Direct Visualization of Mouse Brain Oxygen Distribution by Electron Paramagnetic Resonance Imaging: Application to Focal Cerebral Ischemia

Abstract

Electron paramagnetic resonance imaging (EPRI) is a new modality for visualizing O₂ distribution in tissues, such as the brain after stroke or after administration of drugs of abuse. We have recently shown that 3-acetoxymethoxycarbonyl-2,2,5,5-tetramethyl-1-pyrrolidinyloxyl [1] is a pro-imaging agent that can cross the blood–brain barrier. After hydrolysis by esterases, the anion of 3-carboxy-2,2,5,5-tetramethyl-1-tetramethyl-1-pyrrolidinyloxyl [2] is trapped in brain tissue. In this study, we investigated the feasibility of using this to map the changes of O₂ concentration in mouse brain after focal ischemia. The decrease in tissue O₂ concentration in the ischemic region of mouse brain was clearly visualized by EPRI. The hypoxic zone mapped by EPRI was spatially well correlated with the infarction area in the brain imaged by diffusion-weighted magnetic resonance imaging (MRI). Finally, we observed a decrease in the size of the hypoxic region when the mouse breathed higher levels of O₂. This finding suggests that EPRI with specifically designed nitroxides is a promising imaging modality for visualizing O₂ distribution in brain tissue, especially in an ischemic brain. We believe that this imaging method can be used for monitoring the effects of therapeutic intervention aimed at enhancing brain O₂ supply, which is crucial in minimizing brain injury after stroke.

Introduction

Molecular oxygen is the terminal electron acceptor for many enzymes that are important in brain function, including those involved in the biosynthesis of the neurotransmitters dopamine, serotonin, and norepinephrine. Thus, changes in O₂ levels can have a significant and profound impact on physiology. For example, stroke, which results from occlusion of a major cerebral artery with consequently diminished partial pressure of O₂ (pO₂) in the brain, is the leading cause of long-term disability and a major cause of mortality in the elderly.

For diagnosis of stroke, monitoring the status of cerebral blood flow and O₂ supply in the ischemic region is essential. Different imaging modalities, such as functional magnetic resonance imaging (fMRI), positron emission tomography (PET), and computed tomography (CT), are widely used to assess a range of brain functions, including those associated with stroke. However, these techniques cannot quantitate tissue pO₂, which is essential in determining the physiology and pathophysiology of an affected region after ischemia (Grond et al, 1997; Baird and Warach, 1999; Kidwell et al, 2003; Kida et al, 2004; Hjort et al, 2005; Rauscher et al, 2005). For instance, blood oxygen level-dependent magnetic resonance imaging (MRI) measures blood oxygenation (not tissue oxygenation) by estimating deoxyhemoglobin concentration (Elas et al, 2003; Wedegartner et al, 2005). The accuracy of this measurement depends on three physiologic parameters: (1) local rate of metabolic consumption of O₂, (2) regional cerebral blood volume, and (3) regional cerebral blood flow. Hemodynamic and metabolic changes that influence these parameters can alter the concentration of deoxyhemoglobin and thus affect blood oxygen level-dependent MRI data. Furthermore, perfusion imaging reflects only cerebral blood flow and cerebral blood volume changes and is, therefore, not a direct measure of tissue pO₂. Therefore, development of a novel-imaging methodology for directly quantifying tissue pO₂ in different regions of an ischemic brain is essential in understanding this pathology.

Molecular oxygen, being paramagnetic, broadens the electron paramagnetic resonance (EPR) spectral lines of other paramagnetic species, such as nitroxides or trityl radicals (Elas et al, 2003; Liu et al, 2004). This phenomenon has been developed into an accurate oxygen measurement technique, with the EPR spectral line width for such paramagnetic compounds being used to estimate O₂ concentration in homogeneous solution (Elas et al, 2003; Khan et al, 2003; Shen et al, 2003; Kutala et al, 2004). With the development of low-frequency EPR spectroscopy and imaging and appropriate spin probes, it is now feasible to estimate local O₂ concentrations in situ, in vivo in real time (Elas et al, 2003). An obstacle to the development of minimally invasive electron paramagnetic resonance imaging (EPRI) to quantitate pO₂ in the brain is the difficulty of transporting O₂-sensing probes across the blood–brain barrier.

