|Home | About | Journals | Submit | Contact Us | Français|
Periventricular leukomalacia, PVL, is the leading cause of cerebral palsy in prematurely born infants, and therefore more effective interventions are required. The objective of this study was to develop an ischemic injury model of PVL in mice and to determine the feasibility of in vivo magnetization transfer (MT) magnetic resonance imaging (MRI) as a potential monitoring tool for the evaluation of disease severity and experimental therapeutics. Neonatal CD-1 mice underwent unilateral carotid artery ligation on postnatal day 5 (P5); at P60, in vivo T2-weighted (T2w) and MT-MRI were performed and correlated with postmortem histopathology. In vivo T2w MRI showed thinning of the right corpus callosum, but no significant changes in hippocampal and hemispheric volumes. Magnetization transfer MRI revealed significant white matter abnormalities in the bilateral corpus callosum and internal capsule. These quantitative MT-MRI changes correlated highly with postmortem findings of reduced myelin basic protein in bilateral white matter tracts. Ventriculomegaly and persistent astrogliosis were observed on the ligated side, along with evidence of axonopathy and fewer oligodendrocytes in the corpus callosum. We present an ischemia-induced mouse model of PVL, which has pathologic abnormalities resembling autopsy reports in infants with PVL. We further validate in vivo MRI techniques as quantitative monitoring tools that highly correlate with postmortem histopathology.
Extreme prematurity, defined as a gestational age of 28 weeks, or birth weight <1,500g is a major cause of morbidity and mortality in children worldwide (Wen et al, 2004). Up to 20% of these infants develop cerebral palsy (Hamrick et al, 2004); an additional 25% to 50% experience attention deficit disorder, learning disabilities, or visual cortical impairments (Litt et al, 2005). This group also has a higher incidence of psychiatric morbidities (Johnson, 2007). Perinatal white matter injury (PWMI), also known as periventricular leukomalacia (PVL), is the predominant form of brain injury leading to such neuropsychiatric problems (Efron et al, 2003). Neonatal magnetic resonance imaging (MRI) studies have shown that 70% to 80% of all extremely premature infants develop white matter lesions (Dyet et al, 2006; Woodward et al, 2006). Furthermore, there is a direct relationship between MRI evidence of white matter injury and clinical severity of motor and cognitive impairment (Woodward et al, 2006). In addition to dysmyelination and astrogliosis (Folkerth, 2005), there is increasing evidence for neuroaxonal injury in these patients (Dammann et al, 2001; Volpe, 2005). Back et al (2007) reported that early PWMI lesions involve selective vulnerability and pronounced depletion of oligodendrocyte progenitor cells (OPCs) in humans and rodents. These cells are prominent during the peak period of PVL (i.e., between 24 and 32 weeks gestational age) (Back et al, 2007). There are also data suggesting that surviving OPCs undergo maturational arrest in PWMI lesions: a recent autopsy study of premature infants detected numerous morphologically immature Olig-2-positive cells in PVL lesions, indicative of arrested differentiation (Billiards et al, 2008).
Multiple animal models for PWMI have been established (Johnston et al, 2005). Although injuries in fetal baboons and sheep most closely mimic human disease, these large animal models have significant limitations: these species have more mature brains at birth than do humans, the maintenance of large animals requires costly facilities, and most importantly, the availability of DNA sequence information is often limited. In rodents, glutamate receptor agonists, perinatal hypoxia–ischemia, and lipopolysaccharide exposure have been used to reproduce white matter injury (Johnston et al, 2005). Although a number of rat models of PVL have been established, only two perinatal mouse models with relatively selective white matter injury have been reported. Two separate groups have produced a mouse model of PWMI by exposing newborn pups to chronic hypoxia for the first 3 weeks of life (Kanaan et al, 2006; Schwartz et al, 2004). These pups exhibit hypomyelination and bilateral ventriculomegaly. Unfortunately, this model requires animals to remain in a hypoxia chamber for at least 1 week, which limits the evaluation of therapeutics during the early postnatal period. Hagberg et al have recently reported relatively selective white matter injury in pups that were prenatally exposed to lipopolysaccharide in utero (Wang et al, 2006). The prenatal onset of injury in this latter model again hampers the evaluation of many therapeutics in the acute setting. Neonatal brain injury of varying degrees, depending on the severity and timing of an ischemic or hypoxic–ischemic insult, can be reproduced in rodents (Dammann et al, 2001; Hagberg et al, 2002). Age-related differences in injury severity may reflect changes in neuronal/glial vulnerability during each stage of brain development (Back et al, 2002). In terms of oligodendrocyte maturation, the development of the newborn rodent brain is equivalent to the human fetal brain at 6 to 7 months gestation (Craig et al, 2003). Late OPCs prevail between postnatal days 2 (P2) and P7 in rodents, and at gestational weeks 23 to 32 in humans (Craig et al, 2003). On the basis of these observations, our first experimental question was whether age-related selective vulnerability will result in preferential white matter injury after neonatal ischemia. Therefore, we hypothesized that unilateral carotid artery ligation on P5 (day of birth defined as P1) would lead to relatively selective white matter injury in CD-1 mice.
