Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Evaluating regional blood spinal cord barrier dysfunction following spinal cord injury using longitudinal dynamic contrast-enhanced MRI

Evaluating regional blood spinal cord barrier dysfunction following spinal cord injury using... Background: In vivo preclinical imaging of spinal cord injury (SCI) in rodent models provides clinically relevant information in translational research. This paper uses multimodal magnetic resonance imaging (MRI) to investigate neurovascular pathology and changes in blood spinal cord barrier (BSCB) permeability following SCI in a mouse model of SCI. Methods: C57BL/6 female mice (n = 5) were subjected to contusive injury at the thoracic T11 level and scanned on post injury days 1 and 3 using anatomical, dynamic contrast-enhanced (DCE-MRI) and diffusion tensor imaging (DTI). The injured cords were evaluated postmortem with histopathological stains specific to neurovascular changes. A computational model was implemented to map local changes in barrier function from the contrast enhancement. The area and volume of spinal cord tissue with dysfunctional barrier were determined using semi-automatic segmentation. Results: Quantitative maps derived from the acquired DCE-MRI data depicted the degree of BSCB permeability variations in injured spinal cords. At the injury sites, the damaged barriers occupied about 70% of the total cross section and 48% of the total volume on day 1, but the corresponding measurements were reduced to 55% and 25%, respectively on day 3. These changes implied spatio-temporal remodeling of microvasculature and its architecture in injured SC. Diffusion computations included longitudinal and transverse diffusivities and fractional anisotropy index. Comparison of permeability and diffusion measurements indicated regions of injured cords with dysfunctional barriers had structural changes in the form of greater axonal loss and demyelination, as supported by histopathologic assessments. Conclusion: The results from this study collectively demonstrated the feasibility of quantitatively mapping regional BSCB dysfunction in injured cord in mouse and obtaining complementary information about its structural integrity using in vivo DCE-MRI and DTI protocols. This capability is expected to play an important role in characterizing the neurovascular changes and reorganization following SCI in longitudinal preclinical experiments, but with potential clinical implications. Page 1 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 data collection, post processing and spatial representation Background Damage to blood-spinal cord barrier (BSCB) occurs as a of the BSCB permeability variations in injured SC. consequence of mechanical insult to spinal cord (SC) [1]. The damaged barrier becomes permeable to blood con- Histological studies have demonstrated that, in injured stituents, inflammatory cells and other large molecules, SC, the areas with damaged vasculature overlap with the which collectively activate a cascade of secondary proc- areas of intense neuronal loss [11,12]. Noninvasive diffu- esses harmful to the underlying tissue. With time, vascular sion tensor imaging (DTI) provides information about the bed of the injured SC goes through repair and remodeling neuronal integrity in underlying SC tissue [13,14]. Com- [2]. Providing nutrition and preventing extravasation of bining these, it is likely that in vivo BSCB permeability destructive biochemical compounds in blood protect the and DTI-based measurements, both obtained from the remaining neurons and maintain the existing substrate same region of the injured SC, can provide relevant infor- from further degeneration. At the same time, knowledge mation about the viability of underlying neurovascula- of barrier properties and its status is essential if potential ture, which has so far only been available by performing intravascular drugs with capabilities to improve neurovas- ex vivo histological analysis. Therefore, the third goal of cular protection and to promote repair and recovery will this study is to compare the BSCB permeability and DTI be administered. Such treatment is possible provided that measurements and investigate the nature of their associa- the drugs can pass through the open barriers to reach the tion while assessing the neurovascular response in SCI. destined regions in the SC parenchyma. Therefore, a com- plete understanding of the vascular response to spinal Studies based on postmortem tissue analysis suggest that cord injury (SCI) is required for developing intervention the first 72 hours following SCI are the most critical, offer- strategies aimed at rapidly restoring the barrier integrity as ing a window of opportunity for potential treatments [15- well as blood supply to the ischemic areas of the trauma- 18]. In order to develop effective therapeutic strategies in tized cord. experimental studies, it is important to employ a SCI model that sensitively responds to the treatment at the So far, BSCB permeability changes in injured SC have acute phase of the injury and is also capable of producing been evaluated using a range of tracers and laborious post- measurable vascular and neuropathological changes mortem tissue analysis [3]. Alternatively, in vivo dynamic within the first 72 hr time frame. Therefore, this study contrast-enhanced MRI (DCE-MRI) was proposed to non- focused on the acute phase (postinjury days 1 and 3) of invasively visualize and quantify the changes in BSCB per- the injury with the above pathological properties and meability in a relatively more efficient way [4-6]. The investigated the injured cords using multimodal neuroim- initial DCE-MRI studies were performed using a rat ani- aging (anatomical, DCE-MRI and DTI) noninvasively. In mal model of SCI. Lately, the availability of diverse strains the following, we first describe our imaging protocols, and transgenic varieties has made the mouse a more and then give details of our implementation and basis of attractive model [7-10]. But, to date, DCE-MRI studies on our data processing algorithm and strategies. Next, using mouse are lacking. The first goal of this paper is to dem- these developments, we present results from neuropatho- onstrate the feasibility of performing DCE-MRI on mouse logical evaluations and generate reliable BSCB permeabil- with SCI. ity maps to show the spatiotemporal course of the alterations in vascular permeability within the SC lesion The contrast enhanced data in previous rat studies were and its surroundings. We characterize the area and vol- acquired dynamically over an extended period of time ume of the regions with dysfunctional BSCB. We also test and processed using a pharmacokinetic model with com- the BSCB permeability estimates against the DTI measure- plicated numerical computation routine. The end results ments from the corresponding regions to establish the from the computation included quantitative measure- level of association as determined from statistical analysis. ments that represented the overall exchange of contrast agent between plasma and lesion. Such representation Methods was useful, but yielded limited information, since in prac- All experiments were carried out with twelve-week old tice, rather than global evaluation, more detailed local female C57BL/6 mice (n = 5) in accordance with a proto- variations in BSCB permeability was sought after. In this col approved by the Institutional Animal Care and Use previous approach, the DCE-MRI data acquisition time Committee. All of the mice were subjected to SCI and par- was long since it required covering both the wash-in and ticipated in the MRI scans using DCE-MRI and DTI proto- wash-out phases of the contrast agent. Another weakness cols on postinjury days 1 and 3. was that the parameters were estimated using a time-inef- ficient computation algorithm. Thus, the second goal of Procedures for surgeries and spinal cord injury this study is to provide an improved method in terms of All surgeries were performed in sterile conditions. The mouse was initially anesthetized by a spontaneous inha- Page 2 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 lation of 4% isoflurane in an induction chamber and then TE = 2500/12 ms, field-of-view (FOV) = 26 × 8 mm , moved to a surgery mat. The anesthesia level was reduced matrix = 256 × 128, in-plane pixel resolution = 100 × 63 to 2% isoflurane delivered in a mixture of 40% oxygen μm, slice thickness = 0.5 mm, number of excitations and 58% air through a nose mask. Further adjustments in (NEX) = 2. The corresponding parameters for the axial small increments were made on the percentage of the iso- images were TR/TE = 2500/12 ms, FOV = 12 × 8 mm , flurane level during the surgery. For intravascular delivery matrix = 128 × 128, in-plane pixel resolution = 94 × 63 of the paramagnetic contrast agent with MW = 938 Da μm, slice thickness = 1 mm, number of slices = 14 and (Magnevist, Berlex Imaging, Wayne, NJ) in DCE-MRI NEX = 2. These data constituted the proton density (PD) studies, animals underwent additional jugular vein cathe- weighted images. terization with PE-10 tubing [5]. The catheter ran subcu- taneously and exited the skin at the back between Then, for microstructural imaging, diffusion weighted shoulder blades, then was kept folded behind the animal. images were acquired using diffusion gradient strength = The catheter was flushed with heparin daily to prevent 80 mT/m, width (δ) = 6.5 ms and separation (Δ) = 11 ms clotting. The surgical procedures for inducing SCI to produce a b-value of 534 s/mm . The imaging parame- involved a midline incision posterior from the thoracic ters for these scans were TR/TE = 2000/26 ms, FOV = 12 × levels T10 to T12, followed by dissection of the bilateral 8 mm , image matrix = 128 × 128, in-plane pixel resolu- vertebral muscles to expose the dorsal laminae and tion = 94 × 63 μm, slice thickness = 1 mm and NEX = 2 spinous processes. Laminectomy was performed at the [20-22]. Baseline data acquired in the absence of the dif- T11 level to expose the SC by exercising special care not to fusion weighting constituted the T2-weighted images. damage the dura mater. The spinous processes at T10 and T12 adjacent to the laminectomy were stabilized using Next, for evaluating microvascular response and imaging two hemostatic forceps. A 1-mm diameter injury bit, that BSCB permeability to the contrast agent in the injured ani- was attached to a generic central nervous system injury mal, a DCE-MRI protocol was applied [5]. The contrast device described earlier [19], was positioned perpendicu- agent was delivered as a bolus (< 5 s) at a dose of 0.1 lar to the dorsal surface of the SC. The device consists of mmol/kg while the animal was still in the scanner. To electromechanical components – a linear motor con- detect the contrast enhancement, T1-weighted axial nected to a controller. The controller communicated with images were acquired precontrast and repetitively for up a personal computer through a software program devel- to 2 hr postcontrast using the same parameters as the PD oped in our laboratory to input biomechanical parameter weighted images but with TR = 1000 ms and NEX = 4. The values for inducing a contusion-type SCI. The injury temporal resolution between these acquisitions was 10 parameters used in these experiments were: impact veloc- min. ity of 0.75 m/s; surface displacement depth of 0.5 mm; and compression duration of 85 ms. After the injury, the Quantitative MRI data analysis overlying muscle layers were sutured and skin was closed The MRI data were acquired and visualized using the scan- tightly. Then, the injured mouse was left to recover in a ner's control software VNMRJ (Varian, Palo Alto, CA). heated cage and received postoperative care. DCE-MRI Analysis MRI scans The DCE-MRI data were analyzed off-line using custom- Each injured mouse was scanned on days 1 and 3 using a written software in Matlab (The Mathworks, Inc., Natick, 9.4 T horizontal INOVA Varian system (Varian, Palo Alto, MA). Axial images acquired before and after the contrast CA) and an inductively coupled surface coil [20]. The scan enhancement in a given scanning session were loaded was performed when the mouse was under a general into computer and processed interactively using a graphi- anesthesia which was delivered as a mixture of 2% isoflu- cal user interface. The details of the performed numerical rane, 40% oxygen and 58% air through a nose mask. Vital analysis are described in Appendix. The results are 2-D signs (respiration, heart rate and body temperature) of the maps that quantitatively describe the BSCB permeability anesthetized animal were monitored using a MRI-com- (denoted by the parameter K ) throughout the injured p-sc patible monitoring and gating system (Model 1025, SA SC. Instruments Inc., Stony Brook, NY). Respiratory-gated acquisition was used to increase the image quality by min- Area and volume of dysfunctional BSCB in injured SC imizing breathing related image artifacts. By analyzing the spatial distribution of the permeability changes on K maps, we measured the area and volume p-sc High-resolution anatomical images on all animals were of the injured SC tissue with damaged BSCB. To perform first acquired in sagittal and axial views using a spin-echo the area measurements, a threshold value was required to sequence in multislice and interleaved fashion. The scan differentiate the elevated K values from the background p-sc parameters for the sagittal or horizontal images were TR/ noise. The value of this threshold was determined while Page 3 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 postprocessing the data. To accomplish this task, an algo- DTI analysis rithm was implemented using the theory of signal detec- Diffusion-weighted images were processed and the ele- tion in the presence of additive noise. The procedure ments of diffusion tensor were estimated for each image involved first computing histograms of K maps from voxel using the scanner software VNMRJ (Varian Inc., Palo p-sc two normal sections of the injured cord, but within the Alto, CA). The diffusion tensor represents the statistical masks of the corresponding slices at 4 mm caudal and ros- distribution of microscopic motion of water molecules, tral to the injury epicenter. Because BSCB in these regions and its three eigenvalues (λ , λ and λ ) characterize the 1 2 3 were intact, the contrast agent was retained within the SC principal water diffusivities along its three orthogonal vasculature and no detectable leakage took place. These eigenvectors [21,23,24]. To be consistent with the nomen- histograms depicted K estimates that were normally dis- clature used in previous reports, we denoted λ = λ as an p-sc 1 || tributed around zero, which could be fitted to zero mean expression of longitudinal diffusivity of water molecules Gaussian profiles (please see the results section). The along the axonal fibers, and λ = (λ +λ )/2 as an expres- ⊥ 2 3 larger standard deviation σ from the two Gaussian fits sion of transverse diffusivity for water moving perpendic- (from a caudal and a rostral slice) was used to set a thresh- ular to the direction of neuronal fibers in SC. The at +2σ . This selection was based on the old value for K fractional anisotropy (FA) index was computed algebrai- p-sc property that the Gaussian distribution between -2σ and cally by combining the eigenvalues. FA index is a rotation- +2σ equals to nearly 95% of the total area beneath the ally invariant scalar and quantitatively characterizes the curve. This meant that nearly 97.5% of the pixels in the K degree of anisotropy in the diffusion properties of the p- map from a normal SC were below the threshold and underlying tissue within the voxel. The parameters λ , λ sc 1 ⊥ therefore considered to have intact BSCB. The same and FA were estimated for each voxel throughout the slice threshold value was employed for analyzing the K maps at the epicenter and the corresponding maps were gener- p-sc produced for all the remaining slices. The histograms ated in 2-D. Similar to computing mean K values p-sc from slices near the injury epicenter were observed to be described above, λ , λ and FA measurements from the || ⊥ shifted towards higher K values, as expected, reflecting areas with compromised BSCB at the epicenter were aver- p-sc the presence of compromised BSCB. These histograms aged and the resulting mean values for each animal were had wider spread but still followed the profile of Gaussian recorded in a database along with the corresponding . Group mean and standard deviation of these distribution. mean K p-sc measurements were again computed for days 1 and 3 sep- The number of pixels whose intensity values on the K arately. p-sc maps remained above the threshold was counted to deter- Postmortem tissue analysis mine the area of the SC with compromised BSCB in a given slice. This number was further divided by the total Neurovascular histopathologies of selected injured SCs cross sectional area of the cord in that slice to obtain a were examined postmortem following MRI scan on day 3. measure of normalized area (NA). The slice with largest Each mouse was euthanized by intracardiac perfusion pixel count was considered as representing the injury epi- with 50 mL of phosphate buffered saline (PBS) solution, center. The volume of the compromised BSCB was deter- followed by 50 mL of 4% formaldehyde PBS solution that mined by summing the areas (prior to normalization) in were delivered through a 23-gauge needle connected to a six neighboring slices covering the epicenter. Total cord perfusion pump. The SC was excised and fixed in 4% for- volume was similarly calculated from the total cord area maldehyde. Segments from the injury epicenter or normal in each slice. The total volume with compromised BSCB levels were embedded in paraffin and cut serially in 10 μm was scaled by the total SC volume to obtain normalized thick sections. Representative samples were stained with volume (NV). This normalization of the area and volume standard hematoxylin and eosin (H&E), luxol fast blue measurements compensated for the variations from slice (LFB), vascular endothelial marker (CD34) or neuron spe- to slice and animal to animal, as well as other spatial cific enolase (NSE) for histopathological assessment of scales, such as changes in slice thickness and pixel dimen- the neurovascular pathological changes. Immunohisto- sions. chemistry (IHC) was carried out with CD34 (MY 10 clone, BD Bioscience, Franklin Lakes, NJ) (1:100 dilution) and In addition, K measurements remaining above thresh- NSE (BBS/NC/VI-H14 clone, Dako Cytomation, Carpin- p-sc old within the total area at the epicenter slice were aver- teria, CA) antibodies using a Dako cytomation autostainer aged for each animal. The resulting averages for day 1 and [25]. Briefly, the slides were de-paraffinized by incubation day 3 were then correlated with the corresponding mean in xylene and ascending grades of alcohol. Antigen DTI measurements within the same area, as explained retrieval for CD34 was done by incubation with protein- below. ase Kinase at room temperature for 5 minutes. Antigen retrieval for NSE was done by heating in ethylene diamine tetraacetic acid (EDTA), PH 9.0 buffer (Lab vision, Fre- Page 4 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 mont, CA) at 95°C for 20 min. Both sections were loaded achieve an optimal tuning and matching condition as in the autostainer programmed as follows: 3% hydrogen observed in the frequency response of the coil impedance. peroxide for 10 minutes, blocked with 5% skimmed milk When the animal was placed in the magnet, the catheter for 5 min, incubated with primary antibody for 30 min, attached to it was extended outside at the front end of the followed by incubation with Envision+ system HRP magnet bore. The injured cord was then imaged in trans- (Dako Cytomation, Carpinteria, CA) for 30 minutes. verse and rostral-caudal planes. Acquiring data from dif- Color was developed by incubating samples with diami- ferent orientations allowed better evaluation of the nobenzidine (DAB)+chromogin for 10 min followed by lesion's spatial extent. Dako DAB enhancer for 5 min. Hematoxylin was used as counter stain. The sections were examined using a BX 50 Anatomical axial PD images acquired from one of the Olympus microscope and the photographs were taken injured SC are shown in Figure 1. The intensity contrasts with an attached Olympus DP 70 camera operated with on these images from normal sections delineate gross ana- DP Controller software (Olympus Corporation, Center tomical details of the cord within the white matter (WM) Valley, PA). One section was also incubated with second- and grey matter (GM) as well as the surrounding spinal ary antibody only to check nonspecific binding. structures. The lesion is depicted by an altered intensity contrast in the parenchyma below the laminectomy. The Statistical analysis corticospinal tract in the mouse is anatomically located The quantitative data collected from the measurements between the dorsal horns ventrally next to the central were analyzed statistically. The NA, NV, K , λ , λ and FA canal. In normal cords, the image intensity profile does p-sc || ⊥ values gathered from each animal were listed in two not produce enough contrast to distinguish this tract from groups as day 1 and day 3, and the means and standard the surrounding WM and GM. In this particular injured deviations within each group were computed for each SC, images from the rostral, but not the caudal, sections parameter to understand the intra-group variations. The delineated the corticospinal tract with hyperintensity. The measurements between day 1 and 3 were compared using intensity change was indicative of alterations in the MR the paired Student's t test. This analysis allowed examina- properties of the tract and of a pathological abnormality, tion of inter-group variations and the determination of which was likely associated with Wallerian degeneration the statistical significance of the differences. Statistical sig- [26]. nificance was defined at P < 0.05. Also, statistical depend- encies between K , and measurements λ , λ and FA Figure 2 presents pre- and postcontrast T1-weighted p-sc || ⊥ were determined using Pearson's correlation analysis and images in sagittal and horizontal planes in both postin- the resulting correlation coefficients were reported. jury days 1 and 3. On day 1, the precontrast images revealed a small focal hypointensity close to the dorsal surface reflective of neuropathology. On day 3, the lesion Results The surgical procedures, injuries, and prolonged anesthe- assumed a circular shape and enlarged in size. The post- sia during MRI scans were well tolerated by all mice. For contrast images depicted intensity enhancement at the imaging, each mouse was carefully placed supine over an lesion and its surroundings. The hyperintense regions in inductively coupled radio frequency coil system to postcontrast images represented areas of injured SC tissue Axi Figure 1 al proton-density images of an injured mouse spinal cord Axial proton-density images of an injured mouse spinal cord. The serial images show normal caudal and rostral sec- tions and injury epicenter on postinjury day 1. Arrows point to corticospinal tract (CST). In normal cord, the image intensity profile does not produce enough contrast to distinguish the CST from the surrounding white matter. Interestingly, in this injured SC, the CST at the rostral section, but not the caudal section, has been delineated by slight hyperintensity compared to the background white matter. Page 5 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 Precontrast an m Figure 2 ouse spinal cor d postcontrast d in sagittal and horizontal views T1-weighted images of injured Precontrast and postcontrast T1-weighted images of injured mouse spinal cord in sagittal and horizontal views. The postcontrast images were acquired 130 min (sag- ittal) and 140 min (horizontal) after the IV bolus delivery of the contrast agent. loaded with the contrast agent due to its leakage through the compromised BSCB therein. The postcontrast sagittal images in the figure were acquired 130 min and those in the horizontal views were acquired 140 min after the delivery of the contrast agent. Following its delivery, the contrast agent diffuses passively in the extravascular spaces of the injured SC parenchyma. This leads to the brightness enhancement spreading along the cord in both rostral and caudal directions, as evident in the figure. The brightness enhancement expanding spatially with time was better appreciated in images from axial views in Figure 3. This figure shows strong contrast enhancement Contr con day 3 Figure 3 trast agent with time a ast enhancement followin nd spac g the administration of the e on postinjury day 1 and early on at the lesion and its immediate surroundings fol- Contrast enhancement following the administration of the contrast agent with time and space on postin- lowing the contrast agent delivery. The enhancement was jury day 1 and day 3. The red and blue circles represent mostly localized in the GM than the WM. But as time pro- the regions of interest selected to produce the data in Figure gressed, the SC parenchyma became brighter at the slices 4. Please note that intensity enhancement induced by the distant from the epicenter in both caudal and rostral presence of the contrast agent in cerebrospinal fluid. directions within the normal sections of the injured SC. Although the patterns of contrast enhancements on days Page 6 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 1 and 3 were similar for this exam, the data indicated that Figure 4 compares the temporal profile of the relative the enhancement on day 1 peaked at the epicenter and intensity enhancement (REI, computed using Eq. 1 in the skewed asymmetrically towards the rostral direction, Appendix) within the regions of interest at the injury site unlike more symmetric, but wider, enhancement along on both days 1 and 3. The selected regions were identified the cord as visualized on day 3. In addition, cerebrospinal by circles coded with two different colors in Figure 3. The fluid (CSF) appeared hyperintense on the postcontrast variations in plots with different colors on the same day images. This was another important observation, which indicated REI has spatial dependence. The identical colors was also reported in the previous DCE-MRI study with a showed that the average REI profiles varied between day 1 rat model of SCI [5]. Combining the CSF enhancement in and day 3. According to Eq. 2, REI reflected the amount of rat with the current observation made in mouse, it is likely localized accumulation of the contrast agent, and thereby that rodents may have a unique system of CSF circulation indicated the severity of leakage of the agent through the different than human, since CSF of human does not typi- dysfunctional BSCBs therein. cally enhance after the IV delivery of the contrast agent used in this study. Figure 5 shows quantitative distribution of BSCB permea- -maps in the fig- bility for the data set in Figure 3. The K p-sc For the data analysis, images (precontrast and first post- ure were computed from three slices centered at the contrast) from a given slice location were displayed simul- epicenter using Eq. 4 in Appendix. The permeability esti- taneously. Image alignment was verified visually by mates were overlaid as a new layer with color coding on zooming in on anatomical landmarks. Two out of the ten the precontrast T1-weighted images, which served as the DCE-MRI scans exhibited spatial misalignment between background. In the maps, color coding towards red meant the axial postcontrast and the corresponding precontrast increased BSCB dysfunction. The red zones represented images. In these cases, the sedated mouse reacts to the the regions with greater leakage of the contrast agent and contrast agent, causing its body to move slightly. From the overlapped consistently with the lesion pathology. evaluation of the images, it was evident that the image motions in the misaligned data sets were translational in both cases, but a slight rotation was present in one of them. Neither of the motions was of deformation type. The two misaligned images were registered using a simple postprocessing technique with Matlab's "circshift" and "rotate" functions. A more complicated automated algo- rithm for image registration chould have been used for the registration. But this required complex implementation, which was beyond the scope of the study. sl Figure 5 Color coded ices centered K at the -map epi s computed center using the data set using Eq. 4 from the three in Figure 4 p-sc Color coded K -maps computed using Eq. 4 from p-sc the three slices centered at the epicenter using the data set in Figure 4. The backgrounds are the T1-weighted precontrast images (top and bottom rows). The color from Temporal patterns of averaged within two selected b (diam Figure 4 lue circles in ond) Figure 5) a relati nd on Day ve intensity regions 1 (squar of in enhter ancement ( ee ) and Day 3 st (red an RIE d ) yellow to red indicates linearly increasing compromise in Temporal patterns of relative intensity enhancement blood spinal cord barrier permeability. The areas with color (RIE) averaged within two selected regions of inter- towards red indicate regions with more vascular damage est (red and blue circles in Figure 5) and on Day 1 causing the barriers to become more permeable to blood (square) and Day 3 (diamond). constituents. Page 7 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 Figure 6 shows the precontrast PD and T2-weighted region (window 2) that led to the extravasation of the red images of the same injured SC. The corresponding T1- blood cells. The periphery of the damaged SC region (win- weighted image was reproduced from Figure 3 for a side- dow 3) contained cavities that colocalized with small ves- by-side comparison. Depending on the acquisition proto- sels or endothelial cells. col employed, images delineated the injury pathology with a specific intensity contrast. For this case, almost half Figure 9 shows LFB, NSE and CD34 stained slices showing of the GM appeared hypointense on the T2-weighted the injury and a normal cord section. At the injury, the image (Figure 6-b), but the remaining contra-lateral sec- LFB and NSE stains depicted diffuse demyelination and tion exhibited nearly homogeneous contrast, making it completely disorganized neurons and processes. The difficult to differentiate GM from WM. In addition, a thin CD34 stain from the same areas showed damaged vascu- strip of hyperintensity was observed along the circumfer- lature