Ex vivo Evaluation of ADC Values within Spinal Cord White Matter Tracts

Methods

Region-of-Interest Selection

For each of the normal animals, two consecutive T2-weighted images (b = 0 images) were selected for region-of-interest analysis according to the morphology of the cervical cord. A specific degree of protuberance of lamina IV at the midcervical level was used as a landmark so that the selected images appeared as similar as possible, ensuring a consistent location for analysis between animals. Regions of interest were then drawn on the T2-weighted images corresponding to the dorsal cortical spinal tract (dCST), fasciculus gracilis (FG), fasciculus cuneatus (FC), rubrospinal tract (RST), vestibulospinal tract (VST), and reticulospinal tract (ReST) (Fig 1A) on the basis of current knowledge of rat cervical spinal cytoarchitecture (16–19). On the selected images, regions of interest for each of the six tracts were drawn bilaterally for a total of 12 regions per image. For each animal, the values for two consecutive sections were then averaged. Care was taken not to include any pixels from surrounding buffer. In addition to calculating tADC and lADC values for each tract, the anisotropy index (AI), which is defined as tADC/lADC, was computed. These values from all the tracts were averaged to obtain an overall tADC, lADC, and AI in the white matter for each fixative (Table 1).

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

MR image and corresponding histologic sections from normal rat spinal cord. A, Schematic diagram of major ascending and descending tracts in the rat spinal cord overlaid on a b = 0 (T2-weighted) image of a rat spinal cord. Regions of interest were drawn in the following tracts: rubrospinal tract (RST), reticulospinal tract (ReST), vestibulospinal (VST), fasciculus gracilis (FG), fascisulus cuneatus (FC), and dorsal corticospinal tract (dCST). While the dorsal white matter tracts are homogeneous, the ventral and lateral tracts are mixed with other fibers from the STT-spinothalamic tract and dSCT-dorsal spinocerebellar tract. A significant difference between the human and rat spinal cord is the location of the corticospinal tract. In the human, the corticospinal tract is located in the lateral white matter, while in the rat it is located in the dorsal white matter.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

B–G, 1-μm-thick histologic sections at 40×, stained with toluidine blue of the RST (B), ReST (C), VST (D), FG (E), FC (F), and dCST (G), respectively (black bar = 50 μ). Note that the largest axons appear to be within the VST.

View this table:

• View inline
• View popup

TABLE 1:

Overall transverse ADC, longitudinal ADC and anisotropy indices for PF and GL-PF specimens

To correct for temperature differences, an additional region of interest was placed in the surrounding buffer; the ADC values of the regions of interest from each section were then multiplied by the ratio of the known ADC of water at 20°C (2.04 × 10⁻³ mm²/s) to ADC(buffer). Additionally, by using histogram analysis, pixels with ADC values less than zero were excluded as noise.

Results

Discussion

CNS white matter demonstrates anisotropic water diffusion when evaluated with DW imaging (2, 3, 25). It has been proposed that anisotropic water diffusion in white matter tracts results from myelin sheaths and axonal cellular membranes functioning as barriers to water diffusion in the transverse direction (perpendicular to the long axis of the axon), while allowing relatively unrestricted diffusion in the longitudinal direction (parallel to the long axis of the axon). Alterations in axonal membrane and myelin integrity would therefore be expected to decrease anisotropic water diffusion. Reduced anisotropic water diffusion within white matter following trauma (1, 4, 5, 26–28) or demyelination (29, 30) has been demonstrated both clinically and experimentally, and it has been proposed that measures of anisotropy may be helpful in evaluating the severity (or extent) of white matter injury. However, measurements of anisotropic water diffusion are based on individual directional diffusivities and may be affected by changes in either or both of the directional diffusivities that are measured. There is evidence that each directional diffusivity may reflect a distinct pathologic process, and thus, it may be more helpful to evaluate the individual water diffusivities, as opposed to a general measure of anisotropic water diffusion.

