Design and Synthesis of Gd(III) Chelates

An optimal design for a versatile PET/MRI contrast agent should aim to closely replicate the properties of clinically approved macrocyclic MRI contrast agents, such as high kinetic inertness, high relaxivity, low molecular weight, and fast renal clearance. Equally important is carefully selecting the positron-emitting radionuclide to be incorporated into the molecule for optimal PET detection. Among viable positron-emitting radionuclides, ¹⁸F emerges as the most suitable candidate because (i) it offers minimal molecular alteration, (ii) it has a β⁺ decay half-life of 110 minutes, which is in line with the pharmacokinetics of standard GBCAs, (iii) it has low positron energy, which is optimal for high PET resolution, and (iv) it is widely available and can be produced efficiently and cost-effectively in medical cyclotrons worldwide.²² Finally, the radionuclide should be incorporated into the chelator molecule during the final synthetic step to minimize losses due to decay. This requires direct radiolabeling of a gadolinium chelate, which differs significantly from the conventional organic substrates typically used in PET imaging.

To meet all these requirements in a single molecular design, we hypothesized that a pyridine moiety, known for its strong donor properties, could replace one of the acetate arms in DOTA.²³ This modification would provide an ideal platform for subsequent derivatization, as the pyridine ring is amenable to nucleophilic aromatic substitution,^{24, 25} allowing the substitution of appropriate leaving groups by fluoride anions. Although these reactions have been described for purely organic substrates, their adaptation for successful radiolabeling gadolinium chelates required extensive optimization of structural design and reaction conditions. Firstly, choosing suitable leaving groups was narrowed down to nitro- and fluoro-derivatives. While substitution of the nitro-group by fluoride anion has been reported,²⁶ we explored the possibility of ¹⁹F−¹⁸F isotopic exchange for the first time in gadolinium chelates (Figure 1). Secondly, the coordinated pyridine arm provides multiple positions for placing these substituents, with their reactivity influenced by both first-order factors (electronic activation/deactivation) and second-order effects (coordination/steric hindrance). By thoroughly exploring all positional isomers, we found that only the para-position enables efficient radiolabeling through a combination of strong activation and favorable steric accessibility (Table 1, Table S5, Figures S11, S13). Although the ortho-position is expected to be highly activated, radiolabeling did not occur due to steric hindrance caused by coordination (Table S5, Figure S16).

The synthetic pathways leading to Gd(III) chelates of the best-working para-nitro- and para-fluoro-derivatives are outlined in Scheme 1. The synthesis of the other tested isomers is described in the Supporting Information. For [(Gd(NO₂L¹)], the synthesis begins with the commercially available 2-methyl-4-nitropyridine 1-oxide (1), which was derivatized into the hydroxomethyl-derivative 2 via the Boekelheide rearrangement in trifluoroacetic acid anhydrate (TFAA).²⁷ This compound was then converted to the chloromethyl-derivative 3 in SOCl₂. Following this, 3 was used for the alkylation of tBu₃DO3A, generating the protected chelator 4. Deprotection in trifluoroacetic acid (TFA) yielded NO₂L¹. The subsequent complexation of GdCl₃ in the MOPS buffer, followed by purification through RP-HPLC, generated [Gd(NO₂L¹)] with an overall yield of 45 % relative to tBu₃DO3A (Scheme 1a). The synthesis of [Gd(FL¹)] started with the reduction of methyl 4-fluoro picolinate (5) using NaBH₄, producing alcohol 6. This was then converted into the corresponding mesylate 7. The subsequent steps mirrored those of the nitro-derivative, providing the [Gd(FL¹)] product with an overall yield of 34 % relative to tBu₃DO3A (Scheme 1b). Importantly, it should be noted that the chelator FL¹ exhibits only moderate stability in acidic aqueous solutions; therefore, we used pH-neutral stock solutions for the subsequent steps.

