|Home | About | Journals | Submit | Contact Us | Français|
Diffusion tensor imaging (DTI) is a powerful method to visualize white matter, but its use in patients with acute stroke remains limited because of the lack of corresponding histologic information. In this study, we addressed this issue using a hypoxia–ischemia (HI)-induced thrombotic model of stroke in adult mice. At 6, 15, and 24hours after injury, animals were divided into three groups for (1) in vivo T2- and diffusion-weighted magnetic resonance imaging, followed by histochemistry, (2) ex vivo DTI and electron microscopy, and (3) additional biochemical or immunochemical assays. The temporal changes of diffusion anisotropy and histopathology were compared in the fimbria, internal capsule, and external capsule. We found that HI caused a rapid reduction of axial and radial diffusivities in all three axonal bundles. A large decrease in fractional anisotropy, but not in axial diffusivity per se, was associated with structural breakdown of axons. Furthermore, the decrease in radial diffusivity correlated with swelling of myelin sheaths and compression of the axoplasma. The gray matter of the hippocampus also exhibited a high level of diffusion anisotropy, and its reduction signified dendritic degeneration. Taken together, these results suggest that cross-evaluation of multiple DTI parameters may provide a fuller picture of axonal and dendritic injury in acute ischemic stroke.
White matter (WM), occupying 50% of total brain volume in humans, has a metabolic rate similar to that of gray matter (GM) (Zhang and Sejnowski, 2000; Goldberg and Ransom, 2003). Yet, WM receives a disproportionately small blood supply and little collateral circulation, making it particularly susceptible to hypoxic-ischemic insults (Dewar et al, 1999; Arai and Lo, 2009). However, although the importance of WM injury in cerebrovascular diseases is recognized, it is difficult to distinguish it from GM damage and determine the point of irreversible WM injury in acute stroke using traditional imaging methods.
Diffusion tensor imaging (DTI) is a relatively new magnetic resonance imaging (MRI) technique to visualize axonal organization and to measure directional water diffusion in the nervous system (Basser et al, 1994). In DTI, measurements of water diffusion along multiple axes are fitted into a three-dimensional (3D) model to determine three eigenvalues in the principal (λ1) and two shorter, perpendicular axes (λ2 and λ3) of the ‘diffusion ellipsoid'. The three eigenvalues are used to calculate several DTI indexes, including fractional anisotropy (FA), axial/longitudinal diffusivity (λII), and radial/transverse diffusivity (λ) (Mori and Zhang, 2006). Fractional anisotropy, scaled from 0 (isotropic) to 1 (anisotropic), is a measurement of the relative difference of the three eigenvalues. Axial diffusivity, equivalent to λ1, represents water motion along the length of axons, which can be reduced by structural breakdown or focal enlargement–constriction of the neurites according to theoretical modeling (Beaulieu, 2002; Budde and Frank, 2010). Radial diffusivity, the average of λ2 and λ3, indicates perpendicular water diffusion across the axon, which is hindered by barriers such as myelin sheaths (Song et al, 2002).
In trauma- or axotomy-induced Wallerian degeneration, axial diffusivity steadily declines and radial diffusivity gradually increases, resulting in a progressive reduction of FA (MacDonald et al, 2007; Zhang et al, 2009). In contrast, patients with acute ischemic stroke often exhibit a rapid reduction of both axial and radial diffusivities, accompanied by little to no change of initial FA (Yang et al, 1999; Bhagat et al, 2008; Sakai et al, 2009). A sizable reduction of FA only occurred ~24hours after stroke onset, which was postulated to indicate irreversible axonal injury (Bhagat et al, 2008). However, this hypothesis of important clinical application is yet to be validated, because to date, there are no DTI–neuropathology correlation data in patients with acute ischemic stroke (<24hours).
