Routine MRI sequences do not reliably show abnormalities of the cortical white matter that distinguish patients with primary lateral sclerosis or patients with ALS from healthy individuals. Diffusion tensor imaging, an MRI technique that utilizes the diffusion properties of water molecules, can be used to estimate the overall orientation and integrity of nerve fibres in the white matter. Studies using diffusion tensor imaging in patients with ALS have found reduced fractional anisotropy, indicative of axonal disruption, at various levels of the corticospinal tract (Ellis et al., 1999; Graham et al., 2004; Aoki et al., 2005; Mitsumoto et al., 2007; Iwata et al., 2008). In patients with ALS, reduced fractional anisotropy correlated with differential slowing of central motor conduction in the supra- and infratentorial segments of the corticospinal tract (Iwata et al., 2008). Thus, reduced fractional anisotropy in white matter tracts may be a marker for changes in the physiological properties of axons as well as the structure of white matter tracts.

In the majority of diffusion tensor imaging studies that have been carried out in patients with ALS, diffusion properties of the corticospinal tract were measured within circumscribed regions of interest drawn within anatomical structures known to contain the corticospinal tract at one or a few levels, for example in the posterior limb of the internal capsule. One drawback of visual placement of regions of interest is that the position of the corticospinal fibres varies among individuals (Holodny et al., 2005). The accuracy of positioning a region of interest selectively on the fibres arising from the motor cortex is likely to be further diminished in patients with corticospinal tract degeneration. Diffusion tensor tractography, also known as fibre tracking, allows for the localization of white matter tracts connecting defined regions from diffusion tensor images (Mori et al., 1999) and enables tract-specific measurements of diffusion properties to be made from the reconstructed white matter tracts. We and others have shown that diffusion properties of the corticospinal tract and corpus callosum can be measured reliably and reproducibly in healthy subjects using diffusion tensor tractography (Wakana et al., 2007; Danielian et al., 2010). A disadvantage of diffusion tensor tractography is that it only assesses the selected tracts. Tract-based spatial statistics (TBSS) offers a complementary method for analysing white matter in diffusion tensor images that does not pre-specify tracts or regions of interest. TBSS carries out voxelwise statistical comparisons between co-registered and normalized white matter skeletons of two study populations, providing a non-biased, whole-brain analysis of white matter differences between two groups (Smith et al., 2006).

Diffusion properties normally vary along the length of the corticospinal tract, reflecting differences in the compactness and organization of corticospinal axons as they traverse different structures, splaying out, for example, in the corona radiata and becoming tightly packed in the internal capsule (Reich et al., 2006; Wong et al., 2007). The typical rostrocaudal pattern of diffusion properties was preserved in our patients. Regional differences between patients and controls in the corticospinal tract diffusion properties reflect the disease-related alterations in the tissue microstructure. Experimental studies have shown that fractional anisotropy, a measure of the coherence of directional diffusion, is mainly determined by the organized arrangement of axonal membranes, with a smaller contribution from myelin and relatively little from the intra-axonal neurofibrils; measures of mean diffusivity are affected by intra- and extracellular elements that restrict diffusion of water molecules (Beaulieu, 2002). In patients with ALS, the loss of fractional anisotropy distally, detected in the TBSS analysis, indicates disruption of corticospinal axon integrity. The distal gradient of increasing mean diffusivity and decreasing cross-sectional area below the peduncles in the tract profiles, are compatible with an underlying dying back axonopathy. In patients with primary lateral sclerosis, the combination of reduced fractional anisotropy, increased mean diffusivity, and volume loss is more suggestive of tissue loss, with loss of corticospinal axons and expansion of extracellular space.

One interpretation of the differences between primary lateral sclerosis and patients with ALS in the diffusion properties of the corticospinal tract is that these changes represent the early and late stages in a temporal sequence of structural changes that occurs with axonal breakdown and clearance. However, time alone, i.e. disease duration, is not a sufficient explanation. Previous studies in patients with ALS have reported that corticospinal tract fractional anisotropy declined with disease severity, duration or progression (Ellis et al., 1999; Jacob et al., 2003; Cosottini et al., 2005; Iwata et al., 2008; Nickerson et al., 2009; Roccatagliata et al., 2009; Agosta et al., 2010a). In most studies, the ALSFRS-R score was used to quantify disease severity and to calculate disease progression. In this study, the ALSFRS-R was not significantly correlated with diffusion property measurements. Since the ALSFRS-R measures functions that depend on both lower and upper motor neurons, a drop in the ALSFRS-R is not specific for upper motor neuron dysfunction. We confirmed the correlation between reduced corticospinal tract fractional anisotropy and clinical severity with a scale measuring upper motor neuron dysfunction in patients with ALS. But the relationship between corticospinal tract fractional anisotropy and disease duration differed from previous reports: patients with ALS with longer disease durations exhibited less reduction of corticospinal tract fractional anisotropy than patients with ALS with shorter disease durations. This seeming contradiction may be explained by noting that, in this study, the patients with ALS with shorter disease durations also had a higher upper motor neuron rapidity index, and thus a faster rate of progression. Our ALS cohort may have included more patients with slower progression than in other studies, since our focus was on patients with a comparable severity of upper motor neuron dysfunction to patients with primary lateral sclerosis, and patients with ALS with predominantly upper motor neuron signs are known to have a slower rate of progression (Gordon et al., 2009). An extrapolation of the trend for lesser reduction of corticospinal tract fractional anisotropy in patients with ALS with slower progression could lead to similar findings to what was observed in patients with primary lateral sclerosis, whose corticospinal tract fractional anisotropy values were reduced, intermediate between controls and patients with ALS, and were stable with time. Our data support the proposal that the reduction in corticospinal tract fractional anisotropy is a marker for the rate of disease progression, rather than the severity or the duration of disease (Agosta et al., 2010a, b).

