Enhanced Fluorine-19 MRI Sensitivity using a Cryogenic Radiofrequency Probe: Technical Developments and Ex Vivo Demonstration in a Mouse Model of Neuroinflammation

Neuroinflammation can be monitored using fluorine-19 (19F)-containing nanoparticles and 19F MRI. Previously we studied neuroinflammation in experimental autoimmune encephalomyelitis (EAE) using room temperature (RT) 19F radiofrequency (RF) coils and low spatial resolution 19F MRI to overcome constraints in signal-to-noise ratio (SNR). This yielded an approximate localization of inflammatory lesions. Here we used a new 19F transceive cryogenic quadrature RF probe (19 F-CRP) that provides the SNR necessary to acquire superior spatially-resolved 19F MRI. First we characterized the signal-transmission profile of the 19 F-CRP. The 19 F-CRP was then benchmarked against a RT 19F/1H RF coil. For SNR comparison we used reference compounds including 19F-nanoparticles and ex vivo brains from EAE mice administered with 19F-nanoparticles. The transmit/receive profile of the 19 F-CRP diminished with increasing distance from the surface. This was counterbalanced by a substantial SNR gain compared to the RT coil. Intraparenchymal inflammation in the ex vivo EAE brains was more sharply defined when using 150 μm isotropic resolution with the 19 F-CRP, and reflected the known distribution of EAE histopathology. At this spatial resolution, most 19F signals were undetectable using the RT coil. The 19 F-CRP is a valuable tool that will allow us to study neuroinflammation with greater detail in future in vivo studies.


Methods
Radio frequency coils. The performance of a novel transceive 19 F cryogenic quadrature RF surface probe at 9.4T ( 19 F-CRP, f ~ 376 MHz) was compared to a dual-tunable 19 F/ 1 H volume resonator (φ in = 18.4 mm, l total = 39 mm), previously developed for imaging mouse brain inflammation 11 . The 19 F-CRP has a similar geometry to the existing Bruker 1 H quadrature CryoProbes 24 . The rectangular transceive copper coil elements are overlapping side-by-side on a cylindrical surface (r ~ 11 mm, axis parallel to the main magnetic field direction). The outer dimensions (O.D.) of one coil element are: 16 × 20 mm 2 [arc length (φ × z)] and the total O.D. are: 27 × 20 mm 2 [φ × z]. The 19 F-CRP operates at ~28 K with a dual cooled preamplifier at the base running at ~77 K. Constant cooling is ensured by a closed loop system connected to a remote cryo-cooler. The RF coil is thermally insulated by a vacuum separating it from the surrounding ceramic finger (Fig. 1A). The outer surface of the RF finger is equipped with a temperature sensor and kept at a temperature of choice (35 °C) using a resistive heater. The SNR gain of this CRP relative to a RT coil with similar geometry is expected to be comparable to existing 400 MHz proton CryoProbes 24, 25 . Experimental setup. To evaluate the 19 F-CRP performance, three different phantom-setups were prepared  19 F signal sensitivity as a function of the number of 19 F atoms. Nanoparticles were prepared by emulsifying 1200 mM PFCE (Fluorochem, UK) with Pluronic F-68 (Sigma-Aldrich, Germany) using a titanium sonotrode (Sonopuls GM70, Bandelin, Germany) as previously described 26 . The PFCE nanoparticle stock was then diluted to 25 mM, 50 mM, 100 mM, 200 mM, 400 mM and 600 mM nanoparticle suspensions. NMR tubes containing different nanoparticle concentrations were placed below the CRP using a spacer of 0.75 mm thickness to mimic the distance of the mouse brain from the CRP surface in in vivo applications.
