Hydrophilic fluorinated molecules for spectral 19F MRI

Fluorine-19 (19F) Magnetic Resonance Imaging (MRI) is an emerging modality for molecular imaging and cell tracking. The hydrophobicity of current exogenous probes, perfluorocarbons (PFCs) and perfluoropolyethers (PFPEs), limits the formulation options available for in vivo applications. Hydrophilic probes permit more formulation flexibility. Further, the broad Nuclear Magnetic Resonance (NMR) chemical shift range of organofluorine species enables multiple probes with unique 19F MR signatures for simultaneous interrogation of distinct molecular targets in vivo. We report herein a flexible approach to stable liposomal formulations of hydrophilic fluorinated molecules (each bearing numerous magnetically equivalent 19F atoms), with 19F encapsulation of up to 22.7 mg/mL and a per particle load of 3.6 × 106 19F atoms. Using a combination of such probes, we demonstrate, with no chemical shift artifacts, the simultaneous imaging of multiple targets within a given target volume by spectral 19F MRI.


Results
The molecular design employed fragments of two non-ionic hydrophilic molecules (glycerol and glucose), and 'click' chemistry 12 , as the key synthetic step to couple them to the fluorinated species. The basic requirement for the fluorinated moiety was to have magnetically equivalent 19 F atoms (in order to have a single NMR resonance frequency) per molecule to generate a strong and sharp 19 F MRI signal. Preliminary evaluation of some commercially available organofluorine starting materials at 9.4 T, suggested a minimum peak separation of ~12 ppm between the resonance frequencies of individual species, to generate an image of each without any chemical shift artifacts. The 3,5-bis(trifluoro)phenyl-, trifluoroacetyl-, 1,2-bis(difluoro)methylene-, and disubstituted 2,3,5,6-tetrafluorophenyl-groups showed adequate resonance frequency separation and were selected for the study. Precursors of both the fluorinated and hydrophilic moieties were introduced in the late stage of the synthesis as preformed scaffolds (Fig. 1a) of either the azido or alkynyl derivative. Coupling of the respective hydrophilic and fluorinated moieties followed by deprotection to obtain the final compounds (Fig. 1b) all proceeded with excellent yields, and were optimized to generate gram quantities of each compound. All intermediates and final compounds were characterized by 1 H and 13 C NMR, HRMS, and 19 F NMR (where applicable).
As shown in Fig. 2a, all final products were obtained as solids at room temperature and readily dissolved in aqueous media (water, saline, PBS, and histidine/saline buffers), to give clear solutions (Fig. 2b). 19 F NMR of the solutions each gave a single peak: −75.90, −121.57, −143.39 and −62.65 ppm for ET0863, ET0876, ET0886 and ET0890 respectively (Fig. 2c). The single peaks indicate both the magnetic and chemical purity of each compound (estimated overall chemical purity >95%). 19 F MR images of phantoms of these solutions showed no chemical shift artifacts as expected (Fig. 2d).
Stability testing in plasma at 37 °C showed that ET0863 liposomes leaked more than 5% of their content within a 1.5 hour test period, and were excluded from subsequent experiments. ET0876, ET0886, and ET0890 liposomes all showed less than 5% leak, and were retained for subsequent experiments. Table 1 shows the mean particle size, total lipid concentration, and the 19 F content of each formulation. Assuming a lipid bilayer thickness of 4 nm, the total particles/mL and the total 19 F atoms per particle were also estimated for each formulation based on the total encapsulated volume. 19 F MRI scans were perfomed on phantoms of each neat formulation (Fig. 3a), using similar scan parameters as in the solution phantoms (data rendered in pseudo-color, Fig. 3b). 19 F signal was recorded, with Signal to Noise Ratio (SNR) of 38, 25, and 23 for ET0876, ET0886, and ET0890 formulations respectively. Dilution studies (Fig. 3c), showed that solutions of each formulation with as low as 2.5 × 10 13 particles/mL were detectable by 19 F MRI. A plot of the SNR against particle concentration (Fig. 3d), exhibited a linear relationship for each formulation. Relaxation times T 1 and T 2 (Fig. 3e) were determined for each neat formulation and diluted samples (using similar scan parameters as reported by Tirotta et al. 13 ).
To evaluate the potential of these formulations as probes for spectral 19 F MRI, phantoms of all three were scanned together with a phantom containing the fluorinated inhalable anesthetic, Isoflurane. A single pulse 19 F MR resonance frequency sweep (Fig. 4a), showed peaks corresponding to all the fluorine species in the phantoms: ET0890 (one peak), Isoflurane (two peaks), ET0876 (one peak), and ET0886 (one peak). An image from a 1 H T 2 Rapid Acquisition with Relaxation Enhancement (RARE) scan, representative of an anatomy scan (Fig. 4b), showed the position of each phantom within the bore of the magnet: ET0890 (position 1), Isoflurane (position 2), ET0876 (position 3), and ET0886 (position 4). When the phantoms were subjected to the same 19 F MSME scan sequence as above, with the excitation bandwith centered at the highlighted peak position (Fig. 4c), only that  phantom generated a visible 19 F MR image (shown in pseudo-color, Fig. 4d). An overlay of the 19 F image over the 1 H image allowed visualization of the exact location of that signal in the entire field of view (Fig. 4e). Preliminary evaluation of imaging using these agents in an in vivo environment was performed. 50 µL of each formulation were injected as follows: intramuscularly (ET0876 in right thigh and ET0890 in the left thigh), and subcutaneously (ET0886 in the abdominal area), in C57BL6 mice (n = 4). Each animal was anesthetized with isoflurane, placed in the magnet, and imaged using the same scan protocol as the phantoms. First, a TurboRARE T 2 1 H scan, with the lower torso of the animal in the field of view to show the general anatomy (the numbers 1, 2 and 3 indicate the locations in which the probes were administered, Fig. 5a), was performed. When the animal was imaged using the 19 F MSME scan sequence with the base frequency set to the resonance frequency of highlighted peak on the Single pulse 19 F spectrum (Fig. 5b), only a single 19 F signal, corresponding to that peak was generated on the 19 F MR image (Fig. 5c). Overlay of the 19 F image over the anatomy image showed the exact location of each 19 F spot (Fig. 5d). The average SNR of the porobes within the muscle was determined to be about 15 and about 5 for the probe in the subcutaneous space (Fig. 5e). The more diffuse signal and low SNR in the abdominal area was attributed to faster diffusion of the nanoparticles within the subcutaneous space compared to the muscle. 19 F MRI probes have the potential to simultaneously profile multiple molecular species/disease activity and disease sub-types within a target volume. Currently used PFCs and PFPEs lack flexibility in formulation and images often bear chemical shift artifacts due to the presence of non-identical fluorine nuclei on the probe molecules. Several strategies aimed at addressing some of these limitations are under investigation and well documented in a recent review 5 . Achieving a safely injectable formulation with any of these strategies remains a tall order. Partlow et al. have previously reported the ability to simultaneously track multiple targets in vivo using unique 19 F MR signatures of PFC nanobeacons 14 , the formulation of which have similar limitations as the PFCs and PFPEs.

