[18F]BODIPY-triglyceride-containing chylomicron-like particles as an imaging agent for brown adipose tissue in vivo

Brown adipose tissue (BAT) is present in human adults and the current gold standard to visualize and quantify BAT is [18F]FDG PET-CT. However, this method fails to detect BAT under insulin-resistant conditions associated with ageing and weight gain, such as type 2 diabetes. The aim of this study was to develop a novel triglyceride-based tracer for BAT. For this purpose we designed a dual-modal fluorescent/PET fatty acid tracer based on commercially available BODIPY-FL-C16, which can be esterified to its correspondent triglyceride, radiolabeled and incorporated into pre-synthesized chylomicron-like particles. BODIPY-FL-C16 was coupled to 1,2-diolein with a subsequent radiolabeling step resulting in [18F]BODIPY-C16-triglyceride that was incorporated into chylomicron-like particles. Various quality control steps using fluorescent and radioactive methods were conducted before BAT visualization was tested in mice. Triglyceride synthesis, radiolabeling and subsequent incorporation into chylomicron-like particles was carried out in decent yields. This radiotracer appeared able to visualize BAT in vivo, and the uptake of the radiotracer was stimulated by cold exposure. The here reported method can be used to incorporate radiolabeled triglycerides into pre-synthesized chylomicron-like particles. Our approach is feasible to visualize and quantify the uptake of triglyceride-derived fatty acids by BAT.

uptake of circulating albumin-bound FAs 2-5 . Although direct TRL particle uptake with adjacent FA release has been suggested 4 , more recent findings using glycerol tri[ 3 H]oleate and [ 14 C]cholesteryl oleate double-labeled TRL-like particles showed an approx. 10-fold higher uptake of FAs compared to cholesteryl esters by BAT 5,6 , indicating that the majority of TG-derived FAs is internalized after liberation by lipoprotein lipase (LPL). In fact, TRL-derived FAs were identified as the main supply of TGs in BAT 3,7 and TG-derived FA internalization was shown to be dependent on the presence of lipoprotein lipase (LPL) 4,7,8 , cluster of differentiation 36 (CD36) 4,9 and fatty acid transport proteins (FATP) 10 (see also Fig. 2b).
The variety of quantification approaches of BAT volume and metabolic activity reaches from invasive imaging with fluorescence probes 4 or 3 H-labeled TG 5 to non-invasive experiments with PET [11][12][13][14][15] , SPECT 16,17 , shorwave infrared 18 , targeted MRI with lipoproteins as contrast agent 4 , and MRI [19][20][21] and was reviewed recently by us 22 . [ 18 F] FDG (as a measure for glucose uptake) is frequently used in studies to quantify BAT activity and is currently the gold standard 12,[23][24][25] (Fig. 1). Nevertheless FAs are the main metabolized substance class in BAT, which is not simply mirrored by [ 18 F]FDG scans 14 . In addition, insulin resistance such as type 2 diabetes of BAT occurring with ageing and weight gain will underestimate BAT activity as assessed with [ 18 F]FDG scans, whereas FA and oxidative metabolism is not dependent on insulin sensitivity 26 . It should be mentioned that LPL activity was decreased during insulin resistance in mice 27 and insulin was found to be essential for the lipolytic processing of TGLs by BAT 28 . Therefore BAT visualization with TRLs could become difficult during diabetic conditions, but it might be the most precise way to gain information about BATs lipid turnover and metabolic activity. Taken together,    29,30 ). We here report the development of a FA-tracer based on fluorescent FA BODIPY-FL-C 16 (BDP-FA), which is suitable for both PET and fluorescence imaging, and present a method to esterify the FA into a TG and subsequently incorporate it into a chylomicron-like particle. We hypothesized that by this approach the physiological situation is mimicked where TRL-derived TGs get lipolyzed locally before they are taken up by brown adipocytes. Additionally it will be possible to image from whole body to sub-cellular level and in vitro experiments can be carried out with the same molecule without any radiation dose. BODIPY dyes have been already used to image BAT 4 and it is known that FA transport proteins (FATP) have a preference for BODIPY-FL coupled to a long carbon chain (C ≥ 8) 31 . Furthermore, downstream metabolic reactions in white and brown adipocytes have already been visualized in vitro 32,33 .
