Radiation-induced white matter (WM) damage is a major side effect of whole brain irradiation among childhood cancer survivors. We evaluate longitudinally the diffusion characteristics of the late radiation-induced WM damage in a rat model after 25 and 30 Gy irradiation to the hemibrain at 8 time points from 2 to 48 weeks postradiation. We hypothesize that diffusion tensor magnetic resonance imaging (DTI) indices including fractional anisotropy (FA), trace, axial diffusivity (λ//), and radial diffusivity (λ⊥) can accurately detect and monitor the histopathologic changes of radiation-induced WM damage, measured at the EC, and that these changes are dose and time dependent. Results showed a progressive reduction of FA, which was driven by reduction in λ// from 4 to 40 weeks postradiation, and an increase in λ⊥ with return to baseline in λ// at 48 weeks postradiation. Histologic evaluation of irradiated WM showed reactive astrogliosis from 4 weeks postradiation with reversal at 36 weeks, and demyelination, axonal degeneration, and necrosis at 48 weeks postradiation. Moreover, changes in λ// correlated with reactive astrogliosis (P < 0.01) and λ⊥ correlated with demyelination (P < 0.01). Higher radiation dose (30 Gy) induced earlier and more severe histologic changes than lower radiation dose (25 Gy), and these differences were reflected by the magnitude of changes in λ// and λ⊥. DTI indices reflected the histopathologic changes of WM damage and our results support the use of DTI as a biomarker to noninvasively monitor radiation-induced WM damage. [Cancer Res 2009;69(3):1190–8]
- diffusion tensor imaging
- white matter
Radiation-induced white matter (WM) damage is considered a major contributory factor of treatment-induced neurotoxicity prevalent among childhood cancer survivors who have undergone whole brain irradiation ( 1). This is manifested as cognitive impairment in cancer survivors months to years after treatment and is of increasing effect because of marked improvements in long-term cancer survival ( 2). The identification of an accurate and sensitive marker of WM damage and neurotoxicity is necessary to modify treatment strategies that may prevent or minimize brain damage. Currently, treatment strategies that are being used with the goal to minimize radiation damage include reduction of radiation dose, altered fractionation, use of focal radiotherapy and proton radiotherapy, and improving immobilization so as to reduce planning target volume. Also, radiation is omitted or deferred in children under the age of 3 to 5 years due to the known susceptibility of developing brain tissue to radiation damage ( 3).
Diffusion tensor magnetic resonance imaging (DTI) can be used to map and characterize in vivo the three-dimensional diffusion of water in tissue ( 4, 5) and is a promising method for characterizing microstructural changes in WM. Quantitative diffusion variables can reflect the underlying pathology of WM disease ( 6– 8); for example, increased radial diffusivity (λ⊥) without change in axial diffusivity (λ//) has been found to reflect demyelination or dysmyelination ( 7, 8) and decreased λ// reflects axonal degeneration ( 6). We have shown in clinical studies of childhood medulloblastoma and acute lymphoblastic leukemia survivors after whole brain irradiation that WM fractional anisotropy (FA) is reduced, and it is more sensitive than conventional T2-weighted magnetic resonance imaging (MRI) in the detection of WM injury ( 9, 10). Also, FA reduction correlates with known risk factors of neurotoxicity, including radiation dose and age at irradiation ( 11), and with cognitive outcome ( 12), suggesting that FA is able to reflect WM injury that is associated with treatment-induced neurotoxicity. These similar findings of decreased FA after prophylactic whole brain irradiation have also been found in adult small cell lung cancer patients ( 13) and brain tumor patients ( 14). However, the histologic correlates of these DTI indices remain to be elucidated.
Studies using conventional magnetic resonance imaging sequences in animal radiation injury models found that focal and diffuse increased T2 signal underestimate the extent of histologic lesions and were not consistently seen ( 15– 17). We evaluate longitudinally the diffusion characteristics of late radiation-induced WM damage and test the following hypothesis: (a) DTI indices can accurately detect and monitor radiation-induced WM damage in vivo and can be used as a biomarker for the noninvasive evaluation of radiation-induced brain injury; (b) longitudinal changes of DTI indices are dose and time dependent.