Discussion
There is no previous report of hydrophilic fluorinated molecules designed with the goal of solving the formulation problem associated with current 19 F MRI probes. Previous attempts at liposome formulations have focused on the use of perfluorofatty acids or sulfonate amphiphiles to incorporate flourine in the bilayer of the particle 15,16 , or the use of inorganic fluoride 17 . These approaches suffer from limitted payload capacity. For instance, in a liposome formulation with a mean particle diameter of 150 nm, only a fraction of the ~350,000 total possible molecules in the bilayer (including phospholipids and cholesterol) can be the fluorinated amphiphile. Similarly, the number of water soluble ionic fluorides that can be encapsulated in a liposome is limited by the osmolality of the solution, and potential toxicity of the salt. However, a formulation with a similar particle size distribution and an aqueous interior loaded with nonionic hydrophilic molecules with osmolality of 300 mOsmol/kg can carry ~10 6 hydrophilic molecules per particle.
Our hydrophilic organofluorine molecules approach employs facile synthetic routes to water-soluble molecules with magnetically equivalent fluorine atoms which generate 19 F MRI images with no chemical shift artifacts. When combined with the liposome nanoparticle platform, it can access the broad chemical shift spectrum of organofluorine species, enabling numerous probes with unique 19 F MR signatures. Such probes have the potential to enable noninvasive simultaneous visualization of multiple hot spots within the same region of interest by 19 F MRI as demostrated in both our phantom and preliminary in vivo assessments of the three formulations.
Isoflurane, the commonly used inhalable anesthesia in small animal studies, is fluorinated, highly lipophilic, and readily absorbed by adipose tissue in vivo. The 19 F NMR chemical shifts of its two sets of fluorine nuclei are in the same vicinity as those of most PFCs and PFPEs. These can interfere with any signal from a PFC or PFPE probe, complicating data interpretation. While this problem can be addressed by using non-fluorinated injectable anesthetics such as ketamine and xylazine, these have other problems associated with them including effects on hemodynamics in key organs 18 and a more complicated work flow. 19 F MRI probes that are not affected by isoflurane are highly desirable. This new approach allows for facile access to such probes.
The Rose criterion suggests a SNR of ~4 is needed to be able to distinguish image features at 100% certainty 19 . The ET0876 formulation with a 19 F content of 22.7 mg 19 F atoms/mL, borne on a molecule with a molecular  19 F MR image following a 19 F MSME scan (same scan parameters as in Fig. 3  weight of 472.4 g/mol and generates a phantom image with an SNR of 38. Dimers and oligomers of this molecule as well as the others are synthetically accessible, suggesting that an increase in the 19 F content of each formulation is feasible (though the number of monomers units per oligomer may be limited by viscosity of the corresponding aqueous solution with increasing molecular weight). Moreover, the dilution studies show that a SNR of >4 is obtained from a 10 minutes scan with as few as 2.5 × 10 13 particles/mL of this formulation. This concentration represents an 8-fold dilution of the stock solution or an injection volume of about 250 µL in a 25 g mouse (blood volume ~2 ml), suggesting that these probes may be amenable vascular imaging. The consistently high T 2 values (all >100 ms) allow for several echoes per excitation. This explains the exquisite signals obtained with the MSME scan protocol for all three formulations. They also suggest the possibility of applying both 2D and 3D TurboRARE scan protocols to obtain images with high SNR using these formulations and are subject to subsequent investigations.
In summary, readily available hydrophilic molecules and small organofluorine moieties were condensed to generate nonionic hydrophilic fluorinated molecules with unique 19 F MR signatures. These were used to fabricate stable liposome formulations, and preliminary evaluation in both in vitro and in vivo environments demonstrate that spectral 19 F MRI can be used to report on the location of each probe within the same region of interest without interference from the others or Isoflurane. These results, in addition to the growing interest in 19 F MRI molecular imaging suggest that this is an excellent approach to harness the broad chemical shift spectrum of 19 F molecular species to generate a myriad of unique probes for 19 F MRI. Future efforts will be focused on targeted formulations which can home-in, and report on specific in vivo targets upon intravenous administration. In addition, increased sensitivity on a per particle basis will be sought by utilizing dimers and oligomers of these base molecules.