Recently, we and others showed that 3-acetoxymethoxycarbonyl-2,2,5,5-tetramethyl-1-pyrrolidinyloxyl [1] can cross the blood–brain barrier to enter the brain tissue, in which it is hydrolyzed by esterases to yield 3-carboxy-2,2,5,5-tetramethyl-1-pyrrolidinyloxyl [2], which, being anionic at physiologic pH, becomes trapped in brain tissue (Figure 1; Shen et al, 2006; Miyake et al, 2006; Miura et al, 1997; Sano et al, 2000). After the success of our earlier pharmacokinetic and pharmacodynamic studies (Miyake et al, 2006), we now report that using nitroxide [2] as the O₂ sensor, we can use EPRI to measure O₂ distribution in different regions of a mouse brain after focal cerebral ischemia. Our results indicate that EPRI can reliably map pO₂ in mouse brain in real time. Furthermore, we show that EPRI, in conjunction with MRI, can be used to define specific regions of the brain that are affected by diminished O₂ after an ischemic stroke.

Figure 1

Schematic diagram showing ready diffusion of lipophilic nitroxide [1] into neural cells, in which esterase hydrolysis liberates nitroxide [2]. At physiologic pH, nitroxide [2] is anionic and membrane impermeant, and is, therefore, trapped intracellularly.

Materials and methods

Chemicals

All chemicals and solvents were of ACS grade or better and were used without further purification. The 3-acetoxymethoxycarbonyl-2,2,5,5-tetramethyl-1-pyrrolidinyloxyl [1] and 3-carboxy-2,2,5,5-tetramethyl-1-pyrrolidinyloxyl [2] were synthesized as described in the literature (Rosen et al, 2005). Stock solutions (0.5 M) of the nitroxides were prepared in 95% ethanol and diluted 10-fold with phosphate-buffered saline (PBS) before use in animal experiments. Lithium phthalocyanine (LiPc) particles were a gift from Dr Harold Swartz, Department of Radiology, Dartmouth Medical School, Hanover, NH, USA.

Electron paramagnetic resonance phantom experiments

A Bruker E540L L-band EPR spectrometer equipped with 3D planar gradients was used for phantom experiments. To determine the spatial resolution of L-band EPR spectroscopy, we made a phantom comprising a bundle of 15 capillary tubes (ID, 1.0 mm): eight contained nitroxide [1] (1 mol/L in PBS) and were capped with white critoseal; seven were empty. A photograph of the phantom is shown in Figure 2 (left panel). The phantom was placed in the resonator of the EPR spectrometer, and an EPR spectral-spatial image was recorded with the following acquisition parameters: microwave frequency, 1.12 GHz; microwave power, 5.7 mW; center field, 410 G; modulation frequency, 50 kHz; modulation amplitude, 1 G. The EPR image was reconstructed by back projection using software from Bruker Biospin, Inc. The experiment was repeated with nitroxide [2] in place of nitroxide [1].

Figure 2

Photograph (left) and EPR image (right) of phantom comprising a bundle of 15 capillary tubes: 8 are filled with 1 mol/L nitroxide [1] and 7 are empty. As expected, only the nitroxide-containing tubes are visible in the EPR image. The scale bar represents 1 mm. The pixel size in the EPR image is ∼50 μm. Identical results were obtained with nitroxide [2].

Validation of tissue oxygen measurement with nitroxide by comparison with solid-state electron paramagnetic resonance oximetry probe lithium phthalocyanine

C57 mice were maintained under appropriate lighting conditions for 4 days with free access to food and water, and were used in compliance with the principles set forth in the ‘Guide for Care and Use of Laboratory Animals.’ All animal experimental protocols were approved by the University of New Mexico Animal Care and Use Committee. To test the feasibility of using nitroxides for mapping brain O₂ distribution, LiPc particles (about 0.2 mm in diameter) were implanted into the brains of four mice anesthetized by inhalation of 4% isoflurane in N₂O:O₂ (7:3) 5 days before EPR experiments. Nitroxide [1] (1.95 mmol/kg body weight) was injected into a mouse intraperitoneally (i.p.) 5 min before EPR spectral acquisition. For the experiments, 4% and 1% isoflurane in N₂O:O₂ (7:3) were used, respectively, for induction and maintenance of anesthesia in the mouse. The anesthetized mouse was transferred into the cavity of the Bruker E540L L-band-imaging system. A heating pad was used to maintain the mouse body temperature at 37°C. The mouse sequentially inhaled three gas mixtures containing 30%, 5%, and 0% O₂ (the balance in each mixture was N₂). O₂ concentration in the inhaled gas mixture was constantly monitored with a PA-1B O₂ analyzer. EPR spectra from the mouse head were recorded with the same parameters as those used in the phantom experiments, except that the modulation amplitudes were 0.08 and 0.4 G for experiments with LiPc and nitroxide [2], respectively.