In this study, we describe DTI measurements and corresponding histologic changes in a rodent model of thrombotic stroke induced by cerebral hypoxia–ischemia (HI) (Adhami et al, 2006). We show that a combined hypoxic-ischemic insult triggers a rapid reduction of both axial and radial diffusivities in multiple axonal tracts, similar to the pattern observed in acute stroke patients. In contrast, a large reduction of FA only occurs slowly and is restricted to the nerve tracts with severe axonal destruction. Furthermore, the rapid reduction of radial diffusivity correlates with oligodendrocyte swelling and compression of the axoplasma. Finally, the reduction of anisotropy contrast inside the GM of the hippocampus signifies dendritic degeneration. Taken together, these results suggest that DTI can be used to detect and distinguish multiple aspects of axonal and dendritic injuries in acute ischemic stroke.
Male CD-1 mice (Charles River, Wilmington, MA, USA) and Thy1-YFP mice (Jackson Laboratories, Bar Harbor, ME, USA; stock number 003782; Feng et al (2000)), aged 8 to 12 weeks, were used in this study. Details of the HI-induced thrombotic model of stroke have been described previously (Adhami et al, 2006). In brief, animals were anesthetized using 1% to 2% isoflurane while maintaining respiration at 80 to 120 breaths per minute. Partial cerebral ischemia was established by permanent unilateral (right) common carotid artery occlusion. After carotid ligation, hypoxia was initiated by administering 7.5% O2 balanced by nitrogen through a gas mask for 50minutes under anesthesia. The body temperature of mice was maintained at 37°C±0.5°C using a thermo-controller connected to a rectal probe and heating light. The animal procedures were approved by the Institutional Animal Care and Use Committee and conform to the NIH Guide for Care and Use of Laboratory Animals.
At 2hours after hypoxia, mice were evaluated for neurologic symptoms, and those walking in circles towards the ipsilateral hemisphere were used for analysis in this study. Approximately 80% of HI-challenged animals fulfilled this criterion, and the mortality rate was <10% in CD-1 and Thy1-YFP mice by 24hours recovery. In total, the following six groups of CD-1 mice were used for sequential MRI histology examinations.
Additional CD-1 and Thy1-YFP mice were killed at 1, 6, 15, or 24hours after HI for various biochemical, histochemical, and immunocytochemical evaluations without MRI or EM (the number of animals used for each assay is indicated in text).
All data were collected on a Bruker BioSpec 7-T system (Bruker BioSpec 70/30, Karlsruhe, Germany) equipped with 400mT/m actively shielded gradients. Animals were anesthetized by 1% to 2% isoflurane delivered by oxygen, and the respiration rate was maintained at 60 to 100 breaths per minute. The core temperature was maintained at 37°C by warm air circulated through the magnet bore. All animals were scanned with a custom-built radio frequency coil. Anatomic data were acquired with a 3D, fast spin-echo sequence and the following parameters: echo train length=16, TEeff (echo timeeff)/TR (repetition time)=70.56/1,000msec, FOV (field of view)=32 × 19.2 × 19.2mm3, matrix size=256 × 96 × 96, and 1 average, resulting in a resolution of 125 × 200 × 200μm3, with total acquisition time under 10minutes. The anatomic image was used as a reference for the subsequent diffusion-weighted images. Two-dimensional diffusion-weighted images were collected using a conventional Stejskal–Tanner single spin-echo sequence with a b-value of 800seconds/mm2, FOV=51.2 × 19.2mm2, TE/TR=22/1,000msec, 2 averages, matrix=256 × 96, resolution=200 × 200μm2 in plane, and 600-μm slice thickness. The acquisition time was ~20minutes. Analysis was performed using the ParaVision 4.0 software package (Bruker BioSpec 70/30, Karlsruhe, Germany). Regions of interest were prescribed in the striatum, hippocampus, and rostral and caudal areas of the cortex. Apparent diffusion coefficients were averaged separately for the ipsilateral and contralateral sides at each time point.