with leaky barrier. In the rostral segment, however, ence of the hypointense region on the T2-weighted image. LFB and NSE stains indicated normal myelination and In general, hypointensity on T2-weighted images in the intact neurons, and CD34 stain showed vasculature with early phase of SCI in mouse is associated with the accu- intact barrier. mulation of red blood cells, most likely escaping the vas- culature that was ruptured by the initial mechanical The area and volume of the regions with damaged BSCB impact. Hyperintensity is associated with vasogenic were measured semi-automatically as described above. edema describing the accumulation of fluid containing Figure 10 shows two histograms of K maps from a nor- p-sc plasma proteins into the extracellular spaces of the injured mal section and epicenter. The histogram from the nor- SC. The origin of these contrasts was determined by exam- mal section was used to determine a threshold for ining the corresponding histology slides. spatially segmenting the K maps to measure the areas in p-sc the other slices. The inter- and intra-variations for the Figures 7, 8 and 9 show histologic sections stained with threshold values estimated on days 1 and day 3 were -1 H&E, LFB, NSE or CD34. The slice in Figure 7 matches the within the range 0.03 ± 0.01 min . Figure 11 compares images in Figure 6, but the others in Figures 8 and 9 are NA measurements from K maps of 6 slices centered at p-sc from a different injured SC. The vasculature was depicted the lesion from all animals. These results indicated that by H&E staining in Figure 7 and better outlined in Figure four slices (in 4 mm space) around the epicenter suffered 8 by CD34 immunostaining. The microscopic features of the most compromise in BSCB, covering about 70% of the the neurovascular pathology revealed partially intact GM total cross section of the SC parenchyma on day 1. But, on at the ventral horns, significant damage at the dorsal SC day 3, the regions with damaged BSCB were reduced in with substantial loss of tissue matrix and also small cavi- size to about 55% for the two middle and 20% for the ties distributed throughout the cord, but generally more remaining two slices. Such behavior indicated that the prominent in WM than GM. These regions were evaluated spatial distribution of the areas with compromised BSCB more closely at higher magnification in three selected shrunk with time. This was consistent with the previous regions of interest – windows 1, 2 and 3. As expected, in reports, where the lesion size shrank with time in direc- the normal looking GM (window 1), the vessels were tions both across and along the cord, which was a unique observed to be intact but mostly ruptured in the damaged behavior seen in mice SCI [27,28]. This outcome was fur- ther supported by the normalized volume (NV) measure- ments that decreased from day 1 to 3, as shown in Figure 12. The NV data from all animals indicated that volume of BSCB breakdown occupied 48% of the total volume of the SC on day 1 as compared to 25% on day 3. The differ- ence between the two measured volumes was statistically significant. These results from day 1 and day 3 together implied that dynamic remodeling of the BSCB permeabil- ity took place as part of the ongoing neurovascular repair and recovery processes in injured cord. Figure 13 shows anatomical T2-weighted image of an injured SC and the computed maps (K , λ , λ and FA) p-sc || ⊥ from the same slice orientation. The green circle plotted A p Figure 6 rnior to the injection of atomical images of injur contrast y epicenter on post-injury day agent 3 on the K map represents a region of interest with Anatomical images of injury epicenter on post-injury p-sc increased BSCB permeability within the GM. The mean day 3 prior to the injection of contrast agent: a) PD image, b) T2-weighted image and c) T1-weighted image. values for the parameters K , λ , λ and FA residing p-sc || ⊥ within such selected regions were computed for each ani- Page 8 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 H Figure 7 &E stained histology images of injured SC on day 3 H&E stained histology images of injured SC on day 3: a) at lower magnification and b-d) at higher magnification within the selected square windows marked with numbers 1, 2 and 3 in panel a. The windows were selected based on the MRI- st nd observed pathology in Figure 6. The 1 window was selected in an intact GM region, the 2 window was selected in a signifi- rd cantly damaged region and the 3 window was selected in a region with edema. Black arrows denote normal vessels and capil- laries. Black arrow heads point to disrupted vasculature with damaged BSCB. White arrows point to vessels surrounded by cavities. Circle denotes clusters of extravasated red blood cells. mal. The results from all animals were further averaged The level of the one-to-one dependencies between the and compared in Table 1. The differences between the parameter K and the measurements λ , λ and FA are p-sc || ⊥ measurements on days 1 and 3 were found to be statisti- given in Table 2. From the values in the table, negative cally significant (P < 0.05). Normative λ and λ measure- correlations were found between K and λ and between || ⊥ p-sc || -3 ments from GM of normal mouse SC (λ = 1.25 × 10 K and FA. K and λ were also correlated, but posi- || p-sc p-sc ⊥ 2 -3 2 mm /s, λ = 0.47 × 10 mm /s and FA = 0.61) were tively. Combining these results with the histological find- reported earlier [21]. Comparing these results with the ings (Figures 7, 8 and 9) suggested that a strong corresponding ones in the table indicates significant dif- relationship exists between vascular integrity and the ferences, demonstrating the degree of sensitivity of these structural state of the neuronal tissue as assessed by the parameters to the neuropathology in injured cords. The diffusion measurements. These results were in line with data in the table also shows that the parameters K , λ the prior knowledge that more vascular damage is likely to p-sc || and FA decrease, but λ increases from day 1 to day 3. cause greater neuronal loss following SCI. Such trends are suggestive of vascular restoration, further disruption of axonal integrity, loss of anisotropy and Discussion increased demyelination within the injured SC as time Preclinical neuroimaging methods are required for evalu- progressed. ating the degree of initial mechanical damage and for Page 9 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 H Figure 8 istology images of an injured SC on day 3 Histology images of an injured SC on day 3: a) H&E-stained section with arrows denotes extravasated red blood cells. b- d) higher magnifications of adjacent section stained with the endothelial marker, CD34. The square windows marked with numbers 1, 2 and 3 were selected to show vasculature and BSCB integrity therein. b) Sparse, intact vasculature (arrows). c) Disrupted vasculature with diffusion of immunostaining due to damaged BSCB. d) Disrupted vessels and tissue vacuolization due to edema from the leaky BSCB (arrows). monitoring the subsequent secondary events following of the contrast agent in between [6]. In this model, imme- SCI in experimental research. With this purpose in mind, diately following the IV bolus injection, the transport of we and others have investigated a variety of MRI methods the contrast agent was considered unidirectional from to obtain anatomical, vascular and structural information plasma to the injured SC tissue. This feature allowed asso- from the injured SC [14,21,29-39]. However, to date, no ciating the barrier permeability with the transfer rate con- study has combined DCE-MRI and DTI protocols to stant from plasma-to-spinal cord. Such association was simultaneously evaluate the microvascular and micro- also proved to be critical for simplifying the computa- structural changes in injured SC. The current study is the tional analysis whereby requiring only the early part of the first to jointly apply these techniques, along with anatom- relative intensity enhancement (Figure 4). This ultimately ical imaging, to evaluate SCI in mouse. An additional led to the estimation of localized changes in the barrier achievement that distinguishes this study from the previ- dysfunction throughout the SC (Figure 5). ous works is the computational algorithm implemented for quantitatively mapping the spatial distribution of Our MRI-based data, as in Figures 3, 4 and 5, and histo- BSCB permeability in injured SC. This algorithm was logical analysis, as in Figures 7, 8 and 9, demonstrated the derived from a pharmacokinetic model (Figure 14) that dynamic remodeling of the BSCB as part of the ongoing was originally developed to represent plasma and injured repair and recovery processes in the injured SC tissue. SC by two compartments, and to determine the exchange Using postmortem analysis in mouse, Whetstone et al. [2] Page 10 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 LFB, Figure 9 NSE and CD34 stained contiguous slices at the injury epicenter and at a normal section rostral to the epicenter LFB, NSE and CD34 stained contiguous slices at the injury epicenter and at a normal section rostral to the epi- center. In the rostral segment (top row), LFB (a) and NSE (b) stains show normal myelination and intact neurons, and CD34 stain (c) shows intact vasculature. At the epicenter (bottom row), LFB (d) and NSE (e) stains show diffuse demyelination and damaged neurons, and the corresponding CD34 stain (f) shows damaged vasculature with leaky BSCB. reported that SCI results in a biphasic, temporal pattern of barrier leakage to tracer – luciferase. In their study, the barrier leakage was shown to extend beyond the epicenter into segments that were within 6 mm rostral and caudal to the epicenter. The leakage was seen to be mostly pro- nounced within the first 35 min after the injury followed by a gradual decline within the first 24 hr. A second peak and ep Figure 10 Histograms icenter of th of K -e injury (right) maps from a normal caudal section (left) p-sc Histograms of K -maps from a normal caudal sec- No ce Figure 11 nter rmal ed iz at the e ed area (mean ± std) picenter from measurements from 6 sl all animals on day 1 and day 3 ices p-sc tion (left) and epicenter of the injury (right). Solid lines Normalized area (mean ± std) measurements from 6 -1 are fit to a Gaussian function. A threshold of 0.025 min was slices centered at the epicenter from all animals on determined from the histogram on the left. This value was day 1 and day 3. * denotes statistically significant difference (P < 0.05). used for segmenting the regions with compromised BSCB permeability in the maps from the other slices. Page 11 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 From the application point of view, providing informa- tion about the barrier function should be useful in at least three circumstances. One situation is in characterizing the SCI model where the resulting barrier leakage in a partic- ular SCI model may be another parameter to be defined, and referred back when needed to check against the con- sistency of the newly induced injuries. This would enable the confirmation of whether later injuries have similar properties in terms of the vascular response. Traditional histological analysis was employed to understand the sen- sitivity and specificity of the barrier damage to graded mechanical perturbations [3]. Our approach offers an effi- cient alternative and may become a preferred choice for such purpose. In the second case, quantitatively evaluat- ing the severity of the barrier damage as soon as possible after the injury may provide early indications of the signif- Normali m Figure 12 als on da zed vo y 1 an lume (me d day 3 an ± std) measurements from all ani- icance and level of the expected secondary processes that Normalized volume (mean ± std) measurements are likely to increase neuronal loss. Such capability can from all animals on day 1 and day 3. * denotes statisti- have significant prognostic value. Figure 13 and compara- cally significant difference (P < 0.05). tive analysis in Tables 1 and 2 clearly provided evidence that correlations exist between DCE-MRI and DTI based of abnormal barrier permeability was reported at 3 days measurements. This has important implications in prac- after the injury, which significantly exceeded that tice, especially when the examined SCI produces poor observed at 24 hr. Regarding the barrier permeability to quality of DTI acquisition or the analysis of the resulting the contrast agent, we obtained comparable results that data is not feasible. In such challenging situations, includ- showed stronger intensity enhancement on day 3 than ing the DCE-MRI protocol in the scan may supply infor- day 1. This agreement is encouraging and warrants further mation on the condition of the vasculature, which may exploration, such as investigation of the loss of blood ves- indirectly help to determine the level of neuronal damage sels and revascularization in injured mouse SC using in the underlying SC tissue by using the close associations DCE-MRI in conjunction with MR angiography modality demonstrated in Table 2. Lastly, monitoring the barrier [5,32,40]. function may have important implications in predicting the efficacy of targeted drug or gene delivery through the Requiring only the initial portion of the intensity permeable barriers when potential therapeutic interven- enhancement to assess barrier function offers benefits in tions are considered. While this aspect is foreseeably terms of reducing the imaging time substantially. In the achievable if the drug or transfection vector has molecular current study, we acquired T1-weighted images using a weight comparable to that of the contrast agent used in spin-echo sequence which required relatively long acqui- this study, it remains to be seen if the approach would be sition time. This is one limitation of the approach. Higher viable in cases when the drug or gene has significantly temporal resolution can be achieved between the precon- larger size. trast and postconrast acquisitions by employing faster imaging sequences. Such ability should further improve Conclusion the time and accuracy of estimating the barrier permeabil- This study has demonstrated the potential of DCE-MRI ity. For this purpose, echo-planar imaging appears prom- method to assess the BSCB dysfunction in injured spinal ising [41]. cord noninvasively. The method involved acquiring only Table 1: BSCB permeability and microstructural measurements on day 1 and day 3 (mean ± standard deviation). P values for the measurements between day 1 and 3 are given in the last row. -1 -3 2 -3 2 K (min ) λ (×10 mm /s) λ⊥ (×10 mm /s) FA p-sc || Day 1 0.123 ± 0.008 0.788 ± 0.022 0.579 ± 0.016 0.183 ± 0.025 Day 3 0.076 ± 0.006 0.694 ± 0.026 0.604 ± 0.015 0.097 ± 0.021 P 0.002 0.001 0.002 0.006 Page 12 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 Postinj Figure 13 ury day 1 Postinjury day 1: a) anatomical T2-weighted image of an injured SC, b) K map, c) λ map, d) λ map and e) FA map. p-sc || ⊥ Green circle denotes the region of interest chosen within the GM to compute the mean values for the parameters K , λ , p-sc || λ and FA from the corresponding maps. two T1-weighted images; precontrast and postcontrast, acquired with a minimum time delay following the injec- Two-c