Materials and Methods
Animal model preparation. The experiment was approved by the University Animal Ethics Committee. Forty 12-wk-old female Sprague-Dawley rats (n = 40) with mean body weight 291 ± 20 grams were divided into 2 separate batches to receive irradiation of 25 or 30 Gy to the right hemibrains with a single highly collimated (circular field of 12.5 mm diameter) 6MV photon beam from a linear accelerator (Varian) under anesthesia. This radiation dose induces changes in rat brains with a similar latent period as has been described in humans ( 18) and is the minimum required to produce selective late WM necrosis without deaths or gross neurologic deficits, and is a well-established animal model for the evaluation of late radiation-induced brain injury ( 15, 18). The rats were positioned with the aid of the treatment room lasers, which are well-aligned to within 1-mm accuracy of the isocenter at the right frontoparietal region. The collimated 6MV photon beam has a very steep dose gradient leading to a rapid decrease in dose outside the beam edge, and is therefore suitable for a small irradiated volume. The left hemibrains were confirmed to be excluded from the irradiation field on dose contour maps, and served as the internal control for subsequent analysis. Three rats died due to anesthesia. A total of 37 rats were irradiated with 25 Gy (n = 16) and 30 Gy (n = 21), respectively.
MRI scanning protocol. MRI scans were performed at 2 wk (n = 17), 4 wk (n = 37), 8 wk (n = 29), 16 wk (n = 28), 24 wk (n = 24), 36 wk (n = 19), 40 wk (n = 8), and 48 wk (n = 7) postradiation using a 7T nuclear magnetic resonance scanner (Bruker) and a 38-mm rat brain coil. Rats were anaesthetized using inhalational isoflurane and were prostrated on a custom-made holder with strapping to minimize head motion while respiration was monitored. Coronal T2-weighted MR images were obtained using the following variables: repetition time (TR), 11,189 ms; echo time (TE), 20 ms; field of view (FOV), 2.5 × 2.5 cm2; acquisition matrix, 128 × 128; slice thickness, 1 mm. Then, DTI images were obtained from 2 mm anterior to the corpus callosum to the end of the cerebrum. DTI was acquired with a respiration-gated spin echo 4-shot EPI readout sequence with an encoding scheme of 30 gradient directions homogenously distributed on the unit sphere. The variables used were as follows: TR, 3,000 ms; TE, 32 ms; Δ, 15 ms; δ, 5 ms; FOV, 3.0 × 3.0 cm2; slice thickness, 1 mm; acquisition matrix, 256 × 256; b value, 0 and 1,000 s/mm2. DTI variables including FA, trace, λ// and λ⊥ maps were calculated by previously described equations ( 4) using DTIStudio v2.30 (Johns Hopkins University).
Image analysis. Image analysis (total, 3,400 images) was performed for a final total of 36 rats (25 Gy, n = 15; 30 Gy, n = 21; 1 rat died during MRI scanning). To minimize observer bias, we adopted a semiautomated method for region of interest (ROI) analysis of the external capsule (EC) in both hemispheres. A template was created by transforming a randomly selected FA image to be as symmetrical as possible. To facilitate the comparison between the contralesional and ipsilesional sides, a flipped version of the images was produced. All the images either flipped or not flipped were affinely registered to the template using FSL 4 by matching FA images to the template. ROIs were drawn on the template of the EC on five consecutive slices and were automatically transferred onto FA images of each rat and carefully examined to rule out obvious misregistration; mean values of FA, trace, λ// and λ⊥ of each rat were then quantified on the registered images accordingly ( Fig. 1 ). Images were excluded from analysis if the boundary of the EC was not clearly demarcated due to motion artifact.
Histopathology evaluation. Rats were randomly sacrificed for histologic evaluation of EC at 4 wk (n = 3, 25 Gy; n = 4, 30 Gy), 16 wk (n = 2, 25 Gy; n = 2, 30 Gy), 24 wk (n = 4, 30 Gy), 36 wk (n = 5, 25 Gy; n = 6, 30 Gy) and 48 wk (n = 3, 25 Gy; n = 3, 30 Gy) postradiation. Specimens of rat brains were obtained with standard method ( 6) and were cut in 10-μm-thick coronal slices corresponding to the most posterior and most anterior MRI slices.
