Synthesis and in vivo evaluation of a radiofluorinated ketone body derivative

Design, synthesis, and preliminary validation of the first radiofluorinated ketone body derivative as a PET imaging agent for the study of ketone body metabolism in cancer.

Abstract

The ketone bodies d-beta-hydroxybutyric acid and acetoacetic acid represent the principal oxidative energy sources of most tissues when dietary glucose is scarce. An ¹⁸F-labeled ketone body could be a useful tool for studying ketone body metabolism using positron emission tomography (PET). Here, we report the first radiofluorinated ketone body derivative (3S)-4-[¹⁸F]fluoro-3-hydroxybutyric acid ([¹⁸F]FBHB) as well as its enantiomer and l-beta-hydroxybutyric acid derivative, (3R)-4-[¹⁸F]fluoro-3-hydroxybutyric acid ((R)-[¹⁸F]F3HB). PET imaging in mice showed biodistribution profiles of the radiotracers that were consistent with the biodistribution of the respective endogenous compounds. Moreover, both enantiomers visualized breast cancer xenografts in vivo. Fasting over 24 h showed significantly enhanced brain and heart uptake of [¹⁸F]FBHB and tumor uptake of (R)-[¹⁸F]F3HB. Disorders exhibiting altered energy substrate utilization, such as Alzheimer's disease, epilepsy, diabetes, and cancer may be of interest for PET imaging studies using [¹⁸F]FBHB.

Introduction

Ketosis is a metabolic state induced by fasting or carbohydrate restriction in which scarcity of dietary glucose and a consequent reduction of insulin secretion result in hepatic production of the ketone bodies d-beta-hydroxybutyric acid (d-BHB) and acetoacetic acid (AcAc) as alternative energy sources for the brain, heart, and voluntary muscles (Fig. 1).‡ These acetyl-CoA-derived molecules enter circulation and are converted back to Krebs cycle-initiator acetyl-CoA in extrahepatic tissue.1,2 They have a biodistribution profile in mammalian organs similar to that of glucose, which feeds into acetyl-CoA via glycolysis, and are therefore largely equivalent to glucose as sources of oxidative energy.

Fig. 1. Biochemical pathways of ketolysis (red: d-BHB → acetyl-CoA) and ketogenesis (blue: acetylCoA → d-BHB) with enzymes (grey): BDH, d-beta-hydroxybutyrate dehydrogenase; HMG-CoA, β-hydroxy β-methylglutaryl-CoA; SCOT, succinyl-CoA-3-oxaloacid CoA transferase.

Cancer cells, although metabolically inhomogeneous,3 derive cellular energy in large part through the fermentation of glucose to lactic acid, even when oxygen is available (Warburg effect).4,5 This glycolytic phenotype is the basis for positron emission tomography (PET) imaging of malignancies using the glucose derivative 2-deoxy-2-(¹⁸F)fluoro-d-glucose ([¹⁸F]FDG). Labeled with positron emitter ¹⁸F, [¹⁸F]FDG is widely used for diagnosing, staging, and treatment monitoring of several types of cancer6 and has enabled numerous in vitro and preclinical investigations elucidating the role of glucose in cancer metabolism.

The significance of ketone body metabolism in cancer is under-examined. There is some evidence from in vitro and in vivo studies that ketone bodies and ketosis have antineoplastic effects. For example, the addition of ketone bodies to human colon, breast, and pancreatic cancer cells in vitro has demonstrated growth inhibition.7,8 And whereas d-BHB addition prolonged survival of healthy neurons under glucose restriction in vitro, glioma cells did not survive the same conditions.9 Furthermore, a ketogenic diet, which (among other systemic effects) raises levels of circulating ketone bodies, has been shown in mouse studies10,11 and a few human trials12,13 to inhibit tumor growth and have anticachectic effects. Nevertheless, there are direct challenges to the characterization of ketone bodies as cancer growth suppressors,14,15 and evidence for the efficacy of the ketogenic diet as a cancer therapy from human trials is limited and inconsistent.16

In order to assess the mechanism and efficacy of therapies based on endogenous or exogenous ketone bodies there is a need for clinical tools that help to determine which tumors take up ketone bodies in vivo, the extent of this uptake, and to correlate this uptake with cell growth/inhibition. PET is a convenient, minimally invasive tool for this type of investigation.