Focal cerebral ischemia model

Focal cerebral ischemia was achieved by right-sided endovascular middle cerebral artery occlusion (Longa et al, 1989; Liu et al, 2006). Briefly, in an anesthetized mouse, the right carotid arteries were exposed through a midline cervical incision. The right external carotid artery was dissected free and isolated distally by coagulating its branches (Bipolar Electric Coagulation, GN60; Aesculap AG & Co., Tuttlingen, Germany) and placing a distal ligation before transaction. A piece of 4-O-monofilament nylon suture, with its tip rounded by gentle heating, was introduced through the lumen of the right external carotid artery stump and the right internal carotid artery to embed into the right anterior cerebral artery, so that the right middle cerebral artery was occluded at its origin. An abrupt reduction in regional cerebral perfusion to <30% of the baseline value, as verified by laser Doppler measurement, was considered to achieve focal ischemia. After 30 min of ischemia, MR and EPR images of the heads of sham controls and ischemic mice were obtained.

Magnetic resonance imaging experiments

Mice with middle cerebral artery occlusion were placed in a custom-built holder and moved to the isocenter of the magnet before imaging. Throughout the imaging session, animals were anesthetized with isoflurane (4% for induction, 1% for maintenance), and the heart and respiratory rates were monitored in real time.

MRI was performed on a 4.7T Biospec^® dedicated research MR scanner (Bruker Biospin, Billerica, MA, USA), equipped with a 500 mT/m (rise time 80–120 μs) gradient set for performing small animal imaging, and a small-bore linear RF coil (ID 72 mm). Initial localizer images were acquired using the following parameters: 2D fast low angle shot; TR/TE, 10/3 ms; matrix, 256 × 128; FOV, 6.4 cm; 1 slice per orientation. After acquisition of localizer images, T2-weighted and diffusion-weighted imaging (DWI) were performed with the following parameters: T2-weighted 2D RARE (rapid acquisition with relaxation enhancement), TR/TE, 4000/65 ms; FOV, 2.5 cm × 2.5 cm; slice thickness, 1.0 mm; slice gap, 0.5 mm; number of slices, 10; matrix, 256 × 128; number of averages, 20; receiver bandwidth, 50 kHz; DWI–2D diffusion-weighted RARE, TR/TE; 2000/31.2 ms; FOV, 2.5 cm × 2.5 cm; slice thickness, 1.0 mm; slice gap, 0.5 mm; number of slices, 10; matrix, 64 × 64; number of averages, 20; receiver bandwidth, 100 kHz; d=5 ms; D=20 ms; b-value of 0 and 900 s mm⁻²; diffusion gradient along the slice-select direction. The DWI method was slice matched to the T2-weighted method. The acquired data were postprocessed on a dedicated workstation.

In vivo electron paramagnetic resonance imaging experiments

Mice were placed in a plastic holder, fitted with a plastic tube for inhaled gas anesthesia, and mounted into the cavity of the L-band EPRI system equipped with 3D planar gradient sets.

EPR spectra with and without a field gradient were obtained with the following acquisition parameters: microwave frequency, 1.12 GHz; microwave power, 18 mW; central field, 410 G; modulation frequency, 50 kHz; modulation amplitude, 0.4 G. The image types were either signal intensity images, which report solely on the intensity of the EPR signal in space, or spectral-spatial imaging defining both intensity and the line width of the nitroxide in space. Spectral-spatial-imaging parameters were the same as intensity images, but the field gradients are controlled by proprietary software. Images were reconstructed by filtered back projection, and line width imaging was obtained by further data processing with software provided by Bruker Inc. Moreover, a calibration curve based on the EPR line width of nitroxide [2] as a function of pO₂ (0, 10, 20, 30, 40, 50, and 60 mm Hg) was constructed, as described in our earlier publication (Shen et al, 2006). The calibration curve was used to transform EPR line width-imaging data into pO₂ images.

