Identification of Cellular Infiltrates during Early Stages of Brain Inflammation with Magnetic Resonance Microscopy

A comprehensive view of brain inflammation during the pathogenesis of autoimmune encephalomyelitis can be achieved with the aid of high resolution non-invasive imaging techniques such as microscopic magnetic resonance imaging (μMRI). In this study we demonstrate the benefits of cryogenically-cooled RF coils to produce μMRI in vivo, with sufficient detail to reveal brain pathology in the experimental autoimmune encephalomyelitis (EAE) model. We could visualize inflammatory infiltrates in detail within various regions of the brain, already at an early phase of EAE. Importantly, this pathology could be seen clearly even without the use of contrast agents, and showed excellent correspondence with conventional histology. The cryogenically-cooled coil enabled the acquisition of high resolution images within short scan times: an important practical consideration in conducting animal experiments. The detail of the cellular infiltrates visualized by in vivo μMRI allows the opportunity to follow neuroinflammatory processes even during the early stages of disease progression. Thus μMRI will not only complement conventional histological examination but will also enable longitudinal studies on the kinetics and dynamics of immune cell infiltration.


Introduction
Inflammatory diseases of the central nervous system (CNS) such as multiple sclerosis (MS) involve a recruitment of immune cells during the early stages of pathogenesis, prior to the onset of clinical symptoms [1][2][3]. Normally the blood-brain barrier (BBB) restricts migration of immune cells to the CNS, but during inflammation its function becomes altered. Immune cells gain access to CNS parenchyma via a complex, multi-step process that involves crossing both the vascular endothelium and the glia limitans [4,5]. The indirect detection of contrast-enhancing lesions (CEL) by Magnetic Resonance Imaging (MRI) at the site of BBB disruption as a result of contrast agent leakage into the CNS parenchyma is used as a primary end point in MS clinical trials [6,7] and in the EAE mouse model [8,9]. However, BBB disruption does not provide direct evidence of immune cell trafficking into the CNS [10], and may occur independently of the formation of new lesions [11]. There is therefore a need to pursue supplemental MRI techniques, to gain a more accurate and comprehensive view of the pathogenesis of CNS inflammation. One strategy has been to employ iron oxide nanoparticles [12], particularly as a means of studying immune cell infiltration in the animal model of CNS inflammation, experimental autoimmune encephalomyelitis (EAE) [13][14][15][16]. However the application of paramagnetic nanoparticles is hampered by a number of limitations, including the lack of an a priori knowledge of the specific time of immune cell migration into the brain parenchyma.
Microscopic MRI (mMRI or MR histology) -defined as MRI with a spatial resolution ,100 mm [17,18] -is one means of amplifying image detail in order to observe even minor changes in brain pathology during the course of disease. MRI resolution depends on several factors including magnetic field strength, gradient strengths and digital resolution, but the main limiting factors are RF coil sensitivity and signal-to-noise ratio (SNR) [19]. Upon reducing voxel size to amplify spatial resolution, a loss in SNR is to be expected. This loss can be considerably compensated for by increasing signal averaging: this produces images with an impressive level of microscopic detail as shown in ex vivo fixed brain tissue samples [20]. However, increased signal averaging comes at a cost in scan time, and is hence not practical for studies with anesthetized animals. This along with the presence of movement artifacts makes it inherently difficult to achieve sufficient resolution to visualize brain pathology in vivo.
One approach to maximize effective spatial resolution is to increase signal sensitivity with cryogenically-cooled RF detection devices that boost SNR, the currency spent for image resolution and image quality [21]. Recently, RF coils made of superconducting material have been developed for animal micro-imaging, which reduce coil resistance and thermal noise and therefore increase SNR by up to a factor of up to 2.9 [22]. In essence such a system increases the field strength virtually by a factor of at least 2, according to MR principles, but without the disadvantages associated with higher field strengths, such as stronger susceptibility artifacts, B 0 inhomogeneities and shorter wavelengths. The effectiveness of this approach has been demonstrated in the healthy mouse brain [23].
Realizing the capabilities of mMRI, we focused here on high spatial resolution imaging of mice brain during the course of EAE in an attempt to visualize inflammatory pathology in vivo prior to and during commencement of disease. The ultimate aim of this study was to distinguish cellular infiltrates in microscopic detail at an early phase of disease. In this first study applying cryogenic MR technology to EAE, we demonstrate that brain pathology can be detected even without the use of contrast agents and show excellent correspondence between mMRI findings and conventional histology.