To date, PET studies of ketone body metabolism in humans and animals have relied on [¹¹C]AcAc and [¹¹C]d-BHB.1,17–19 These ¹¹C radiotracers enable the differentiation of oxidative-catabolic versus anabolic metabolism of ketone bodies in vitro or in vivo by monitoring for release of [¹¹C]CO₂.20,21 However, the drawbacks of the use of the ¹¹C radioisotope include a short half-life (t_1/2 = 20 min) and an incapacity for metabolic intracellular trapping. ¹⁸F is better suited for many diagnostic imaging applications due to its lower positron energy (β⁺_max = 634 keV) and convenient half-life (t_1/2 = 109.8 min) that enable multi-hour scanning protocols with excellent spatial resolution. Moreover, derivatization of metabolites with ¹⁸F can, as in the case of [¹⁸F]FDG, facilitate desirable metabolic trapping.

In the present work we describe the design, synthesis and preliminary validation of the first radiofluorinated ketone body derivative as a PET imaging agent and its potential to study ketone body metabolism in cancer.

Results and discussion

The C-2 methylene group of AcAc and the C-4 methyl group of d-BHB (Fig. 2A) were considered to be reasonable sites for ¹⁸F incorporation because these modifications preserve the carboxylic acid functionality necessary for recognition by the monocarboxylate transporters (MCTs) that, along with diffusion, are involved in ketone body cell entry. Likewise, the beta-keto/hydroxy motifs would be left unaltered to potentiate interactions with ketolytic enzymes. [¹⁸F]1 and [¹⁸F]FBHB (Fig. 2C) were therefore selected as target radiofluorinated derivatives of AcAc and BHB, respectively.

Fig. 2. (A) Endogenous metabolites relevant to this work: AcAc, d-BHB, and l-3-hydroxybutyric acid (l-3HB); (B) ¹⁹F reference compounds; (C) target ¹⁸F ketone bodies and initial (unproductive) precursors.

To facilitate confirmation of the identities of [¹⁸F]1 and [¹⁸F]FBHB, we developed syntheses for reference compounds 1 (chemical half-life ≈90 h at pH 7.4) and racemate 2 (Fig. 2B).22

As a potential route to [¹⁸F]1 (refer to retrosyntheses in Fig. 2C) ester derivatives of AcAc were functionalized with a bromide or tosylate leaving group installed at C2 (compounds 3 and 4). Unfortunately, bromo esters 3, under standard radiofluorination conditions with [¹⁸F]KF/K222 gave labeled side products rather than the desired esters of [¹⁸F]1. Tosylate 4 also resisted ¹⁸F incorporation under standard reaction conditions (K222, K₂CO₃) as well as less basic conditions (NBu₄HCO₃). It is possible that alpha proton abstraction by [¹⁸F]F^– is favored over nucleophilic substitution in this case. Otherwise, the energy barrier to S_N2 at C2 could be the limiting factor; radiofluorinations of secondary aliphatic carbons (e.g. [¹⁸F]FDG, [¹⁸F]FLT) usually employ better leaving groups like triflates or nosylates.23 Nevertheless, increasing the inductive properties of a leaving group on a precursor such as 4 could exacerbate a competing acid–base reaction.

Compound [¹⁸F]1 appeared to be an appealing target based on the relative ease of preparation of precursors like 3 and 4via functionalization of the active methylene of AcAc. Unfortunately, the electronic and steric properties at C2 of precursors like 3 and 4 render them unsuitable for traditional nucleophilic radiofluorination. Nevertheless, [¹⁸F]1 remains an interesting target given its alpha carbon labeling position. A synthetic route incorporating an electrophilic fluorinating reagent (e.g. [¹⁸F]Selectfluor,24 prepared from [¹⁸F]F₂) would likely be a more suitable approach for future attempts.