Radiolabeling with ¹⁸F

The reaction conditions for the nitro-to-fluoro substitution were initially tested using the non-radioactive tetrabutylammonium fluoride (TBAF), adopting conditions previously used for purely organic pyridine substrates,²⁶ modified to be compatible with the chelate [(Gd(NO₂L¹)]. Specifically, we used DMSO as a solvent to guarantee complete solubility of the Gd(III) chelate and excess TBAF to drive the reaction to completeness. The fact that the conversion to [Gd(FL¹)] can be easily observed on HPLC (Figure S9) has facilitated this work. Next, we explored and optimized two methods for preparing fresh [¹⁸F]TBAF, a radiolabeling reagent whose quality proved crucial for successful radiolabeling. Both methods start from [¹⁸F]HF, the primary product of the cyclotron production of the ¹⁸F isotope. In Method A, the [¹⁸F]HF solution was loaded onto a QMA (quaternary methyl ammonium) SPE (solid-phase extraction) cartridge preconditioned with NaHCO₃. The aforementioned [¹⁸F]TBAF was generated from eluting the immobilized [¹⁸F]fluoride from the QMA SPE with a solution of tetrabutylammonium bicarbonate (TBAHCO₃), diluted with an equal volume of acetonitrile. The solution containing [¹⁸F]TBAF was then evaporated to dryness under a stream of argon. For Method B, the QMA SPE cartridge was pre-treated with KOTf²⁸ (potassium trifluoromethylsulfonate), loading the [¹⁸F]HF solution, washing the resin with dry MeOH, and drying it under an argon stream for 3 minutes. Elution of [¹⁸F]TBAF was conducted with TBAOTf (tetrabutylammonium trifluoromethylsulfonate) in dry MeOH, followed by drying at 100 °C for 5 minutes under an argon stream. Reaction conditions for the conversion of [(Gd(NO₂L¹)] to [¹⁸F][Gd(FL¹)] were tested manually using aliquots of thoroughly dried [¹⁸F]TBAF and small solvent volumes. Extensive screening (Table S3) found an optimum at 90 °C and a reaction time of 5 minutes with 3 eq. of [¹⁹F]TBAF. Notably, [¹⁹F]TBAF serves the dual purpose of enhancing the chemical conversion and generating the non-radioactive product required for the PET/MRI formulation applications—high specific activity is not necessary in this instance.

The [¹⁸F]TBAF produced through Method B demonstrated superior results, with a radiochemical yield (RCY) of 17 % and a purity of 87 %, surpassing the results obtained by Method A, which yielded an RCY of 11 % with a purity of 74 % (Table S3). Interestingly, using [Gd(FL¹)] as the substrate in the isotopic exchange ¹⁸F/¹⁹F reaction under identical conditions resulted in further purity improvement (Table 1). This isotopic exchange has been previously reported for simple fluoro-pyridine compounds^{26, 29, 30} but never for metal chelates. Furthermore, it is worth noting that our attempts to label the free ligand FL¹ with [¹⁸F]TBAF were unsuccessful (Figure S17, Table S5). This could be attributed to the interaction of the [¹⁸F]F⁻ anion with the protonated tertiary amines of the macrocyclic ligand, rendering the anion unavailable for reaction. Data suggest that the presence of the chelated metal ion in [Gd(FL¹)] is crucial for successful radiolabeling (Table 1).

Scaling up the [¹⁸F][Gd(FL¹)] synthesis to higher activities for imaging experiments required the use of a custom-built automated synthesis module (GE). Starting from 16 GBq of [¹⁸F]HF, the automated synthesis yielded 1340 MBq of the final product (8 % RCY decay uncorrected) in 32 minutes, exhibiting 97 % chemical and 100 % radiochemical purity (Figure 2). For in vitro and in vivo PET/MRI experiments, the product was further combined with non-radioactive [Gd(FL¹)] to adjust the activity/Gd ratio to levels compatible with both modalities. The concerted efforts in rational design, precursor screening, and radiolabeling optimization provided scalable synthesis as a prerequisite to the practical applicability of the hybrid PET/MRI agent. Notably, this first reported case of directly labeling a gadolinium chelate with ¹⁸F emphasizes the active role of the chelated metal in enabling radiolabeling and directing regioselectivity.