distrib cord tissue Figure 14 u otion mpartment p of the con harmocokine trast agent in plasma and tic model representing the injured spinal Two-compartment pharmocokinetic model repre- tion of contrast agent. This simple data acquisition strat- senting the distribution of the contrast agent in egy provides sufficient temporal information on the plasma and injured spinal cord tissue. The subscripts p contrast-enhancement that was necessary to map the spa- and sc denote plasma and spinal cord. [C] (t) and [C] (t) are p sc tial distribution of BSCB permeability. The BSCB dysfunc- time-dependent concentrations in mmol/L in plasma and spi- tion correlated strongly with the degree of axonal loss and nal cord compartments, respectively. The parameters k p-sc demyelination. The relationship between the vascular and k denote transfer rate constants in l/min for forward sc-p damage and neuronal loss in injured SC has been well- transfer of the contrast agent from plasma-to-spinal cord and established previously, but by using histological tech- reverse transfer from spinal cord-to-plasma compartments, niques. Reconfirming this association using DCE-MRI respectively. and DTI data acquired from living animals has important implications in translational research. Preclinical efforts are currently focused on developing BSCB permeable Competing interests drugs for improving neurovascular function by repairing The authors declare that they have no competing interests. vascular network, delaying neurodegenerative processes or promoting neuronal recovery. Research efforts are also Authors' contributions underway for understanding the role of specific genes in IT performed the statistical analysis. PCC postprocessed vascular reorganization, neuronal repair and recovery the acquired DCE-MRI data. HES contributed with the from an injury on experimental test systems involving dif- histology and staining. MMD evaluated the pathology. ferent strains of "transgenic" or gene knock-out mice. On MB was responsible for the conception and design of the the basis of multiple neuroimaging methodologies devel- experiments as well as the analysis and interpretation of oped in this study, scanning the same animal model with the data collectively. All authors read and approved the anatomical, DCE-MRI and DTI protocols provides meas- final manuscript. urable parameters that can serve as sensitive and specific in vivo neurovascular biomarkers for comprehensively Appendix evaluating the pathological state of the injured SC, may The presence of paramagnetic contrast agent at concentra- offer a prognostic value for functional recovery from SCI tion [C] (in mmol/L) in neuronal tissue of an injured SC and potentially serve as a monitoring tool for evaluating was detected by relative intensity enhancement (RIE) on the efficacy of a treatment efficacy. T1-weighted magnetic resonance images. Postcontrast RIE at time t was calculated for each pixel at a spatial position (x, y) and a slice location z using the formula Table 2: Pearson's correlation coefficient results between the BSCB permeability and microstructural measurements on day 1 and day 3. It (,x,y,z)−= It ( 0,x,y,z) RIE(, t x, y, z) = xy,( ∈ Mask z) ). It(, =0 x,y,z) ρ(K , λ ) ρ(K ,λ ) ρ(K , FA) p-sc || p-sc ⊥ p-sc (1) Day 1 -0.87 0.64 -0.77 Here, I(t = 0, x, y, z) and I(t, x, y, z) represent intensities in th the precontrast and t postcontrast images of a given slice Day 3 -0.92 0.97 -0.70 z, respectively. The mask Mask(z) was obtained by using Page 13 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 manual segmentation applied on the postcontrast images where the SC was delineated best. The mask covered the 1 dRIE(, t x,y,z) kx (,y,z) = xy,( ∈ Mask z). ps − c cord but excluded cerebrospinal fluid, which also exhib- Tr [C] () t =0 dt 10 1 p t0 = ited postcontrast enhancement. The total cross-sectional (3) area of the SC at that particular slice was measured by the area covered by the mask. Considering that the factor T r has negligible depend- 10 1 ence on spatial location, we reorganize Eq. [3] to obtain a When a spin-echo sequence is used for DCE-MRI acquisi- new parameter that is linearly proportional to k : p-sc tion and a small dose of contrast agent is delivered as a bolus, as in this study, the SC concentration [C] and RIE dRIE(, t x,y,z) sc K (,x yz , )== T r [C] (t 0)k (,x yz , )= xxy,( ∈ Mask z). ps−− c 10 1 p psc dt are related by the formula [4] t0 = (4) RIE(, t x,y,z) This equation indicates that the initial slope of REI pro- [C] (, t x, y, z)=∈ xy,( Mask z). (2) sc Tr 10 1 vides a measurable quantity representing the BSCB per- meability at a spatial location (x, y, z). Specifically, Here, T denotes relaxation time in second (s) before numerical computation of Eq. [4] involves discrete deriv- administering the contrast agent and r is relaxivity intro- duced by the contrast agent to that of the underlying SC RIE(, t+− ΔT x, y, z) RIE(t, x, y, z) ative operation for t = 0 and ΔT -1 -1 ΔT tissue and expressed in mM s . For mice scanned at 9.4 T, T was reported to be 1730 ms for the gray and 1690 = 10 min – the temporal resolution in our DCE-MRI ms for the white matters of the SC [21], but r associated acquisitions. By noticing that RIE (t = 0, x, y, z) = 0, the with the contrast agent for these tissue types has yet to be derivative operation can be simplified further. In this determined. RIE(, t +ΔT x, y, z) study, was calculated for each x, y ∈ ΔT Equation [2] plays an important role when constructing a Mask(z) and the results were put together to form a 2-D pharmacokinetic model to theoretically represent the quantitative K map delineating the regions of the p-sc uptake of contrast agent in an injured SC [4]. Construc- injured SC with compromised BSCB permeability for each tion of a proper pharmacokinetic model requires at least slice. two compartments – one for the plasma and one for the cord, as shown in Figure 1[6]. In this model, two rate con- Acknowledgements This work was funded in part by NIH grants NS052610 and NS054019. The -1 stants (min ) quantitatively describe the transport of con- authors thank Dr. Fan Yang for helping with animal surgeries and care. trast agent from plasma-to-SC and SC-to-plasma. Following an IV bolus injection, the contrast agent first References 1. Tator CH, Koyanagi I: Vascular mechanisms in the pathophysi- leaks from plasma to the injured SC tissue through the ology of human spinal cord injury. J Neurosurg 1997, 86:483-492. compromised BSCBs. This thereby allows the representa- 2. Whetstone WD, Hsu JY, Eisenberg M, Werb Z, Noble-Haeusslein LJ: Blood-spinal cord barrier after spinal cord injury: relation to tion of the barrier permeability with the rate constant k p-sc revascularization and wound healing. J Neurosci Res 2003, governing the unidirectional transfer of the contrast agent 74:227-239. 3. Maikos JT, Shreiber DI: Immediate damage to the blood-spinal from plasma-to-SC. This rate constant is then estimated cord barrier due to mechanical trauma. J Neurotrauma 2007, using the formula 24:492-507. 4. Bilgen M, Narayana PA: A pharmacokinetic model for quantita- d[C] (,tx,y,z) 1 sc tive evaluation of spinal cord injury with dynamic contrast- kx (,y,z) = [4,6]. In this ps − c [C] () t =0 dt p enhanced magnetic resonance imaging. Magn Reson Med 2001, t0 = 46:1099-1106. expression, [C] (t = 0) denotes the initial concentration of 5. Bilgen M, Abbe R, Narayana PA: Dynamic contrast-enhanced MRI of experimental spinal cord injury: in vivo serial studies. the contrast agent in mouse plasma and computed from Magn Reson Med 2001, 45:614-622. 6. Bilgen M, Dogan B, Narayana P: In vivo assessment of blood-spi- the injected dose and the total plasma volume. The oper- nal cord barrier permeability: serial dynamic contrast d(.) enhanced MRI of spinal cord injury. Magn Reson Imaging 2002, ator denotes time derivative evaluated immedi- dt 20:337. t0 = 7. Guertin PA: Paraplegic mice are leading to new advances in ately after the delivery of the contrast agent. Using Eq. [2], spinal cord injury research. Spinal Cord 2005, 43:459-461. 8. Rosenzweig ES, McDonald JW: Rodent models for treatment of we can write spinal cord injury: research trends and progress toward use- ful repair. Curr Opin Neurol 2004, 17:121-131. 9. Steward O, Schauwecker PE, Guth L, Zhang Z, Fujiki M, Inman D, Wrathall J, Kempermann G, Gage FH, Saatman KE, et al.: Genetic approaches to neurotrauma research: opportunities and potential pitfalls of murine models. Exp Neurol 1999, 157:19-42. Page 14 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 10. Kigerl KA, McGaughy VM, Popovich PG: Comparative analysis of 33. Gaviria M, Bonny JM, Haton H, Jean B, Teigell M, Renou JP, Privat A: lesion development and intraspinal inflammation in four Time course of acute phase in mouse spinal cord injury mon- strains of mice following spinal contusion injury. J Comp Neurol itored by ex vivo quantitative MRI. Neurobiol Dis 2006, 2006, 494:578-594. 22:694-701. 11. Pannu R, Christie DK, Barbosa E, Singh I, Singh AK: Post-trauma 34. Kim JH, Trinkaus K, Ozcan A, Budde MD, Song SK: Postmortem Lipitor treatment prevents endothelial dysfunction, facili- delay does not change regional diffusion anisotropy charac- tates neuroprotection, and promotes locomotor recovery teristics in mouse spinal cord white matter. NMR Biomed 2007, following spinal cord injury. J Neurochem 2007, 101:182-200. 20:352-359. 12. Pearse DD, Chatzipanteli K, Marcillo AE, Bunge MB, Dietrich WD: 35. Huang ZL, He YY, Bilgen M: In vivo tracing of corticospinal tract Comparison of iNOS inhibition by antisense and pharmaco- in mouse using manganese-enhanced contrast. J Neurol Sci logical inhibitors after spinal cord injury. J Neuropathol Exp Neu- [Turk] 2007, 24:38-44. rol 2003, 62:1096-1107. 36. Loy DN, Kim JH, Xie M, Schmidt RE, Trinkaus K, Song SK: Diffusion 13. Elshafiey I, Bilgen M, He R, Narayana PA: In vivo diffusion tensor tensor imaging predicts hyperacute spinal cord injury sever- imaging of rat spinal cord at 7 T. Magn Reson Imaging 2002, ity. J Neurotrauma 2007, 24:979-990. 20:243-247. 37. Bilgen M: Magnetic resonance microscopy of spinal cord injury 14. Kim JH, Loy DN, Liang HF, Trinkaus K, Schmidt RE, Song SK: Nonin- in mouse using a miniaturized implantable RF coil. J Neurosci vasive diffusion tensor imaging of evolving white matter Methods 2007, 159:93-97. pathology in a mouse model of acute spinal cord injury. Magn 38. Bilgen M, Al-Hafez B, Alrefae T, He YY, Smirnova IV, Aldur MM, Fes- Reson Med 2007, 58:253-260. toff BW: Longitudinal magnetic resonance imaging of spinal 15. Aimone JB, Leasure JL, Perreau VM, Thallmair M: Spatial and tem- cord injury in mouse: changes in signal patterns associated poral gene expression profiling of the contused rat spinal with the inflammatory response. Magn Reson Imaging 2007, cord. Exp Neurol 2004, 189:204-221. 25:657-664. 16. Ousman SS, David S: MIP-1alpha, MCP-1, GM-CSF, and TNF- 39. Clark CA, Werring DJ, Miller DH: Diffusion imaging of the spinal alpha control the immune cell response that mediates rapid cord in vivo: estimation of the principal diffusivities and phagocytosis of myelin from the adult mouse spinal cord. J application to multiple sclerosis. Magn Reson Med 2000, Neurosci 2001, 21:4649-4656. 43:133-138. 17. Parker D: Pharmacological approaches to functional recovery 40. Bilgen M, Al-Hafez B, He YY, Brooks WM: Magnetic resonance after spinal injury. Curr Drug Targets CNS Neurol Disord 2005, angiography of rat spinal cord at 9.4 T: a feasibility study. 4:195-210. Magn Reson Med 2005, 53:1459-1461. 18. Pearse DD, Bunge MB: Designing cell- and gene-based regener- 41. Callot V, Duhamel G, Cozzone PJ: In vivo mouse spinal cord ation strategies to repair the injured spinal cord. J Neuro- imaging using echo-planar imaging at 11.75 T. Magma 2007, trauma 2006, 23:437-452. 20:169-173. 19. Bilgen M: A new device for experimental modeling of central nervous system injuries. Neurorehabil Neural Repair 2005, Pre-publication history 19:219-226. 20. Bilgen M: Simple, low-cost multipurpose RF coil for MR The pre-publication history for this paper can be accessed microscopy at 9.4 T. Magn Reson Med 2004, 52:937-940. here: 21. Bilgen M, Al-Hafez B, Berman NE, Festoff BW: Magnetic resonance imaging of mouse spinal cord. Magn Reson Med 2005, 54:1226-1231. http://www.biomedcentral.com/1471-2342/9/10/prepub 22. Bilgen M: Imaging corticospinal tract connectivity in injured rat spinal cord using manganese-enhanced MRI. BMC Medical Imaging 2006, 6:15. 23. Bilgen M, Elshafiey I, Narayana PA: Mohr diagram representation of anisotropic diffusion tensor in MRI. Magn Reson Med 2002, 47:823-827. 24. Bilgen M, Narayana PA: Mohr diagram interpretation of aniso- tropic diffusion indices in MRI. Magn Reson Imaging 2003, 21:567-572. 25. Wei H, Desouki MM, Lin S, Xiao D, Franklin RB, Feng P: Differential expression of metallothioneins (MTs) 1, 2, and 3 in response to zinc treatment in human prostate normal and malignant cells and tissues. Mol Cancer 2008, 7:7. 26. Bilgen M, Rumboldt Z: Neuronal and vascular biomarkers in Syringomyelia: investigations using longitudinal MRI. Biomar- kers in Medicine 2008, 2:113-124. 27. Inman D, Guth L, Steward O: Genetic influences on secondary degeneration and wound healing following spinal cord injury in various strains of mice. J Comp Neurol 2002, 451:225-235. 28. Inman DM, Steward O: Physical size does not determine the unique histopathological response seen in the injured mouse Publish with Bio Med Central and every spinal cord. J Neurotrauma 2003, 20:33-42. scientist can read your work free of charge 29. Wamil AW, Wamil BD, Hellerqvist CG: CM101-mediated recov- ery of walking ability in adult mice paralyzed by spinal cord "BioMed Central will be the most significant development for injury. Proc Natl Acad Sci USA 1998, 95:13188-13193. disseminating the results of biomedical researc h in our lifetime." 30. Bonny JM, Gaviria M, Donnat JP, Jean B, Privat A, Renou JP: Nuclear Sir Paul Nurse, Cancer Research UK magnetic resonance microimaging of mouse spinal cord in vivo. Neurobiol Dis 2004, 15:474-482. Your research papers will be: 31. Stieltjes B, Klussmann S, Bock M, Umathum R, Mangalathu J, Letellier available free of charge to the entire biomedical community E, Rittgen W, Edler L, Krammer PH, Kauczor HU, et al.: Manganese- enhanced magnetic resonance imaging for in vivo assess- peer reviewed and published immediately upon acceptance ment of damage and functional improvement following spi- cited in PubMed and archived on PubMed Central nal cord injury in mice. Magn Reson Med 2006, 55:1124-1131. 32. Bilgen M, Al-Hafez B: Comparison of spinal vasculature in yours — you keep the copyright mouse and rat: investigations using MR angiography. Neuro- BioMedcentral anatomy 2006, 5:12-18. Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 15 of 15 (page number not for citation purposes) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Medical Imaging Springer Journals

Evaluating regional blood spinal cord barrier dysfunction following spinal cord injury using longitudinal dynamic contrast-enhanced MRI

Download PDF
15 pages

Loading next page...
 
/lp/springer-journals/evaluating-regional-blood-spinal-cord-barrier-dysfunction-following-Mp76KKGXDE

References (41)

Publisher
Springer Journals
Copyright
Copyright © 2009 by Tatar et al; licensee BioMed Central Ltd.