Early detection of EAE pathology with high resolution cryogenic imaging
Using a cryogenically-cooled RF coil, we performed high spatial resolution brain imaging in EAE mice prior to and upon commencement of disease ( Figure S1). The mMRI performed enabled us to detect lesions (as defined in Methods) already prior to clinical manifestation of disease. Using a T 2 weighted (T2W) TurboRARE sequence that results in better contrast between grey and white matter boundaries, we observed lesions in multiple regions of the EAE mouse brains, particularly the cerebellum, cerebral cortex and subcortical regions (thalamus and striatum). Figure 1A shows the anatomical distribution of the detected lesions as well as the time point of their first detection (relative to the starting point of disease). Out of 9 immunized mice, all of which developed EAE, 7 mice exhibited lesions in the brain ( Figure S2). The average EAE score for all 9 mice was 1.760.7 (6 S.D.) and the average day of onset was 11.261.6 days (6 S.D.) with a range of 9-14 days post immunization. All mice were scanned 5 days post immunization; thus between 4-9 days prior to disease onset. Notably, lesions could be detected as early as 3 days (d-3) prior to the appearance of disease manifestations ( Figure 1A). No lesions were observed earlier than 3 days prior to clinical manifestations. Lesions in the cerebellum were the most common, occurring between 1 to 3 days prior to onset of disease ( Figure 1A, Figure 1B and Figure S3). The T2W images shown in Figure 1B and Figure  S3 represent the structural changes occurring in the cerebellum in a mouse during the 4-5 days prior to disease onset (d-5 -d0) in comparison to baseline measurements (d-14) that were carried out prior to immunization. Hyperintense lesions in the white matter of the cerebellar Arbor vitae were already evident 2 days prior to neurological symptoms (d-2, upper and middle arrow head) and prominent on d-1 (along upper and lower arrowheads). Signal extinction that is indicative of cellular involvement became apparent on d-1 (dotted islands) but was more prominent upon onset of clinical symptoms (d0). To illustrate the potential for quantitative assessment of the lesions, we generated images in which we subtracted control baseline scans from scans showing brain pathology as changes in signal intensity ( Figure S4A). Hypointense lesions on T2W images represent cellular infiltrates Early during disease progression, even prior to the onset of clinical manifestations, we also observed lesions in the cortex. Figure 2 illustrates a representative mouse at onset of clinically observable signs of disease. Although major anatomical structures including the corpus callosum, striatum, ventricular system and hippocampus could be clearly resolved with T 1 weighted (T1W) scans (Figure 2A), no inflammatory pathology was observed with this method. T2W imaging, on the other hand, revealed focal hypo-intense lesions in multiple regions of the inflamed brains, mostly in the cerebellum ( Figure 1B) but also in the somatosensory cortex ( Figure 2B). T 2 * weighted (T2*W) imaging revealed the signal extinctions observed by T2W imaging as punctate lesions, with greater detail and in association with intracortical vessels ( Figure 2C and Figure 3A). We observed venous irregularities within the cortex in 3 out of the 7 mice that exhibited lesions in the brain. The irregularities were observed one day prior to symptoms or upon initiation of symptoms. Processing of T2*W phase maps for susceptibility weighted imaging (SWI) [24] has been employed in MS lesions to reveal superior detail in small anatomical structures such as the venous vasculature and structural variations within gray and white matter [25]. SWI of the phase maps derived from T2*W imaging yielded excellent visualization of the small venous irregularities and enhanced contrast between normal tissue, focal inflammatory lesions and the microvasculature as highlighted in Figure 3B. To make a more objective assessment of these structural changes we subtracted the images showing pathology from images obtained at baseline, prior to EAE immunization ( Figure S4B). Notably, the structural irregularities observed in T2W or T2*W images corresponded with areas of inflammatory pathology -both in cerebrum ( Figure 3C) and cerebellum ( Figure 3D) -as revealed by conventional hematoxylin and eosin (H&E) histology. H&E stains corroborate the mMRI data and conform to the expected pattern of pathology in the EAE model. The EAE model has been extensively characterized, and it is well established that the lesions are comprised of immune cells, especially CD4+ T cells. Indeed, the lesions that we could visualize with mMRI show the same appearance and localization characteristics of CD4+ T cells ( Figure 3D). The clear correlation between the MRI data and the histology underscores the utility of microscopic MRI to reveal histologically relevant pathology in vivo [24,25].
Contrast enhanced lesions in close proximity to ventricular system Gadolinium contrast agents are frequently used in EAE studies, and Gd CEL have a typically diffuse appearance. T1W imaging post contrast (0.2 mmol/kg Gd-DTPA, Magnevist, Bayer-Schering Pharma AG, Berlin, Germany) revealed diffuse CEL ( Figure 4A and Figure 4B) in all 7 mice exhibiting hyper-or hypo-intense lesions in the brain. These CEL, although more diffuse, corresponded in location to the T2W lesions shown by mMRI ( Figure 1, Figure 2, Figure 3). This result corroborates the interpretation that the changes in signal intensities detected by T2W and T2*W microstructural MRI demonstrates the potential of microscopic MRI to reveal brain pathology with greater precision and detail than conventional contrast agents.
In EAE mice with lesions close to the ventricles, administration of Gd-DTPA enabled the visualization of the ventricular system ( Figure 4C). The contrast-enriched ventricular system was visualized three-dimensionally together with the brain vasculature by performing a 29-minute 3D FLASH sequence with flow compensation (GEFC), typically employed for MR angiography, with an isotropic spatial resolution of 59 mm ( Figure 4C). An overlay of the maximum intensity projections (MIP) of the 3D GEFC images with the MIP of post contrast T1W MDEFT CEL images shows the spatial relationship of the contrast-enhanced parenchyma (due to BBB disruption) with the ventricles and cerebral vasculature ( Figure 4D). The visualization of the ventricular system together with the vasculature post contrast was only visible in EAE mice with periventricular lesions and was not visible in healthy non-immunized mice ( Figure S5). While the precise mechanism still needs to be determined, these results suggest that an inflammation-induced disruption of the bloodcerebrospinal fluid barrier might lead to extravasation of contrast agent into the ventricular system.