The objectives of this study were (1) to develop a mouse model for PWMI that could serve as a testbed for the study of experimental therapeutics and (2) to use the model to evaluate in vivo MRI as a monitoring tool for the assessment of white matter injury in the living animal. Magnetization transfer (MT) imaging is a relatively new technique, in which the image contrast is generated using an off-resonance spectrally selective radiofrequency pulse that saturates protons associated with macromolecules, such as myelin (Smith et al, 2005). We have previously shown that MT imaging can enhance the detection of white matter lesions in patients with demyelinating disorders (Fatemi et al, 2005; Smith et al, 2005). An important experimental question is whether the MT signal is a direct in vivo correlate of myelin density. Therefore, our second hypothesis was that MT-MRI will be able to detect and quantify white matter injury in mice with neonatal ischemia.
This study was approved by the Johns Hopkins Animal Care and Use Committee (protocol no. MO06M326 and MO09M422). In all, 13 CD-1 litters of 10 pups each were purchased from Charles River Laboratories (Wilmington, MA, USA) at P2. The day of birth was defined as P1.
On P5, mice were placed in an incubator at 35°C for 15 to 30minutes and were then anesthetized with isoflurane (4% induction 1% to 1.2% maintenance), and the right common carotid artery was ligated. Pups recovered at 36°C for 30 to 60minutes and were returned to the dam. Rectal temperatures were 36°C±0.5°C before surgery and 34.5°C±1°C postoperatively; surgery time was standardized at 12 to 15minutes.
Studies were performed on a Bruker 9.4-T vertical bore spectrometer (Bruker BioSpin Corp., Billerica, MA, USA) at P60. Low-dose isoflurane was used for sedation; heart and respiratory rates were monitored; and the isolette was maintained at 32°C. T2-weighted (T2w) MRI was performed with the following parameters: echo time=40milliseconds, repetition time=2,000milliseconds, field of view=16 × 13mm2, matrix size of 192 × 160, and total 40 slices with slice thickness 0.8mm. For MT-MRI, coronal MT-weighted images (i.e., two averages, offset=1.6kHz, saturation field strength=2.2μT, 5.5minutes scan time per slice, field of view, and matrix size as above) were obtained, at the level of the anterior hippocampus and at the most anterior and most posterior parts of the corpus. Images were then collected without an MT prepulse to obtain M0 images at the same coronal level as the MT-weighted images. The total scan time was ~40 to 45minutes per animal.
Magnetic Resonance Imaging Studies: Images were reconstructed to 256 × 256 using Paravision 3 (Bruker BioSpin Corp.). For volumetric analysis, data were converted into image format and imported into MIPAV software (NIH, Bethesda, MD, USA). The corpus callosum, the hippocampus, and the whole hemisphere were manually outlined in sequential coronal T2w images, and volumes of interest were calculated based on the thickness of each slice. For analysis of the MT-MRI data, MT ratio (MTR) maps, defined by were created and saved as image files and processed further in DTI Studio (Susumu Mori, Johns Hopkins University). The corpus callosum, at the level of the anterior hippocampus, and the internal capsule were manually outlined in the MTR maps in each hemisphere and mean MTR values calculated in the DTI studio.
