Although MRI technique often shows rather high spatial resolution, high natural contrast, high penetration depth and is innocuous to the patient, it is commonly criticized for a lack of sensitivity. This drawback becomes of critical importance at low polarization fields, in small samples or animals, or in specific research areas such as molecular imaging, where poorly concentrated species are to be detected in restrained anatomical regions. A possible way to greatly enhance NMR signals is to use hyperpolarized molecules [1, 2] or gases [3, 4]. These methods have proved to be of great value but hyperpolarization is short-lived in this case, and long-term experiments in animals would require serial injections of hyperpolarized moiety. Another approach is to enhance the water signal measured on conventional MR systems by the so-called Overhauser effect  in the presence of an unpaired electron . The principle is to saturate an Electron Paramagnetic Resonance (EPR) transition, which leads to dynamic nuclear polarization (DNP) of water protons present in the vicinity of the unpaired electrons . The Overhauser effect in liquids has been used for MRI purposes, known as PEDRI (proton electron double resonance imaging) [8-10] or OMRI (Overhauser magnetic resonance imaging) [11, 12].
In vivo, OMRI was first performed by Grucker and colleagues  in the abdomen (kidney) of rats at 6.8 mT after an intraperitoneal injection of Fremy's salt. Images were acquired in two dimensions without slice selection. In further studies several free radicals were used, depending on applications (bio-distribution, pH or oxymetry measurements), but images of mouse or rats were again acquired in two dimensions with slice thicknesses that were not lower than 5 mm [11, 13-18]. Molecular imaging applications using OMRI detection of free radicals would obviously require much higher spatial resolution in 3D with nearly isotropic voxels. This can be achieved at the expense of signal-to-noise ratio (SNR), which is usually enhanced by increasing the polarization field. High magnetic fields cannot be used during electron saturation because of excessively high irradiation frequencies that would induce sample heating and low penetration depths. Field-cycling [9, 12, 19-23] with a low field for EPR and a higher field for MRI was used to limit power deposition, but such an approach, although promising, remains to be validated in well-resolved 3D images. Alternatively, an original mouse-compatible setup for OMRI at constant magnetic field (0.2 T) was developed in a previous work . It used a 5.43 GHz cavity for electron excitation in a clinical C-shaped MRI system. In vitro, Overhauser enhancements in nitroxide solutions of more than 50 were observed without significant heating of the samples. It was proposed to use OMRI to generate a specific contrast with free-radical probes whose EPR signature depends on a particular biochemical activity, in order to follow in vivo biological unknown processes such as proteolysis. As stated above, such a molecular imaging application requires high-quality images.
Here the feasibility of 3D-OMRI at 0.2 T in vivo was evaluated in the brain of living mice with implanted glioma after TOPCA (1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrole-3-carboxylic acid) injection. This free radical was chosen because it is well tolerated by living mice (our observations) and its EPR spectrum can also be modulated for future molecular imaging purposes. The usefulness of the method for kinetics studies was also investigated.
MATERIALS AND METHODS
TOPCA was purchased from Sigma-Aldrich (France) and dissolved to a final concentration of 240 m m in phosphate-buffered saline. The longitudinal relaxivity (r1) of TOPCA was evaluated at 0.194 T in water using inversion-recovery measurements as a function of radical concentration. The r1-value of TOPCA in water at 20 °C was 0.35 ± 0.04 s−1 m m−1.
Brain tumors were implanted in nude mice (18–20g, n = 6, Charles River, L'Arbresle, France) by stereotaxic injection in the striatum site of 2 × 105 C6 glioma cells, derived from N-nitrosomethylurea-induced rat glioblastoma. Tumor cells were implanted around 10 days before the MRI study. This protocol has been fully described in a previous work  and was approved by the University Animal Ethics Committee. Tumor localization was checked using high-resolution MRI at 4.7 T  before performing OMRI at 0.2 T.
For MRI, mice were anesthetized with 1.5–2% isoflurane (Centravet, La Palisse, France) mixed in air. Mice were placed in a supine position, lying on a home-made thermo-regulated bed that maintained body temperature. Temperature was continuously monitored with a rectal probe. In order to evaluate hyperfrequency-induced heating, an additional probe was introduced in the mouth of two mice (STF Fast Response Immersion Probe; Luxtron, Santa Clara, CA, USA). A respiratory sensor was placed on the abdomen to monitor animal state throughout the experiment.
A heparin-coated catheter was placed in the tail vein for free radical injection and 200 µl of TOPCA at 240 m m was manually injected for 30 s prior to OMRI acquisition. TOPCA injection did not induce any animal loss. Mice were then sacrificed 2 days after OMRI experiments to avoid animal pain due to tumor growth.