Subject
Medicine & Public Health; Imaging / Radiology
eISSN
1471-2342
DOI
10.1186/1471-2342-9-10
pmid
19519898
Publisher site
See Article on Publisher Site

Abstract

Background: In vivo preclinical imaging of spinal cord injury (SCI) in rodent models provides clinically relevant information in translational research. This paper uses multimodal magnetic resonance imaging (MRI) to investigate neurovascular pathology and changes in blood spinal cord barrier (BSCB) permeability following SCI in a mouse model of SCI. Methods: C57BL/6 female mice (n = 5) were subjected to contusive injury at the thoracic T11 level and scanned on post injury days 1 and 3 using anatomical, dynamic contrast-enhanced (DCE-MRI) and diffusion tensor imaging (DTI). The injured cords were evaluated postmortem with histopathological stains specific to neurovascular changes. A computational model was implemented to map local changes in barrier function from the contrast enhancement. The area and volume of spinal cord tissue with dysfunctional barrier were determined using semi-automatic segmentation. Results: Quantitative maps derived from the acquired DCE-MRI data depicted the degree of BSCB permeability variations in injured spinal cords. At the injury sites, the damaged barriers occupied about 70% of the total cross section and 48% of the total volume on day 1, but the corresponding measurements were reduced to 55% and 25%, respectively on day 3. These changes implied spatio-temporal remodeling of microvasculature and its architecture in injured SC. Diffusion computations included longitudinal and transverse diffusivities and fractional anisotropy index. Comparison of permeability and diffusion measurements indicated regions of injured cords with dysfunctional barriers had structural changes in the form of greater axonal loss and demyelination, as supported by histopathologic assessments. Conclusion: The results from this study collectively demonstrated the feasibility of quantitatively mapping regional BSCB dysfunction in injured cord in mouse and obtaining complementary information about its structural integrity using in vivo DCE-MRI and DTI protocols. This capability is expected to play an important role in characterizing the neurovascular changes and reorganization following SCI in longitudinal preclinical experiments, but with potential clinical implications. Page 1 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 data collection, post processing and spatial representation Background Damage to blood-spinal cord barrier (BSCB) occurs as a of the BSCB permeability variations in injured SC. consequence of mechanical insult to spinal cord (SC) [1]. The damaged barrier becomes permeable to blood con- Histological studies have demonstrated that, in injured stituents, inflammatory cells and other large molecules, SC, the areas with damaged vasculature overlap with the which collectively activate a cascade of secondary proc- areas of intense neuronal loss [11,12]. Noninvasive diffu- esses harmful to the underlying tissue. With time, vascular sion tensor imaging (DTI) provides information about the bed of the injured SC goes through repair and remodeling neuronal integrity in underlying SC tissue [13,14]. Com- [2]. Providing nutrition and preventing extravasation of bining these, it is likely that in vivo BSCB permeability destructive biochemical compounds in blood protect the and DTI-based measurements, both obtained from the remaining neurons and maintain the existing substrate same region of the injured SC, can provide relevant infor- from further degeneration. At the same time, knowledge mation about the viability of underlying neurovascula- of barrier properties and its status is essential if potential ture, which has so far only been available by performing intravascular drugs with capabilities to improve neurovas- ex vivo histological analysis. Therefore, the third goal of cular protection and to promote repair and recovery will this study is to compare the BSCB permeability and DTI be administered. Such treatment is possible provided that measurements and investigate the nature of their associa- the drugs can pass through the open barriers to reach the tion while assessing the neurovascular response in SCI. destined regions in the SC parenchyma. Therefore, a com- plete understanding of the vascular response to spinal Studies based on postmortem tissue analysis suggest that cord injury (SCI) is required for developing intervention the first 72 hours following SCI are the most critical, offer- strategies aimed at rapidly restoring the barrier integrity as ing a window of opportunity for potential treatments [15- well as blood supply to the ischemic areas of the trauma- 18]. In order to develop effective therapeutic strategies in tized cord. experimental studies, it is important to employ a SCI model that sensitively responds to the treatment at the So far, BSCB permeability changes in injured SC have acute phase of the injury and is also capable of producing been evaluated using a range of tracers and laborious post- measurable vascular and neuropathological changes mortem tissue analysis [3]. Alternatively, in vivo dynamic within the first 72 hr time frame. Therefore, this study contrast-enhanced MRI (DCE-MRI) was proposed to non- focused on the acute phase (postinjury days 1 and 3) of invasively visualize and quantify the changes in BSCB per- the injury with the above pathological properties and meability in a relatively more efficient way [4-6]. The investigated the injured cords using multimodal neuroim- initial DCE-MRI studies were performed using a rat ani- aging (anatomical, DCE-MRI and DTI) noninvasively. In mal model of SCI. Lately, the availability of diverse strains the following, we first describe our imaging protocols, and transgenic varieties has made the mouse a more and then give details of our implementation and basis of attractive model [7-10]. But, to date, DCE-MRI studies on our data processing algorithm and strategies. Next, using mouse are lacking. The first goal of this paper is to dem- these developments, we present results from neuropatho- onstrate the feasibility of performing DCE-MRI on mouse logical evaluations and generate reliable BSCB permeabil- with SCI. ity maps to show the spatiotemporal course of the alterations in vascular permeability within the SC lesion The contrast enhanced data in previous rat studies were and its surroundings. We characterize the area and vol- acquired dynamically over an extended period of time ume of the regions with dysfunctional BSCB. We also test and processed using a pharmacokinetic model with com- the BSCB permeability estimates against the DTI measure- plicated numerical computation routine. The end results ments from the corresponding regions to establish the from the computation included quantitative measure- level of association as determined from statistical analysis. ments that represented the overall exchange of contrast agent between plasma and lesion. Such representation Methods was useful, but yielded limited information, since in prac- All experiments were carried out with twelve-week old tice, rather than global evaluation, more detailed local female C57BL/6 mice (n = 5) in accordance with a proto- variations in BSCB permeability was sought after. In this col approved by the Institutional Animal Care and Use previous approach, the DCE-MRI data acquisition time Committee. All of the mice were subjected to SCI and par- was long since it required covering both the wash-in and ticipated in the MRI scans using DCE-MRI and DTI proto- wash-out phases of the contrast agent. Another weakness cols on postinjury days 1 and 3. was that the parameters were estimated using a time-inef- ficient computation algorithm. Thus, the second goal of Procedures for surgeries and spinal cord injury this study is to provide an improved method in terms of All surgeries were performed in sterile conditions. The mouse was initially anesthetized by a spontaneous inha- Page 2 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 lation of 4% isoflurane in an induction chamber and then TE = 2500/12 ms, field-of-view (FOV) = 26 × 8 mm , moved to a surgery mat. The anesthesia level was reduced matrix = 256 × 128, in-plane pixel resolution = 100 × 63 to 2% isoflurane delivered in a mixture of 40% oxygen μm, slice thickness = 0.5 mm, number of excitations and 58% air through a nose mask. Further adjustments in (NEX) = 2. The corresponding parameters for the axial small increments were made on the percentage of the iso- images were TR/TE = 2500/12 ms, FOV = 12 × 8 mm , flurane level during the surgery. For intravascular delivery matrix = 128 × 128, in-plane pixel resolution = 94 × 63 of the paramagnetic contrast agent with MW = 938 Da μm, slice thickness = 1 mm, number of slices = 14 and (Magnevist, Berlex Imaging, Wayne, NJ) in DCE-MRI NEX = 2. These data constituted the proton density (PD) studies, animals underwent additional jugular vein cathe- weighted images. terization with PE-10 tubing [5]. The catheter ran subcu- taneously and exited the skin at the back between Then, for microstructural imaging, diffusion weighted shoulder blades, then was kept folded behind the animal. images were acquired using diffusion gradient strength = The catheter was flushed with heparin daily to prevent 80 mT/m, width (δ) = 6.5 ms and separation (Δ) = 11 ms clotting. The surgical procedures for inducing SCI to produce a b-value of 534 s/mm . The imaging parame- involved a midline incision posterior from the thoracic ters for these scans were TR/TE = 2000/26 ms, FOV = 12 × levels T10 to T12, followed by dissection of the bilateral 8 mm , image matrix = 128 × 128, in-plane pixel resolu- vertebral muscles to expose the dorsal laminae and tion = 94 × 63 μm, slice thickness = 1 mm and NEX = 2 spinous processes. Laminectomy was performed at the [20-22]. Baseline data acquired in the absence of the dif- T11 level to expose the SC by exercising special care not to fusion weighting constituted the T2-weighted images. damage the dura mater. The spinous processes at T10 and T12 adjacent to the laminectomy were stabilized using Next, for evaluating microvascular response and imaging two hemostatic forceps. A 1-mm diameter injury bit, that BSCB permeability to the contrast agent in the injured ani- was attached to a generic central nervous system injury mal, a DCE-MRI protocol was applied [5]. The contrast device described earlier [19], was positioned perpendicu- agent was delivered as a bolus (< 5 s) at a dose of 0.1 lar to the dorsal surface of the SC. The device consists of mmol/kg while the animal was still in the scanner. To electromechanical components – a linear motor con- detect the contrast enhancement, T1-weighted axial nected to a controller. The controller communicated with images were acquired precontrast and repetitively for up a personal computer through a software program devel- to 2 hr postcontrast using the same parameters as the PD oped in our laboratory to input biomechanical parameter weighted images but with TR = 1000 ms and NEX = 4. The values for inducing a contusion-type SCI. The injury temporal resolution between these acquisitions was 10 parameters used in these experiments were: impact veloc- min. ity of 0.75 m/s; surface displacement depth of 0.5 mm; and compression duration of 85 ms. After the injury, the Quantitative MRI data analysis overlying muscle layers were sutured and skin was closed The MRI data were acquired and visualized using the scan- tightly. Then, the injured mouse was left to recover in a ner's control software VNMRJ (Varian, Palo Alto, CA). heated cage and received postoperative care. DCE-MRI Analysis MRI scans The DCE-MRI data were analyzed off-line using custom- Each injured mouse was scanned on days 1 and 3 using a written software in Matlab (The Mathworks, Inc., Natick, 9.4 T horizontal INOVA Varian system (Varian, Palo Alto, MA). Axial images acquired before and after the contrast CA) and an inductively coupled surface coil [20]. The scan enhancement in a given scanning session were loaded was performed when the mouse was under a general into computer and processed interactively using a graphi- anesthesia which was delivered as a mixture of 2% isoflu- cal user interface. The details of the performed numerical rane, 40% oxygen and 58% air through a nose mask. Vital analysis are described in Appendix. The results are 2-D signs (respiration, heart rate and body temperature) of the maps that quantitatively describe the BSCB permeability anesthetized animal were monitored using a MRI-com- (denoted by the parameter K ) throughout the injured p-sc patible monitoring and gating system (Model 1025, SA SC. Instruments Inc., Stony Brook, NY). Respiratory-gated acquisition was used to increase the image quality by min- Area and volume of dysfunctional BSCB in injured SC imizing breathing related image artifacts. By analyzing the spatial distribution of the permeability changes on K maps, we measured the area and volume p-sc High-resolution anatomical images on all animals were of the injured SC tissue with damaged BSCB. To perform first acquired in sagittal and axial views using a spin-echo the area measurements, a threshold value was required to sequence in multislice and interleaved fashion. The scan differentiate the elevated K values from the background p-sc parameters for the sagittal or horizontal images were TR/ noise. The value of this threshold was determined while Page 3 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 postprocessing the data. To accomplish this task, an algo- DTI analysis rithm was implemented using the theory of signal detec- Diffusion-weighted images were processed and the ele- tion in the presence of additive noise. The procedure ments of diffusion tensor were estimated for each image involved first computing histograms of K maps from voxel using the scanner software VNMRJ (Varian Inc., Palo p-sc two normal sections of the injured cord, but within the Alto, CA). The diffusion tensor represents the statistical masks of the corresponding slices at 4 mm caudal and ros- distribution of microscopic motion of water molecules, tral to the injury epicenter. Because BSCB in these regions and its three eigenvalues (λ , λ and λ ) characterize the 1 2 3 were intact, the contrast agent was retained within the SC principal water diffusivities along its three orthogonal vasculature and no detectable leakage took place. These eigenvectors [21,23,24]. To be consistent with the nomen- histograms depicted K estimates that were normally dis- clature used in previous reports, we denoted λ = λ as an p-sc 1 || tributed around zero, which could be fitted to zero mean expression of longitudinal diffusivity of water molecules Gaussian profiles (please see the results section). The along the axonal fibers, and λ = (λ +λ )/2 as an expres- ⊥ 2 3 larger standard deviation σ from the two Gaussian fits sion of transverse diffusivity for water moving perpendic- (from a caudal and a rostral slice) was used to set a thresh- ular to the direction of neuronal fibers in SC. The at +2σ . This selection was based on the old value for K fractional anisotropy (FA) index was computed algebrai- p-sc property that the Gaussian distribution between -2σ and