To avoid decreasing the FA characteristic properties of BDP-FA by introduction of another chelator molecule, we performed 18 F/ 19 F exchange reactions used to transform fluorescent dyes into dual-modality PET/fluorescent imaging dyes 4,[34][35][36][37] . Neither an increased steric demand, nor lowering of the targeting efficiency of the FA towards proteins responsible for FA uptake is expected as BDP-FA is only modified at the end of the carbon chain and therefore keeps its FA characteristics. We recently described the synthesis and radiolabeling of [ 18 F]BDP-TG and conducted first in vitro experiments with primary human adipocytes where [ 18 F]BDP-FA uptake could be modified by different BAT activating and blocking agents 33 . The aim of the presented study was to incorporate the [ 18 F]BDP-TG into chylomicron-like particles to obtain one of the first TG-based PET imaging agents and to test the new developed tracer in mice. Different to other used FA-based imaging tracers, the here presented TG will be first lipolyzed on-site before it can be internalized by BAT. This reflects the physiological situation in a better way than other FA-based tracers do and will help to quantify BATs consumption of lipids and its contribution to whole body energy expenditure.

Results
Synthesis of BDP-TG. Synthesis of BDP-TG was conducted as described before 33 . The resulting TG was obtained with a yield of 45 ± 8% SD after HPLC purification. BDP-TG has a t r of 12.3 min and NMR and ESI-MS confirmed the identity of BDP-TG. Synthesis of chylomicron-like particles and characterization. Chylomicron-like particles were synthesized essentially as reported before 5,6,38 . After synthesis size of the particles was determined by DLS, showing a mean diameter of 164 ± 20 nm and a polydispersity index of 0.181 (n = 4). Additionally, particles were analyzed by transmission electron microscopy. Polydisperse particles could be visualized, all of which showed an encapsulated lipid core and a distinct surface shell with a mean particle diameter of 156 ± 55 nm (n = 25) (Fig. S1).

Ex vivo incorporation of [ 18 F]BDP-TG into chylomicron-like particles and characterization.
After synthesis the chylomicron-like particles were loaded with BDP-TG or [ 18 F]BDP-TG. Interestingly, once the BDP-TG is encapsulated, particles show a dark band within their lipid core (compare Fig. 3a with Fig. S1). Incorporation of [ 18 F]BDP-TG into chylomicron-like particles in time was analyzed by TLC. After 25 min more than 99% of the TG was incorporated (Fig. 3b). Different temperatures did not affect the incorporation speed, where r.t. showed the best result after 60 min (99.5%) compared to 0 °C (98.9%) and 38 °C (99.1%) (Fig. 3c).
To test whether the [ 18 F]BDP-TG was truly incorporated, chylomicron-like particles loaded with BDP-TG were measured on a fluorescence microplate reader. Excitation was compared to chylomicron-like particles or BDP-TG alone. A significant increase in intensity (>1,400 fold) was observed for the chylomicron-like particles loaded with BDP-TG in comparison to the particles or the BDP-TG alone (Fig. 3d). Additionally, chylomicron-like particles incubated with [ 18 F]BDP-TG did not show any impurities of free fluorine-18 or free [ 18 F]BDP-TG after 60 min of incubation, analyzed by gel electrophoresis and radio-TLC (Fig. 3e,f).

Animal experiments. [ 18 F]BDP-TG-chylomicron-like particles (1-5 MBq) were injected into female
C57Bl/6 mice, which were fasted for 4 h either at r.t. or at 4 °C. After scanning for 1 h the animals were euthanized and the organs were harvested. Analysis of the PET images showed highest uptake in liver and heart at r.t. and at 4 °C ( Fig. 4a,b). A rapid increase with a slow washout in both organs could be visualized (Fig. S2a,b). In bone a constant increase in signal was observed (Fig. S2c,d), which probably indicates a defluorination process of the tracer in vivo, as reported in literature 37 . Lung showed a fast increase with a fast washout and stayed constant at later time points under both temperature conditions. Brain as a negative control showed negligible uptake. BAT uptake increased significantly when comparing cold exposed animals vs. control animals at 50-60 min p.i. (Mann-Whitney U, n = 10, two tailed p < 0.0001) (Fig. S2c,d).
PET images are supported by the results of the biodistribution. Highest uptake values in animals kept at r.t. were found in heart (17.9 ± 3.3% ID/g), liver (16.8 ± 2.6% ID/g), bone (9.0 ± 3.0% ID/g) and spleen (8.5 ± 2.2% ID/g). After 1 h only 4.5 ± 1.1% ID/g were found in the blood, indicating a fast blood clearance during the scanning time. In total 6.9 ± 1.3% ID was found in the blood calculated by an assumed total blood volume of 5.85 mL/100 g 39 . Uptake by BAT (3.5 ± 0.7% ID/g) was approximately 3-fold higher than uptake by WAT (1.1 ± 0.6% ID/g; Mann-Whitney U, n = 5, two tailed p < 0.001).