Myelin Sheath Preservation and Transverse Water Diffusion

Evaluations of water diffusivity in the spinal cords of mice with genetic myelin deficiencies have shown that the absence of a myelin sheath increases transverse water diffusion in white matter tracts as compared with transverse diffusion in the axons of normally myelinated mice, but not to such an extent as to eliminate anisotropic water diffusion (31, 32). The preservation of anisotropic diffusion confirms that myelination is not necessary for anisotropic water diffusion in white matter (33). On the basis of these findings, one may expect lower tADC values in animals fixed with GL-PF, assuming that stronger fixation of the myelin sheath results in a more restrictive barrier to transverse water diffusion. Our findings, however, indicate that better preservation of myelin with GL-PF fixation does not affect tADC values. One explanation for our findings is provided by Takahashi et al (34), who noted that increasing myelination in developing nerves did not result in increased tADC values and suggested that increasing myelination past a certain point did not increase restricted diffusion with the MR imaging parameters they used. Although fixation with PF at physiological pH results in greater splitting of the myelin sheath than that of fixation with GL-PF (14), tADC values may not change significantly without actual loss of myelin.

In our comparison of individual spinal cord tracts, we found that the VST had greater tADC values than that of the other tracts. Although the etiology of this difference is unknown, Takahashi et al (33) reported increased tADC values in spinal cord tracts with larger axons, and they suggest that increased axonal diameter and decreased axonal density may play a role in increasing tADC.

Longitudinal Water Diffusion in Axons

The superior preservation of intra-axonal cytoskeletal structures in specimens fixed with GL-PF, as compared with PF only, may explain the differences in ADC values seen in our study, as intra-axonal water diffusion could be dependent on structural integrity of neurofilaments and microtubules. Beaulieu and Allen (35) evaluated intra-axonal ADC values in the giant axon of the squid and demonstrated isotropic water diffusion, a finding confirmed by Takahashi et al (33) in the larval lamprey. Therefore, the longitudinal orientation of the neurofilaments and microtubules does not appear to be responsible for anisotropic water diffusion (35). Beaulieu and Allen (36) also studied anisotropic water diffusion in garfish nerves following the application of vinblastine, which acts to depolymerize microtubules. The administration of vinblastine for 48 hours decreased longitudinal water diffusion within the nerves, but anisotropic water diffusion was preserved. In addition, these authors observed a reduction in longitudinal ADC values in degenerated nerves in vitro. These findings suggest that the breakdown of axoplasmic structures results in decreased water diffusivity. Our findings support this hypothesis, because the PF specimens, with greater breakdown of axoplasmic structures, had lower lADC values than those of GL-PF specimens. Beaulieu and Allen also noted a decrease in transverse water diffusivity in those nerves treated with vinblastine. We also noted a decrease in tADC values in those spinal cords fixed with PF only, perhaps suggesting that the integrity of cytoskeletal proteins may play a role in transverse water diffusion; however, the difference was minimal and not statistically significant.

The density, as well as the integrity, of cytoskeletal proteins may affect longitudinal water diffusion. Kinoshita et al (37) evaluated anisotropic water diffusion in the optic nerve of rats following methylmercury administration, which decreases the overall density of microtubules and neurofilaments within axons, without affecting the myelin sheath. They demonstrated increased longitudinal water diffusion in the injured optic nerves and inferred that the presence of neurofilaments and microtubules restricts longitudinal water diffusion. Their findings may partially explain our results that show increased lADC values in the VST, which contains larger axons as compared with the dCST and ReST. Larger axons have more neurofilaments; however, the interfilament spacing increases and a lower density of neurofilaments results (38, 39). Additionally, microtubules have been shown to decrease in density in larger axons (40, 41). Thus, it may be the decreased density of cytoskeletal proteins seen in larger axons that explains the increased lADC values within the VST.

Studies in rodents with genetic myelin deficiencies have produced somewhat discordant results concerning longitudinal water diffusion. Decreased longitudinal water diffusion was seen in nerves of the twitcher mouse, which is a model of Krabbe disease where demyelination and degeneration of axons are observed (42). The authors suggest that waviness of the demyelinated axons may result in decreased longitudinal water diffusion; however, axonal degeneration and subsequent disorganization of cytoskeletal proteins may also contribute to decreased water diffusion. Song et al (32) noted that diffusion tensor (DT) imaging of the shiverer mouse spinal cord showed no decrease in longitudinal water diffusion despite decreased axonal caliber and increased axonal microtubule and neurofilament density (43, 44). Gulani et al (31) showed increased longitudinal water diffusivity in DT imaging findings of myelin-deficient rats as compared with diffusivity in normal controls. Therefore, axonal diameter and cytoskeletal protein status may not be the only factors influencing longitudinal water diffusion, especially in the presence of other pathologic processes such as demyelination.