Standardized H&E stain was used to detect morphologic characteristics of brain tissue. Luxol fast blue (LFB) stain was performed to detect the myelin in the WM. Tissue sections were processed as free floating and were incubated in the following primary antibodies: monoclonal antibody to pan-axonal neurofilament marker (NF; SMI-312; 1:1,000; Covance) and monoclonal antibody to glial fibrillary acidic protein (GFAP; SMI-22, 1:1,000, Covance) for immunohistochemistry staining of axon and glial cell, respectively. Appropriate secondary antibodies were used at a dilution of 1:200 in 0.1 mol/L PBS and incubated in the secondary antibody goat anti-mouse IgG FITC (Chemicon AP124F).
All sections were examined using a light microscope (Carl Zeiss). Histologic images were recorded by digital photomicrography (Spot advanced), and staining intensity was measured from sections stained for LFB, NF, and GFAP using imageJ (NIH; ref. 19). The stain intensity was measured at ×200 histologic digital images in both symmetrical ECs at the regions corresponding to ROI analysis of DTI images.
Statistical analysis. All results were expressed as mean ± SD. Ratios of injury/control DTI indices (e.g., FA I/FA C) and injury/control staining intensity of histologic evaluations (LFBI/LFBC, NFI/NFC and GFAPI/GFAPC) were calculated for statistical analysis. Paired t test was used to detect statistical differences in the DTI quantitative indices and staining intensity of LFB, NF, and GFAP between the injury and control EC. To evaluate the dose effects in two radiation groups (30 Gy versus 25 Gy), injury/control DTI indices of both dose cohorts were compared using Mann-Whitney U test at every time point. Based on the plots of the DTI indices against time post radiation, we observed that the changes of DTI indices were better described by a quadratic mixed model as opposed to a linear model. We constructed the model in SPSS (Version 15; SPSS) with DTI indices as dependant variables and dose cohort (belonging to 25 or 30 Gy) as a fixed factor, the linear and quadratic terms of time postradiation as covariates. The interaction terms between dose cohort and the linear and quadratic terms of time postradiation were also modeled, as different dose of radiation may be associated with different curves for longitudinal changes of DTI indices. In another words, for both dose cohort, each of the DTI indices was modeled as a quadratic function of time postradiation with the formula, y = a·× 2+b·×+c, where y = DTI indices, x = weeks postradiation, and a, b, c modeled the coefficients of the quadratic function. It is noteworthy that the coefficients of the quadratic function of time postradiation were calculated for both dose cohorts, and this was accomplished simultaneously in one model by including the interaction terms mentioned above. The vertex of the trend line for quadratic function between DTI indices and weeks postradiation was calculated as −b/2a. The steepness of the parabola, or the sharpness at the vertex was determined by the coefficient a, and the larger its absolute value, the steeper the parabola. Based on the vertex of the trend line, the “turn point” (vertex) of the DTI indices could be determined. Finally, Pearson's correlation test was used to evaluate correlations between injury/control DTI indices and injury/control histologic staining intensity. All statistical analyses were performed using the statistical package SPSS for Windows. A P value of <0.05 was considered to indicate statistical significance.
General Results and Neurologic Findings
One rat in the 30 Gy cohort was found to have abnormal signs of gait and movement from 40 weeks postradiation. No gross neurologic abnormalities were observed in other rats.
Longitudinal Trend of DTI Indices and Dose Effect
Of all DTI slices (n = 3,400), 98.6% (n = 3,352) of the slices were satisfactory in image quality and were included into ROI analysis. Forty-eight slices belonging to nine rats were excluded due to motion artifacts, as determined visually. None of the rats were excluded from image analysis. FA, trace, λ//, and λ⊥ of the EC in both dose cohorts are shown in Table 1 . Mixed model was successfully applied for evaluating the longitudinal changes of the DTI indices, and Fig. 2 shows the trend lines of the DTI indices as predicted by the model.