Setup 3 (mouse brain): Ex vivo tissues from fixed EAE mice embedded in 15-ml tubes, for comparing 19 F signal sensitivity and anatomical detail. All experiments were conducted in accordance with procedures approved by the Animal Welfare Department of the State Office of Health and Social Affairs Berlin (LAGeSo), and conformed to national and international guidelines to minimize discomfort to animals (86/609/EEC). EAE was induced as described previously 11 in SJL/J mice (n = 6, female, 6-8 weeks old). Five days following EAE induction, mice were administered nanoparticles (10µmol PFCE) intravenously each day for 5 d as described previously 11 . EAE mice were transcardially perfused with 20 ml PBS followed by 20 ml 4% paraformaldehyde (PFA) following terminal anesthesia. Mice were cleared from external pelt, extremities, and abdominal tissues. Brain, spinal cord and neck lymphoid organs were preserved in situ within the skull and vertebral column. The tissues were transferred into a 15 ml tube filled with 4% PFA and stored at 4 °C.  29 . For single channel RF coils, intensity values of MR images follow a Rician distribution 30,31 . For a two-receiver, quadrature system ( 19 F-CRP), they follow a non-central chi distribution 32 . We estimated the true SNR from the S m and background σ m using where c σ is 0.655 (Rician) and 0.687 (chi), and the correction function f S is derived from the respective distribution's mean 30,32 . For Setup 2, a single SNR value was determined from the mean signal intensity over a central circular region-of-interest covering ~90% of pixels. The number of atoms per image pixel was estimated from nanoparticle concentration and voxel size.
Ex vivo mouse brain 19 19 F images (RT and CRP) to be spatially aligned with the RT 1 H images. For this, three repetitions of the RT 19 F scan were averaged to achieve sufficient 19 F signal with the RT-coil and an effective registration. Co-registration was applied using affine diffeomorphic image registration (12 degrees of freedom) by explicit B-spline regularization 33 , which is part of the Advanced Normalisation Tool (ANTs) 34 . Registration of the Allen brain atlas 35 to the 1 H image was achieved as follows: (1) 1 H image and atlas template were segmented in grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF) probability maps with SPMMouse (http://www. spmmouse.org/) 36 , (2) two synthetic images were generated with signal intensity in each voxel I(x,y,z)

Results
Transmit field characteristics of the 19 F-CRP. Since transceive surface coils do not achieve a spatially uniform excitation like volume resonators 24 , we assessed the B 1 + characteristics of the 19 F-CRP ( Fig. 2A) and quantified changes in FA. A profile plot of the FA along the vertical axis (Fig. 2B) reveals a strong FA decrease with increasing distance from the CRP surface. Across a distance of 10.4 mm the measured FA varies between 152° and 0°. From the nominal FA of 90° the actual FA deviates up to 50% within a range of 6.0 mm (1.5-7.5 mm from CRP surface).

SNR assessment in phantoms.
To study the SNR performance of the 19 F-CRP, we first used a high 19 F concentration (33% TFE solution) (Fig. 3A). The transversal spin-echo 19 F MR images demonstrate a homogenous SNR for the RT coil and a spatially varying SNR for the CRP (Fig. 3A). We adjusted the reference pulse power in order to avoid substantial signal loss at the dorsal side of the brain. Using this reference pulse power, the SNR reached its peak at a distance of 2.4 mm, where it was ~15-fold higher than the SNR of the RT coil (Fig. 3B). The SNR of both RF coils are approximately equal at a distance of 8.6 mm from the CRP.
We next investigated the detection limits for both coils by measuring 19 F nanoparticles, as a biologically relevant preparation. We employed concentrations of PFCE (25 mM-200 mM) yielding a range of 10 15 -10 18 19 F atoms per voxel (Fig. 4A). Qualitatively, we reached a detection limit in the order of 10 15 fluorine atoms using the 19 F-CRP, compared to 10 16 fluorine atoms with the 19 F RT-coil. Specifically, an SNR of 3.0 was achieved with (0.1 × 0.1 × 0.4) mm³ voxels of a 25 mM PFCE concentration (equating to 5.2 × 10 15 fluorine atoms) when using the 19 F-CRP. In contrast an SNR of 2.4 was achieved with (0.1 × 0.1 × 1.2) mm³ voxels of a 100 mM PFCE concentration (equating to 6.2 × 10 16 fluorine atoms) when using the 19 F-RT-coil. In both cases the measurement time was 36 s. MR images with an SNR value below 2 were not sharply defined. To estimate SNR provided by the 19 F-CRP compared to the 19 F/ 1 H RT-coil, we used SNR = 2 as a cutoff equating to ~5 × 10 16 (RT) and ~4 × 10 15 (CRP) fluorine atoms per voxel. Next we prepared higher concentrations of 19 F nanoparticles (200 mM to 1200 mM) to achieve SNR values well above 2, spanning a range of 10 17 -10 19 atoms per voxel. From these experiments we calculated an SNR gain of ~16 for the 19 F-CRP when compared to the 19 F/ 1 H RT-coil (Fig. 4B).