Discussion
Experimental autoimmune encephalomyelitis is an animal model widely employed to examine the basic biology of CNS inflammatory processes, and to evaluate the effectiveness of nascent therapeutic approaches for multiple sclerosis [26][27][28][29][30][31]. In particular, EAE studies focusing on the early stages of brain inflammation have revealed new insights on the mechanisms involved in the different waves of immune cell entry during evolution of disease [32].
To gain a comprehensive and longitudinal view of brain inflammation -particularly during the early stages of EAEmethods employing high resolution brain imaging would be advantageous. Several studies in EAE employing MRI have significantly contributed to our understanding of the pathology of brain inflammation [8,16,33,34]. However, in longitudinal MRI studies (eg. to follow pathology during the course of EAE) one main limiting factor is a restriction in the number of animals that can be employed due to the length of each scanning procedure. To overcome this, methods that increase SNR will be beneficial to enable shorter scans. The gain in SNR achieved by cryogenicallycooled systems that reduce thermal noise [22] is a major advantage for studies with animal models for two main reasons: (i) it enables the acquisition of images with greater spatial resolution within reasonable scan times; and (ii) it facilitates the imaging of larger numbers of animals with standard spatial resolution but within shorter times. Depending on the design of the study it is possible to compensate between both strengths.
The advantages and feasibility of cryogenically cooled coilsparticularly to image mouse brain anatomy -has already been demonstrated at 4.7T [35] and 9.4T [22,23]. The scan time required using a conventional room-temperature coil to observe early pathological changes in the EAE brain (23 min for T2W imaging, 24 min for T2*W imaging) was at least three times longer than that of the currently employed cryogenically cooled coil (5 min for T2W imaging, 8 min for T2*W imaging) using the same magnetic field strength, considering the necessity to increase the number of excitations (NEX = 16 for RT coil compared with NEX = 1 for cryo coil) ( Figure S6).
Handling of the currently available MRI cryogenic RF coilswhich evolved from NMR-spectroscopy [21] -has significantly improved in recent years, and indeed is quite similar to any conventional RT RF-coil. One major drawback of the cryogenic probe used in this study is the transmission(TX)/reception(RX)coil design, which leads to an inhomogeneous excitation field [22,35]. The non-uniform excitation field needs to be taken into account when applying quantification techniques such as relaxometry or spectroscopy, as well as magnetisation transfer ratio (MTR). Due to the variation of the flip-angle throughout the sample, the reference gain has to be carefully adjusted for the volume of interest to minimize these changes. In this case a RXonly design with a conventional actively de-tuned TX volume resonator for a homogeneous excitation would be beneficial [36].
The effective anatomical resolution we employed in this study is almost an order of magnitude higher than that attainable in human brain imaging using clinical MRI scanners. Using this technology we observed structural changes in EAE brains -in regions including the cerebellum and cerebral cortex -prior to the onset of neurological symptoms. These detailed and highly resolved morphological changes were corroborated by subsequent histological examination, which showed clear evidence of infiltrated immune cells. The general trend in lesion distribution - with the cerebellum being the predominant site of damage in the brain -is in agreement with the ascending pattern of EAE. Disease progression in EAE shows a typical course of ascending paralysis, accompanied by inflammatory pathology that proceeds from the spinal cord initially affecting the tail and hind limbs, and subsequently precipitating complete paralysis. With this cryogenically-cooled system we were able to apply imaging protocols for high spatial resolution (as small as 356356400 mm), good grey matter-white matter distinction and high SNR with scan times not exceeding 15 min. The entire scanning protocol presented herein requires at most 60 min pre contrast and 60 min post contrast, and is therefore suitable for both longitudinal and cross-sectional in vivo studies, with acceptable sample sizes. Importantly, we were able to identify the presence of inflammatory cells within the brain parenchyma without the necessity of Gd-DTPA contrast agents. In contrast to the typical diffuse lesions revealed by Gd enhancement, the lesions detected by mMRI were more focused and more clearly demarcated. This represents a clear advantage of the cryogenic coil over the previous technology, as the ability to detect focal lesions by mMRI offers the potential for more precise quantification of inflammation severity, leading to more quantitatively robust EAE studies. However contrast agents are still powerful tools since they provide clear evidence of BBB disruption, although leakage of contrast agent in the parenchyma does not give definitive evidence of parenchymal damage. In situations where structural changes observed in the brain with mMRI are not accompanied by contrast enhancement, it is reasonable to speculate that immune cells may have penetrated the vascular endothelium thereby entering the pervascular spaces, but have not yet breached the glia limitans and infiltrated into the brain parenchyma. Such perivascularrestricted lesions are widely seen at the histological level in EAE but have not yet been detected by MRI. The possibility of mMRI with the cryogenic coil to detect such lesions, which are by definition not detectable by Gd leakage into the parenchyma, offers the potential for future investigations to observe pathological processes that would otherwise be invisible, and underscores the importance of this technology for MS research.
Another distinct advantage of the cryogenic RF coil for MS research is the possibility to include magnetic resonance angiography in the scanning protocol, in a quick and efficient manner, to monitor changes in the vasculature over the course of disease. Indeed, the present study represents the first application of MR angiography in the mouse EAE model. The observation of leakage of contrast agent into the ventricular system early in disease is interesting, in light of recent reports highlighting the importance of the choroid plexus [30] and the involvement of the circumventricular organs in EAE [37].
In summary, the visualization of cellular infiltrates by in vivo mMRI provides an opportunity to follow neuroinflammatory processes throughout disease progression. Thus in parallel to conventional histological examination and in combination with contrast agents, mMRI will be invaluable for longitudinal studies investigating immune cell infiltration during brain inflammation and evaluation of novel therapeutics. Future directions for mMRI studies in EAE will involve the application of techniques -such as T2 mapping and MTR -that provide valuable information on water/lipid content and myelin integrity during lesion development. These techniques have already been successfully employed in a model of oligodendrogliopathy [38,39] and will complement MR methods typically employed in EAE studies since they provide further knowledge regarding the neurodegenerative component of the disease.

