O in DMSO (1 mg/mL, 0.15 μmol, 1 equiv.) were added and the mixture was stirred for 30 min at room temperature. The resulting 18 F-C1 was added to a pre-loaded Eppendorf containing 800 μL of cisplatin in ultrapure water (62.5 μg/mL, 2 equiv.). The reaction mixture was incubated during 5 min. Quality control was performed via analytical radio-HPLC (tr = 8.6 min, Figure S21, chromatogram c). Afterwards, 160 μL of saline were added to reach a final volume of 1 mL for in vivo injections.

In vivo and ex vivo biodistribution studies
Animals: Female mice (BALB/cJRj, 8 weeks, Janvier; 9 animals) weighing 22 � 2 g were used to conduct the biodistribution studies. The animals were maintained and handled in accordance with the Guidelines for Accommodation and Care of Animals (European Convention for the Protection of Vertebrate Animals Used for Experimental and Other Scientific Purposes) and internal guidelines. All experimental procedures were approved by the internal committee and the local authorities.
Biodistribution studies: Mice were anesthetized by inhalation of 3 % isoflurane in pure O 2 and maintained by 1.5-2 % isoflurane in 100 % O 2 . With the animal under anesthesia, 18 F-L1, 18 F-C1 or cisplatin loaded 18 F-C1 were injected intravenously via one of the lateral tail's veins (110 μL, 1.48 � 0.74 MBq, n = 3 per compound). Dynamic whole body 60-min PET scans were started immediately after administration using MOLECUBES β-CUBE (PET) scanner. After each PET scan, whole-body high-resolution CT acquisitions were performed on the MOLECUBES X-CUBE (CT) scanner to provide anatomical information of each animal as well as the attenuation map for later image reconstruction. Random and scatter corrections were automatically applied during image reconstruction (3D OSEM reconstruction algorithm). PET-CT images of the same mouse were co-registered and analyzed using the PMOD image processing tool. Volumes of interest (VOIs) were manually delineated on selected organs (brain, heart, lungs, liver, kidneys, and bladder). Time-activity curves (decay corrected) were obtained as cps/cm 3 in each organ. Curves were transformed into real activity (Bq/cm 3 ), and finally injected dose normalization was applied to express the results as percentage of injected dose per cm 3 of tissue (% ID/cm 3 ).
Ex vivo studies: After the imaging session, animals were sacrificed, organs of interest were collected and weighed, and the radioactivity was measured in a gamma-counter (Wallach Wizard, PerkinElmer, Waltham, MA, USA). The uptake was calculated as a percentage of the injected dose per gram of tissue (% ID/g). Then, the weighted organs were immersed in digest solution of HNO 3 /HCl (4 : 1, 5 mL) and heated to boiling until complete dissolution. The solution was subsequently analyzed by ICP-MS to determine the concentration of Pd in each sample.