Discussion
Exploring BAT and its metabolism has become an interesting and fast developing topic in endocrine research. A variety of different imaging approaches have been used in the past reaching from in vitro experiments 40 over invasive imaging with fluorescence probes 4 or 3 H-compounds 5 to non-invasive experiments with PET 11-15 , SPECT 16,17 and MRI [19][20][21] . In a clinical environment most often [ 18 F]FDG scans are used for BAT imaging but it only shows glucose-related uptake and has therefore the chance to misinterpret BAT activity by underestimating lipid uptake and metabolism. Additionally, [ 18 F]FDG uptake is dependent on insulin sensitivity and therefore might not reflect the real activation state of BAT. In studies with [ 18 F]FTHA it was observed that radiolabeled FAs showed an increased uptake in BAT under cold stimulation in humans 14 . However, these results with a free FA-based tracer might be less relevant because the majority of FAs is TRL derived 3,7 where FAs are transported as TGs. With our developed tracer [ 18 F]BDP-TG which is incorporated into chylomicron-like particles we can overcome these limitations and gain new insights in BATs lipid metabolism.
BDP-TG was produced in a decent yield (45%) like previously published 33 and radiolabeling was carried out with a decay corrected radiochemical yield of 44% which is in accordance to radiolabeling yields reported in literature 36,41 .
Chylomicron-like particles have been synthesized with a mean diameter of 164 nm (DLS) and 156 nm (TEM). Those sizes are in accordance with previously described particles 5,6,38 .
After synthesis of the particles, loading with either BDP-TG or [ 18 F]BDP-TG was performed. A pre-purification of [ 18 F]BDP-TG from free fluorine-18 and SnCl 4 before incubation with chylomicron-like particles is of immense importance. In a previous approach [ 18 F]BDP-TG received a single wash with H 2 O (500 µL) before it was incubated with the particles. After 60 min the sample was purified by centrifugal filtration with an Amicon Ultra centrifugal filter (10 KDa). This caused two problems: 1) remaining SnCl 4 caused a co-precipitation of the particles 2) filtration destroyed the particles yielding large lipid emulsions. This could be overcome by an intensified washing procedure where all the SnCl 4 as well as free fluorine-18 was washed out before [ 18 F]BDP-TG was added to the chylomicron-like particles.
An incorporation of the TG during the formation of the chylomicron-like particles in the sonicator has been tested but resulted in a breakdown of the boron-fluoride bond. Incorporation speed of [ 18 F]BDP-TG into chylomicron-like particles was found to be fast (>99% after 60 min (Fig. 3b)) and temperature seemed to have no effect on the incorporation yield and speed (Fig. 3c). Purity of radiolabeled chylomicron-like particles was found to be >96% and therefore suitable for in vivo applications. (Fig. 3e,f).
Additionally, we evaluated the incorporation of the BDP-TG into chylomicron-like particles by fluorescence. We previously found that the fluorescence intensity is strongly related to the environment of the compound, meaning that only if the BDP-TG is dissolved it will give a fluorescence signal 33 . BDP-TG incorporated into chylomicron-like particles in HEPES solution showed a huge increase in signal (1413 fold) compared to BDP-TG or chylomicron-like particles alone. This indicates that once the BDP-TG is incorporated into a chylomicron-like particle it regains its fluorescence because it is in a lipophilic environment.
In vivo preferential uptake of [ 18 F]BDP-TG in BAT compared to WAT was observed. Exposure to cold during fasting, thereby activating BAT 2,25 , pronounced the difference, indicating the ability of our tracer to visualize BATs metabolic activity and FA consumption. Although not visible on the microPET images, biodistribution data showed that PVAT followed a similar trend. Muscle tissue, which may become activated in the cold due to shivering 42 showed only low uptake of [ 18 F]BDP-TG and no increased uptake due to cold-exposure. This does not exclude a higher metabolic activity of muscle, as muscle may preferentially use glucose under conditions of cold exposure 43 .