Ex vivo DTI was performed to assess the damage in the WM structures. Animals were killed immediately after the in vivo MRI scan by transcardial perfusion of fixatives. The brains were excised, stored in phosphate-buffered saline at 4°C for <2 weeks, and equilibrated to the ambient bore temperature for ~30minutes. All ex vivo images were acquired using a custom-built solenoid transmit/receive coil. Whole-brain data were collected using a 3D, conventional spin-echo DTI sequence. The DTI parameters were: b=800seconds/mm2 gradient separation/duration (D/δ)=12/4msec, FOV=32 × 12 × 12mm3, TE/TR=22/1,000msec, 1 average, matrix=256 × 96 × 96, resolution=125μm isotropic. The total scan time was ~18hours. Quantification of DTI measures including FA, λII, and λ was performed using DTIstudio software (free Sharewave reported in Jiang et al, 2006).
Brain samples for lipid peroxidation were taken from the cerebral cortex on either side of the brain, and homogenized in the phosphate buffer. The amount of malondialdehyde was quantified using a commercial kit following the manufacturer's instructions (Oxis Research, Portland, OR, USA).
All immunohistochemistry procedures were performed on frozen brain sections following transcardial perfusion of 4% paraformaldehyde and sucrose cryoprotection. The following antibodies were used: a rabbit polyclonal antibody against fibrinogen (a gift from Dr J Degen), MBP (Chemicon, Temecula, CA, USA), Olig2 (Chemicon), MAP2 (Sigma, St Louis, MO, USA), and NeuN (Chemicon). Biotinylated secondary antibodies and streptavidin conjugated to Alexa Fluor 488 or Alexa Fluor 594 (Molecular Probes, Carlsbad, CA, USA) were used to amplify immunosignals. The TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling) was performed as described previously (Yang et al, 2009). Cerebral perfusion analysis was performed by injecting fluorescein isothiocyanate-dextran (5μg in 100μL phosphate-buffered saline of 2 × 106 molecular weight; Sigma) into the left ventricle of animals at 4hours after HI. For detection of superoxide, the unfixed mouse brains were removed and snap frozen on dry ice. Brain sections of 10-μm thickness were then incubated with DHE (dihydroethidium) (5μmol/L; Invitrogen, Carlsbad, CA, USA) in phosphate buffer for 30minutes at 37°C in a light-protected humidified chamber. The in vivo Turnbull's reaction histochemistry for detection of ferrous ion [Fe(II)] was performed as described previously (Yu et al, 2001). A published mouse atlas was used to determine brain locations and anatomic terminology (Franklin and Paxinos, 2008).
At indicated times after HI, animals were perfused transcardially with 4% paraformaldehyde and 1% glutaraldehyde in 0.1mol/L phosphate buffer. The brains were removed and horizontal 100-μm-thick sections were cut using a vibratome. Areas of the fimbria, external capsule, and the hippocampus from the HI-challenged and contralateral hemispheres were dissected out and postfixed with 1% OsO4. The tissues were embedded in epoxy resin (Ted Pella, Redding, CA, USA). Thin sections of 70-nm thickness were cut using a Leica UC6 ultramicrotome (Leica, Deerfield, IL, USA) and stained with uranyl acetate and lead citrate. The sections were analyzed in a JEM 1200 EXII electron microscope (JEOL, Peabody, MA, USA).
All data were analyzed using SPSS for Windows version 12.0 (SPSS, Chicago, IL, USA). Data were first analyzed using a general linear model analysis with time, side (contralateral or ipsilateral), and location as factors. Side and time were significant factors for all locations, but there was no significant side × location effect. There was a significant time × side effect for axial diffusivity and a significant location × time effect for FA. To determine the specific interactions, all data were analyzed for each side using one-way ANOVA (analysis of variance) with time as the factor. Contrasts were defined in this analysis comparing the 6- and 15-hour time points and the 6- and 24-hour time points. After that analysis, each time point was analyzed comparing the contralateral and ipsilateral sides by Student's t-test. Quantitative indices for the hippocampal subfields were not calculated because the contrast between these structures disappeared on the anisotropy maps at 15 and 24hours.