The first efforts to synthesize labeling target [¹⁸F]FBHB (Fig. 2C) centered on ring opening radiofluorination. At first, a precursor containing a cyclic sulfate (as in compound 5) seemed promising because this dual-purpose leaving/protecting group has been used to achieve [¹⁸F]fluorohydrins in syntheses of [¹⁸F]FDG25 and [¹⁸F]fluoroestradiol.26 Unfortunately, after preparing and isolating compound 5,27 it polymerized upon solvent removal.22 The next approach was directed to potential epoxide precursors. 6a (ref. 28) and 6b (ref. 22) (Fig. 2C) were prepared as benzyl and p-methoxybenzyl esters, respectively, so that labeling intermediates could be compared with UV-detectable reference standards by HPLC. Yet precursors 6 showed no ¹⁸F incorporation and appeared, by radio-TLC analysis, to break down under the heating (80–100 °C) and slightly basic conditions of radiofluorination using either the convenient methodology (K222, K₂CO₃) or milder (NBu₄HCO₃) conditions. Base sensitivity and isomerization is documented for epoxyesters like 6,29 but the reported synthesis30 of a radiofluorinated lactate derivative from epoxyester 7 (Fig. 3A), which differs from precursor 6a by a single methylene carbon, had inspired the attempt. Encouragingly, a small amount of ¹⁸F incorporation with compounds 6a and 6b (Fig. 2C) was achieved when applying an enantioselective epoxide opening protocol disclosed by Graham et al.31 that involved pre-formation of a (salen)Co[¹⁸F]F species (similar to a Jacobsen epoxidation catalyst). An example from this work is depicted in Fig. 3B. Maximum incorporations of ¹⁸F for epoxides 6a and 6b of 10% and 12% were achieved, respectively, as estimated by radio-TLC analysis, and the labeling intermediates were confirmed by radio-HPLC by comparison with the corresponding ¹⁹F reference standards (benzylic esters) prepared in our previous work.22 Nevertheless, these reactions were less efficient than the reported decay-corrected radiochemical yield (RCY) of 23% for naphthyl ester [¹⁸F]10 (Fig. 3B) prepared by this protocol,31 a difference that could result from our choice of more labile carboxylate protecting groups. It is worth noting that Verhoog et al. recently reported a strategy for epoxide radiofluorination using [¹⁸F]FeF derived from [¹⁸F]HF.32 The method tends to incorporate ¹⁸F at the more substituted side of the epoxide and thus was not applied to the present study.

Fig. 3. Reported syntheses relevant to this work. (A) Synthesis of lactate derivative [¹⁸F]-FLac [¹⁸F]8;30 (B) selected synthesis from an enantioselective ring opening radiofluorination;31 (C) preparation of [¹¹C]-beta-hydroxybutyric acid [¹¹C]13.33.

Thorell and coauthors were the first to report a synthesis for [¹¹C]-beta-hydroxybutyric acid (compound [¹¹C]13, Fig. 3C).33 This was accomplished in three steps from chiral propylene oxide 11 and proceeded through nitrile intermediate [¹¹C]12. Applying that approach to the present work, we explored the use of (S)-[¹⁸F]epifluorohydrin ([¹⁸F]15, Scheme 1A) as a potential building block for a three-step synthesis of compound [¹⁸F]FBHB. Gratifyingly, this synthetic approach afforded [¹⁸F]FBHB in sufficient quantity for preliminary biological evaluations using PET. (2S)-(+)-Glycidyl tosylate 14 was converted to (2S)-[¹⁸F]epifluorohydrin [¹⁸F]15 analogously to published methods.34 This volatile intermediate (bp = 85 °C) was not isolated but was reacted directly with cyanide in a ring opening reaction to form nitrile [¹⁸F]16. Following HPLC purification of [¹⁸F]16, hydrolysis was achieved by enzymatic conversion with E. coli-derived nitrilase, which catalyzes the conversion of nitriles to carboxylic acids.35 We departed from the acid-based deprotection applied by Thorell et al. in their synthesis of [¹¹C]13 (Fig. 3C) in order to avoid the potentials of dehydration or partial racemization of the target compound.