FA. There was a significant decrease in FA in the injury EC compared with the control EC, of earlier onset in the 30 Gy cohort (at 2 weeks) compared with the 25 Gy cohort (at 4 weeks), and this continued at all the following time points ( Fig. 2A). However, the differences did not reach statistical significance. Trend line for FAI/FAC was described as downward parabola in both radiation dose cohorts, and this was significant quadratically with time postradiation (P = 0.019). The vertex was around 24 weeks for the 25 Gy cohort and 16 weeks for the 30 Gy cohort and the rate of decline in FA I/FA C became distinctly steeper from 16 weeks for the 30 Gy cohort. The steepness of the parabola of FA I/FA C for the 25 Gy and 30 Gy dose cohort was −4.59 × 10−5 and −4.61 × 10−5, respectively, and there was no significant interaction between the trend of FA I/FA C and dose (P = 0.996). The maximum difference between injury and control EC was at the final 48-week time point in both 25 Gy cohort and 30 Gy cohort, with a reduction of 6% ± 4% and 13% ± 3%, respectively.
Trace. No significant differences were found in trace in both dose cohorts at the first 2 week time point, but this became significantly reduced in the injury EC compared with the control EC from 4 to 36 weeks in the 30 Gy cohort, and only at 36 weeks and 40 weeks in the 25 Gy cohort ( Fig. 2B). The reversal in trend showed a steeper return to normal in the 30 Gy cohort compared with the 25 Gy cohort. Trace was increased in the final 48-week time point in the 30 Gy cohort, but this was not statistically significant. Longitudinal trend of trace I/trace C was described as upward parabola in both dose cohorts with a decline followed by increase, and the trend was significant quadratically with time post radiation (P < 0.001). The vertex was around 36 weeks for the 25 Gy cohort and 24 weeks for the 30 Gy cohort. The steepness of the parabola of trace I/trace C for the 25 Gy and 30 Gy dose cohort was 3.84 × 10−5 and 9.76 × 10−5, respectively, and there was no significant interaction between the trend of trace I/trace C and dose (P = 0.157).
λ//. This was significantly reduced in the injury EC compared with the control EC in both dose cohorts at all time points except the final 48-week time point ( Fig. 2C). Although the degree of reduction was larger in the 30 Gy cohort compared with the 25 Gy cohort from 2 to 36 weeks, this difference did not achieve statistical significance. Longitudinal trend of λ//I/λ// c was described as upward parabola in both dose cohorts with a decline followed by increase, and the trend was significant quadratically with time postradiation (P = 0.009). The vertex was around 36 weeks for the 25 Gy cohort and 24 weeks for the 30 Gy cohort such that the reduction in λ// became progressively less after these time points. The difference between injury EC and control EC was no longer significant for both cohorts at the final 48-week time point. The steepness of the parabola of λ//I/λ// c for the 25 Gy and 30 Gy dose cohort was statistically significant at 8.00 × 10−6 and 1.19 × 10−4 (P = 0.023), respectively, indicating a significant interaction between trend of λ// I/λ// C and dose.
λ⊥. There were no significant differences in the λ⊥of the EC in both dose cohorts at all time points except at the final 48-week time point and for the 30 Gy cohort only, where λ⊥ was significantly increased in the injury EC compared with the control EC (10% ± 2%, P = 0.038; Fig. 2D). Longitudinal trend of the λ⊥ I/λ⊥C was described as upward parabola with vertex around 24 weeks for the 25 Gy cohort and 16 weeks for the 30 Gy cohort, and the trend was significant quadratically with time postradiation (P < 0.001). The steepness of the parabola of λ⊥ I/λ⊥ C for the 25 Gy and the 30 Gy dose cohort was 8.39 × 10−5 and 1.22 × 10−4, respectively, and there was no significant interaction between trend of λ⊥ I/λ⊥ C and dose (P = 0.418).