The experimental setup was already described . A C-shaped MRI system, Magnetom Open Viva operating at 0.194 T (Siemens, Erlangen, Germany) was used. The proton frequency was 8.24 MHz. Gradient strength was 15 mT m−1. Electron spin excitation was induced by a resonant hyperfrequency (HF) cylindrical cavity (Bruker, Wissembourg, France) running in TE011 mode, positioned at the center of the MRI magnet bore . Its geometry (240 mm diameter, 29 mm width) was designed to reduce the electric component of the electromagnetic HF field in its center, thus minimizing sample heating upon microwave emission. The usable magnetic component was concentrated at the center of this cavity where an opening from both sides allowed sample access. The sample area at the center of the cavity was 28 mm in diameter and 29 mm in length. The original microwave cavity was modified, replacing bulky copper parts close to the imaging volume with synthetic materials (polyvinylchloride, polyepoxyde), coated with 35 µm copper inside the cavity. As a result, eddy currents were strongly reduced, allowing the use of standard fast MR gradient-echo sequences in 2D and 3D. Tuning and matching adjustments of the EPR cavity were carried out with a network analyzer (Wiltron, Anritsu, Kanagawa, Japan) measuring the standing wave ratio. As assessed from the analog meters of the final stage microwave amplifier, the reflected power never exceeded 1.5 W. Actual power deposition was evaluated by measuring temperature elevation with a temperature probe placed in a phosphate-buffered saline phantom (25 mm in diameter). By neglecting conduction, a heat equation was used to fit the initial linear temperature changes upon microwave excitation. Peak power was estimated to be in the range of 4 W, corresponding to an averaged value of about 1.2 W in 3D-OMRI experiments in mice (see below). The quality factor of the loaded (QL) and unloaded (QU) cavity was also measured and was used to evaluate the absorbed power, which was in the range of 10% of the incident power. The forward power value displayed by the power meter was in the range of 60 W, leading to a maximal absorbed power of 6 W. This value, even though in agreement with temperature measurements, is probably overestimated because of power losses brought by the transmission line (co-axial cable and tuning/matching circuitry).
A home-made MRI transmit–receive saddle-shaped coil (20 mm diameter and 26 mm length) dedicated to mouse brain was centered in the EPR cavity and connected to the MRI system. Both EPR saturation and MRI acquisition were synchronized by an external pulse generator. The electron frequency used in OMRI experiment was 5435 MHz corresponding to the central EPR line of the TOPCA nitroxide at 0.194 T. All MRI adjustments were done manually. Water line width at half-height was around 20 Hz for the whole mouse head.
3D OMRI in vivo
Optimized positioning of the mouse head for OMRI experiments was assessed from scout views acquired with standard 2D multi-slices, multi-orientations and fast 3D gradient echo sequences. High-resolution anatomical images without DNP were then acquired at 0.194 T with a 16 min duration 3D FLASH gradient echo sequence using the following parameters: TE/TR = 14/60 ms, flip angle 30°, receiver bandwidth 26 Hz per pixel, matrix size 64 × 64 × 32; field of view (FOV) 26 × 26 × 16 mm, actual spatial resolution 410 × 410 × 500 µm.
OMRI was performed with a 3D gradient echo sequence with the following parameters: TE = 10 ms/TR = 200 ms, flip angle 70°, receiver bandwidth 52 Hz per pixel, echo-shift to 25% of readout period, FOV 26 × 26 × 16 mm, matrix size 64 × 64 × 16, actual spatial resolution 410 × 410 × 1000 µm. Details of the OMRI acquisition protocol are given in Fig. 1. When applied, EPR saturation occurred for 120 ms immediately prior to the MRI acquisition (Fig. 1A).
In order to limit animal heating, the acquisition of the 3D data set was split into three blocks separated by 1 min rest interval. To ensure steady state of Overhauser-enhanced 1H magnetization in the second and the third block, 10 dummy scans with EPR excitation were carried out for 2 s before NMR acquisition. Moreover only the central lines of the k-space (32 phase-encoding steps out of the 64 in the phase-encoding direction) benefitted from electron irradiation (Fig. 1C). The loop counter of slice encoding was incremented prior to the loop counter of phase encoding.
The complete acquisition of an OMRI 3D data set needed less than 5 min. A control image without DNP was acquired before TOPCA injection. TOPCA was injected at the beginning of the OMRI protocol depicted in Fig. 1(C), when no electron saturation occurred. The receiver gain was constant throughout the whole experiment.