cally by combining the eigenvalues. FA index is a rotation- +2σ equals to nearly 95% of the total area beneath the ally invariant scalar and quantitatively characterizes the curve. This meant that nearly 97.5% of the pixels in the K degree of anisotropy in the diffusion properties of the p- map from a normal SC were below the threshold and underlying tissue within the voxel. The parameters λ , λ sc 1 ⊥ therefore considered to have intact BSCB. The same and FA were estimated for each voxel throughout the slice threshold value was employed for analyzing the K maps at the epicenter and the corresponding maps were gener- p-sc produced for all the remaining slices. The histograms ated in 2-D. Similar to computing mean K values p-sc from slices near the injury epicenter were observed to be described above, λ , λ and FA measurements from the || ⊥ shifted towards higher K values, as expected, reflecting areas with compromised BSCB at the epicenter were aver- p-sc the presence of compromised BSCB. These histograms aged and the resulting mean values for each animal were had wider spread but still followed the profile of Gaussian recorded in a database along with the corresponding . Group mean and standard deviation of these distribution. mean K p-sc measurements were again computed for days 1 and 3 sep- The number of pixels whose intensity values on the K arately. p-sc maps remained above the threshold was counted to deter- Postmortem tissue analysis mine the area of the SC with compromised BSCB in a given slice. This number was further divided by the total Neurovascular histopathologies of selected injured SCs cross sectional area of the cord in that slice to obtain a were examined postmortem following MRI scan on day 3. measure of normalized area (NA). The slice with largest Each mouse was euthanized by intracardiac perfusion pixel count was considered as representing the injury epi- with 50 mL of phosphate buffered saline (PBS) solution, center. The volume of the compromised BSCB was deter- followed by 50 mL of 4% formaldehyde PBS solution that mined by summing the areas (prior to normalization) in were delivered through a 23-gauge needle connected to a six neighboring slices covering the epicenter. Total cord perfusion pump. The SC was excised and fixed in 4% for- volume was similarly calculated from the total cord area maldehyde. Segments from the injury epicenter or normal in each slice. The total volume with compromised BSCB levels were embedded in paraffin and cut serially in 10 μm was scaled by the total SC volume to obtain normalized thick sections. Representative samples were stained with volume (NV). This normalization of the area and volume standard hematoxylin and eosin (H&E), luxol fast blue measurements compensated for the variations from slice (LFB), vascular endothelial marker (CD34) or neuron spe- to slice and animal to animal, as well as other spatial cific enolase (NSE) for histopathological assessment of scales, such as changes in slice thickness and pixel dimen- the neurovascular pathological changes. Immunohisto- sions. chemistry (IHC) was carried out with CD34 (MY 10 clone, BD Bioscience, Franklin Lakes, NJ) (1:100 dilution) and In addition, K measurements remaining above thresh- NSE (BBS/NC/VI-H14 clone, Dako Cytomation, Carpin- p-sc old within the total area at the epicenter slice were aver- teria, CA) antibodies using a Dako cytomation autostainer aged for each animal. The resulting averages for day 1 and [25]. Briefly, the slides were de-paraffinized by incubation day 3 were then correlated with the corresponding mean in xylene and ascending grades of alcohol. Antigen DTI measurements within the same area, as explained retrieval for CD34 was done by incubation with protein- below. ase Kinase at room temperature for 5 minutes. Antigen retrieval for NSE was done by heating in ethylene diamine tetraacetic acid (EDTA), PH 9.0 buffer (Lab vision, Fre- Page 4 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 mont, CA) at 95°C for 20 min. Both sections were loaded achieve an optimal tuning and matching condition as in the autostainer programmed as follows: 3% hydrogen observed in the frequency response of the coil impedance. peroxide for 10 minutes, blocked with 5% skimmed milk When the animal was placed in the magnet, the catheter for 5 min, incubated with primary antibody for 30 min, attached to it was extended outside at the front end of the followed by incubation with Envision+ system HRP magnet bore. The injured cord was then imaged in trans- (Dako Cytomation, Carpinteria, CA) for 30 minutes. verse and rostral-caudal planes. Acquiring data from dif- Color was developed by incubating samples with diami- ferent orientations allowed better evaluation of the nobenzidine (DAB)+chromogin for 10 min followed by lesion's spatial extent. Dako DAB enhancer for 5 min. Hematoxylin was used as counter stain. The sections were examined using a BX 50 Anatomical axial PD images acquired from one of the Olympus microscope and the photographs were taken injured SC are shown in Figure 1. The intensity contrasts with an attached Olympus DP 70 camera operated with on these images from normal sections delineate gross ana- DP Controller software (Olympus Corporation, Center tomical details of the cord within the white matter (WM) Valley, PA). One section was also incubated with second- and grey matter (GM) as well as the surrounding spinal ary antibody only to check nonspecific binding. structures. The lesion is depicted by an altered intensity contrast in the parenchyma below the laminectomy. The Statistical analysis corticospinal tract in the mouse is anatomically located The quantitative data collected from the measurements between the dorsal horns ventrally next to the central were analyzed statistically. The NA, NV, K , λ , λ and FA canal. In normal cords, the image intensity profile does p-sc || ⊥ values gathered from each animal were listed in two not produce enough contrast to distinguish this tract from groups as day 1 and day 3, and the means and standard the surrounding WM and GM. In this particular injured deviations within each group were computed for each SC, images from the rostral, but not the caudal, sections parameter to understand the intra-group variations. The delineated the corticospinal tract with hyperintensity. The measurements between day 1 and 3 were compared using intensity change was indicative of alterations in the MR the paired Student's t test. This analysis allowed examina- properties of the tract and of a pathological abnormality, tion of inter-group variations and the determination of which was likely associated with Wallerian degeneration the statistical significance of the differences. Statistical sig- [26]. nificance was defined at P < 0.05. Also, statistical depend- encies between K , and measurements λ , λ and FA Figure 2 presents pre- and postcontrast T1-weighted p-sc || ⊥ were determined using Pearson's correlation analysis and images in sagittal and horizontal planes in both postin- the resulting correlation coefficients were reported. jury days 1 and 3. On day 1, the precontrast images revealed a small focal hypointensity close to the dorsal surface reflective of neuropathology. On day 3, the lesion Results The surgical procedures, injuries, and prolonged anesthe- assumed a circular shape and enlarged in size. The post- sia during MRI scans were well tolerated by all mice. For contrast images depicted intensity enhancement at the imaging, each mouse was carefully placed supine over an lesion and its surroundings. The hyperintense regions in inductively coupled radio frequency coil system to postcontrast images represented areas of injured SC tissue Axi Figure 1 al proton-density images of an injured mouse spinal cord Axial proton-density images of an injured mouse spinal cord. The serial images show normal caudal and rostral sec- tions and injury epicenter on postinjury day 1. Arrows point to corticospinal tract (CST). In normal cord, the image intensity profile does not produce enough contrast to distinguish the CST from the surrounding white matter. Interestingly, in this injured SC, the CST at the rostral section, but not the caudal section, has been delineated by slight hyperintensity compared to the background white matter. Page 5 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 Precontrast an m Figure 2 ouse spinal cor d postcontrast d in sagittal and horizontal views T1-weighted images of injured Precontrast and postcontrast T1-weighted images of injured mouse spinal cord in sagittal and horizontal views. The postcontrast images were acquired 130 min (sag- ittal) and 140 min (horizontal) after the IV bolus delivery of the contrast agent. loaded with the contrast agent due to its leakage through the compromised BSCB therein. The postcontrast sagittal images in the figure were acquired 130 min and those in the horizontal views were acquired 140 min after the delivery of the contrast agent. Following its delivery, the contrast agent diffuses passively in the extravascular spaces of the injured SC parenchyma. This leads to the brightness enhancement spreading along the cord in both rostral and caudal directions, as evident in the figure. The brightness enhancement expanding spatially with time was better appreciated in images from axial views in Figure 3. This figure shows strong contrast enhancement Contr con day 3 Figure 3 trast agent with time a ast enhancement followin nd spac g the administration of the e on postinjury day 1 and early on at the lesion and its immediate surroundings fol- Contrast enhancement following the administration of the contrast agent with time and space on postin- lowing the contrast agent delivery. The enhancement was jury day 1 and day 3. The red and blue circles represent mostly localized in the GM than the WM. But as time pro- the regions of interest selected to produce the data in Figure gressed, the SC parenchyma became brighter at the slices 4. Please note that intensity enhancement induced by the distant from the epicenter in both caudal and rostral presence of the contrast agent in cerebrospinal fluid. directions within the normal sections of the injured SC. Although the patterns of contrast enhancements on days Page 6 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 1 and 3 were similar for this exam, the data indicated that Figure 4 compares the temporal profile of the relative the enhancement on day 1 peaked at the epicenter and intensity enhancement (REI, computed using Eq. 1 in the skewed asymmetrically towards the rostral direction, Appendix) within the regions of interest at the injury site unlike more symmetric, but wider, enhancement along on both days 1 and 3. The selected regions were identified the cord as visualized on day 3. In addition, cerebrospinal by circles coded with two different colors in Figure 3. The fluid (CSF) appeared hyperintense on the postcontrast variations in plots with different colors on the same day images. This was another important observation, which indicated REI has spatial dependence. The identical colors was also reported in the previous DCE-MRI study with a showed that the average REI profiles varied between day 1 rat model of SCI [5]. Combining the CSF enhancement in and day 3. According to Eq. 2, REI reflected the amount of rat with the current observation made in mouse, it is likely localized accumulation of the contrast agent, and thereby that rodents may have a unique system of CSF circulation indicated the severity of leakage of the agent through the different than human, since CSF of human does not typi- dysfunctional BSCBs therein. cally enhance after the IV delivery of the contrast agent used in this study. Figure 5 shows quantitative distribution of BSCB permea- -maps in the fig- bility for the data set in Figure 3. The K p-sc For the data analysis, images (precontrast and first post- ure were computed from three slices centered at the contrast) from a given slice location were displayed simul- epicenter using Eq. 4 in Appendix. The permeability esti- taneously. Image alignment was verified visually by mates were overlaid as a new layer with color coding on zooming in on anatomical landmarks. Two out of the ten the precontrast T1-weighted images, which served as the DCE-MRI scans exhibited spatial misalignment between background. In the maps, color coding towards red meant the axial postcontrast and the corresponding precontrast increased BSCB dysfunction. The red zones represented images. In these cases, the sedated mouse reacts to the the regions with greater leakage of the contrast agent and contrast agent, causing its body to move slightly. From the overlapped consistently with the lesion pathology. evaluation of the images, it was evident that the image motions in the misaligned data sets were translational in both cases, but a slight rotation was present in one of them. Neither of the motions was of deformation type. The two misaligned images were registered using a simple postprocessing technique with Matlab's "circshift" and "rotate" functions. A more complicated automated algo- rithm for image registration chould have been used for the registration. But this required complex implementation, which was beyond the scope of the study. sl Figure 5 Color coded ices centered K at the -map epi s computed center using the data set using Eq. 4 from the three in Figure 4 p-sc Color coded K -maps computed using Eq. 4 from p-sc the three slices centered at the epicenter using the data set in Figure 4. The backgrounds are the T1-weighted precontrast images (top and bottom rows). The color from Temporal patterns of averaged within two selected b (diam Figure 4 lue circles in ond) Figure 5) a relati nd on Day ve intensity regions 1 (squar of in enhter ancement ( ee ) and Day 3 st (red an RIE d ) yellow to red indicates linearly increasing compromise in Temporal patterns of relative intensity enhancement blood spinal cord barrier permeability. The areas with color (RIE) averaged within two selected regions of inter- towards red indicate regions with more vascular damage est (red and blue circles in Figure 5) and on Day 1 causing the barriers to become more permeable to blood (square) and Day 3 (diamond). constituents. Page 7 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 Figure 6 shows the precontrast PD and T2-weighted region (window 2) that led to the extravasation of the red images of the same injured SC. The corresponding T1- blood cells. The periphery of the damaged SC region (win- weighted image was reproduced from Figure 3 for a side- dow 3) contained cavities that colocalized with small ves- by-side comparison. Depending on the acquisition proto- sels or endothelial cells. col employed, images delineated the injury pathology with a specific intensity contrast. For this case, almost half Figure 9 shows LFB, NSE and CD34 stained slices showing of the GM appeared hypointense on the T2-weighted the injury and a normal cord section. At the injury, the image (Figure 6-b), but the remaining contra-lateral sec- LFB and NSE stains depicted diffuse demyelination and tion exhibited nearly homogeneous contrast, making it completely disorganized neurons and processes. The difficult to differentiate GM from WM. In addition, a thin CD34 stain from the