Histologic evaluation of radiation-induced WM damage. Histologic changes of EC are illustrated in the Fig. 3 . Quantitative analysis of staining intensity of GFAP, LFB, and NF in injury and control sides of EC are shown in the Table 2 .
H&E. In the 25 Gy cohort, necrosis was not observed till 48 weeks (n = 1/3). A slight change of vascular dilation was observed at 16 weeks (n = 1/2) and this effect was more obvious at 36 weeks (n = 1/5) and 48 weeks (n = 2/3). Vacuolation changes in the EC were only observed at 48 weeks (n = 1/3). In the 30 Gy cohort, necrosis, vascular dilation and WM vacuolation were observed from 16 weeks (n = 1/2). Vessels around the necrotic region showed markedly dilated vessels with thickened walls. These effects were more marked at the subsequent time points (n = 2/4 at 24 weeks and n = 2/6 at 36 weeks), and finally at 48 weeks when all 3 rats were observed to have severe necrosis and vacuolation ( Fig. 3A and B). Cerebral swelling was not observed in all time points.
GFAP. In both dose cohorts, GFAP-immunoreactive cells with hypertrophic cytoplasm and process were observed in the injury EC compared with the control EC and the GFAP-immunoreactive cells were markedly increased in the injury EC at 4 weeks ( Fig. 3C and D), 16 weeks, and 24 weeks for all rats. Staining intensity of GFAP was significantly increased in the injury EC compared with control EC at all these time points ( Table 2). At 36 weeks, astrocytes in the injury EC were observed with hypertrophy in all rats but staining intensity was not significantly increased in both dose cohorts ( Table 2). Finally, at 48 weeks, GFAP-immunoreactive cells were not significantly increased and disappeared in the area of necrotic core.
LFB. In the 25 Gy cohort, staining intensity was similar between both sides of EC from 4 to 36 weeks ( Table 2). At 48 weeks, weaker LFB staining intensity was found in one of 3 rats in the injury EC. There were no significant differences in LFB staining intensity at these time points ( Table 2). In the 30 Gy cohort, decreased LFB intensity was observed at the injury EC at 16 and 36 weeks (n = 1/2 at 16 weeks and n = 2/4 at 36 weeks). Finally, at 48 weeks, all 3 rats showed much weaker LFB staining and severe disrupted and thinner myelin fibers in the injury EC especially in the area near the necrotic core ( Fig. 3E and F). Quantitative analysis confirmed significantly decreased staining intensity in the injury EC compared with control EC at the 48-week time point in the 30 Gy cohort only (P < 0.001; Table 2).
NF. In both dose cohorts, NF staining intensity and distribution were similar between both sides of EC from 4 to 36 weeks, and at 48 weeks, there was much weaker NF staining intensity in the injury EC (n = 2/3, 25 Gy; n = 3/3, 30 Gy; Fig. 3G and H). Quantitative analysis of staining intensity confirmed significantly decreased NF intensity at this last time point ( Table 2).
Correlation between injury/control DTI indices and injury/control histologic staining intensity. Using the Pearson's correlation test, the strength of a linear relationship between the injury/control DTI indices and injury/control staining intensity of histologic variables was evaluated ( Table 3 ). GFAPI/GFAPC intensity was significantly correlated with λ// I/λ// C (r = −0.451; P = 0.01) and FA I/FAC (r = −0.482; P = 0.005). LFB I/LFB C intensity was significantly correlated with λ⊥ I/λ⊥ C (r = −0.493; P = 0.004) and FA I/FAC (r = 0.382; P = 0.031). There was no significant correlation between NF I/NF C staining intensity and DTI indices.
To the best of our knowledge, this is the first study to evaluate the longitudinal correlation between DTI variables and histology of radiation-induced WM damage. This translational model of late WM injury resulting from whole brain irradiation showed a reduction in WM FA in keeping with previously reported findings in children ( 9, 10), and the reduction was found to be progressive at the later time points. In addition, we found that reduction in FA was driven mainly by reduction in λ// in the earlier phase and an increase in λ⊥ and return to baseline in λ// in the late phase. Moreover, directional diffusivities were correlated with radiation-induced pathologic changes of reactive astrogliosis (reduced λ//) in the earlier phase and demyelination (increased λ⊥) at the late time point. With regards to dose effect, our results showed that the higher radiation dose (30 Gy) induced earlier and more severe histologic changes in the WM than the lower dose (25 Gy), and these differences could be reflected by the magnitude of change in DTI variables to some extent.