Hyperoxic treatment

To show the effects of normobaric hyperoxic treatment on pO₂ in the focal ischemic mouse brain in vivo, four mice were anesthetized with isoflurane, injected with nitroxide [1] (1.95 mmol/kg body weight, i.p.), and subjected to middle cerebral artery occlusion. First, EPR spectra were collected from the head of the mouse when the mouse breathed 30% O₂ (the balance of the gas was 65% N₂ and 5% CO₂). Thereafter, the experiment was repeated with the mouse breathing 95% O₂ and 5% CO₂. The two EPR data sets were used to construct pO₂ images of the brain under normoxic and hyperoxic conditions, respectively.

Results

Phantom study for determining spatial resolution of electron paramagnetic resonance images

An initial series of experiments focused on establishing the spatial resolution of EPRI. We used a phantom comprising a bundle of 15 tubes, eight of which contained 1 mol/L nitroxide [1] in PBS (see Materials and methods for details). Figure 2 shows a photograph of the phantom (left panel) and the corresponding EPR image (right panel). It can be seen that the intensity EPR image accurately localized the eight nitroxide-containing tubes, whereas the empty tubes were invisible, as expected. We estimate that the EPR image of nitroxide [1] in Figure 2 has a spatial resolution of 0.1–0.2 mm. Essentially identical results were obtained with nitroxide [2] (data not shown).

Oxygen sensitivity of nitroxides in mouse brain

Data depicted in Figure 2 indicate that the spatial resolution of EPR spectral-spatial images may be adequate for mapping variations in brain O₂ levels. We next evaluated nitroxides as oxygen sensors for EPRI in vivo. Earlier, we reported that the EPR spectral line width of LiPc particles, stereotaxically implanted into specific sites in a mouse brain, was an accurate measure of local O₂ concentration at the site of implantation (Liu et al, 1995; Ilangovan et al, 2004; Hou et al, 2004; Liu et al, 2004; Hou et al, 2005). Therefore, as a reference standard for in vivo O₂ measurement, LiPc particles were implanted into the brains of mice (n=6) 3 days before administration of nitroxide [1] (1.95 mmol/kg body weight, i.p.). The EPR spectrum of nitroxide [2] and LiPc can be acquired simultaneously from the mouse brain, and, importantly, the two spectra are clearly resolved from each other (Figure 3A). Therefore, as each animal breathed 30%, 5%, or 0% O₂, changes in the EPR spectral line widths of nitroxide [2] and LiPc in the brain can be evaluated independently. As shown in Figure 3B, over the range of 0–30% O₂, the EPR spectral line widths of LiPc and nitroxide [2] were linear functions of O₂ concentration. More importantly, as implanted LiPc particles have been shown to be accurate reporters of O₂ concentration in mouse brain, the excellent linear correlation between LiPc and nitroxide [2] line widths (Figure 3C) shows the feasibility of using nitroxides to quantitate changes in pO₂ in different brain regions.

Figure 3

Variation of EPR line width of nitroxide [2] and LiPc with O₂ concentration in mouse brain in vivo. LiPc particles were implanted in the mouse brain 3 days before EPR measurement, whereas nitroxide [1] (1.95 mmol/kg body weight, i.p.) was administered 5 min before EPR measurement. Sequentially, the mouse inhaled gas mixtures containing 30%, 5%, and 0% O₂. (A) A representative EPR spectrum acquired from the mouse head shows that the spectral lines of nitroxide [2] and LiPc was clearly resolved. (B) Variation of EPR spectral line widths of nitroxide [2] and LiPc as the mouse inhaled gas mixtures with different O₂ concentrations. The reported line width of nitroxide [2] is that of the first (most downfield) peak in the three-line spectrum. Solid lines are linear least-square fits to the data. The line widths of nitroxide [2] and LiPc were linear functions of O₂ concentration, with slopes of 8.40±0.97 and 7.60±0.45, respectively. Wherever not shown, error bars are smaller than the plot symbol. (C) Linear correlation of the line widths of LiPc and nitroxide [2], showing excellent correlation of the line widths of nitroxide [2] and LiPc (r=0.9912).