To induce stroke, mice were treated with unilateral ligation of the common carotid artery, followed by 50minutes systemic hypoxia (7.5% O2) with the core temperature kept at 37.5°C±0.5°C. This model produced thrombosis and infarction on the HI-challenged side of the brain, while causing no discernible damage to the contralateral side of the brain (Adhami et al, 2006). The undamaged contralateral side of the brain thus served as the internal control for MRI and additional analyses. In Supplementary Figure 1, we showed that at 4hours after injury, there were still many punctate fibrin clots and pockets of perfusion deficits on the HI-challenged side of the brain but not on the contralateral hemisphere, suggesting an extended period of thrombosis in small arteries.
For in vivo 3D T2-weighted MRI correlated with ADC maps, different groups of mice were imaged at 6hours (Figures 1A to 1D), 15hours (Figures 1E to 1H), or 24hours (Figures 1I to 1L) after the HI insult (n=4 for each time point). This analysis showed a gradual increase of T2 signals (an indication of water accumulation) and expansion of affected areas on the HI-injured hemisphere from 6 to 24hours recovery. Consistent with the carotid artery-supplying territories, the damaged regions included the striatum, cerebral cortex, hippocampus, and thalamus, but spared the cerebellum and the brainstem that receive blood supply from the vertebral arteries. Interestingly, T2-weighted MRI occasionally showed hyperintensity along the nerve tracts, suggesting axonal edema, on the HI-challenged hemisphere at 15 to 24hours recovery. For example, Figure 1K shows increased T2 signals in the external capsule (ec), the molecular layer of the dentate gyrus (MoDG), and the alveus and stratum oriens (alv/Or) in the ipsilateral hippocampus.
The ADC map showed a rapid reduction of isotropic diffusion signals (caused by water accumulation intracellularly or in a tortuous extracellular space) in the striatum and cerebral cortex at all time points (P<0.05 compared with contralateral). However, the ADC map did not differentiate WM and showed little temporal progression of signal reduction. These results are consistent with the notion that ADC maps have high sensitivity in detecting cellular injury, but low specificity is resolving WM injury in acute stroke (Bhagat et al, 2008; Sakai et al, 2009).
Next, we used ex vivo DTI to derive 3D directionally encoded color (DEC) maps of the brains from mice killed at 6, 15, or 24hours after HI (n=3 for each time point). The DEC maps distinguished all major nerve bundles in the forebrain, including the anterior commissure (ac), external capsule (ec), internal capsule (ic), the fimbria (fm)/fornix, and the optic tract (opt) (Figure 2A). For quantitative analysis, regions of interest were drawn in the fimbria, internal capsule, and external capsule to calculate λII, λ, and FA on both sides of the brain. Previous studies have shown that although formalin fixation does not alter diffusion anisotropy, it reduces the magnitude of water diffusion coefficient in the brain (Guilfoyle et al, 2003; Sun et al, 2003, 2005). Thus, we calculated the ratio of λII, λ, and FA in specific axonal tracts on the HI-injured hemisphere relative to their counterparts on the contralateral hemisphere to standardize lesion-associated changes of directional diffusivities (Table 1 ).
This analysis showed that axial diffusivity exhibited a steady decline in all three axonal tracts after cerebral HI, decreasing to 63%±8% (mean±s.d., relative to the contralateral side) in the fimbria, to 73%±16% in the internal capsule, and to 60%±7% in the external capsule at 24hours recovery. These values were also significantly smaller than those at 6hours on the ipsilateral hemisphere (P<0.01 by ANOVA). Similarly, radial diffusivity showed a gradual reduction to 78%±10% in the fimbria and to 80%±9% in the internal capsule at 24hours recovery (P<0.01 compared with 6hours) and a more blunted attenuation to 83%±13% at 24hours recovery in the external capsule (P=0.965 compared with 6hours). Our finding of concomitant reduction of axial and radial diffusivities following HI is similar to those detected in patients with acute ischemic stroke (Yang et al, 1999; Bhagat et al, 2008; Sakai et al, 2009).