Glial hypothesis and vascular hypothesis are considered two possible mechanisms of radiation-induced WM damage ( 20). Glial hypothesis suggests that progressive demyelination and necrosis in WM is related to a gradual loss of oligodendrocytes or their precursors postradiation ( 20). Kurita and colleagues ( 21) found rapid apoptotic depletion of oligodendroglial population in WM within 24 hours postradiation. The vascular hypothesis considers damage to the vasculature, causing increased permeability, blood-brain barrier disruption, and edema as a key step in the development of WM necrosis ( 22). Endothelial cell density was found to be reduced within 24 hours postradiation, with later development of pathologic alterations in blood vessels ( 23).
Previous studies using animal models have shown dynamic histopathologic changes with time in radiation-induced WM damage ( 18, 24– 27). Initial acute changes were characterized as lymphocyte infiltration, reduction of vascular endothelium and oligodendrocyte and their progenitors ( 24). In the subacute phase, there was vascular dilation ( 25) and reactive astrogliosis ( 25, 27). Some studies have found that reactive astrogliosis is a predominant histologic change as early as 3 to 24 hours postradiation ( 28) until 24 weeks postradiation ( 18). This finding is in agreement with our results, which showed reactive astrogliosis in the earlier phase post radiation (4–24 weeks). In the late phase (48 weeks postradiation), there is vascular dilation or rarefaction ( 26), diffuse demyelination ( 18, 19), axonal degeneration ( 18, 19), and necrosis ( 18), similar to our results. Late onset demyelination is a finding characteristic of radiation-induced WM damage. In a similar animal model, Rhinhold and colleagues showed demyelination several months postradiation (52 weeks) and the 50% effect time for demyelination after 25 Gy was 35 weeks postradiation ( 18). In another study using 25-Gy radiation dose in rats, a significantly decreased myelin basic protein was observed 36 weeks postradiation ( 19). Consistent with this, delayed demyelination was also shown in the late time point (48 weeks) in our experiment.
DTI has been used in a broad spectrum of central nervous system (CNS) applications, especially in WM disease ( 9, 10, 13). This is because water diffusion in WM is highly sensitive to WM microstructural architecture including myelin sheath, axonal transportation, and direction of neural fibers ( 6– 8). Decreased FA has been shown in many types of CNS injury such as ischemia ( 29), dysmyelination ( 7, 8), and axonal damage ( 6, 30). In this regard, FA is not specific for the characterization of histologic changes of WM injury. As expected, significantly decreased FA was found in injured WM from 4 to 48 weeks postradiation and it correlated with both astrogliosis (GFAP) and demyelination (LFB).