Serial 2D-OMRI in vivo
Kinetic measurements of signal enhancement were achieved within a single slice using a 2D NMR gradient echo sequence synchronized with electron saturation. The acquisition parameters were: EPR pulse length 160 ms, TE = 10 ms, TR = 200 ms, flip angle 70°, receiver bandwidth 52 Hz per pixel, matrix size 64 × 64, FOV 22 × 22 mm, slice thickness 3 mm, spatial resolution 344 × 344 × 3000 µm, acquisition time 15 s per image. TOPCA injection was immediately followed by the first OMRI acquisition. Images were then repeatedly acquired at a maximum rate of one per minute for 6 min and at increasing time intervals thereafter. A control image without EPR was acquired before TOPCA injection. A similar acquisition protocol was also used in one experiment with intra-peritoneal injection. The acquisition parameters were then EPR pulse length = 420 ms, TE = 28 ms, TR = 500 ms, flip angle 90°, receiver bandwidth 78 Hz per pixel, matrix size 64 × 32; FOV 20 × 20 mm, slice thickness 5 mm, spatial resolution 313 × 625 × 5000 µm, acquisition time 20 s per image. One minute after the end of TOPCA injection five images were acquired every minute. Then images were acquired every 3 min for a total duration of 1 h. The EPR saturation time of 420 ms was chosen to achieve a better electron saturation in order to favor detection of enhancements generated by very low plasmatic radical concentrations expected with an intra-peritoneal injection route.
Parametric image calculations were carried out with IGOR Pro (Wavemetrics, Lake Oswego, OR, USA). Signal-to-noise and signal enhancement were evaluated with OSIRIX open-source imaging software. In a given region of interest (ROI), signal enhancement was calculated as the ratio Son/Soff with Son the signal measured on images acquired with EPR and Soff the signal measured without EPR. Mean ROI size was 2.1 ± 0.7 mm2 in the tumor and 2.1 ± 0.9 mm2 in the brain vessels area. Data are given as means ± standard deviation.
Animal care and specific absorption ratio considerations
The rectal temperature elevation owing to microwave effects was 1.0 ± 0.6 °C during a 3D OMRI experiment and returned to baseline within a few minutes. Under such conditions the head temperature, as assessed from the mouth probe, could be acutely elevated by 2–4 °C during microwave excitation for maximum time periods of 30 s. An increase in respiration rate was noticeable at the end of EPR irradiation and returned to basal value within 1 min. No observable sign of pain or reduced activity was present when mice were returned to their cages.
Figure 2 shows anatomical slices extracted from a typical OMRI 3D data set at 0.2 T of a mouse brain after TOPCA injection (Fig. 2A–C) and the corresponding slices extracted from the high-resolution anatomical image (Fig. 2D–F) measured at the same field. In the latter the SNR was 18, which allowed a correct delineation of major brain structures. The signal enhancement after TOPCA injection was easily observed in tumor region and in large deeply seated vessels as shown in Fig. 2(A–C). Hypersignal appeared in the tumor and was surrounded by a hypo-intense boundary, in which the Overhauser effect magnitude was low. Mean SNR in the tumor and in carotid arteries were respectively 27.1 ± 3.7 and 52.7 ± 23.6 in six mice. In reference images without DNP, the proton signal was low and mean SNR at the same locations were 10.8 ± 1.2 and 7.8 ± 0.8, respectively. Mean signal enhancements were then 2.5 ± 0.3 in tumor and 6.8 ± 2.9 in carotid arteries. Of note, free radicals could be detected in other regions such as salivary glands or in the area of the orbital cavities. Thanks to the high contrast, it was possible to calculate parametric images of signal enhancement from the 3D data sets measured with and without EPR. Examples of slices extracted in the tumor and the vessel regions are depicted in Fig. 3. Interestingly, tumor sizes assessed from Overhauser-enhanced imaging were in agreement with those observed with conventional high-resolution imaging at 4.7 T.
Two-dimensional OMRI in living mice could be performed at a time resolution of 15 s. For animal safety considerations, 2D images were acquired every minute after TOPCA injection to assess the time-course of signal enhancement in an axial slice encompassing healthy brain, tumor and large vessels. Figure 4 shows the resulting curves in a mouse 14 days after tumor cell implantation. Before TOPCA injection SNR in tumor and arteries were quite similar: respectively 3.7 and 3.0. As expected, high enhancements (6-fold increase) appeared in arteries immediately after injection and TOPCA was rapidly cleared by renal elimination with a half-time of about 3 min. In the tumor, the signal reached its maximum value (2-fold increase) about 3 min after injection and decayed more slowly. When TOPCA was injected intraperitoneally, signal enhancement occurred simultaneously in the tumor and in the major brain vessels, as expected in this case (images not shown). Maximum signal was observed 19 min after the end of TOPCA injection. Enhancement values were 2.3 and 1.4 in the tumor and in major vessels, respectively.