BAT was however not the only tissue with a high uptake of [ 18 F]BDP-TG. Uptake of the tracer was in fact highest in liver, heart, spleen and bone, and a significant increase by cold exposure is also demonstrated in liver, heart and spleen. Hepatocytes have a low LPL expression 6 . High uptake values can be explained by increased uptake of remnants of the chylomicron-like particles, which still contain TGs 44 , or by spill over of FAs generated during lipolysis 45 . This might also explain the elevated uptake due to cold activation because more remnants and FA are produced in this situation. Since the heart has high LPL expression 46,47 a high tracer accumulation was expected. Increased LPL activity due to cold stimulation was already reported in cardiomyocytes 48,49 which explains the increased heart uptake. The spleen, as an organ of the mononuclear phagocytic system, contains high numbers of macrophages. Those macrophages are able to engulf large particles which might explain the marked uptake observed in the spleen 50 . Lung uptake might be the result of a polydisperse particle distribution, as large particles are prone to get entrapped in lungs capillaries 51 . Indeed, in previous experiments with filtrated particles, where particle coagulation was frequent and sizes >1000 nm diameter could easily be reached, we experienced very high uptake values of >200% ID/g in the lung.
We also noted a high uptake of radioactivity in the bone. Although ex vivo plasma stability tests with [ 18 F] BDP-FA showed >99% intact compound after 4 h 33 , this may not be valid in vivo. We were unable to demonstrate free fluorine-18 in vivo in the plasma, due to the rapid blood clearance and the difficulty to perform analyses on radiochemical purity of 18 F-labeled chylomicrons from within a blood sample. Still, in recent literature it was shown that in vivo the stability is not ensured for radiolabeled BODIPY-compounds, making it likely that our compound is defluorinated as well 37 . In a former publication 1,3-diolein was coupled to BDP-FA 33 . We speculated (2019) 9:2706 | https://doi.org/10.1038/s41598-019-39561-z www.nature.com/scientificreports www.nature.com/scientificreports/ that FAs on position 1 or 3 on the glycerol backbone might have a higher chance to get released during lipolysis. No significant differences in chemical and radiochemical yields during synthesis nor changes in the biodistribution can be reported for 1,3-diolein-BDP-FA vs. 1,2-diolein-BDP-FA (data not shown).
In comparison to our data, similar particles loaded with glycerol tri[ 3 H]oleate previously showed higher uptake of [ 3 H]oleate by BAT (approx. 5 fold 6 , 6 fold 5 and 10 fold 52 ) and lower [ 3 H]oleate uptake by liver (approx. 0.6 fold 6 , 0.6 fold 5 and 0.5 fold 52 ) and heart (approx. 0.3 fold 6 , 0.3 fold 5 and 0.5 fold 52 ). Also no other organs showed an increased uptake when BAT was stimulated. Different experimental conditions (e.g. use of anesthetized vs. non-anesthetized mice) could be a reason for this difference in the biodistributions as gaseous anesthetics such as isoflurane are known to suppress adrenergic signaling 53 . In general these results may point to a reduced BAT LPL-activity due to anesthesia which results in increased uptake by liver and heart.

Conclusion
In the current manuscript, we presented a dual-modal fluorescent and PET active TG which was successfully incorporated into chylomicron-like particles. With different quality control methods we showed incorporation of the radiolabeled TG into chylomicron-like particles. In vivo animal studies showed that the resulting tracer was able to reach BAT but was also taken up by other tissues which employ LPL-mediated FA uptake. BAT uptake of the tracer was increased in cold exposed animals. The here presented technique is able to visualize TRL-derived FA BAT uptake after TG-lipolysis which is advantageous in comparison to conventional FA-based tracers which do not reflect the physiological situation and are mainly taken up by the liver. We anticipate that [ 18 F]BDP-TG-chylomicron-like particles are a promising step forward to visualize and quantify BATs lipid metabolism and gain more information about BATs contribution to whole body energy expenditure in the future.
All HPLC purifications (1.0 mL/min, solvent A; 0.1% TFA in H 2 O, solvent B; CH 3 CN, 50 °C) were performed on a Shimadzu UFLC HPLC system equipped with a DGU-20A 5 degasser, a SPD-M20A UV detector, a LC-20AT pump system, a CBM-20A communication BUS module, a CTO-20AC column oven, and a Scan-RAM radio-TLC/HPLC-detector from LabLogic using an Aeris ™ Widepore column (C4, 3.6 μm, 4.6 mm × 250 mm) for the BODIPY-triglyceride (BDP-TG). ESI-MS was performed on an Applied Biosystems SCIEX API 150 EX electrospray ionization quadrupole (ESI-Q) mass spectrometer with the method of McAnoy et al. 54 . Briefly, 0.1 M aqueous ammonium acetate solution was added to the sample to observe the ammonium salt of the synthesized TG in the MS. 1 H-NMR spectra were carried out on a Bruker UltrashieldTH 400 plus at 400 MHz. Tol-d 8 was used as solvent with TMS as internal standard. Chemical shifts are reported in parts per million (ppm) relative to the internal standard.