Electron paramagnetic resonance imaging of brain O₂ during focal ischemia

In Figure 3, the changes in nitroxide EPR spectral line width reflect the average changes of pO₂ across the entire brain, whereas similar data from LiPc only report pO₂ in the tissue immediately surrounding the implanted LiPc (spherical volume with ∼1 mm diameter). It would, however, be advantageous to measure oxygenation in tissues with inhomogeneous distribution, as occurs after ischemic stroke or traumatic brain injury. The addition of a magnetic field gradient to a static magnetic field (Elas et al, 2003; Matsumoto et al, 2005; Matsumoto et al, 2006) enables spatial encoding of the EPR spectrum of nitroxide molecules within each local region (i.e., pixel). Line width data from such a spatially localized spectrum can then be converted into pO₂ for that local region.

To test the validity of our approach, L-band EPR spectroscopy, with nitroxide [1] as the pro-imaging agent, was used to map O₂ distribution in a focal brain ischemic mouse model. Each mouse (n=5) received nitroxide [1] (1.95 mmol/kg body weight, i.p.) and EPR spectra were recorded before and after induction of focal cerebral ischemia. Representative slices from three-dimensional EPR intensity images from a sham control and a focal cerebral ischemic mouse are shown in Figures 4A and 4B, respectively. The EPR intensity image of the sham control mouse (Figure 4A) depicts a homogeneous distribution of the nitroxide signal, whereas the image of a focal cerebral ischemic mouse (Figure 4B) reveals heterogeneous distribution, with a bright yellow area in the right hemisphere, indicating a dramatic increase in EPR signal intensity. This increased intensity most likely results from a reduction of the EPR spectral line width (i.e., a sharper spectral line) in response to decreased O₂ levels. The MR diffusion image of the brain from the same mouse was used to anatomically localize and identify the ischemic region of the brain (Figure 4C). The region of intense yellow in the EPR image (Figure 4B) matched well with the ischemic area depicted in the MR diffusion-weighted image (Figure 4C). However, although the intensity images alone showed the potential of EPRI to map oxygen in the brain, they cannot be used directly to provide tissue oxygenation levels without making assumptions about nitroxide uptake (which may also be heterogeneous).

Figure 4

Comparison of EPR intensity image and MR diffusion image in ischemic mouse brain is shown. (A) EPR intensity image of nitroxide [2] in sham control mouse brain. (B) EPR intensity image of nitroxide [2] in the same mouse brain after focal ischemia. (C) MR diffusion image in the focal ischemic mouse brain. The results show that the hypoxic region of the spectral-spatial image matched well with the infarction area as revealed by the MR diffusion image.

To definitely show the oxygen distribution in the ischemic brain, spatial spectral images were obtained, and were then transformed into pO₂ images by using the line width of the EPR signal (Figures 5A and 5B). As expected, O₂ concentration in the head of the control mouse was relatively uniform throughout the brain (Figure 5A). However, in parallel with the EPR spectral-spatial image of the focal ischemic brain (Figure 4B), O₂ concentrations in the left and right hemispheres were markedly different (Figure 5B). By comparing the EPR and MR images, the low-O₂ region in the EPR image matched well with the corresponding infarction area identified with the MR diffusion image (compare images in Figures 5B and 5C). Importantly, the EPR line width image indicated that, even within the same ischemic hemisphere, the pO₂ distribution was heterogeneous. Interestingly, the hypoxic region with pO₂ <5 mm Hg was relatively small (darkest region in Figure 5B), even though the MRI diffusion image showed a relatively large infarction area induced after 30 min of focal cerebral ischemia (Figure 5C). To examine the pO₂ distribution in the ischemic brain, we measured the spatial profile of pO₂ along the black line in Figures 5A and 5B. The spatial pO₂ profiles are shown in Figure 5D and reveal that only a few points showed pO₂ <10 mm Hg at the particular brain locus. The pO₂ values across the brain section were also plotted in Figure 5E as a histogram, which shows that the distribution of tissue pO₂ values was centered at ∼40 mm Hg in the nonischemic brain, whereas several loci in the ischemic brain had pO₂ <20 mm Hg. These findings suggest that the anoxic regions are smaller than the infarct area, perhaps offering a window of opportunity for therapeutic intervention aimed at correcting these intensely hypoxic foci.