In Vitro Physicochemical Measurements

The relaxivity of GBCAs (r₁) depends on several physicochemical parameters.^{31, 32} The number of bound water ligands (hydration, q) was assessed using the ¹⁷O NMR lanthanide-induced shift method with Dy(III) complexes.³³ This approach confirmed the coordination of a single water molecule in solution (Figure S1. The exchange rate (k_ex) of the bound water with the bulk in solution was determined by temperature-dependent ¹⁷O NMR relaxometric studies (Figure S2, 2). A value of k_ex=1.7×10⁶ s⁻¹ was found, approximately 2.4 times slower than that of [Gd(DOTA)].³⁴ This observation is consistent with the slightly lower longitudinal relaxivity of 3.43 mM⁻¹ s⁻¹ (0.47 T, 37 °C) for [Gd(FL¹)]. However, it should be noted that this value falls well within the typical range of 3–4 mM⁻¹ s⁻¹ observed in clinically approved MRI contrast agents (Table 2).³⁵

Another critical factor for potential clinical use is the kinetic inertness of the metal chelate. Macrocyclic Gd(III) chelates, particularly those of the DOTA type, exhibit significantly higher inertness when compared to their acyclic counterparts.^{32, 37, 38} We have used an acid-assisted decomplexation method in 0.1 M HCl at 37 °C to monitor the kinetic inertness of [Gd(FL¹)] in solution.³⁹ Under these conditions, the chelate breaks down, releasing a fully hydrated Gd(III) ion into the solution, which has a higher r₁ relaxivity. The overall mono-exponential increase in relaxation rate (R₁) can be described by the half-life (t_1/2, Figure S3) as a measure of the dechelation rate. We observed that while the acyclic clinical MRI contrast agent [Gd(DTPA)] dissociated rapidly (t_1/2<6 s), [Gd(FL¹)] showed a remarkable kinetic inertness with a dissociation half-life of 45 hours, comparable to the clinically approved [Gd(DOTA)] (Table 2).

In Vitro PET/MRI Phantom Imaging

Following the successful radiolabeling and comprehensive in vitro analysis of [Gd(FL¹)], we designed a phantom imaging study for PET/MRI examination. We prepared a formulation containing [¹⁸F][Gd(FL¹)] from automated radiosynthesis and a 0.5 mM stock solution of [Gd(FL¹)] in PBS, to provide a reference solution of known molar activity and chemical concentration. This mixture was aliquoted into test tubes to create an array of samples with concentrations ranging from 0.048 to 0.489 mM (quantified by ICP-OES analysis) (Figure 3a). The quantification of PET/MR images acquired in a 7T preclinical scanner exhibited a remarkable linear correlation between PET-measured radioactivity and signal derived from T₁-weighted MR images (R²=0.99, Figure 3b). This correlation was mirrored by the relaxation rate derived from MRI T₁-maps (R²=0.99, Figure 3c). The relationship was further validated by quantifying ¹⁸F activity using gamma counting and Gd(III) via ICP-OES (R²=0.99, Figure S21). These findings show the potential of [¹⁸F][Gd(FL¹)] to provide quantitative PET/MRI.