A subset of animals were examined at P6 (24hours survival after ligation, n=8) or at P7 (48hours survival after ligation, n=8) and compared with P6 controls (n=7), whereas all other animals (29 ligated, 33 controls) were killed at P60. Mice were anesthetized and perfused with phosphate-buffered saline, followed by 4% paraformaldehyde in phosphate buffer postfixed with paraformaldehyde for 12 to 18hours. Brains were embedded in paraffin, sectioned at 20μm, and serial sections were used for immunohistochemical detection of glial fibrillary acidic protein (GFAP) and for hematoxylin and eosin, cresyl violet, and Luxol fast blue staining. For other immunohistochemical stains, brains were fixed with paraformaldehyde, cryoprotected in sucrose, and cryostat sectioned at 40μm. Slides were incubated in blocking solutions, followed by primary antibody incubation at 4°C: anti-myelin basic protein (MBP, Millipore, Billerica, MA, USA, 1:2,500), anti-CC1 antibody (Millipore, 1:1,000) for detection of mature oligodendrocytes, anti-platelet derived growth factor receptor-α (PDGFRα) antibody (BD Biosciences, San Jose, CA, USA, 1:500) for detection of OPCs, anti-GFAP (Dako, Carpinteria, CA, USA, 1:2,500) for detection of astrocytes, anti-iba-1 (Wako, Richmond, VA, USA, 1:2,000) for detection of microglia, anti-non-phosphorylated neurofilament SMI32 (Covance, Princeton, NJ, USA, 1:200) for detection of pathologic axons, anti-Pan-Neurofilament (PanNF, Covance, 1:200) for detection of all axons, and anti-activated caspase3 (aCasp3, Cell Signaling Technologies, Danvers, MA, USA, 1:200) for detection of apoptotic cells. The antigen–antibody complex was visualized using an ABC ELITE kit (Vector Labs, Burlingame, CA, USA). For aCasp3 staining, a 0.05% toluidine blue counterstain was used to outline anatomic structures. For double staining of aCasp3- and PDGFRα-positive cells, fluorescent secondary antibodies were used.
For quantification of MBP immunostaining, semiquantitative densitometry was performed using MCID Core (InterFocus Imaging Ltd, Cambridge, UK) after calibration with optical density standards. The right and left anterior, mid and posterior corpus callosum, and the right and left internal capsule/cerebral peduncles were manually delineated in three sections (outlined in red in Supplementary Figures A to C): at the level of the anterior commisure, anterior hippocampus, and posterior corpus callosum (forceps major), and the average density values for each of these areas were calculated in each animal and used for further statistical comparisons. For quantification of GFAP and iba-1 staining, three to four fields within the same regions were sampled using a × 20 objective. Threshold-based target detection was performed; images were reviewed, artifacts excluded, and fractional area stained was calculated (referred to as proportional area). This method has been shown to be a reproducible quantitative tool for assessment of monochrome histologic stains with less user dependency that unbiased stereology (Donnelly et al, 2009). For PanNF and SMI32 densitometry, the midsagital corpus callosum (outlined in yellow in Supplementary Figure) was selected in a minimum of three coronal sections per brain, and the average density value for each brain was used for further analysis. For quantification of aCasp3 staining, multiple images were taken under a × 40 objective along the left and right corpus callosum from the most anterior to the most posterior section for each animal, and aCasp3-positive cells were manually counted within each side of the corpus callosum and the average count used for analysis. For quantification of mature oligodendrocytes, images were taken under a × 40 objective along the corpus callosum on a CC1-stained section at the level of the anterior corpus callosum. Images were stitched together using the mosaic tool of the Zeiss AxioVision software (Zeiss Microimaging, LLC, Thornwood, NY, USA), and threshold-based automated cell counting was performed in MCID, and the total number of CC1 positive cells was counted on each side of the section from the midsagittal point to the medial edge of the lateral ventricle. Target discrimination was visually verified for each single image, and target selections that appeared to be artifactual were manually deleted.
One-way ANOVA (analysis of variance) with Bonferroni's post hoc comparison were performed to assess for significant differences in MRI (volume of interest and MTR values) and histologic variables within identical regions between ligated and controls. Regression analysis was performed to evaluate the correlation between measured MTR values and MBP density measures.
Right-sided unilateral carotid artery ligation was performed in 60 mice on P5; 40 littermates were used as normal controls. The mortality rate of the ligation procedure was 22% (13/60) and was caused by either rupture of the carotid artery or failure to recover from anesthesia. Postsurgical death occurred in 2 of 60 (3%) mice, both within the first 48hours, increasing the total mortality to 25%.