In contrast to the trend of gradual decrease of axial and radial diffusivities in all three axonal tracts, a significant reduction of FA after cerebral HI was only detected in the external capsule, in which FA decreased to 67%±14% at 24h recovery (P<0.05 compared with 6hours). In the fimbria, FA was only reduced to 92%±11% at 24hours recovery (P=0.876 compared with 6hours). In the internal capsule, FA was decreased to 96%±3% at 24hours recovery (P=0.403 compared with 6hours). The significant decrease in FA only in the external capsule, despite general reduction of axial diffusivity in all nerve tracks examined, suggests that the more severe axonal destruction in stroke may manifest as a sizable reduction of FA.
To examine this possibility, we used EM and Thy1-YFP mice to assess the extent of axonal injury in this model of thrombotic stroke. Thy1-YFP is a line of transgenic mice that expresses the yellow fluorescent protein in a soluble form to fill the entire cellular architecture of most cortical layer V and hippocampal neurons, thus offering a global survey of axonal injury after cerebral HI (Feng et al, 2000). The analysis using Thy1-YFP mice showed severe breakdown of axonal fascicles in the external capsule (n=5; Figures 3A to 3D), but not in the fimbria/fornix and internal capsule (Figures 3A, 3B, 3G, and 3H) at 15 to 24hours recovery. On EM following ex vivo DTI (n=3), the external capsule on the HI-injured hemisphere showed great reduction of nerve fibers and expansion of empty space between axonal bundles at 15hours after HI (Figure 3F). In contrast, there was minimal reduction of axonal bundles in the fimbria on the HI-challenged hemisphere at 15hours recovery (Figures 3I and 3J). Taken together, these data suggested a correlation between large FA reduction and severe axonal damage in specific nerve fiber tracts in acute cerebral ischemia.
The observed early reduction of radial diffusivity in patients and our animal model of stroke stands in contrast to the slow increase in radial diffusivity in trauma- or axotomy-induced Wallerian degeneration (MacDonald et al, 2007; Zhang et al, 2009). Chronic demyelination often causes an increase in radial diffusivity (Song et al, 2002). Thus, the rapid reduction of radial diffusivity after cerebral HI may be caused by retention of myelin sheaths but expansion of the less-anisotropic extraaxonal components in WM (Beaulieu, 2002). To test this hypothesis, we used immunohistochemistry and EM to examine the relationship between myelin and axoplasma after cerebral HI.
We found that HI caused greater reduction of the immunoreactivity against MBP inside the cerebral cortex than in the external capsule or the fimbria at 15hours recovery (Figures 4A and 4D; n>6). On EM, besides severe axonal degeneration and edema (Figure 3F), two types of abnormalities were frequently observed in the less-destructed WM on the HI-challenged hemisphere. One is the swelling and separation of myelin sheaths (M in Figure 4E). The other is the appearance of large vesicles (black asterisks in Figure 4F) often containing myelin on the surface (arrows in Figure 4F), to compress the axoplasma (colored in pink in Figures 4C and 4F). There was also a trend of increased extraaxonal empty space in the WM (red asterisks in Figure 4F). These morphologic anomalies, remarkably similar to those previously described in isolated optic nerves following anoxic injury (Waxman et al, 1990, 1992), suggest rapid oligodendrocyte injury after HI.