Ethics Statement
Animal experiments were carried out in accordance with the guidelines provided and approved by the Animal Welfare Department of the LAGeSo State Office of Health and Social Affairs Berlin (Permit G0172/10).

In vivo MRI
Shortly before and during the MR session, mice were anesthetized using a mixture of isoflurane as inhalation narcosis (0.5-1.5%), pressurized air and oxygen. All MRI-scans were performed on a Bruker Biospec 9.4T USR94/20 (Bruker, Ettlingen, Germany). Mice were imaged using a Transceiver Mouse Brain CryoProbe (Z106543, 400 MHz, Bruker BioSpin MRI, Ettlingen, Germany). The CryoProbe is a half-cylindricalshaped 2-channel transmit/receive quadrature-driven surface coil, for full mouse-brain coverage with a maximum field of view of c. 30620620 mm. The temperature of the mice was regulated at 37uC. The breathing rate and temperature was monitored by a remote monitoring system (Model 1025, SA Instruments Inc., Syracuse, New York, USA). Images were acquired using T 1weighted (T1W), T 2 -weighted (T2W) and T 2 *-weighted (T2*W) techniques. Slice positioning was kept fixed through longitudinal brain examination: axial slices were positioned parallel to the base of the brain, coronal slices were positioned perpendicular to axial slices and covering the brain from the olfactory bulb/frontal lobe fissure to the cervical spinal cord.