In vivo MRI studies were performed on P60 in 34 animals. Three imaged animals had sustained large hemispheric strokes with cavitation of the injured hemisphere that could be identified on the localizer MRI sequence; these were excluded from further imaging and are not reported here. From the remaining 31 animals, 7 studies were discarded for excessive motion artifacts, and 12 ligated and 12 control animals were used for further analysis.
T2-weighted MRI in the ligated animals showed ventriculomegaly on the side of the ligation and subtle T2-signal hyperintensity in the bilateral corpus callosum (Figure 1). Volumetric analysis (Figure 2) detected unilateral thinning of the corpus callosum on the ligated side, compared with controls (P=0.03).
Visual inspection of the MTR maps revealed varying degrees of signal reduction bilaterally in the corpus callosum and internal capsule in 10 of 12 ligated animals (Figure 3A). Quantitative analysis showed that this reduction of MTR values in the corpus callosum and internal capsule was significant compared with controls (Figure 3B). Although two of the ligated animals showed a diffuse MTR signal hypointensity throughout most of the ligated hemisphere (Figure 3A, ligated 2), there was no significant asymmetry when the entire ligated group was analyzed.
Neuropathological evaluation of hematoxylin and eosin- and Nissl-stained sections of P6, P7, or P60 animals revealed a varying degree of unilateral ventriculomegaly on the ligated side, but otherwise no evidence of any necrosis or major structural lesions in the majority of animals. However, 5 of 45 ligated animals that survived ligation showed a large hemispheric stroke (including the 3 animals that were identified with this injury during the MRI studies reported above and 1 of the pups in the 48-hours survival group). These animals were excluded from further analysis.
Evaluation of Apoptosis: Immunohistochemical staining of aCasp3 performed 24 and 48hours after unilateral carotid artery ligation showed a significant bilateral increase in the number of aCasp3-positive cells in the corpus callosum of ligated mice (Figure 4A), whereas only 1 to 2 cells per section were positively stained in the gray matter (cortex, hippocampus, striatum), and a comparable number of such cells was also seen in the control gray matter. The majority of apoptotic cells in the white matter had a small round soma, from which several slender primary processes grew, bifurcating multiple times (Figure 4B). This morphology was suggestive of oligodendrocyte progenitors or immature oligodendrocytes; therefore, we pursued double staining with aCasp3 and PDGFRα, a marker for OPCs. As shown in Figure 4C, the morphology of PDGFRα-positive cells is very similar to that of aCaps3-positive cells (Figure 4D), and many of these cells coexpressed both markers (Figure 4E), suggesting that oligodendrocyte progenitors were undergoing caspase-dependent apoptotic cell death.
Oligodendrocyte Cell Count: Mature oligodendrocytes were counted using sections stained with the CC1 antibody in the anterior corpus callosum at P60 (ligated n=6, controls n=6). There were significantly fewer mature oligodendrocytes in the ligated corpus callosum than those in controls (left ligated versus left control: 190±20 estimated counts versus 228±25, P=0.02; right ligated versus right control: 184±21 versus 222±25 estimated counts, P=0.03, see Figure 4F). This reduction was bilateral and there were no significant interhemispheric differences (Figures 4G and 4H).
Myelin Basic Protein Densitometry: Immunohistochemical studies at P60 showed a highly significant bilateral reduction of MBP density in the anterior, mid, and posterior corpus callosum and in the internal capsule of ligated animals, compared with controls. However, there was no significant difference between the ligated and contralateral hemispheres (Figures 5A and 5B). There was no significant reduction in the fimbriae of the hippocampi between ligated versus contralateral side or ligated versus control animals.
Analysis of Astrocytes and Microglia: Proportional area analysis of P60 animals showed a highly significant increase in GFAP staining in the anterior corpus callosum, external capsule, and internal capsule on the ligated side, when compared with controls or with the contralateral side, suggesting persistent unilateral gliosis. Morphologically, cells appeared to have larger cytoplasm and more prominent processes (Figure 6A). Interestingly, the contralateral hemisphere of the injured animals did not show any significant difference in GFAP staining when compared with controls (Figure 6B). The gliosis appeared most prominent in the periventricular white matter. Analysis of microglial staining revealed no significant difference in the proportional area measurements of iba-1-stained sections at P60 in the corpus callosum (0.05±0.008 (s.d.) ligated side versus 0.06±0.008 (s.d.) contralateral; 0.05±0.008 (s.d.) ligated side versus 0.06±0.005 (s.d.) control animal) and in the internal capsule (0.06±0.007 (s.d.) ligated side versus 0.06±0.006 (s.d.) contralateral; 0.07±0.007 (s.d.) ligated side versus 0.06±0.005 (s.d.) control animal).