High spatially-resolved 19 F MRI. An important utilization of the SNR gain is to localize cell infiltrates in the brain with more detail. Previously areas of inflammation were detected using spatial resolutions greater than 600 μm 11 . Here we exploited the superior SNR of the 19 F-CRP, and used an isotropic spatial resolution of 150 μm. Ex vivo MR images obtained with the 19 F-CRP from an exemplary EAE mouse (day 10 following EAE induction, score = 1.25) show a more precise distribution of intraparenchymal inflammation. At this spatial resolution, the majority of the 19 F signals obtained by the 19 F-CRP were not detected with the RT coil ( Fig. 5A-C). In addition we show similar inflammatory patterns in a pre-symptomatic mouse, also sacrificed on day 10 following EAE induction (Supplementary Figure). Within the cerebellum, inflammatory infiltrates were mostly localized within the white matter of the arbor vitae, particularly near deep cerebellar nuclei (Fig. 5B). Clearly delineated inflammatory areas were found in grey matter regions running adjacent to white matter tracts in the cerebellum (Fig. 5A). This is consistent with the expected patterns of inflammation in the EAE model 39,40 , also as observed in our own prior studies 11,41,42 . Using the 19 F-CRP, we also observed strong 19 F signals in the cerebrum emanating from the striatum and pallidum appearing continuous with 19 F signals from the third ventricle (Fig. 5A). Additionally, clear extraparenchymal meningeal inflammation could be seen, consistent with recent reports [43][44][45] . Especially strong inflammatory signals were observed along the dorsal surface of the brain, including meningeal regions lining fissures between the cerebellar lobules. These inflammatory regions extended ventrally to the prepyramidal fissure, parafloccular sulcus and lateral recess of the fourth ventricle. A dominant 19 F signal was observed around the meninges lining the ventral part of the retrosplenial area of the cerebral cortex (Fig. 5B), spreading caudally towards the cerebellum, running in parallel to the superior sagittal sinus, and eventually the retroglenoid vein (Fig. 5C). In these experiments we focused on highly resolved inflammation imaging in the EAE brain, employing long acquisition times in order to compensate for the considerably lower 19 F signal sensitivity of the 19 F/ 1 H RT-coil. Since these acquisition times (11 h) are not applicable for in vivo studies, we performed further experiments in which we reduced the scan time. Upon reducing the scan time from 11 h to 0.5 h we could still detect 19 F signals with the 19 F-CRP (Fig. 6). Despite the clear differences we were nevertheless still able to detect a considerable 19 F signal, even with a scan of only 2 h, which is amenable for in vivo MRI.

Discussion
In this study we show first 19 F MR images obtained with a 19 F-CRP driven in quadrature mode. Compared to the 19 F/ 1 H RT-coil we previously developed 11 , we show that the 19 F-CRP facilitates superior ex vivo images of brain  inflammation in an animal model of MS. At the current stage of development the 19 F-CRP cannot yet be employed for in vivo imaging due to incompatibilities with conventional 1 H RT coils, as discussed later. Nevertheless the results are encouraging, and offer proof-of-concept demonstration of the potential for this technology. After introducing the concept of cryogenically-cooled RF coil hardware to reduce thermal noise and thus increase SNR 46 , CRP technologies were developed for small animal MRI, particularly for anatomical 1 H MRI of mouse brain 41,[47][48][49][50] . Introducing a quadrature CRP design, enabled further SNR gains (~2.5) at 400 MHz 24, 25 compared to RT coils with similar geometries. The SNR gain prediction for the 19 F-CRP is expected to be equivalent due to the close Larmor frequency (376 MHz at 9.4T).