Figure 5

pO₂ distribution in sham control and ischemic mouse brain obtained by EPR spectral-spatial imaging. (A) pO₂ map in the head of a sham control mouse head obtained by EPRI of nitroxide [2]. (B) pO₂ map in the head of the same mouse after ischemia. (C) MR diffusion image of the focal ischemic mouse head. (D) Representative spatial profiles of pO₂ in sham control and ischemic mouse brain. The mice were inhaled with 30% O₂ and the pO₂ values were measured along the black lines shown in A and B. (E) Frequency of pO₂ values along the black lines shown in A and B. The results show that the hypoxic region of the pO₂ image matched well with the infarction area of the MR diffusion image.

Effects of hyperoxic treatment on brain pO₂ in focal cerebral ischemic mice

It has been reported that normobaric hyperoxia is beneficial in reducing brain injury in animal models of stroke, when O₂ is administered during transient cerebral ischemia (Singhal et al, 2002a, 2002b; Liu et al, 2006). Our earlier study using single-site pO₂ measurement with LiPc indicates that normobaric hyperoxic treatment with 95% O₂ significantly increased cerebral tissue pO₂ in mouse brains during ischemia (Liu et al, 2006). We now wish to ascertain the overall effect of hyperoxic treatment on brain oxygenation by mapping O₂ distribution in the entire ischemic brain through EPRI, and to determine whether hyperoxic treatment could increase cerebral pO₂. To this end, nitroxide [1] (1.95 mmol/kg body weight, i.p.) was administered to three mice before initiating focal cerebral ischemia. EPR spectra were acquired when the focal cerebral ischemic mice breathed 30% O₂. Then, O₂ in the inspired gas mixture was increased to 95%. Figure 6 shows representative brain pO₂ maps in the focal cerebral ischemic mouse inhaling either 30% or 95% O₂. It is important to note that when O₂ in the inspired gas was increased from 30% to 95%, pO₂ significantly increased in the ischemic brain, and the size of the hypoxic area in the ischemic hemisphere was markedly reduced.

Figure 6

(A) EPRI of nitroxide [2] in the head of a focal cerebral ischemic mouse inhaling 30% O₂, acquired as in Figure 5. (B) EPRI of the head of the same mouse inhaling carbogen gas containing 95% O₂.

Discussion

Continued advances in neuroimaging technology, such as functional MRI, computed tomography, or positron emission tomography, have made it possible to map anatomic and functional changes during brain infarction, thus providing therapeutic opportunities to ameliorate the consequences of a stroke (Baird and Warach, 1999). However, despite these remarkable advances in stroke diagnosis and evaluation, there are no reliable, minimally invasive-imaging modalities that can accurately measure brain O₂ distribution in vivo, in situ, and in real time. In earlier publications, we evaluated the ability of structurally different nitroxides to cross the blood–brain barrier and enter the brain tissue, in which local enzymatic action caused entrapment of these paramagnetic probes in the brain tissue. Furthermore, we showed the ability of nitroxides to accurately report local O₂ concentrations using low-frequency EPR spectroscopy as an imaging modality (Miyake et al, 2006; Shen et al, 2006). The experiments described here are among the first reports of O₂ mapping in an ischemic mouse brain in real time. The change of O₂ concentration in this experimental paradigm of brain ischemia was clearly visualized by EPR spectral-spatial imaging of the nitroxide line width. The hypoxic zone in the EPR images matched well with the infarction area in the brain as revealed by diffusion-weighted MRI. EPRI indicated that the hypoxia region with the pO₂ <5 mm Hg was relatively small, even though the MR diffusion image showed a relatively large infarction area after 30 min of focal cerebral ischemia, suggesting that not all tissues in the ‘infarcted area,’ as defined by DWI or TTC staining, are highly hypoxic. Normobaric hyperoxia treatment significantly increased overall tissue pO₂ and reduced the hypoxic region in the ischemic mouse brain.