MRI sequences
Pre and post contrast T1W imaging was done with a Modified Driven-Equilibrium Fourier Transformation sequence (MDEFT) (3D MDEFT: TR/TE/TI: 3000/3.9/950 ms, FA 20u, matrix 3846384). Horizontal sections of the entire mice brain were performed in 11 min at a spatial resolution of (476476400) mm 3 . Coronal sections were performed in 15 min at a spatial resolution of (476476500) mm 3 . As T 1 -enhancing contrast agent gadolinium diethylenetriamine penta-acetate (Gd-DTPA, Magnevist, Bayer-Schering Pharma AG, Berlin, Germany) was administered intravenously at a concentration of 0.2 mmol/kg. T2W scans (2D TurboRARE: TR/TE: 3000/43 ms, matrix 3846384) were performed in 5 min with geometry and spatial resolution identical to that of T1W imaging. To image brain tissue in association with the microvasculature we applied a T2*W multislice fast low angle shot (2D FLASH: TR/TE: 473/18 ms, FA 40u, matrix 5126512) sequence with an in plane resolution of (35635) mm 2 . For horizontal scans we used a slice thickness of 400 mm and 16 slices for full brain coverage with a scan time of 8 min and for coronal scans we used 22 slices of 500 mm with a scan time of 11 min to image the whole mouse brain. For mice showing leakage of contrast agent, a 29 min 3D FLASH sequence with flow compensation (GEFC: TR/TE: 30/5.9 ms, matrix 51262566 256) was applied, for visualization of the vasculature and ventricular system 1 hr after contrast injection at an isotropic spatial resolution of 59 mm.
The following MRI sequences were used for the baseline scans, prior to immunization: TurboRARE, MDEFT, 2D FLASH and 3D GEFC. Five days post-immunization, the mice underwent daily horizontal and coronal T2W (TurboRARE) scans. If no lesions or clinical signs were observed, the mouse was returned to the home cage and scanned again the next day. If hypo-or hyperintense lesions were detected, then 2D FLASH and MDEFT precontrast scans of that mouse were done immediately. The mouse was then injected i.v. with Gd-DPTA, and post-contrast MDEFT, TurboRARE and 3D GEFC scans were done (15,45, and 55 minutes following contrast injection, respectively). If the mouse did not show clinical signs, regardless of the observation of lesions, it was returned to the home cage and scanned again the next day. If the mouse did show clinical signs, this was the pre-defined endpoint, the mouse was sacrificed on the same day, and the brain extracted for histological comparison. Given the inherent variability in the time of onset of EAE, we elected to use this design ( Figure S1), customized for each individual animal, rather than use a pre-selected timetable.

Image analysis
Images from longitudinal experiments in mouse EAE were analyzed by three individuals who were blinded towards disease activity of corresponding mice. Lesions on T2W images were defined as described for T2W lesions in MS patients [40]. Briefly, lesions had to be clearly visible, non-artifactual areas of change in the signal intensity (signal enhancement or extinction on the grey background) of T2W images compared to baseline T2W images (taken prior to EAE immunization). To generate susceptibility weighted images post processing of the phase maps derived from T2*W imaging for was performed by fully-automated reconstruction using ParaVision 5.1 (Bruker, Ettlingen, Germany).

Histology
After terminal anesthesia, mice were transcardially perfused with 20 ml PBS, then with 20 ml zinc fixation solution (0.5% zinc acetate, 0.5% zinc chloride, 0.05% calcium acetate). Brains were then extracted and subsequently post-fixed in zinc solution for 3 d at room temperature. The tissues were then cryoprotected by incubation overnight at 4 degrees in 30% sucrose in PBS, then embeded in O.C.T. and frozen in methylbutane with dry ice. The tissues were cut into 12 um sections on a cryostat, and stained with hematoxylin and eosin according to standard procedures. For immunostaining, tissue sections were blocked with avidin, biotin, and normal goat serum, then incubated overnight at 4uC with rat anti-mouse CD4 antibody (Invitrogen). The sections were then incubated with biotinylated goat anti-rat IgG antibody (Vector Laboratories), then streptavidin-conjugated peroxidase, and visualized with Vector NovaRED Peroxidase substrate (Vector Laboratories) and counterstained with hematoxylin. Figure S1 Flow chart illustrating the design of the mMRI study in EAE mice. Following a baseline scan, mice were immunized and scanned daily (horizontal and coronal T2W MRI) 5 d post-immunization. More intensive scans were performed and contrast (Gd-DPTA) was i.v. applied when lesions were detected with T2W MRI. (TIF) Figure S2 EAE scores compared with day of lesion occurrence. The neurological score of each mouse is plotted against the day of first lesion occurrence (d-1 denotes that lesions were first observed one day prior to onset of symptoms).