In Vivo PET/MRI Studies in Mice

The in vivo performance of [Gd(FL¹)] as a hybrid PET/MRI contrast agent was investigated in healthy C57BL/6 mice (n=6). A formulation of [Gd(FL¹)] and its radiolabeled counterpart [¹⁸F][Gd(FL¹)] in saline was administered to anesthetized mice via a single tail vein bolus injection. The administered dose was 0.09±0.01 mmol/kg of Gd, equivalent to 0.91±0.07 MBq. Two cohorts of three mice each underwent different imaging protocols (Scheme S1). In the first group, we conducted a continuous 62-minute session of simultaneous dynamic contrast-enhanced (DCE) MRI and dynamic PET scanning. The second group underwent 11 sequential T₁ maps acquired over 55 minutes, followed by a 4.6-minute static PET scan. After completing the imaging procedures, all animals were euthanized 62 minutes post-injection, and organs were collected for ex vivo analysis. The [¹⁸F][Gd(FL¹)] agent exhibited a consistent pharmacokinetic profile in both MRI (Figure 4a) and PET (Figure 4b) modalities, demonstrating swift systemic distribution and rapid blood clearance through the kidneys into the bladder—a common pathway for most non-targeted Gd-based agents. The agent exhibited minimal liver uptake and blood retention, indicating negligible hepatobiliary excretion (Figure 4c).

The representative PET/MRI images (Figure 4d) clearly showed immediate enhancement of the renal cortex and medulla at the time of injection, followed by a rapid transit of the contrast agent to the renal pelvis within 10 minutes and its almost complete elimination to the bladder within 60 minutes (Figure 4b). PET/MRI showed no tissue retention of [¹⁸F][Gd(FL¹)]. While PET and MRI signal enhancement generally correlated perfectly, an intriguing discrepancy was noted in the pelvis of healthy kidneys at 10 minutes post-injection (Figure 4e). Here, the freshly formed urine had a high local concentration of the contrast agent, resulting in hypo-intense regions on the MRI due to potent T₂ and T₂* relaxation mechanisms that quenched the signal in T₁-weighted contrast.⁴⁰ This highlights the complex non-linear relationship between GBCA concentration and its impact on the MRI signal, making it challenging to quantify in vivo MRI-based GBCA reliably. However, the hybrid [¹⁸F][Gd(FL¹)] allows an independent and quantitative evaluation of the local concentration of the contrast agent through the PET signal. This was particularly evident as the hypo-intense areas observed on the MRI correlated perfectly with the regions of high PET signal (Figure 4e).

Kidney tissue from the healthy mice was sectioned and further analyzed by autoradiography (for ¹⁸F) and laser ablation ICP-MS (for Gd). As previously reported, the combination of laser ablation-inductively coupled plasma mass spectrometry (LA-ICP-MS) and autoradiography provides complementary measurements, offering detailed insights into both metal distributions and radioactivity.⁴¹ Both methods confirmed the observed in vivo distribution of [Gd(FL¹)], showing the lowest concentration of the contrast agent in the cortex and the highest concentration in the pelvis (Figure 4f). Furthermore, all tissue samples from euthanized animals were analyzed to quantify radioactivity using a gamma counter and the gadolinium content using ICP-OES (Figure 4g–h). The biodistribution pattern remained consistent across both methods, revealing the highest concentration of [Gd(FL¹)] in the urine and bladder, followed by the kidneys. These biodistribution patterns are consistent with the study by Le Fur et al. where [⁸⁶Y][Y(DOTA)] was used to track [Gd(DOTA)] biodistribution.⁴² All other tissues had significantly lower [Gd(FL¹)] contents. Importantly, radioactivity in the femur was low (0.86±0.12 %ID/g), suggesting no significant release of the bone-seeking [¹⁸F]F⁻ anion from [¹⁸F][Gd(FL¹)].⁴³ Tissue concentrations determined by gamma counting and ICP-OES covered several orders of magnitude. A back-to-back comparison on a logarithmic scale revealed an excellent correlation (R²=0.96), underlining the robustness and stability of the ¹⁸F/Gd ratio across various concentrations and tissue types (Figure 4i). Furthermore, we performed an LC–MS analysis of a urine sample, which confirmed that the agent was excreted intact into the bladder, indicating an optimal in vivo pharmacokinetic and pharmacodynamic profile (Figure S8).