With regard to the mechanisms of HI-induced oligodendrocyte injury, previous studies have suggested that oligodendrocytes are very susceptible to oxidative stress owing to an insufficient antioxidant system and a high content of cellular iron (Thorburne and Juurlink, 1996; Irving et al, 1997; McCracken et al, 2000). To test this possibility, we used DHE stain to detect superoxide and found intense signals along the blood vessels on the HI-challenged hemisphere at as early as 1hour recovery (Figures 5A; n>4). Measurement of malondialdehyde also suggested an increase in lipid peroxidation on the ipsilateral hemisphere at 6hours after HI (Figure 5C; P<0.05 by ANOVA). These findings raised the possibility that HI-induced oxidative stress may reduce the ferric ion [Fe(III)] in oligodendrocytes to ferrous ion [Fe(II)], which in turn acts as an oxidizing agent to generate more superoxide and hydroxyl radical species (Dewar et al, 2003). Consistent with this possibility, an in vivo Turnbull's reaction histochemistry stain detected many more [Fe(II)]+ cells in the corpus callosum on the ipsilateral hemisphere (48±6 per visual field) than on the contralateral hemisphere (6±3 per visual field) at 15hours recovery (Figures 5D; n=6) (Yu et al, 2001). In addition, the ipsilateral corpus callosum showed reduced immunoreactivity against Olig2, a marker of oligodendrocytes and their progenitors, when compared with the contralateral hemisphere (Figures 5E; n=3). On double labeling, many TUNEL+ nuclei were found adjacent to Olig2+ nuclei in the ipsilateral corpus callosum (arrows in Figure 5H), suggesting that the dying cells were oligodendrocytes in view of their typical tandem disposition in WM. On EM, the HI-challenged hemisphere also displayed many pyknotic nuclei in the corpus callosum at 15hours after HI (Figures 5F and 5I). Taken together, these results suggest that cerebral HI rapidly induces oxidative stress in oligodendrocytes, the swelling and compression of which onto the axoplasma contributes to acute reduction of radial diffusivity in WM.
Although axons have the highest FA signals in the nervous system, in theory, any tissue with a polarized and orderly microstructure could display diffusion anisotropy (Basser et al, 1994). For example, a previous study using ex vivo DTI detected the dendrite-enriched stratum radiatum of the hippocampus in mice (Zhang et al, 2002). Consistent with this earlier report, our DTI analysis also appeared to detect the stratum radiatum inside the hippocampus (rad in Figure 2A). To verify the identities of the structures seen on DEC maps in the hippocampal subfield, we compared Nissl stain, ADC, FA, and DEC maps in the same imaging plane (Figure 6A). This cross-comparison revealed that neuronal cell bodies in the stratum pyramidal (CA (Cornu Ammonis)) and stratum granulosum (DG (dentate gyrus)) have a high isotropic diffusion (ADC) and a low anisotropic diffusion (FA) value, whereas myelinated axons in the external capsule (ec) and the molecular layer of the dentate gyrus (MoDG) behave oppositely. Examining the DEC map, we concluded that the stratum radiatum, a dendrite-rich hippocampal subfield, shows relatively high diffusion anisotropy, presumably attributed to an orderly array of apical dendrites in this region.
Interestingly, close inspection of the post-HI DEC maps indicated a rapid reduction of diffusion anisotropy contrast in the stratum radiatum on the ipsilateral hemisphere (Figure 2A). By 15hours after HI, DEC maps showed that the ipsilateral stratum radiatum was almost completely black, coinciding with severe dendritic degeneration shown in Thy1-YFP mice (Figures 6B and 6C). This finding suggests that dendrites of hippocampal neurons are highly sensitive to HI injury. To further examine this possibility, in Supplementary Figure 2, we used TUNEL, Golgi, anti-MAP2, and NeuN immunestains to compare the timing of somal-versus-dendritic injury in the hippocampus. This analysis indicated that dendritic destruction precedes the apoptosis of hippocampal neurons after cerebral HI.
Finally, to evaluate how early HI-induced dendritic degeneration occurs, we used T2-weighted MRI and EM to examine mouse brains collected at 6hours after HI (n=3). For EM, we compared the pathology in the stratum radiatum (consisting of apical dendrites), the pyramidal cell layer (the nucleus of pyramidal neuron), and the fimbria (axons of hippocampal neurons). This analysis showed widespread dissolution of the dendritic cytoskeleton (Den in Figure 6I) and protrusion of vesicles in axons (asterisks in Figure 6J), when no obvious change was observed in the hippocampus on T2-weighted MRI (Figure 6D) and no nuclear pyknosis in pyramidal neurons on EM (Figures 6E and 6H). Taken together, these data suggest that dendrites of hippocampal neurons undergo rapid degeneration after cerebral HI, for which DTI has higher detection sensitivity than does T2-weighted MRI.