Scheme 1. Synthesis of (A) [¹⁸F]FBHB and (B) its enantiomer, (R)-[¹⁸F]F3HB from (2S)-(+)- or (2R)-(–)-glycidyl tosylate, respectively. RCY = decay-corrected radiochemical yield. Reagents and conditions: (a) [¹⁸F]KF/K222, CH₃CN, 90 °C, 25 min; (b) KCN aq., CH₃CN, 90 °C, 15–20 min, two-step one-pot RCY estimated by radio-TLC: 85%; (c) E. coli nitrilase, H₂O, Tris, pH 9, 30 °C, 60 min, isolated RCY (step c): 10% (for [¹⁸F]FBHB), 35% (for (R)-[¹⁸F]F3HB). Total RCY from cyclotron-generated ¹⁸F to isolated injectable solution: 1.3% ([¹⁸F]FBHB), 13% ((R)-[¹⁸F]F3HB). Total synthesis time: 4 h.

Given the lack of UV active groups on the labeled nitrile and final product acid, the identities of each species were confirmed with HPLC using a mixture of purified labeled and unlabeled reference compounds by collecting fractions of eluent and verifying by TLC that the fractions comprising the radiopeak coincided with the appearance of a stain (KMnO₄) with the retention factor characteristic of the reference standards.

Molar activity could not be determined due to the lack of UV absorbance for the cold standard acid. Even so, molar activity is of little relevance for a tracer such as [¹⁸F]FBHB. Transporter saturation is immaterial since MCTs are ubiquitous and transport is passive and unsaturable (in at least brain tissue) over a wide range of physiologically normal ketone body concentrations.1

The three-step approach to give [¹⁸F]FBHB involving radiofluorination, alkylation, and hydrolysis afforded several unforeseen advantages. First, it obviated the need for intricate labeling precursors. All of the reagents for this method are commercially available, and the use of [¹⁸F]epifluorohydrin as a labeling prosthetic group is well-established, having been applied, for example, to the synthesis of hypoxia imaging agent [¹⁸F]FMISO.36 Second, rigorous stereocontrols were avoided given the availability of enantiopure starting material. Third, the approach enabled production of two different radiotracers, the ketone body relevant isomer, [¹⁸F]FBHB, and its enantiomer, (R)-[¹⁸F]F3HB, (Scheme 1A and B) depending on the isomer of starting material employed (14versus17). A comparison of these tracers would give insight into the role that the C-3 stereocenter plays in their in vivo distribution and utilization. Lastly, the use of an enzyme in the final step of the reaction sequence effectively functioned as a quality control; by monitoring the reaction of nitrile intermediates [¹⁸F]16 and [¹⁸F]19 with nitrilase it was possible to indirectly confirm 1) the presence of nitrile functionality and, judging by the rate of conversion, 2) non-racemization at the chiral carbon, by virtue of the noticeable stereopreference37,38 of the commercial enzyme for R nitrile [¹⁸F]19.

Although not a primary labeling target of this work, compound (R)-[¹⁸F]F3HB (Scheme 1B) was convenient to produce and merits independent study; this stereoisomer could be viewed as a radiofluorinated form of l-3-hydroxybutyric acid (l-3HB, Fig. 2A), a metabolite found in small quantities in humans that is thought to be produced by the liver or cardiomyocytes39 and shunted to lipid/sterol synthesis pathways of the nervous system.40

Ketone body derivative [¹⁸F]FBHB and, for comparison, enantiomer (R)-[¹⁸F]F3HB were first assessed by PET for their uptake and clearance profiles in normal BALB/c mice. Dynamic PET scans over 60 min revealed an overall uptake profile for [¹⁸F]FBHB consistent with that of the structurally related ketone body d-BHB, including moderate heart (SUV_mean,60min 1.23) and brain (SUV_mean,60min 1.48) uptake and a renal clearance pathway (see ESI,† Fig. S1 and S2). The uptake pattern in the brain, in particular, are suggestive of some form of metabolic trapping. The enantiomer (R)-[¹⁸F]F3HB displayed similar renal clearance but a distinct uptake pattern (ESI,† Fig. S1 and S2), notably demonstrating reduced retention (Δ17%) in the heart (blood pool) and considerably less in the brain (Δ33%) 60 min post injection. Also particular to the R isomer was prominent joint and spine uptake which will require further investigation to explain. Interestingly, in a study in rats with ¹⁴C-labeled l-3HB, the spinal cord contained the highest radioactivity of the organs analyzed.40