Calculations of Anisotropy

In this study of normal spinal cord white matter, we have shown that differences in fixation result in changes in longitudinal water diffusion, although transverse water diffusion is not affected. It is, therefore, not surprising that there is a subsequent increase in anisotropic water diffusion. Such a change in anisotropy could be produced either by an increase in longitudinal, or a decrease in transverse, diffusion coefficients. Thus, it may be more important to evaluate specific directional diffusivities, as each may reflect distinct pathologic processes.

Our findings suggest that the organization and density of axoplasmic cytoskeletal proteins, such as neurofilaments and microtubules, are important, although not the sole determinants, in influencing longitudinal water diffusion. Therefore, it may be possible to detect disease that disrupts axoplasmic structures without disrupting the myelin sheath or cellular membrane integrity. Conversely, it may be possible to detect disruption of the myelin sheath that does not alter axoplasmic proteins.

The importance of evaluating individual directional diffusivities is supported in a study of water diffusivity in a model of optic nerve ischemia in which Song et al (45) observed that decreased longitudinal water diffusion occurred earlier than increases in transverse water diffusion, reflecting axonal degeneration, whereas increases in transverse diffusion occurred later with loss of myelin. Interestingly, relative anisotropy did not appear to significantly change when only increased longitudinal diffusion was seen, again supporting the greater utility of evaluating individual directional diffusivities rather than a calculated measure of anisotropy.

Research in animal models of spinal cord injury also demonstrates the utility of evaluating individual directional diffusivities. In a rodent model of syringomyelia, DW imaging depicted cyst formation earlier than did conventional MR imaging owing to increases in both longitudinal and transverse ADC values, and the data indicate no changes in the ratio of transverse to longitudinal ADC values (46). Also, in an evaluation of spinal cord transplant function, decreasing transverse and longitudinal ADC values within fibroblast transplants reflect increased axonal in-growth, without demonstrating anisotropic water diffusion (1).

Methodological Issues

A possible confound to our comparisons of fixatives could be differences in the degree of tissue shrinkage, which are observed on histologic sections. Shrinkage of fixed tissue, however, occurs mostly in the processing required for sectioning following perfusion fixation. The perfusion fixation with either PF or GL-PF does not result in CNS tissue shrinkage of the gross specimen before histologic evaluation (47). Because our imaging is performed before histologic preparation and sectioning, there should be no differences in tissue volume between the GL-PF and PF fixed specimens.

We imaged the differently fixed specimens with slightly different imaging parameters owing to our observation (unpublished data) that the T2 relaxation time of GL-PF specimens is shorter than the T2 relaxation time of PF specimens. Shorter TE and shorter diffusion times for GL-PF specimens were required to increase signal-to-noise ratio and allowed differentiation between gray matter and white matter. The difference in diffusion times should not adversely affect our comparison of tADC values, because the diffusion times employed indicate that the values should be on the asymptotic curve for restricted water diffusion with both sets of parameters (48). Our lADC values also should not be affected, because both sets of parameters have diffusion times short enough to permit essentially unrestricted water diffusion with minimal water exchange (49).

A comparison between live and formalin-fixed brains showed no differences in anisotropy ratios, supporting the usefulness of optimizing data analysis on fixed tissue before imaging live subjects (50). We have shown, however, that there are significant differences in lADC and AI between specimens fixed with PF or GL-PF, without changes in tADC. Therefore, care must be used in directly extrapolating specimen data to live tissue without taking into account fixative type. Additionally, separate controls are required in ex vivo experiments if different fixation methods are used.

To better characterize the histologic correlates of each directional diffusivity, we are developing an automated method of segmenting high-resolution histologic images, and obtaining data for parameters such as axonal diameter, axonal density, myelin sheath thickness, axonal spacing, and volume fractions of myelin, intracellular space and extracellular space. These variables can then be correlated with ADC values (51). A noninvasive method can then be developed to determine functionally important pathohistologic parameters in the injured CNS white matter.

Conclusion

Our study with differently fixed normal spinal cord specimens indicates that better preservation of cytoskeletal structures, such as the longitudinally oriented neurofilaments, results in increased lADC values. Despite the expected splitting of the myelin sheath in PF fixed specimens, the tADC values were the same as those in GL-PF specimens, suggesting that actual myelin loss may be necessary to decrease tADC values.