Histopathology is the gold standard for establishing the diagnostic utility of any imaging method; however, the systematic imaging–histology correlation is not always feasible in patients. Hence, studies of animal models of the disease may illuminate the neuropathological basis of imaging findings in similar clinical settings. Previous studies have characterized the temporal evolution of diffusion tensor MRI in rodent or nonhuman primate models of cerebral ischemia at 3 and 24hours and beyond (Carano et al, 2000; Liu et al, 2007). This study uses ex vivo DTI combined with histology and biochemical assays to examine the neuropathological basis of several DTI parameters, including axial and radial diffusivities and fraction anisotropy. Our results suggest that cross-evaluation of multiple DTI parameters may provide a fuller picture of WM and dendritic injury in acute ischemic stroke.
Previous studies in patients of ischemic stroke showed a rapid reduction of eigenvalues in all diffusion axes within 24hours of symptom onset (Yang et al, 1999; Bhagat et al, 2008; Sakai et al, 2009). As FA is a measurement of the relative difference of the three eigenvalues, there was little-to-no decrease in the FA value in the hyperacute (2.5 to 7hours) phase of ischemic stroke. Only at ~24hours after stroke onset did patients show reduction of FA in subcortical WM. This secondary reduction of FA has been suggested to indicate severe axonal damage, but this hypothesis of important clinical application has not been validated by histology (Bhagat et al, 2008). If validated, changes of FA values in individual axonal tracts can be used to monitor WM injury following ischemic stroke.
In this study using a murine model of HI-induced thrombotic stroke, we have reproduced the unique pattern of DTI alterations in ischemic stroke patients. This pattern includes acute reduction of eigenvalues in all three axes (thus, the concomitant reduction of axial and radial diffusivities), as well as a smaller and more variable change of FA in different axonal tracts. We found that the fimbria, internal capsule, and external capsule all had 25% to 40% reduction of axial diffusivity at 24hours after HI, but showed a great variation in the degree of FA reduction. In the fimbria and internal capsule, there was <10% decrease in FA at 24hours after HI, whereas the external capsule showed >30% reduction at 24hours recovery. Importantly, a severe structural breakdown of the axons was only detected in the external capsule, but not in the fimbria or internal capsule. These results support the hypothesis that a large reduction of FA is an imaging marker of severe WM injury in acute ischemic stroke (Bhagat et al, 2008). However, more studies of the DTI–histopathology correlation are necessary to determine the threshold of FA reduction for irreversible axonal injury in acute stroke.
The cellular basis of alterations of radial and axial diffusivities remains incompletely understood (Beaulieu, 2002). Axial diffusivity is believed to represent water diffusion in the lengths of axons, and its reduction could be caused by increased viscosity in the intraaxonal space (owing to cytoskeletal breakdown) or focal enlargement constriction of neuritis (‘beading') (Budde and Frank, 2010). Microtubules and myelin sheaths are generally agreed to be principally responsible for radial diffusivity, because they constrain transverse water diffusion across the axons. In trauma- or axotomy-induced Wallerian degeneration of axons, radial diffusivity slowly increases as a consequence of demyelination (MacDonald et al, 2007; Zhang et al, 2009). In contrast, clinical studies of stoke patients and our data in animal models all demonstrated an acute, concomitant reduction of axial and radial diffusivities. This disparity of DTI alterations suggested unique pathologic changes in ischemic stroke. We found that edema, separation of myelin sheaths, and protrusion of oligodendrocyte-derived vesicles compressing the axoplasma are typical WM pathologies in acute ischemic stroke. Our data of large, intramyelinic vesicles are strikingly similar to those previously described in isolated optic nerves after anoxic insults (Waxman et al, 1992), and consistent with the notion that oligodendrocyte swelling is one of the earliest morphologic changes in cerebral ischemia (Pantoni et al, 1996). It is conceivable that an increase in extraaxonal water content and compression of the axoplasma by swollen myelin sheaths may lead to reduction of radial diffusivity and even axial diffusivity in acute cerebral ischemia.