[¹⁸F]FBHB and (R)-[¹⁸F]F3HB were next assessed by PET in immunodeficient NIH-III mice bearing subcutaneous human MDA-MB231 triple-negative breast cancer tumors. This cell line was selected because a prior in vitro study7 demonstrated substantial (2-fold) growth inhibition of MDA-MB231 cells in response to the addition of 10 mM AcAc to the cell growth media, a response which implied uptake of the ketone body. Furthermore, MDA-MB231 cells are reported to express MCT2,41 the transporter isoform with the highest affinity for ketone bodies.42

There was higher accumulation of both [¹⁸F]FBHB and (R)-[¹⁸F]F3HB in MDA-MB231 breast tumors as compared to muscle tissue (Fig. 4B). Ketone body derivative [¹⁸F]FBHB showed higher overall tumor uptake compared to (R)-[¹⁸F]F3HB (Fig. 4B). In comparison, [¹⁸F]FDG uptake in the same tumor model43 exhibited lower muscle uptake and higher tumor accumulation than [¹⁸F]FBHB, particularly after 15 minutes post injection (Fig. 4C).

Fig. 4. (A) Representative PET images of 3 MDA-MB231 xenograft mice 1 h post injection (p.i.) of [¹⁸F]FBHB, (R)-[¹⁸F]F3HB, or [¹⁸F]FDG;43 (B) time-activity curves (TACs) of tumor to muscle ratios for [¹⁸F]FBHB versus (R)-[¹⁸F]F3HB over 60 min; (C) TACs of tumor to muscle ratios for [¹⁸F]FBHB versus [¹⁸F]FDG43 over 60 min. MIP, maximum intensity projection; SUV, standardized uptake value.

As a control, nitrile intermediates [¹⁸F]16 and [¹⁸F]19 (Scheme 1), were assessed in the MDA-MB231 tumor-bearing mice. Neither were taken up by tumor tissue (see ESI,† Fig. S3). Rather, both compounds had minimal overall retention with renal clearance and no distinct organ accumulation. There was a slightly higher baseline whole-body retention of S isomer [¹⁸F]16. These results suggest that the carboxylate functionality is essential for tumor uptake of [¹⁸F]FBHB and (R)-[¹⁸F]F3HB and for their distinctive brain and heart accumulation.

The uptake and utilization of endogenous ketone bodies in humans and other mammals is positively correlated with their plasma concentrations.44–47 It was therefore hypothesized that fasting of the tumor-bearing mice in the present study might result in an increase in [¹⁸F]FBHB uptake. To analyze this phenomenon, the PET studies with the MDA-MB231 mice were conducted A) under standard diet conditions (free feeding of a carbohydrate rich diet) and B) after 24 hour fasting of the same mice. Fasting-induced effects on blood glucose and ketone bodies were measured using a test strip meter after blood sampling post fasting and the PET scan. The measurements showed a significant decrease in mean blood glucose (8.48 ± 1.00 to 4.62 ± 0.53 mmol L^–1, n = 5; p < 0.01) and a substantial, fourfold increase in ketone body concentration (0.54 ± 0.08 to 2.16 ± 0.33 mmol L^–1, n = 5; p < 0.01) (see ESI,† Fig. S4).

Time-activity curves for tissue-specific accumulation of [¹⁸F]FBHB and (R)-[¹⁸F]F3HB showed divergent impacts of the 24 h fasting condition (summarized in Table 1, detailed in Fig. 5). [¹⁸F]FBHB uptake in both the heart and brain were elevated significantly post fasting (both +Δ9%, p < 0.05). This finding supports the idea that [¹⁸F]FBHB indeed acts as a ketone body analogue in vivo; however, further studies are needed to fully prove this, regarding, for example, the analysis of specific transporters involved (e.g. MCT isoforms 1, 2, and 4) and the role of competitive inhibition by endogenous ketones present in the blood plasma. And whereas it was shown that the heart and brain accumulated additional [¹⁸F]FBHB (or radiometabolites thereof) during short-term nutrient deprivation, the MDA-MB231 tumor tissue did not take up significantly more [¹⁸F]FBHB post fasting (Fig. 6A). Only (R)-[¹⁸F]F3HB showed a moderate and statistically significant (+Δ15%) increase in tumor accumulation after fasting (Fig. 6B), which may indicate biosynthetic utilization of this specific isomer.