Studies have suggested that the amplitudes and combinations of the eigen values reflected more specific information about histopathologic changes in WM. Radial diffusivity (λ⊥) seems to be modulated by myelin in WM; impaired myelin will increase λ⊥ due to increased water diffusion in the perpendicular direction. In studies using animal models of dysmyelination ( 8) and demyelination ( 7), a significantly increased λ⊥ was shown in the injured side of WM compared with control side. In our study, significantly decreased LFB staining intensity was not shown until 48 weeks postradiation (in the 30 Gy group) when concurrent significant increase in λ⊥ was found. We also found a significant correlation between λ⊥ and myelination as shown by LFB staining intensity, further supporting the notion that λ⊥ reflects myelination. Axial diffusivity (λ//) represents the water diffusion parallel to the axonal fibers. In a mouse model of retinal ischemia, Song and colleagues ( 31) found that a significantly decreased λ// with similar λ⊥ reflected axonal degeneration without demyelination in optic nerves. This phenomenon is proposed to be due to several reasons including disordered microtubule arrangement, cellular debris from the breakdown of axonal structure ( 6), impaired axonal transportation ( 32), and shift of water from the extracellular to intracellular space due to axonal injury–associated cellular swelling ( 33). Although λ// was found to be significantly decreased at the early time points in our study, we did not observe axonal degeneration indicating that another mechanism might play a key factor in changing λ//. Instead, we found predominantly reactive astrogliosis, also reflected in the increased GFAP staining, and this correlated significantly with λ//. At the late time point, there was reduction in astrogliosis (and GFAP staining) as necrosis set in and this was paralleled by normalization of λ//. Reactive astrogliosis occurs as a response to diverse neurologic insults, and this has been shown after various insults to the CNS ( 34), including WM damage induced by radiation ( 18, 28). Reactive astrogliosis is generally characterized by astrocytic proliferation and extensive cellular hypertrophy. To date, information is scarce in the literature about how reactive astrogliosis affects MRI signal or DTI indices, but there is evidence from animal experiments measuring the extracellular space in brain injury models that diffusion is reduced in astrogliosis ( 35, 36). It was proposed that reactive astrogliosis imposes diffusion barriers in the CNS due to hypertrophy of astrocytic processes, and an increased production of extracellular matrix components are causes of restricted diffusion. This was further evaluated in a diffusion-weighted MRI study that found reduced apparent diffusion coefficient at the injury site ( 37). However, in another study using a mouse model with dysmyelination, Harsan and colleagues ( 38) found that astrocytic hypertrophy was manifested as amplified magnitudes of λ// and λ⊥ contradictory to our findings. They hypothesized that cytoplasmic processes of astrocytes wrapped unmyelinated axons or bundles of fibers along the axonal pathway longitudinally and might be the reason for increased diffusion. The key difference between their model and our model is that both reactive astrogliosis and dysmyelination was occurring concurrently in their model, and these processes may interact to affect the directionality of diffusion. We found no significant change in λ⊥ during this period of reactive astrogliosis (from 4–24 weeks postradiation), and this may be due to preservation of the myelin sheath and intact axonal membrane, which are the main factors restricting water diffusion in the perpendicular direction.
Variation of trace is mainly due to the dynamic distribution of water between the extracellular space and intracellular space or cellular density. Vasogenic edema increases the free water in the extracellular space and increases trace, whereas cytotoxic edema or cell proliferation restricts diffusion and decreases trace. In our results, decreased trace was found in the injured EC at earlier time points, in particular for the higher dose cohort. It is may be due to cytotoxic edema from glial hypertrophy and increased numbers of reactive glial cells, which decrease the free diffusion of water. At the late time points, increased trace in the injured WM might reflect destruction of WM structure due to necrosis and demyelination.
Radiation-induced brain damage is well-recognized as dose and time dependent ( 17, 25). Although our results showed no significant differences in the reduction of DTI indices between both radiation dose cohorts, there was a trend of earlier and more severe change in the DTI indices in the higher dose cohort. Also, the rates of longitudinal changes in DTI indices between the two dose cohorts were different with steeper curves in the higher dose cohort and a significant interaction was observed between λ// and dose. In the 30 Gy cohort, there was earlier significant reduction in FA and trace, and finally, a significant increase in λ⊥ was found only in this higher dose cohort. Thus, our results support the known risk factor that radiation-induced WM damage is dose dependent, and suggest that DTI indices are able to reflect this.
In conclusion, combining the analysis of DTI indices of FA, trace, λ// and λ⊥, provided specific information about histopathologic changes of radiation-induced WM damage and our results support the use of DTI as a biomarker to noninvasively monitor radiation-induced WM damage. This experimental model may be used to assess the neurotoxic adverse effects of irradiation treatment and to test the effectiveness of potential neuroprotective therapies.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: University of Hong Kong Committee on Research and Conference grants (HKU7587/06M).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We are grateful for assistance of staff at the laboratory of the Department of Pediatric and Adolescent Medicine, The University of Hong Kong.
- Received July 11, 2008.
- Revision received October 19, 2008.
- Accepted November 3, 2008.
- ©2009 American Association for Cancer Research.