Axonal Integrity Assessment: Densitometric analysis of SMI32 immunohistochemical in the mid corpus callosum at P60 showed a significant increase in staining in ligated animals, compared with control littermates (P<0.05), suggesting abnormal phosphorylation of axonal neurofilaments (Figures 7A and 7B). In contrast, PanNF sections in the same animals within the same region showed no significant difference in total neurofilament staining (Figure 7C), compared with controls, suggesting that there was no overall loss of axons.
Magnetization transfer ratio values were randomly sampled in 25 different regions of interest (regions of interest defined as in the ‘Materials and methods' section) in ligated and control animals at P60 and compared with their respective postmortem MBP densitometry values. In vivo MTR values, were highly correlated with postmortem MBP density, as shown in Figure 5C (r2=0.695, P<0.0001).
Periventricular leukomalacia is the leading cause of cerebral palsy and currently no effective treatment exists. Ischemia–reperfusion and inflammatory processes are considered the underlying culprits for the cascade of injury that eventually leads to abnormal myelination and axonal injury. A mouse model for PVL can be helpful for further understanding of pathologic mechanisms involved in the molecular cascade leading to white matter injury and serves as a testbed for the evaluation of experimental therapeutics. In this study, we present an ischemia model of PVL, induced by performing unilateral carotid artery ligation at a very early postnatal age.
The first important finding in our study is that unilateral carotid artery ligation when performed on P5 in CD-1 mouse pups leads to quite selective injury to the white matter in most of the animals. Our group and others have reported unilateral ischemia combined with hypoxia in neonatal mice, which typically leads to unilateral striatal, hippocampal, and cortical injuries (Johnston et al, 2005). In all these studies, hypoxia–ischemia was induced in older mice typically between P7 and P12 and typically adds a period of hypoxia, leading to more severe injury. Our group has shown that C57Bl/6 and C3HeB/FeJ mice were much less vulnerable to seizures and brain injury after unilateral carotid ligation than were CD-1 mice at P12 (Comi et al, 2009). Differences in vascular anatomy and other genetic strain-related differences are likely to contribute to different vulnerability. Furthermore, Back et al (1998) have shown age-selective vulnerability of OPCs, which along with data regarding the development of glutamate receptors in the cortex (Johnston, 2009) was the basis for selection of P5 as a ligation time point for our white matter injury model.
The majority of apoptotic cells that we found during the acute phase in our study were in the white matter, and had a morphology similar to oligodendrocyte progenitors or immature oligodendrocytes and many coexpressed aCasp3 and PDGFRα, leading to the hypothesis that these cells are selectively targeted by ischemia or the downstream excito-oxidative cascade known to be triggered by ischemia, and ultimately the apoptosis of these oligodendrocyte lineage cells leads to dysmyelination. We also found evidence of axonal injury in the corpus callosum in our model, similar to recent reports from autopsy cases of PVL (Billiards et al, 2008). Therefore, an alternative hypothesis can be offered here: The process of final differentiation and myelin synthesis of OPCs is orchestrated by a complex interaction between axons and these cells (Bozzali and Wrabetz, 2004). These immature cells are known to express a high level of glutamate receptors that are excitatory and respond to stimuli received from axons (Bergles et al, 2010). Altered signaling between axons and oligodendrocyte progenitors can activate an apoptotic pathway in these progenitors. Hence, we postulate that ischemia or related downstream effects will primarily compromise axons, which then leads to abnormal signaling to oligodendrocyte progenitors, resulting in their apoptosis. This phenomenon could explain the bilateral dysmyelination observed in our study, even though arterial ligation was unilateral. A third hypothesis would be that ischemia and related downstream effects lead to a differentiation arrest or alternative differentiation of OPCs into astrocytes. In line with this theory, we found a striking increase in the proportional target area of GFAP-stained sections, suggesting increased astrocytic cell body size, or increased number of astrocytes, or both. To answer this question, further studies using transgenic mice with cell-specific inducible reporter genes would be necessary, which could allow the detection of the fate of these cells after ischemia by inducing the reporter gene at the time of injury. As our model uses mice, we plan to conduct these studies in the future.