Table 1. Fasting-induced changes in radiotracer retention, SUV_mean,60min (n = 3), * = p < 0.05.

	[18F]FBHB	[18F]FBHB	(R)-[18F]F3HB	(R)-[18F]F3HB
		24 h fasted		24 h fasted
Tumor	0.91 ± 0.02	0.95 ± 0.04	0.77 ± 0.06	0.89 ± 0.05*
Muscle	0.57 ± 0.01	0.06 ± 0.03	0.50 ± 0.05	0.54 ± 0.02
Heart	0.96 ± 0.01	1.05 ± 0.02*	0.84 ± 0.09	0.95 ± 0.03n.s.
Brain	1.08 ± 0.02	1.17 ± 0.02*	0.91 ± 0.07	1.02 ± 0.05n.s.
Joint	1.10 ± 0.08	1.07 ± 0.08	1.30 ± 0.10	1.58 ± 0.02p=0.0501
Spine	1.10 ± 0.07	1.13 ± 0.04	1.39 ± 0.03	1.56 ± 0.03*
Liver	0.74 ± 0.01	0.80 ± 0.02*	0.66 ± 0.08	0.74 ± 0.02n.s.
Kidneys	1.44 ± 0.05	1.38 ± 0.16	1.96 ± 0.17	2.36 ± 0.25n.s.
Harderian glands	1.10 ± 0.01	1.15 ± 0.02	1.56 ± 0.13	1.39 ± 0.11n.s.

Fig. 5. TACs for different organ uptake after injection of [¹⁸F]FBHB or (R)-[¹⁸F]F3HB under fed (control) and 24 h fasted conditions. Data are shown as mean ± SEM from 3 experiments per condition.

Fig. 6. (A) (Top) Representative PET images at 60 min post injection of [¹⁸F]FBHB in a fed state (left) or after 24 h fast (right) and TACs (bottom) for tumor and muscle uptake; (B) the same is shown after injection of (R)-[¹⁸F]F3HB. Semi-quantitative data are shown as mean ± SEM from 3 experiments per condition. MIP, maximum intensity projection.

Experimental

All animal experiments were carried out in accordance with guidelines of the Canadian Council on Animal Care (CCAC) and approved by the local Animal Care Committee of the Cross Cancer Institute. Normal BALB/c mice or breast cancer xenografts were used for the PET imaging experiments. Human MDA-MB231 cells (3–5 × 10⁶ cells in 100 μL PBS) were injected subcutaneously into 8–12 weeks old anesthetized female NIH-III nude mice (Charles River, Saint-Constant, QC, Canada). Tumors were grown for 3–4 weeks, reaching sizes of 300–500 mm³. For fasting experiments the food was removed from the mice for 24 h prior to the PET experiment.

Mice were anesthetized with isoflurane (40% O₂/60% N₂) and their body temperature was kept constant at 37 °C. They were positioned and immobilized in prone position into the centre of the field of view of an INVEON® PET scanner (Siemens Preclinical Solutions, Knoxville, TN, USA). A transmission scan for attenuation correction was not acquired. Radioactivity present in the injection solution (0.5 mL insulin syringe) was determined using a dose calibrator (AtomlabTM 300, Biodex Medical Systems, New York, NY, USA). After emission scan was started, radioactivity (4–8 MBq in 100–150 μL saline) was injected with a delay of ∼15 s through a tail vein catheter. Dynamic PET data acquisition was performed in 3D list mode for 60 min. Dynamic list mode data were sorted into sinograms with 54 time frames (10 × 2 s, 8 × 5 s, 6 × 10 s, 6 × 20 s, 8 × 60 s, 10 × 120 s, 5 × 300 s). Frames were reconstructed using the maximum a posteriori (MAP) reconstruction mode. No correction for partial volume effects was performed. Image files were further processed using the ROVER v2.0.51 software (ABX GmbH, Radeberg, Germany). Masks defining 3D ROI were set and defined by 50% thresholding. Mean standardized uptake values [SUV = (activity/mL tissue)/(injected activity/body weight)] were calculated for each ROI. Time-activity curves (TAC) were generated from the dynamic scans using GraphPad® Prism 5.04 (GraphPad Software, La Jolla, CA, USA).