Materials and Methods
Chemical synthesis and characterization of intermediates and final compounds are described in supplementary information S1.  extruded on a Lipex thermoline extruder (Northern Lipids Inc., Canada), beginning with five passes through a 400 nm Nuclepore membrane (Waterman, Newton, MA) followed by eight passes through a 100 nm membrane. The resulting preparation was subjected to diafiltration through a 500 kD membrane (Spectrum Labs, Rancho Dominguez, CA) for 10 volume exchanges to practically eliminate any unencapsulated compound. Mean particle size was determined by Dynamic Light Scattering on a BI-90 goniometer/autocorrelator system at 90° using a 532 nm solid state laser source (Brookhaven Instruments Corp, Holtville, NY), and the final 19 F content determined by comparing 19 F NMR integrals against a standard solution prepared from analytical grade trifluoroacetic acid (Sigma Aldrich).

Preparation of liposomes.
Leak Test in Bovine Plasma. The percentage leak in bovine plasma (Sigma-Aldrich, St. Louis, MO) was determined by diluting 70 µL of the liposomal prep in 500 µL of bovine plasma. The solution was then incubated for 90 ± 5 minutes at 37 ± 1 °C. After the incubation was completed, 400 µL of the sample was diluted in 2.4 mL of 0.9% saline. 2 mL of the sample was then transferred into a 10,000 MWCO Vivaspin-2 centricon tube (Sartorius, Bohemia, New York) and centrifuged at 1100 rpm for 15 minutes. The filtrate was then analyzed for fluorine by NMR integration against a standardized trifluoroacetic acid solution. The % leak was determined by comparing the concentration of fluorine that leaked out to the total concentration of fluorine in the formulation.
MRI acquisition and data processing. All MRI scans were performed on a 9.4 T Bruker small animal MR scanner equipped with a 1 H/ 19 F dual-tunable volume RF coil (35 mm inner diameter, 50 mm length; Rapid Biomed, Würzburg, Germany), located in the Small Animal Imaging Facility (SAIF) at Texas Children's Hospital. 19 F images of both phantoms and mice were acquired with an MMSE scan protocol (Excitation bandwidth = 2000 Hz, TR = 2000 ms, TE = 8.95 ms, scan time = 10 min 40 s). 1 H images were acquired with a TurboRARE T 2 scan protocol (TR = 2500 ms, TE = 11 ms, RARE factor = 4, scan time = 5 min 20 s). Mice were anesthetized by exposure to Isoflurane prior to injection of probes and maintained under anesthesia for the duration of the experiment, as well as a temperature of 37 °C using a temperature controlled air-flow system. Dicoms obtained from scans were processed using the OsiriX v.5.8.5 software (Pixmeo SARL, Bernex, Switzerland).
Ethical approval. All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. Animal handling procedures were carried out following approved protocols by the Baylor College of Medicine Institutional Animal Care and Use Committee (IACUC).