The second important finding in our study is that the MTR maps can nicely detect the severity of white matter injury in a quantitative manner in the living mouse. Although postmortem histology can be used as the ultimate outcome, it does not allow the assessment of evolving changes of function within an individual. In terms of functional testing of rodents, it is believed that neurologic abnormalities are very subtle in mouse models of perinatal brain injury and that these models usually do not exhibit prominent motor impairment, as seen in patients with PVL-associated cerebral palsy (Johnston et al, 2005). Ment et al have shown locomotor hyperactivity in the chronic sublethal hypoxia mouse model, and others have shown abnormal open-field testing after neonatal hypoxia–ischemia in mice (Weiss et al, 2004). Rather than using these limited functional studies, which often require animal adaptation to a new environment and are time consuming, one could use the in vivo imaging studies we report here as surrogate markers for disease progression and outcome. However, one weakness in our study is the lack of functional assessment in our model. We plan to conduct these studies in the future and correlate the findings with the MRI results.
Interestingly, there were bilateral reductions in MTR, MBP density, and in the number of mature oligodendrocytes (CC1+), but the reduction observed in corpus callosum volume was limited to the ligated side. Abnormal SMI32 staining at P60, combined with normal PanNF staining, is consistent with persistent axonopathy without axonal loss. We speculate that although there is equal bilateral reduction of myelin content in the corpus callosum, pathologic changes in axons originating in the ligated hemisphere may result in smaller caliber axons, and a thinner corpus callosum on the ligated side. Similarly, others have shown that in cuprizone-induced demyelination, the loss of oligodendrocytes is associated with SMI32 staining of axons and a reduction in axon caliber (Lindner et al, 2009; Mason et al, 2001). The thinning of axons would not affect the MT-MTR signal, because the main molecules that are believed to affect MT are the large lipid moieties of myelin. We also note that the lack of axonal loss is not consistent with reported human autopsy cases in which there is loss of axons and neurons (Billiards et al, 2008). A number of hypotheses can be speculated here. (1) Our injury model may be milder than the reported human autopsy cases, because these cases are likely to reflect the more severe end of PVL; (2) It is possible that there is a greater degree of axonal regeneration in our rodent model than in human infants with PVL. Future electron microscopy experiments at different time points after ischemia are required to further delineate the axonal pathology in this model.
A third and perhaps most important finding in our study is the striking degree of correlation between the in vivo MT-MRI and the postmortem myelin density. Magnetization transfer imaging has been widely used in multiple sclerosis and other demyelinating conditions as a sensitive tool to detect demyelination (Fatemi et al, 2005; Horsfield, 2005; McGowan et al, 2000). Here, we find that in vivo MTR is an excellent predictor of myelin density. A strong correlation has been shown between MT and the severity of demyelination in cuprizone-induced white matter injury (Zaaraoui et al, 2008) and in inflammatory-induced demyelination models (Gareau et al, 2000); however, these other injury models show quite severe demyelination and inflammation. Our model shows a mild-moderate degree of dysmyelination, and although inflammation may have an important role shortly after ischemia, at the time MRI was conducted (P60), there was no evidence of edema or microglial activation. Therefore, we conclude that the MTR decrease is likely a direct correlate of myelin content. One limitation of our study is the fact that precise coregistration between in vivo imaging and postmortem histology is extremely difficult to achieve. However, one would expect that limitations of coregistration should reduce the amount of correlation, rather than artificially increase it.
In summary, we report an ischemia-induced neonatal mouse model of PVL in which we find selective vulnerability of the white matter and histologic abnormalities that resemble autopsy reports in infants with PVL. We further validate in vivo MRI techniques as quantitative monitoring tools for assessment of injury severity and predictors of histopathology. This model has an advantage over previously reported mouse models for PVL, because the injury is induced at a single time point postnatally, hence opening a window for acute interventions. This model, therefore, combined with the techniques presented can serve as a feasible testbed for evaluation of experimental therapeutics.
The authors thank Patrice Carr and Karen Connor for their kind assistance with immunohistochemical studies. The contents of this report are solely the responsibility of the authors and do not necessarily represent the official views of the National Institute of Health, DHHS.
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 study was funded by the National Institute of Health (NINDS R01NS028208 (MVJ), NINDS K08NS063956 (AF), NIBIB K01EB006394 (MTM), NIBIB R01EB003543 (SM)) and the Miracle for Megan Foundation (AF).