All semi-quantified PET data are expressed as means ± SEM. Where applicable, statistical differences were tested by unpaired Student's t test and were considered significant for p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***).

Determination of blood glucose and ketone concentrations in mice

Blood samples from the same MDA-MB231 bearing NIH-III nude mice as used for the PET experiments (unfasted and after 24 h fasting) were collected through tail vein puncture of the anesthetized mice. Blood drops were loaded onto test strips for glucose and ketone monitoring and concentrations were measured using a Freestyle Precision Neo® reading meter (Abbot Diabetes Care Ltd, Witney, UK). Data were measured as mmol L^–1 and are shown as mean ± SEM from n mice. GraphPad® Prism 5.04 (GraphPad Software, La Jolla, CA, USA) was used to generate the diagrams. Statistical differences were tested using unpaired Student's t test and was considered significant for p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***).

General

Kryptofix® 222 (K222) (>98.0%) was purchased from TCI chemicals. Anhydrous CH₃CN (99.9+%, extra dry, acroseal) was purchased from Acros organics. All other chemicals and reagents were purchased from Millipore Sigma and used without further purification. Buffer pH was measured on a Fisher Scientific Accumet basic ab15 pH meter. Thermoshaking was performed on an Eppendorf Thermomixer R. TLC plates were purchased from Millipore (TLC Silica gel 60 F254). RadioTLCs were read on a Bioscan AR-2000 plate reader.

Synthetic methods and full characterizations of compounds 1, 2, 5,22,27 6a,28 and 6b are provided in our previous work.22 Compounds 3a,483b, 3c, 4,49 and nitrile reference standard50 FCH₂CHOHCH₂CN were prepared according to published methods; see ESI† for details and characterization.

[¹⁸F]fluoride drying

No-carrier-added [¹⁸F]fluoride was produced from [¹⁸O]H₂O (¹⁸O(p,n)¹⁸F) on an ACSI TR-19/9 cyclotron and provided as an aqueous solution (1–2 GBq). [¹⁸F]fluoride was captured on a strong anion-exchange extraction cartridge (Waters, Sep-Pak Accell Plus QMA Plus Light) after priming with aq. K₂CO₃ (0.5 M, 8 mL) and water (10 mL). [¹⁸F]fluoride was eluted with 1.5 mL of a solution (14% H₂O, 86% CH₃CN) containing K222 (26.6 mM) and K₂CO₃ (13.3 mM). The [¹⁸F]fluoride was dried by azeotropic distillation at 90–100 °C under N₂ flow using 3 sequential 1 mL additions of anhydrous CH₃CN.

Preparation of nitrile intermediates [¹⁸F]16 and [¹⁸F]19

In a typical labeling procedure, to the residue of dried [¹⁸F]KF/K222 was added 0.01 g (0.044 mmol) of either (2S)-(+)-glycidyl tosylate (for compound [¹⁸F]16) or (2R)-(–)-glycidyl tosylate (for compound [¹⁸F]19) as a solution in anhydrous CH₃CN (0.4 mL). The vial was sealed tightly and heated for 25 min at 95 °C. CAUTION! The radioproduct, compound [¹⁸F]15/[¹⁸F]18, is volatile (epifluorohydrin b.p. 85 °C). RadioTLC of this intermediate is uninformative due to its volatility. After allowing the vial to cool to rt, an aq. solution (50 μL, 3.0 M) of KCN was directly added, the vial resealed, and the mixture briefly vortexed and heated at 95 °C for 15–20 min. RadioTLC of nitrile intermediates [¹⁸F]16 or [¹⁸F]19 on silica gel showed incorporations of 85 ± 9% (M ± SD, n = 7), R_f = 0.4, 1 : 2 hexanes : EtOAc; however, radioTLC at this stage is not quantitative owing to the possibility of residual [¹⁸F]epifluorohydrin evaporating prior to plate reading. The reaction mixture containing intermediate [¹⁸F]16 or [¹⁸F]19 was filtered through a neutral alumina cartridge (Waters Sep-Pak Alumina N Plus Light Cartridge), the cartridge flushed with THF, and the combined eluates (containing CH₃CN, H₂O, and THF) condensed to a residue under N₂ stream at rt. The residue was resuspended in water and [¹⁸F]16/[¹⁸F]19 was purified by HPLC on a C18 column eluting with isocratic water. The chemical identities of intermediates [¹⁸F]16/[¹⁸F]19 were confirmed by co-elution with non-UV active racemic reference standard FCH₂CHOHCH₂CN (TLC detection by KMNO₄ stain), synthesized according to Liu et al.50