The potential applications of 19 F MR methods to image inflammation have long been recognized [11][12][13][14][15][16][17] . For several years, neuroinflammation has been studied using gadolinium-based contrast agents. However, gadolinium-enhancing lesions are diffuse, and lack spatial precision. Improvements have been realized with the use of alternative contrast agents, such iron oxide nanoparticles, although their effects on magnetic susceptibility limit their discrimination from endogenous confounding artifacts. 19 F MR methods abrogate this, since 19 F signals derive exclusively from exogenously applied 19 F nanoparticles. Efforts have been made to boost 19 F signal e.g. by promoting 19 F nanoparticle cellular uptake 20 . Nevertheless, major challenges of signal sensitivity constraints remain. Improving 19 F sensitivity with the 19 F-CRP will be essential to realizing the full potential of 19 F MR.
Our motivation to investigate the 19 F-CRP was to increase the sensitivity to detect neuroinflammation. Considering the geometrical differences between both coils, it was imperative to measure SNR at locations below the CRP that correspond to the mouse brain, using phantoms spanning the entire coronal view, as a basis for future in vivo studies. We performed SNR measurements for both 19 F-CRP and control 19 F/ 1 H RT-coil using a spin echo sequence (RARE), commonly used for 19 F MRI due to its high SNR per unit time compared to spoiled gradient echo sequences.
The sensitivity of the 19 F-CRP is spatially dependent. Given that the CRP is a transceive quadrature surface coil array, both transmit field (B 1 + ) and receive sensitivity (B 1 − ) diminish with increasing distance from the RF coil -a factor that must be accounted for in quantitative imaging by measuring the actual B 1 and correcting the signal intensities using the signal equation of the employed pulse sequence. This is absolutely essential when signal quantification is necessary in order to ascertain the level of inflammation over the entire region of the brain during EAE. Nevertheless, this characteristic is shared by all transceive surface coils. This adverse effect is counterbalanced by an SNR gain, up to ~15-fold in the practical comparison made within this study. This SNR gain can be attributed to factors including cooling (in the range of 2-3 for 1 H 24, 25 ), differences in RF coil design (birdcage vs. surface coil; quadrature versus linear), RF coil sample loading, and the specific RF pulse power adjustments. Here, pulse power was adjusted in order to avoid substantial signal loss at the dorsal part of the brain, which is observed when using a RARE sequence with excessive RF power. Predicting the sensitivity and detection limits of 19 F measurements for specific hardware setups 51 will help facilitate further 19 F-CRP studies with other fluorinated compounds.