PET/MRI Renography

The renal system has a robust compensatory mechanism that effectively compensates one kidney for the dysfunction of the other kidney. However, detecting kidney disease early is challenging. The standard diagnostic test, which measures serum creatinine concentration, only indicates a significant stage of renal dysfunction⁴⁴ and is not sensitive to unilateral or partial kidney disease.^{45, 40} Imaging techniques provide more nuanced information but present specific limitations. Planar SPECT renography is obscured by organ overlap, while SPECT/CT and PET/CT have significant exposure to ionizing radiation.⁴⁴ Thus, MRI using renally excreted GBCAs emerges as an attractive alternative despite its limitations in quantification.⁴⁰ While imaging is not typically employed for early diagnosis, it enables radiologists to identify unforeseen conditions when conducted for any other primary reason. Hence, PET/MRI using [¹⁸F][Gd(FL¹)] emerges as a potent and versatile tool for studying renal function due to its quantitative nature.

During the in vivo studies, we identified abnormal renal filtration profiles in two mice, M1 and M6, that underwent two different PET/MRI protocols (details in Scheme S1). Both mice presented unilateral retention of [¹⁸F][Gd(FL¹)] in the right kidney. For mouse M6, this was evident on the static PET image at 55 min post-injection, revealing a strong and persistent signal in the right renal pelvis, whereas no probe accumulation was observed in the left kidney (Figure 5a). The whole PET signal in the right kidney was approximately 10-fold higher than in the left (Figure S24c). Conversely, T₁-weighted MR images were less conclusive, showing hypo-intense regions in the right kidney due to the strong T₂ and T₂* effects (Figure 5a, left image), but no substantial difference between the kidneys in the mean relaxation rate (Figure S24d). Thus, relying solely on MRI data might result in overlooking evidence of abnormal function in the right kidney. However, by leaveraging PET as a reporter, we were able to quantify the signal and perform a detailed ex vivo analysis of tissue sections with autoradiography (¹⁸F) and laser ablation ICP-MS (Gd). This approach confirmed the heterogeneous accumulation of [¹⁸F][Gd(FL¹)] in the right kidney compared to the left (Figure 5b–c). Histology revealed dilated tubules and luminal proteinaceous material in the right kidney, confirming the renal pathological dysfunction (Figure 5d).

In the second detected abnormal case (mouse M1), the imaging protocol included a 62-minute simultaneous PET/MRI acquisition (Figure 6), allowing us to follow the dynamic behavior of the probe and evaluate the renal function in detail. The left kidney and the kidneys of two other mice from the same cohort (M2, M3) served as references for normal renal function. We used an innovative voxel-based analysis of co-registered images from both modalities to calculate probe wash-out parameters (Figure 6b–e). The resulting parametric maps derived independently from the MRI and PET data showed heterogeneous clearance patterns within the abnormal right kidney, which correlated well with the hot-spot areas of impaired function (Figure 6d–e). While these areas showed signal loss on MRI, incorrectly suggesting rapid probe wash-out (Figure 6b–d), the areas showed substantial probe accumulation in the PET imaging (Figure 6c–e). In a further voxel-based analysis, we calculated the area under the curve (AUC) and compared two regions of interest (ROI) in the impaired kidney, one with a highly abnormal excretion pattern and the other with a near-normal excretion pattern. The near-normal region was characterized by a narrow distribution of the AUC values, whereas the abnormal ROI showed a broad distribution in both MRI- and PET-derived data (Figure 6f–g). Compared to the abnormal case, all healthy kidneys generally showed relatively homogeneous excretion patterns in both modalities (Figures S26–27).

Overall, PET/MRI with the hybrid contrast agent [¹⁸F][Gd(FL¹)] offers significant advantages for the assessment of renal function, with both modalities providing crucial diagnostic value. The PET channel quantifies the agent independently of the wide filtration dynamics in the renal tissue, allowing for the precise detection of abnormal excretion patterns. On the other hand, MRI provides anatomical context and detailed spatial localization. As shown in Figure 6, MRI alone can be confounding due to negative contrast enhancement, whereas PET alone may lack the resolution to discern individual hot spots of impaired tissue function. To the best of our knowledge, this is the first report using such PET/MRI analysis obtained with a hybrid PET/MRI probe.