For future biological application of OMRI it seems mandatory to achieve high spatial resolution for proper anatomical localization of radical species that generate Overhauser enhancement. In small animals image quality requires resolution greater than 1 mm in three dimensions. Conventional high-resolution MRI uses both high-field gradients (greater than 100 mT m−1) and high polarization fields (7.0T or more) to overcome the lack of proton signal. Such high fields cannot be used for OMRI in vivo because of hyperfrequency penetration depth considerations. Conversely MRI at very low fields is unable to provide well-defined images of the mouse body. This is the reason why an intermediate constant field of 0.2 T was chosen, for both conventional MRI and dynamic nuclear polarization in living mice.
In this study, Overhauser-enhanced 3D images of mouse brain were acquired in vivo with a resolution of 400 × 400 × 1000 µm3 in 5 min at 0.2 T. Such images could be superimposed on 3D anatomical images acquired at 500 µm resolution in each dimension.
The combination of good resolution and Overhauser enhancement was made possible with a microwave cavity whose geometry was designed to minimize sample heating upon microwave emission . At this microwave frequency, 5435 MHz, the penetration depth was in the range of the anatomical region explored, and it was shown here that nitroxide present in deeply seated arteries could be detected. The HF mode used here, namely TE011, was employed because it concentrated the magnetic field component in the central part of the cavity where the mice were positioned. It had also the advantage of minimizing the electrical component of the HF field in this area, thus minimizing animal heating. As assessed from phantom experiments, the average power deposition was in the range of 1 W in OMRI experiments, taking into account the duty cycle of microwave irradiation. Mean rectal temperature elevation was about 1 °C in mice and was in agreement with previous studies [15, 16]. Temperature elevation in the mouse head was of course higher (2–4C) over short periods, but this did not seem to compromise animal safety. In other studies high incident power levels (up to 90W) were also used to perform OMRI in living mice [12, 26], without apparent animal overheating.
Here, signal enhancement was simply calculated as the ratio of the signal measured with EPR irradiation over the signal observed without EPR in a reference image acquired before TOPCA injection. The possible role of increased longitudinal relaxation rate induced by the presence of the radical should be taken into account to calculate actual enhancements provided by the Overhauser effect. Systematic acquisition of control images in the presence of free radials without EPR saturation was not possible in order to limit time delays between radical injection and OMRI acquisition. The T1-weighting contribution of the free radical in the overall observed signal enhancements could be calculated from the relaxivity value, the sequence parameters and supposed radical concentration. Taking into account the rapid clearance of nitroxide radicals , radical concentrations were probably lower than 1 m m, inducing a positive contrast without EPR of 20% or less (simulated data not shown). Thus, the signal enhancements measured here were predominantly due to Overhauser effect.
Overhauser enhancements measured in the present study were quite reproducible and were in the range of literature data [15-17], despite the use of various free radicals and/or injection protocols. For instance an average signal enhancement of 11 measured in major abdominal vessels with triarylmethyl (injection of 800 µl at 30 m m)  can be compared with the mean value of 8 or the maximum value of 13 found here in brain vessels with TOPCA (injection of 200 µl at 240 m m). In an more recent study using TOPCA (injection of 500 µl at 100 m m), an enhancement of 2–3 was measured in the same region . It could be emphasized that OMRI images displayed here were acquired with a spatial resolution in the slice direction that was at least 5 times higher than in any other OMRI experiments.
In tumor environment, a low-signal boundary surrounding hyper-intense area was observed (Fig. 2C). This phenomenon was also observed around the bladder and in some cases in kidneys in previous studies [16, 27]. This was probably due to a very low concentration of free radical in this area. In this case, the low magnitude of negatively signed Overhauser effect due to electron–proton dipolar interaction may cancel the magnetization.
Although the nitroxide used here had no specificity towards glioma, tumor labeling was observed. This was probably because the blood–brain barrier was disrupted and passive retention of the spin label could be helped by tumor angiogenesis.
As in other studies [16, 18, 23], OMRI kinetics was performed in two dimensions. Here kinetic data were measured at high temporal and spatial resolutions, thanks to the use of relatively high polarization fields.
To conclude, high-resolution OMRI at a constant field of 0.2 T in mice can now be applied to follow a biological process that modulates Overhauser enhancement of designed spin probes. As mentioned elsewhere , protease activity can be revealed by DNP in vitro and the use of high-resolution OMRI opens the way to in vivo molecular imaging of disease-associated proteolysis.
This work was supported by the Agence Nationale de la Recherche (Blanc_09_434420).