An SNR gain of 15 can be exploited in several ways -by reducing scan time by a factor 225 (e.g. from 1 h to ~15 s), or doubling 3D spatial resolution (e.g. from 600 µm to 300 µm) while still gaining SNR (~2.5). In this study we made use of the superior SNR, employing isotropic spatial resolutions of 150 μm to study neuroinflammation. Using the 19 F-CRP at this resolution, we gained more precise information regarding inflammatory cell localization in the brain, compared to our previous study 11 . The 19 F MR images with the CRP showed excellent correspondence with the typical pattern of histopathology 39,40 . A robust accumulation of inflammatory lesions, especially in the white matter tracts of the cerebellum, is a hallmark of EAE in SJL mice, which we also observed in our prior studies using high resolution 1 H MR 41,42 and low resolution 19 F MR 11 . The pathology also extends into the cerebrum, as shown both prior to the occurrence of clinical symptoms (Supplementary Figure) and also during ongoing clinical disease (Figs 5 and 6). The 19 F-CRP MR images also enabled discrimination of extraparenchymal meningeal inflammation, consistent with recent reports highlighting the relevance of inflammatory cell trafficking via the blood meningeal barrier 43,44 and extravasation via leptomeningeal microvessels into the subarachnoid space 45 . This also reflects the situation in MS [3][4][5] . Recent studies have argued for the presence of a lymphatic circulation in the meninges in association with these vessels, capable of draining immune cells from meningeal spaces 8 and brain parenchyma 7 into deep cervical lymph nodes. Therefore, the capacity to perform non-invasive longitudinal investigations with fidelity 19 F MRI to monitor the dynamics and distribution of infiltrating immune cells will be directly relevant for experimental neuroimmunologists. The gradient in the B 1 field of the 19 F-CRP leads to a gradual decline in 19 F MR signal with increasing distance from the probe head. This results in reduced signal in ventral regions. Studies of the EAE model are, in general, more focused on imaging of the CNS, and less so on imaging of the superficial lymph nodes. When imaging of the lymph nodes in the ventral regions is necessary, one could consider measuring the mouse brain in the supine and prone positions, in order to ensure coverage of the dorsal sides comprising the whole brain as well as ventral sides to include the draining lymph nodes. Other possible workarounds include adding an anterior 19 F RT RF coil to the mouse bed or combining 19 F images from RT and CRP. These approaches could help to overcome this inherent limitation of the 19 F-CRP, while still utilizing its superior SNR. While the spatial dependency poses a constraint for studies investigating the involvement of the draining lymph nodes, the translational applications of the 19 F-CRP are not limited to EAE. The 19 F-CRP will also be useful for studying brain inflammation in animal models of tumour growth (especially those tumours implanted in the cortex or striatum), and studies on the middle cerebral artery occlusion model of stroke. Inflammation in these preclinical models could readily be imaged, since the focus of pathology in these models is located in regions where the 19 F-CRP clearly outperforms the 19 F/ 1 H RT-coil.
In vivo 19 F MRI studies require acquisition of anatomical 1 H MR images within a reasonable time frame. A dual-tunable RF probe would be most ideal, in order to avoid inaccurate co-registration of both signals 52 . Despite the clear improvement in SNR of the 19 F-CRP, the quadrature design prohibits the presence of a dual resonant MR signal that would be needed for anatomical 1 H MRI. Furthermore conventional 1 H RF resonators cannot be used in combination with the 19 F-CRP due to coupling between both RF coils. To avoid this, the 19 F-CRP would need to be removed while the in vivo 1 H images are acquired. This would cause changes in the alignment of the mouse within the scanner during in vivo measurements that are serious enough to constitute a major hindrance. Even with the use of reference markers, any slight shift in the position of the markers with respect to the mouse during the procedure will result in an incorrect registration between 19 F and 1 H images. The current procedure of registering the 19 F images of the CRP with those of the RT RF coil is complicated and time consuming, requires sufficient SNR and is an impediment for in vivo experiments. A proposed solution to this limitation could be to construct an anterior 1 H RT RF coil, specifically designed to be added to the mouse bed while the 19 F-CRP remains installed, in order to provide anatomical guidance. A dual-tunable 1 H/ 19 F RT RF coil would also take into account the above approach (implementation of a 19 F RF-coil below the mouse head).
This study presents the first demonstration of the performance of a quadrature 19 F-CRP tailored for small rodents, showing superior SNR and 19 F MR image quality. The logical extension of this work will be to translate these results into in vivo studies, such as those studying pathological changes during neuroinflammatory disease. While the results of the current study are highly encouraging, a challenging road still lies ahead for the application of the 19 F-CRP in in vivo studies. Previous studies using 19 F MR have been seriously hampered by the low SNR, and compensating for this limitation by using low spatial resolution has generally yielded images with rather poor definition, and therefore limited scientific utility. The current study aims to improve this situation, bringing 19 Figure 6. High spatial resolution 19 F MRI using acquisition times feasible for in vivo imaging. 19 F MR images were acquired with the 19 F-CRP using acquisition times between 30 min and 11 h. The 19 F images were scaled to units of SNR, thresholded at SNR = 4, and overlayed onto the 1 H MR images using a pseudocolor scale.