same areas showed damaged vascu- strip of hyperintensity was observed along the circumfer- lature with leaky barrier. In the rostral segment, however, ence of the hypointense region on the T2-weighted image. LFB and NSE stains indicated normal myelination and In general, hypointensity on T2-weighted images in the intact neurons, and CD34 stain showed vasculature with early phase of SCI in mouse is associated with the accu- intact barrier. mulation of red blood cells, most likely escaping the vas- culature that was ruptured by the initial mechanical The area and volume of the regions with damaged BSCB impact. Hyperintensity is associated with vasogenic were measured semi-automatically as described above. edema describing the accumulation of fluid containing Figure 10 shows two histograms of K maps from a nor- p-sc plasma proteins into the extracellular spaces of the injured mal section and epicenter. The histogram from the nor- SC. The origin of these contrasts was determined by exam- mal section was used to determine a threshold for ining the corresponding histology slides. spatially segmenting the K maps to measure the areas in p-sc the other slices. The inter- and intra-variations for the Figures 7, 8 and 9 show histologic sections stained with threshold values estimated on days 1 and day 3 were -1 H&E, LFB, NSE or CD34. The slice in Figure 7 matches the within the range 0.03 ± 0.01 min . Figure 11 compares images in Figure 6, but the others in Figures 8 and 9 are NA measurements from K maps of 6 slices centered at p-sc from a different injured SC. The vasculature was depicted the lesion from all animals. These results indicated that by H&E staining in Figure 7 and better outlined in Figure four slices (in 4 mm space) around the epicenter suffered 8 by CD34 immunostaining. The microscopic features of the most compromise in BSCB, covering about 70% of the the neurovascular pathology revealed partially intact GM total cross section of the SC parenchyma on day 1. But, on at the ventral horns, significant damage at the dorsal SC day 3, the regions with damaged BSCB were reduced in with substantial loss of tissue matrix and also small cavi- size to about 55% for the two middle and 20% for the ties distributed throughout the cord, but generally more remaining two slices. Such behavior indicated that the prominent in WM than GM. These regions were evaluated spatial distribution of the areas with compromised BSCB more closely at higher magnification in three selected shrunk with time. This was consistent with the previous regions of interest – windows 1, 2 and 3. As expected, in reports, where the lesion size shrank with time in direc- the normal looking GM (window 1), the vessels were tions both across and along the cord, which was a unique observed to be intact but mostly ruptured in the damaged behavior seen in mice SCI [27,28]. This outcome was fur- ther supported by the normalized volume (NV) measure- ments that decreased from day 1 to 3, as shown in Figure 12. The NV data from all animals indicated that volume of BSCB breakdown occupied 48% of the total volume of the SC on day 1 as compared to 25% on day 3. The differ- ence between the two measured volumes was statistically significant. These results from day 1 and day 3 together implied that dynamic remodeling of the BSCB permeabil- ity took place as part of the ongoing neurovascular repair and recovery processes in injured cord. Figure 13 shows anatomical T2-weighted image of an injured SC and the computed maps (K , λ , λ and FA) p-sc || ⊥ from the same slice orientation. The green circle plotted A p Figure 6 rnior to the injection of atomical images of injur contrast y epicenter on post-injury day agent 3 on the K map represents a region of interest with Anatomical images of injury epicenter on post-injury p-sc increased BSCB permeability within the GM. The mean day 3 prior to the injection of contrast agent: a) PD image, b) T2-weighted image and c) T1-weighted image. values for the parameters K , λ , λ and FA residing p-sc || ⊥ within such selected regions were computed for each ani- Page 8 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 H Figure 7 &E stained histology images of injured SC on day 3 H&E stained histology images of injured SC on day 3: a) at lower magnification and b-d) at higher magnification within the selected square windows marked with numbers 1, 2 and 3 in panel a. The windows were selected based on the MRI- st nd observed pathology in Figure 6. The 1 window was selected in an intact GM region, the 2 window was selected in a signifi- rd cantly damaged region and the 3 window was selected in a region with edema. Black arrows denote normal vessels and capil- laries. Black arrow heads point to disrupted vasculature with damaged BSCB. White arrows point to vessels surrounded by cavities. Circle denotes clusters of extravasated red blood cells. mal. The results from all animals were further averaged The level of the one-to-one dependencies between the and compared in Table 1. The differences between the parameter K and the measurements λ , λ and FA are p-sc || ⊥ measurements on days 1 and 3 were found to be statisti- given in Table 2. From the values in the table, negative cally significant (P < 0.05). Normative λ and λ measure- correlations were found between K and λ and between || ⊥ p-sc || -3 ments from GM of normal mouse SC (λ = 1.25 × 10 K and FA. K and λ were also correlated, but posi- || p-sc p-sc ⊥ 2 -3 2 mm /s, λ = 0.47 × 10 mm /s and FA = 0.61) were tively. Combining these results with the histological find- reported earlier [21]. Comparing these results with the ings (Figures 7, 8 and 9) suggested that a strong corresponding ones in the table indicates significant dif- relationship exists between vascular integrity and the ferences, demonstrating the degree of sensitivity of these structural state of the neuronal tissue as assessed by the parameters to the neuropathology in injured cords. The diffusion measurements. These results were in line with data in the table also shows that the parameters K , λ the prior knowledge that more vascular damage is likely to p-sc || and FA decrease, but λ increases from day 1 to day 3. cause greater neuronal loss following SCI. Such trends are suggestive of vascular restoration, further disruption of axonal integrity, loss of anisotropy and Discussion increased demyelination within the injured SC as time Preclinical neuroimaging methods are required for evalu- progressed. ating the degree of initial mechanical damage and for Page 9 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 H Figure 8 istology images of an injured SC on day 3 Histology images of an injured SC on day 3: a) H&E-stained section with arrows denotes extravasated red blood cells. b- d) higher magnifications of adjacent section stained with the endothelial marker, CD34. The square windows marked with numbers 1, 2 and 3 were selected to show vasculature and BSCB integrity therein. b) Sparse, intact vasculature (arrows). c) Disrupted vasculature with diffusion of immunostaining due to damaged BSCB. d) Disrupted vessels and tissue vacuolization due to edema from the leaky BSCB (arrows). monitoring the subsequent secondary events following of the contrast agent in between [6]. In this model, imme- SCI in experimental research. With this purpose in mind, diately following the IV bolus injection, the transport of we and others have investigated a variety of MRI methods the contrast agent was considered unidirectional from to obtain anatomical, vascular and structural information plasma to the injured SC tissue. This feature allowed asso- from the injured SC [14,21,29-39]. However, to date, no ciating the barrier permeability with the transfer rate con- study has combined DCE-MRI and DTI protocols to stant from plasma-to-spinal cord. Such association was simultaneously evaluate the microvascular and micro- also proved to be critical for simplifying the computa- structural changes in injured SC. The current study is the tional analysis whereby requiring only the early part of the first to jointly apply these techniques, along with anatom- relative intensity enhancement (Figure 4). This ultimately ical imaging, to evaluate SCI in mouse. An additional led to the estimation of localized changes in the barrier achievement that distinguishes this study from the previ- dysfunction throughout the SC (Figure 5). ous works is the computational algorithm implemented for quantitatively mapping the spatial distribution of Our MRI-based data, as in Figures 3, 4 and 5, and histo- BSCB permeability in injured SC. This algorithm was logical analysis, as in Figures 7, 8 and 9, demonstrated the derived from a pharmacokinetic model (Figure 14) that dynamic remodeling of the BSCB as part of the ongoing was originally developed to represent plasma and injured repair and recovery processes in the injured SC tissue. SC by two compartments, and to determine the exchange Using postmortem analysis in mouse, Whetstone et al. [2] Page 10 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 LFB, Figure 9 NSE and CD34 stained contiguous slices at the injury epicenter and at a normal section rostral to the epicenter LFB, NSE and CD34 stained contiguous slices at the injury epicenter and at a normal section rostral to the epi- center. In the rostral segment (top row), LFB (a) and NSE (b) stains show normal myelination and intact neurons, and CD34 stain (c) shows intact vasculature. At the epicenter (bottom row), LFB (d) and NSE (e) stains show diffuse demyelination and damaged neurons, and the corresponding CD34 stain (f) shows damaged vasculature with leaky BSCB. reported that SCI results in a biphasic, temporal pattern of barrier leakage to tracer – luciferase. In their study, the barrier leakage was shown to extend beyond the epicenter into segments that were within 6 mm rostral and caudal to the epicenter. The leakage was seen to be mostly pro- nounced within the first 35 min after the injury followed by a gradual decline within the first 24 hr. A second peak and ep Figure 10 Histograms icenter of th of K -e injury (right) maps from a normal caudal section (left) p-sc Histograms of K -maps from a normal caudal sec- No ce Figure 11 nter rmal ed iz at the e ed area (mean ± std) picenter from measurements from 6 sl all animals on day 1 and day 3 ices p-sc tion (left) and epicenter of the injury (right). Solid lines Normalized area (mean ± std) measurements from 6 -1 are fit to a Gaussian function. A threshold of 0.025 min was slices centered at the epicenter from all animals on determined from the histogram on the left. This value was day 1 and day 3. * denotes statistically significant difference (P < 0.05). used for segmenting the regions with compromised BSCB permeability in the maps from the other slices. Page 11 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 From the application point of view, providing informa- tion about the barrier function should be useful in at least three circumstances. One situation is in characterizing the SCI model where the resulting barrier leakage in a partic- ular SCI model may be another parameter to be defined, and referred back when needed to check against the con- sistency of the newly induced injuries. This would enable the confirmation of whether later injuries have similar properties in terms of the vascular response. Traditional histological analysis was employed to understand the sen- sitivity and specificity of the barrier damage to graded mechanical perturbations [3]. Our approach offers an effi- cient alternative and may become a preferred choice for such purpose. In the second case, quantitatively evaluat- ing the severity of the barrier damage as soon as possible after the injury may provide early indications of the signif- Normali m Figure 12 als on da zed vo y 1 an lume (me d day 3 an ± std) measurements from all ani- icance and level of the expected secondary processes that Normalized volume (mean ± std) measurements are likely to increase neuronal loss. Such capability can from all animals on day 1 and day 3. * denotes statisti- have significant prognostic value. Figure 13 and compara- cally significant difference (P < 0.05). tive analysis in Tables 1 and 2 clearly provided evidence that correlations exist between DCE-MRI and DTI based of abnormal barrier permeability was reported at 3 days measurements. This has important implications in prac- after the injury, which significantly exceeded that tice, especially when the examined SCI produces poor observed at 24 hr. Regarding the barrier permeability to quality of DTI acquisition or the analysis of the resulting the contrast agent, we obtained comparable results that data is not feasible. In such challenging situations, includ- showed stronger intensity enhancement on day 3 than ing the DCE-MRI protocol in the scan may supply infor- day 1. This agreement is encouraging and warrants further mation on the condition of the vasculature, which may exploration, such as investigation of the loss of blood ves- indirectly help to determine the level of neuronal damage sels and revascularization in injured mouse SC using in the underlying SC tissue by using the close associations DCE-MRI in conjunction with MR angiography modality demonstrated in Table 2. Lastly, monitoring the barrier [5,32,40]. function may have important implications in predicting the efficacy of targeted drug or gene delivery through the Requiring only the initial portion of the intensity permeable barriers when potential therapeutic interven- enhancement to assess barrier function offers benefits in tions are considered. While this aspect is foreseeably terms of reducing the imaging time substantially. In the achievable if the drug or transfection vector has molecular current study, we acquired T1-weighted images using a weight comparable to that of the contrast agent used in spin-echo sequence which required relatively long acqui- this study, it remains to be seen if the approach would be sition time. This is one limitation of the approach. Higher viable in cases when the drug or gene has significantly temporal resolution can be achieved between the precon- larger size. trast and postconrast acquisitions by employing faster imaging sequences. Such ability should further improve Conclusion the time and accuracy of estimating the barrier permeabil- This study has demonstrated the potential of DCE-MRI ity. For this purpose, echo-planar imaging appears prom- method to assess the BSCB dysfunction in injured spinal ising [41]. cord noninvasively. The method involved acquiring only Table 1: BSCB permeability and microstructural measurements on day 1 and day 3 (mean ± standard deviation). P values for the measurements between day 1 and 3 are given in the last row. -1 -3 2 -3 2 K (min ) λ (×10 mm /s) λ⊥ (×10 mm /s) FA p-sc || Day 1 0.123 ± 0.008 0.788 ± 0.022 0.579 ± 0.016 0.183 ± 0.025 Day 3 0.076 ± 0.006 0.694 ± 0.026 0.604 ± 0.015 0.097 ± 0.021 P 0.002 0.001 0.002 0.006 Page 12 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 Postinj Figure 13 ury day 1 Postinjury day 1: a) anatomical T2-weighted image of an injured SC, b) K map, c) λ map, d) λ map and e) FA map. p-sc || ⊥ Green circle denotes the region of interest chosen within the GM to compute the mean