In earlier studies, EPRI has been used to map local distribution of O₂in vivo, including in tumor-bearing animals (Matsumoto et al, 2006; Krishna et al, 1998; Elas et al, 2003; Kuppusamy and Zweier, 2004; Matsumoto et al, 2008). However, a major obstacle in EPRI for O₂ mapping in the brain has been the synthesis of paramagnetic probes that (1) can cross the blood–brain barrier, (2) have high O₂ sensitivity, and (3) can be retained in brain tissue at sufficiently high concentration to permit imaging of O₂ over time. As detailed in recent publications (Kao and Rosen, 2004; Shen et al, 2006; Miyake et al, 2006; Burks et al, 2008), we have designed and synthesized a family of nitroxides that can cross the blood–brain barrier to enter the brain tissue, in which local metabolic conversion traps a derivative of the parent nitroxide. In this study, we tested the feasibility of using nitroxide [1] as a pro-imaging agent to map O₂ distribution in mouse brain. We first performed intensity imaging, which provided an EPR intensity for each pixel in an image. As the spatial and spectral dimensions are fully separable, local line width, and hence local O₂ concentration, can in principle be derived independently (Kuppusamy and Zweier, 2004) by spectral-spatial imaging. As EPR line broadening is induced by the spin interaction of O₂ with a nitroxide and is linearly proportional to tissue pO₂, the spectral-spatial images of line width can be transformed into line width images and pO₂ images. The pO₂ images directly showed that the O₂ distribution in the hypoxic region of the ischemic mouse brain matched well with the infarction area in the diffusion-weighted MRI.

An important finding is the ability to quantitate O₂ in the brain after introducing nitroxide [1] into a mouse (Figure 5). Before inducing focal ischemia, EPRI revealed relatively uniform pO₂ across brain tissue (Figure 5A). After focal ischemia was induced, however, a marked decrease in pO₂ was noted (Figure 5B). This is the combined result of disruption of O₂ delivery into, and continued metabolic consumption of O₂ in, the occluded region. After 30 min of focal cerebral ischemia, the significant decrease in pO₂, seen in Figure 5B, must undoubtedly have had profound effects on brain function. We note that the very hypoxic area (<5 mm Hg) was quite small and that pO₂ in most of the hypoxic region was between 10 and 25 mm Hg. This significant variability in tissue oxygenation within the DWI-defined core of the stroke is likely because of heterogeneity of the stroke with penumbra within the infarction core, as has been observed in cerebral blood flow images. Thus, there is a time window for therapeutic intervention to save the hypoxic brain, in which therapy may lead to a positive outcome.

It is noteworthy that breathing 95% O₂ significantly increased the pO₂ level in the focal cerebral ischemic brain (Figure 6)—overall pO₂ was markedly increased relative to when 30% O₂ was inhaled. This result is direct evidence that normobaric hyperoxia treatment can improve O₂ delivery to the infarction area. In our earlier reports, we suggested that normobaric hyperoxia treatment could increase penumbral pO₂, reduce infarction volume, improve neurologic function, decrease the generation of reactive O₂ species, and reduce matrix metalloproteinase-9 expression and caspase-8 cleavage in the ischemic brain (Liu et al, 2006; Liu et al, 2009). Thus, normobaric hyperoxia could be a valuable therapeutic approach for saving the ischemic brain.

Our long-term objective is to map O₂ distribution for diagnosis of brain hypoxia and for evaluation of therapeutic approaches in ischemic stroke in animal experiments and clinical trials. Despite the success of the experiments reported here, there are several issues that need to be resolved before EPRI is applicable for clinical application in cerebral stroke: (1) the EPR line width of nitroxides, such as nitroxides [2], is significantly larger than that of LiPc; (2) the EPR image acquisition time is relatively long for real-time measurement of pO₂ (acquisition time for 3-D spectral-spatial imaging with useful resolution is generally 30 to 40 min; therefore, at present, images with only two spatial dimensions (the third dimension being the EPR spectra) can be acquired in useful time periods. For longer data acquisition time, supplemental administration of the compound has to be given. (3) EPRI provides pO₂, but not anatomic information in the ischemic brain. To address these limitations, we need to improve the sensitivity of the oximetry probes by synthesizing ¹⁵N- and deuterium-substituted nitroxides, which will have much narrower EPR spectral lines. Furthermore, improvements in low-frequency EPR spectrometers, and in data acquisition and analysis procedures (such as rapid scanning and selective scanning), will undoubtedly improve data quality and the rate of data acquisition. Efforts in both directions are under way.