Unlike Wallerian degeneration in which the injury of oligodendrocytes is a slow process (Vargas and Barres, 2007), our data suggested that oligodendrocytes undergo rapid degeneration after hypoxic-ischemic insults. Although the exact causes of this hypersensitivity remain to be determined, our results support the notion that reactive oxidative stress and a high iron content in oligodendrocytes contribute to their increased sensitivity to hypoxic-ischemic insults (Thorburne and Juurlink, 1996; Irving et al, 1997; McCracken et al, 2000; Dewar et al, 2003). Importantly, recent studies have suggested that oligodendrocytes have a critical role in maintaining axonal function and survival (Kassmann and Nave, 2008). Hence, supply of antioxidants to oligodendrocytes may be a useful strategy for WM protection in acute ischemic stroke.
This study also suggests the potential of DTI for detecting dendritic degeneration in hippocampal neurons, which may have implications beyond acute ischemic stroke. Diffusion anisotropy results from directional water flow in any tissue of a polarized intracellular structure and an orderly architecture (Basser et al, 1994). Although neural fibers are the prime example for structures with diffusion anisotropy, they are not the only such structures in the body. For example, a previous study using ex vivo DTI identified the stratum radiatum (consisting of dendrites of hippocampal neurons) in the mouse hippocampus (Zhang et al, 2002). Consistent with the earlier report, we also identified the stratum radiatum in the hippocampus by prominent diffusion anisotropy, most readily visualized on DEC maps and presumably attributed to an orderly array of apical dendrites. Furthermore, we showed that the loss of contrast on DEC maps in the stratum radiatum is associated with the degeneration of dendrites. Importantly, color-coded DEC maps have a greater sensitivity than T2-weighted MRI in detecting dendritic degeneration. Although ex vivo DTI was used in our study to increase the signal-to-noise ratio in small rodent brains, the advance of MRI technology may enable the use of in vivo DTI to detect dendritic degeneration, especially in laminated brain structures such as the hippocampus or cerebellum, in patients of ischemic stroke.
Moreover, using T2-weighted MRI followed by EM analysis, we have shown that the degeneration of hippocampal dendrites is a very early event in ischemic stroke. This conclusion is consistent with a recent in vivo two-photon confocal imaging study showing rapid degeneration of dendrites of cortical neurons in cerebral ischemia (Li and Murphy, 2008). The similarity of findings in these two studies suggests that dendrites are highly susceptible to stroke-related injury. On the basis of our results, we suggest the following sequence of pathologic events and associated imaging presentations in acute ischemic stroke:
It is interesting to note that WM-associated T2-hyperintensity in our study resembles the hallmark of leukoaraiosis, an incident MRI finding in the elderly and an important contributor of vascular dementia (Hachinski et al, 1986; Schmidt et al, 2007). Moreover, a recent study suggested that leukoaraiosis-like WM lesions are caused by chronic hypoxia hypoperfusion (Fernando et al, 2006), which could also injure dendrites. Taken together, our results and the relevant literature raise an intriguing possibility that the decline of cognitive functions in leukoaraiosis-like vascular dementia may be in part caused by the ‘invisible' dendritic degeneration (and the consequential loss of synaptic inputs) instead of strict WM injury. If this hypothesis is correct, DTI may be a useful tool for early diagnosis of dendritic degeneration and vascular dementia in the future.
In conclusion, we have used an animal model to study the utility of anisotropic diffusion imaging in acute ischemic stroke. Although our findings with ex vivo imaging have to be confirmed by in vivo imaging, the results of our study suggest that cross-evaluation of multiple DTI parameters may provide a fuller picture of WM and dendritic injury in acute ischemic stroke.
The authors thank Drs Ton Degrauw, Joseph Broderick, Lawrence Wechsler, and Charles Dumoulin, the anonymous reviewers for critical reading of this manuscript, and Dr Steve Danzer for providing Thy1-YFP-H mice.
The authors declare no conflict of interest.
Supplementary Information accompanies the paper on the Journal of Cerebral Blood Flow & Metabolism website (http://www.nature.com/jcbfm)
This work was supported by the National Institute of Health grant NS 074559 and the Alzheimer's Association (to C-YK) and an American Heart Association fellowship (to DY).