values for the parameters K , λ , p-sc || λ and FA from the corresponding maps. two T1-weighted images; precontrast and postcontrast, acquired with a minimum time delay following the injec- Two-c distrib cord tissue Figure 14 u otion mpartment p of the con harmocokine trast agent in plasma and tic model representing the injured spinal Two-compartment pharmocokinetic model repre- tion of contrast agent. This simple data acquisition strat- senting the distribution of the contrast agent in egy provides sufficient temporal information on the plasma and injured spinal cord tissue. The subscripts p contrast-enhancement that was necessary to map the spa- and sc denote plasma and spinal cord. [C] (t) and [C] (t) are p sc tial distribution of BSCB permeability. The BSCB dysfunc- time-dependent concentrations in mmol/L in plasma and spi- tion correlated strongly with the degree of axonal loss and nal cord compartments, respectively. The parameters k p-sc demyelination. The relationship between the vascular and k denote transfer rate constants in l/min for forward sc-p damage and neuronal loss in injured SC has been well- transfer of the contrast agent from plasma-to-spinal cord and established previously, but by using histological tech- reverse transfer from spinal cord-to-plasma compartments, niques. Reconfirming this association using DCE-MRI respectively. and DTI data acquired from living animals has important implications in translational research. Preclinical efforts are currently focused on developing BSCB permeable Competing interests drugs for improving neurovascular function by repairing The authors declare that they have no competing interests. vascular network, delaying neurodegenerative processes or promoting neuronal recovery. Research efforts are also Authors' contributions underway for understanding the role of specific genes in IT performed the statistical analysis. PCC postprocessed vascular reorganization, neuronal repair and recovery the acquired DCE-MRI data. HES contributed with the from an injury on experimental test systems involving dif- histology and staining. MMD evaluated the pathology. ferent strains of "transgenic" or gene knock-out mice. On MB was responsible for the conception and design of the the basis of multiple neuroimaging methodologies devel- experiments as well as the analysis and interpretation of oped in this study, scanning the same animal model with the data collectively. All authors read and approved the anatomical, DCE-MRI and DTI protocols provides meas- final manuscript. urable parameters that can serve as sensitive and specific in vivo neurovascular biomarkers for comprehensively Appendix evaluating the pathological state of the injured SC, may The presence of paramagnetic contrast agent at concentra- offer a prognostic value for functional recovery from SCI tion [C] (in mmol/L) in neuronal tissue of an injured SC and potentially serve as a monitoring tool for evaluating was detected by relative intensity enhancement (RIE) on the efficacy of a treatment efficacy. T1-weighted magnetic resonance images. Postcontrast RIE at time t was calculated for each pixel at a spatial position (x, y) and a slice location z using the formula Table 2: Pearson's correlation coefficient results between the BSCB permeability and microstructural measurements on day 1 and day 3. It (,x,y,z)−= It ( 0,x,y,z) RIE(, t x, y, z) = xy,( ∈ Mask z) ). It(, =0 x,y,z) ρ(K , λ ) ρ(K ,λ ) ρ(K , FA) p-sc || p-sc ⊥ p-sc (1) Day 1 -0.87 0.64 -0.77 Here, I(t = 0, x, y, z) and I(t, x, y, z) represent intensities in th the precontrast and t postcontrast images of a given slice Day 3 -0.92 0.97 -0.70 z, respectively. The mask Mask(z) was obtained by using Page 13 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 manual segmentation applied on the postcontrast images where the SC was delineated best. The mask covered the 1 dRIE(, t x,y,z) kx (,y,z) = xy,( ∈ Mask z). ps − c cord but excluded cerebrospinal fluid, which also exhib- Tr [C] () t =0 dt 10 1 p t0 = ited postcontrast enhancement. The total cross-sectional (3) area of the SC at that particular slice was measured by the area covered by the mask. Considering that the factor T r has negligible depend- 10 1 ence on spatial location, we reorganize Eq. [3] to obtain a When a spin-echo sequence is used for DCE-MRI acquisi- new parameter that is linearly proportional to k : p-sc tion and a small dose of contrast agent is delivered as a bolus, as in this study, the SC concentration [C] and RIE dRIE(, t x,y,z) sc K (,x yz , )== T r [C] (t 0)k (,x yz , )= xxy,( ∈ Mask z). ps−− c 10 1 p psc dt are related by the formula [4] t0 = (4) RIE(, t x,y,z) This equation indicates that the initial slope of REI pro- [C] (, t x, y, z)=∈ xy,( Mask z). (2) sc Tr 10 1 vides a measurable quantity representing the BSCB per- meability at a spatial location (x, y, z). Specifically, Here, T denotes relaxation time in second (s) before numerical computation of Eq. [4] involves discrete deriv- administering the contrast agent and r is relaxivity intro- duced by the contrast agent to that of the underlying SC RIE(, t+− ΔT x, y, z) RIE(t, x, y, z) ative operation for t = 0 and ΔT -1 -1 ΔT tissue and expressed in mM s . For mice scanned at 9.4 T, T was reported to be 1730 ms for the gray and 1690 = 10 min – the temporal resolution in our DCE-MRI ms for the white matters of the SC [21], but r associated acquisitions. By noticing that RIE (t = 0, x, y, z) = 0, the with the contrast agent for these tissue types has yet to be derivative operation can be simplified further. In this determined. RIE(, t +ΔT x, y, z) study, was calculated for each x, y ∈ ΔT Equation [2] plays an important role when constructing a Mask(z) and the results were put together to form a 2-D pharmacokinetic model to theoretically represent the quantitative K map delineating the regions of the p-sc uptake of contrast agent in an injured SC [4]. Construc- injured SC with compromised BSCB permeability for each tion of a proper pharmacokinetic model requires at least slice. two compartments – one for the plasma and one for the cord, as shown in Figure 1[6]. In this model, two rate con- Acknowledgements This work was funded in part by NIH grants NS052610 and NS054019. The -1 stants (min ) quantitatively describe the transport of con- authors thank Dr. Fan Yang for helping with animal surgeries and care. trast agent from plasma-to-SC and SC-to-plasma. Following an IV bolus injection, the contrast agent first References 1. Tator CH, Koyanagi I: Vascular mechanisms in the pathophysi- leaks from plasma to the injured SC tissue through the ology of human spinal cord injury. J Neurosurg 1997, 86:483-492. compromised BSCBs. This thereby allows the representa- 2. Whetstone WD, Hsu JY, Eisenberg M, Werb Z, Noble-Haeusslein LJ: Blood-spinal cord barrier after spinal cord injury: relation to tion of the barrier permeability with the rate constant k p-sc revascularization and wound healing. J Neurosci Res 2003, governing the unidirectional transfer of the contrast agent 74:227-239. 3. Maikos JT, Shreiber DI: Immediate damage to the blood-spinal from plasma-to-SC. This rate constant is then estimated cord barrier due to mechanical trauma. J Neurotrauma 2007, using the formula 24:492-507. 4. Bilgen M, Narayana PA: A pharmacokinetic model for quantita- d[C] (,tx,y,z) 1 sc tive evaluation of spinal cord injury with dynamic contrast- kx (,y,z) = [4,6]. In this ps − c [C] () t =0 dt p enhanced magnetic resonance imaging. Magn Reson Med 2001, t0 = 46:1099-1106. expression, [C] (t = 0) denotes the initial concentration of 5. Bilgen M, Abbe R, Narayana PA: Dynamic contrast-enhanced MRI of experimental spinal cord injury: in vivo serial studies. the contrast agent in mouse plasma and computed from Magn Reson Med 2001, 45:614-622. 6. Bilgen M, Dogan B, Narayana P: In vivo assessment of blood-spi- the injected dose and the total plasma volume. The oper- nal cord barrier permeability: serial dynamic contrast d(.) enhanced MRI of spinal cord injury. Magn Reson Imaging 2002, ator denotes time derivative evaluated immedi- dt 20:337. t0 = 7. Guertin PA: Paraplegic mice are leading to new advances in ately after the delivery of the contrast agent. Using Eq. [2], spinal cord injury research. Spinal Cord 2005, 43:459-461. 8. Rosenzweig ES, McDonald JW: Rodent models for treatment of we can write spinal cord injury: research trends and progress toward use- ful repair. Curr Opin Neurol 2004, 17:121-131. 9. Steward O, Schauwecker PE, Guth L, Zhang Z, Fujiki M, Inman D, Wrathall J, Kempermann G, Gage FH, Saatman KE, et al.: Genetic approaches to neurotrauma research: opportunities and potential pitfalls of murine models. Exp Neurol 1999, 157:19-42. Page 14 of 15 (page number not for citation purposes) BMC Medical Imaging 2009, 9:10 http://www.biomedcentral.com/1471-2342/9/10 10. Kigerl KA, McGaughy VM, Popovich PG: Comparative analysis of 33. Gaviria M, Bonny JM, Haton H, Jean B, Teigell M, Renou JP, Privat A: lesion development and intraspinal inflammation in four Time course of acute phase in mouse spinal cord injury mon- strains of mice following spinal contusion injury. J Comp Neurol itored by ex vivo quantitative MRI. Neurobiol Dis 2006, 2006, 494:578-594. 22:694-701. 11. Pannu R, Christie DK, Barbosa E, Singh I, Singh AK: Post-trauma 34. Kim JH, Trinkaus K, Ozcan A, Budde MD, Song SK: Postmortem Lipitor treatment prevents endothelial dysfunction, facili- delay does not change regional diffusion anisotropy charac- tates neuroprotection, and promotes locomotor recovery teristics in mouse spinal cord white matter. NMR Biomed 2007, following spinal cord injury. J Neurochem 2007, 101:182-200. 20:352-359. 12. Pearse DD, Chatzipanteli K, Marcillo AE, Bunge MB, Dietrich WD: 35. Huang ZL, He YY, Bilgen M: In vivo tracing of corticospinal tract Comparison of iNOS inhibition by antisense and pharmaco- in mouse using manganese-enhanced contrast. J Neurol Sci logical inhibitors after spinal cord injury. J Neuropathol Exp Neu- [Turk] 2007, 24:38-44. rol 2003, 62:1096-1107. 36. Loy DN, Kim JH, Xie M, Schmidt RE, Trinkaus K, Song SK: Diffusion 13. Elshafiey I, Bilgen M, He R, Narayana PA: In vivo diffusion tensor tensor imaging predicts hyperacute spinal cord injury sever- imaging of rat spinal cord at 7 T. Magn Reson Imaging 2002, ity. J Neurotrauma 2007, 24:979-990. 20:243-247. 37. Bilgen M: Magnetic resonance microscopy of spinal cord injury 14. Kim JH, Loy DN, Liang HF, Trinkaus K, Schmidt RE, Song SK: Nonin- in mouse using a miniaturized implantable RF coil. J Neurosci vasive diffusion tensor imaging of evolving white matter Methods 2007, 159:93-97. pathology in a mouse model of acute spinal cord injury. Magn 38. Bilgen M, Al-Hafez B, Alrefae T, He YY, Smirnova IV, Aldur MM, Fes- Reson Med 2007, 58:253-260. toff BW: Longitudinal magnetic resonance imaging of spinal 15. Aimone JB, Leasure JL, Perreau VM, Thallmair M: Spatial and tem- cord injury in mouse: changes in signal patterns associated poral gene expression profiling of the contused rat spinal with the inflammatory response. Magn Reson Imaging 2007, cord. Exp Neurol 2004, 189:204-221. 25:657-664. 16. Ousman SS, David S: MIP-1alpha, MCP-1, GM-CSF, and TNF- 39. Clark CA, Werring DJ, Miller DH: Diffusion imaging of the spinal alpha control the immune cell response that mediates rapid cord in vivo: estimation of the principal diffusivities and phagocytosis of myelin from the adult mouse spinal cord. J application to multiple sclerosis. Magn Reson Med 2000, Neurosci 2001, 21:4649-4656. 43:133-138. 17. Parker D: Pharmacological approaches to functional recovery 40. Bilgen M, Al-Hafez B, He YY, Brooks WM: Magnetic resonance after spinal injury. Curr Drug Targets CNS Neurol Disord 2005, angiography of rat spinal cord at 9.4 T: a feasibility study. 4:195-210. Magn Reson Med 2005, 53:1459-1461. 18. Pearse DD, Bunge MB: Designing cell- and gene-based regener- 41. Callot V, Duhamel G, Cozzone PJ: In vivo mouse spinal cord ation strategies to repair the injured spinal cord. J Neuro- imaging using echo-planar imaging at 11.75 T. Magma 2007, trauma 2006, 23:437-452. 20:169-173. 19. Bilgen M: A new device for experimental modeling of central nervous system injuries. Neurorehabil Neural Repair 2005, Pre-publication history 19:219-226. 20. Bilgen M: Simple, low-cost multipurpose RF coil for MR The pre-publication history for this paper can be accessed microscopy at 9.4 T. Magn Reson Med 2004, 52:937-940. here: 21. Bilgen M, Al-Hafez B, Berman NE, Festoff BW: Magnetic resonance imaging of mouse spinal cord. Magn Reson Med 2005, 54:1226-1231. http://www.biomedcentral.com/1471-2342/9/10/prepub 22. Bilgen M: Imaging corticospinal tract connectivity in injured rat spinal cord using manganese-enhanced MRI. BMC Medical Imaging 2006, 6:15. 23. Bilgen M, Elshafiey I, Narayana PA: Mohr diagram representation of anisotropic diffusion tensor in MRI. Magn Reson Med 2002, 47:823-827. 24. Bilgen M, Narayana PA: Mohr diagram interpretation of aniso- tropic diffusion indices in MRI. Magn Reson Imaging 2003, 21:567-572. 25. Wei H, Desouki MM, Lin S, Xiao D, Franklin RB, Feng P: Differential expression of metallothioneins (MTs) 1, 2, and 3 in response to zinc treatment in human prostate normal and malignant cells and tissues. Mol Cancer 2008, 7:7. 26. Bilgen M, Rumboldt Z: Neuronal and vascular biomarkers in Syringomyelia: investigations using longitudinal MRI. Biomar- kers in Medicine 2008, 2:113-124. 27. Inman D, Guth L, Steward O: Genetic influences on secondary degeneration and wound healing following spinal cord injury in various strains of mice. J Comp Neurol 2002, 451:225-235. 28. Inman DM, Steward O: Physical size does not determine the unique histopathological response seen in the injured mouse Publish with Bio Med Central and every spinal cord. J Neurotrauma 2003, 20:33-42. scientist can read your work free of charge 29. Wamil AW, Wamil BD, Hellerqvist CG: CM101-mediated recov- ery of walking ability in adult mice paralyzed by spinal cord "BioMed Central will be the most significant development for injury. Proc Natl Acad Sci USA 1998, 95:13188-13193. disseminating the results of biomedical researc h in our lifetime." 30. Bonny JM, Gaviria M, Donnat JP, Jean B, Privat A, Renou JP: Nuclear Sir Paul Nurse, Cancer Research UK magnetic resonance microimaging of mouse spinal cord in vivo. Neurobiol Dis 2004, 15:474-482. Your research papers will be: 31. Stieltjes B, Klussmann S, Bock M, Umathum R, Mangalathu J, Letellier available free of charge to the entire biomedical community E, Rittgen W, Edler L, Krammer PH, Kauczor HU, et al.: Manganese- enhanced magnetic resonance imaging for in vivo assess- peer reviewed and published immediately upon acceptance ment of damage and functional improvement following spi- cited in PubMed and archived on PubMed Central nal cord injury in mice. Magn Reson Med 2006, 55:1124-1131. 32. Bilgen M, Al-Hafez B: Comparison of spinal vasculature in yours — you keep the copyright mouse and rat: investigations using MR angiography. Neuro- BioMedcentral anatomy 2006, 5:12-18. Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 15 of 15 (page number not for citation purposes)

Journal

BMC Medical Imaging – Springer Journals

Published: Jun 11, 2009

There are no references for this article.