Gel electrophoresis was used to determine the amount of free fluorine-18 and [ 18 F]BDP-TG in the solution containing [ 18 F]BDP-TG incorporated in chylomicron-like particles. Gel electrophoresis was carried out under native running conditions where the sample was mixed (1:2) with native sample buffer and loaded into an any kD TGX gel (20 kBq per lane). For visualization phosphor screens were exposed for 10 h to the gel and analyzed by a Typhoon FLA 7000 phosphor imager (GE Healthcare).

Synthesis of chylomicron-like particles.
Synthesis of chylomicron-like particles was performed as reported before 38,55 . Briefly, emulsion particles were prepared from triolein (70 mg), egg yolk phosphatidylcholine (Lipoid) (22.7 mg), lysophosphatidylcholine (2.3 mg), cholesteryl oleate (3.0 mg), and cholesterol (2.0 mg). Sonification was performed using a Soniprep 150 (MSE Scientific Instruments, UK) that was equipped with a water bath for temperature (54 °C) maintenance, at 10 μm output. The emulsion was fractionated by density gradient ultracentrifugation steps in a Beckman SW 40 Ti rotor. After centrifugation for 30 min at 17,850 rpm at 20 °C, an emulsion fraction containing chylomicron-like particles was removed from the top of the tube by aspiration. Characterization of chylomicron-like particles was done by DLS and transmission electron microscopy. Chylomicron-like particles were stored at 4 °C and were used within 5 days following preparation.
Dynamic Light Scattering. The particle sizes were measured by photon correlation spectroscopy performed at an angle of 90°; with a setup consisting of an ALV-SP8 goniometer, an ALV-SIPC photomultiplier, a multiple τ digital real-time ALV-7004 correlator, and a solid state laser (Koheras) with a red laser (λ = 633 nm) as light source. The time resolved signal of two Single Photon Counting Modules was cross-correlated. To prevent multiple scattering highly diluted chylomicron-like particle solutions of 0.1 mg/mL in bi-distilled and filtered H 2 O (1.2 μm poly(tetrafluoroethylene) membrane filters) were prepared. Sample cuvettes were immersed in a toluene bath and tempered within an error of ± 0.1 °C. Autocorrelation functions of intensity fluctuations g 2 (q, t) are converted by the Siegert relation and give the field autocorrelation function f(q,t): The intensity-weighted decay-time τ distributions obtained from the field autocorrelation function by cumulant analysis were analyzed in respect to multimodality. For each diffusive mode the decay rate Γ = 1/τ was plotted against the squared length of the scattering vector q 2 . The slope gave the Z-average translational diffusion coefficient D 0 and results in the hydrodynamic radius R h after use of the Stokes Einstein equation: in the reaction solution 500 μL H 2 O were added and centrifuged for 5 min. The aqueous phase was aspirated and another 500 μL H 2 O were added to precipitate the remaining SnCl 4 . The mixture was heated to 100 °C, the organic phase was evaporated and the aqueous phase was taken off. [ 18 F]BDP-TG was reconditioned in 20 μL EtOH and another radio -TLC was performed. [ 18 F]BDP-TG could be obtained with a radiochemical purity of >96% and an overall decay corrected radiochemical yield of 21%. 400 μL chylomicron-like particles in HEPES were added (1.5 mg TG content) and incubated for 1 h at r.t. [ 18 F]BDP-TG-chylomicron-like particles were obtained (overall decay corrected RCY: 18%, 19 MBq) with a radiochemical purity of >99% analyzed by gel electrophoresis and radio-TLC (Fig. 3e,f). Shorter incubation time points and different temperatures were tested by radio-TLC (Fig. 3b, Animal experiments. Experimental protocols were approved by the "Centrale Commissie Dierproeven" and all animal experiments and procedures were performed in accordance with the guidelines set of this institution. From 13:00 p.m. on, female C57Bl/6 mice (fasted for 4 h at r.t. or fasted and cold exposed at 4 °C for 4 h) were anesthetized (Isoflurane, 1.5-2.0% at 2 mL/min in oxygen), mice kept at r.t. were put on a heating pad and both groups were injected with [ 18 F]BDP-TG-chylomicron-like particles (1)(2)(3)(4)(5) in HEPES (100 μL) via the tail vein. Mice were scanned dynamically for 1 h on a microPET (Focus 120, Siemens). Images were analyzed using Pmod V3.707. After the scanning time animals were killed and organs harvested, weighed wet and counted using a WIZARD 2 automatic γ-counter from Perkin Elmer.