Preparation of (3S)- or (3R)-4-[¹⁸F]fluoro-3-hydroxybutyric acid, [¹⁸F]FBHB and (R)-[¹⁸F]F3HB

The aqueous solution of HPLC purified nitrile [¹⁸F]16/[¹⁸F]19 was concentrated to a volume of ≈0.5 mL by rotary evaporation under moderate heating (bath temp ≈60 °C) and transferred to a lo-bind snap cap microcentrifuge tube. E. coli-derived nitrilase enzyme (1 mg) was added to the aqueous solution along with a small volume of TRIS buffer (10–20 μL of 10 mM, pH 9) to achieve a solution pH of 9. The enzymatic reaction was placed on a thermoshaker at 30 °C, 750 rpm for 60 min. Conversion was monitored by radioTLC (silica gel, carboxylic acid [¹⁸F]FBHB/(R)-[¹⁸F]F3HB R_f = 0.1–0.2, EtOAc + 1% AcOH). The final product acid [¹⁸F]FBHB/(R)-[¹⁸F]F3HB was purified and prepared as a concentrated injectable solution using a strong anion-exchange cartridge (Waters, Sep-Pak Accell Plus QMA Plus Light Cartridge); after loading the enzyme reaction mixture onto the QMA, the cartridge was washed sequentially with EtOH (3 mL) and water (3 mL), and eluted with NaOAc buffer (0.25 M, pH 5.5) as an injectable solution. Additional notes are provided in ESI.† Radiochemical purity of the final products were assessed by a Shimadzu UFLC using a Kinetex 2.6 μm F5 100 Å LC column 150 × 4.6 mm (Phenomenex), gradient elution, 0–30% CH₃CN in H₂O, 20 min. Molar activity could not be determined due to a lack of UV absorbance for the cold standard acid. Nitrilase activity showed stereopreference for conversion of R nitrile [¹⁸F]19. Isolated RCY (enzyme reaction) 10% (compound [¹⁸F]FBHB), 35% (compound (R)-[¹⁸F]F3HB), radiopurity >98%.

Total RCY from cyclotron-generated ¹⁸F to isolated injection solution: 13% ((R)-[¹⁸F]F3HB), 1.3% ([¹⁸F]FBHB). Total synthesis time from cyclotron-generated ¹⁸F to purified, injectable solution: 4 h.

Conclusions

The described synthetic strategy for preparing [¹⁸F]FBHB and (R)-[¹⁸F]F3HB can be considered a first generation method applicable to preliminary radiopharmacological studies; however, the method will likely require revision for scale-up and clinical translation.

Preliminary in vivo investigations of [¹⁸F]FBHB and (R)-[¹⁸F]F3HB demonstrated distinct uptake patterns. These promising first in vivo results warrant future experiments with [¹⁸F]FBHB to determine: 1) its metabolic stability and reactivity toward ketolytic enzymes, 2) the transporter isoforms responsible for its cellular import, and 3) the mode of its metabolic trapping in normal and cancer cells. These experiments will help to validate [¹⁸F]FBHB as a “real” d-BHB analogue, and as a useful tool for investigations of ketone body metabolism in cancer and diseases like diabetes, Alzheimer's, epilepsy, and ischemic heart disease.

Conflicts of interest

There are no conflicts to declare.