Clinical ratings of upper motor neuron dysfunction correlated with reduced corticospinal tract fractional anisotropy in patients with ALS, and with increased corticospinal tract mean diffusivity in patients with primary lateral sclerosis in both diffusion tensor tractography and TBSS analyses. Increased corticospinal tract mean diffusivity in patients with ALS was detected in the fibre tracking analysis, with somewhat more prominence in the distal portions of the corticospinal tract profile. Increased mean diffusivity seems to correspond to regions of white matter areas with more chronic changes, as would be compatible with an evolution of diffusion properties as degeneration progresses. Longitudinal imaging studies and pathological correlation will be critical to refine interpretations of imaging findings in ALS and primary lateral sclerosis. Pathological correlations are needed to determine whether fluctuations in fractional anisotropy correspond to degeneration or loss of axons and crossing fibres, and whether mean diffusivity reflects tissue loss and gliosis.

Changes in the diffusion properties of the corpus callosum were observed in ALS and patients with primary lateral sclerosis by fibre tracking and TBSS analyses. The changes occurred in the mid-posterior portion of the corpus callosum, an area containing fibres originating from the sensory and motor cortex, leaving the genu and splenium unaffected. This localization within the callosum is consistent with images presented in other studies (Bartels et al., 2008; Ciccarelli et al., 2009; Filippini et al., 2010; Unrath et al., 2010), but differs from the callosal regions affected in dementia syndromes (Zhang et al., 2007; Zhuang et al., 2010) and hereditary spastic paraparesis (Unrath et al., 2010). Of note is that none of the patients in this study was demented. The changes in the corpus callosum signify that the degeneration is not selective for corticospinal neurons in ALS and primary lateral sclerosis. Neurons whose axons form the corpus callosum are distinct from those that form the corticospinal tract. Although some corticospinal neurons transiently extend collaterals through the corpus callosum during development, by adulthood, neurons forming the corpus callosum are separate from corticospinal neuron population (Catsman-Berrevoets et al., 1980; Koester and O'Leary, 1993). This differentiation is driven by expression of neuronal subtype specific genes (Molyneaux et al., 2007, 2009). In primates, the pyramidal neurons that form the corpus callosum primarily reside in cortical layer IIIb, whereas corticospinal neurons reside in cortical layer V (Jones et al., 1979). Although callosal neurons are excitatory pyramidal neurons, their major targets are inhibitory interneurons in the contralateral hemisphere (Li and Pleasure, 2011). Callosal activation using transcranial stimulation produces contralateral inhibition of the motor cortex (Avanzino et al., 2007). In ALS, transcallosal inhibition is impaired at an early stage of the disease, before the appearance of upper motor neuron signs (Wittstock et al., 2007). Loss of transcallosal inhibition is likely to underlie the clinical phenomenon of mirror movements observed in patients with ALS, which have been shown to correlate with reduced fractional anisotropy in the corpus callosum in diffusion tensor imaging images (Bartels et al., 2008). Mirror movements were not systematically assessed in this study. Reduction in callosal fractional anisotropy may thus be an early marker for motor neuron disease. Reductions in fractional anisotropy values of the corpus callosum and corticospinal tract were correlated. The concurrent, or earlier, impairment of the shorter callosal axons argues against a strictly length-dependent susceptibility to degeneration. However, concurrent degeneration would be compatible with the hypothesis that degeneration in motor neuron disease begins focally in a region of the brain and spreads to contiguous or homotopic regions (Ravits and La Spada, 2009).

In this study, diffusion tensor tractography, or fibre tracking, was used to define the white matter tracts selectively for quantitation. The major findings from the fibre tracking analysis were confirmed by TBSS, a whole-brain method for analysing white matter tracts. Both methods identified altered diffusion properties of the corticospinal tract and the motor portion of the corpus callosum, as well as correlations of the clinical severity of upper motor neuron impairment with corticospinal tract fractional anisotropy in ALS and with corticospinal tract mean diffusivity in primary lateral sclerosis. However, fibre tracking and TBSS yielded slightly different results. The discrepancies were largely failures of detection, rather than contradictory findings. Diffusion tensor tractography detected increased corticospinal tract mean diffusivity in ALS that was not seen by TBSS. TBSS found correlations between small clusters of voxels with reduced fractional anisotropy in the subcortical white matter and the upper motor neuron impairment score in primary lateral sclerosis that were not detected by fibre tracking. The differing assumptions and computational procedures of the two methods account for these differences. Each method has strengths and weaknesses. Diffusion tensor tractography has the advantage that the white matter tracts visualized are known to connect defined brain regions in individual subjects, whereas the identification of tracts in TBSS is based on images registered to a standardized atlas. Given individual differences in the localization of corticospinal axons, their identification is less precise for TBSS than for diffusion tensor tractography. On the other hand, fibre tracking algorithms only connect contiguous voxels that have fractional anisotropy values above a threshold, creating a bias towards relatively intact axonal tracts. Fibre tracking can truncate prematurely in regions where fibres become less compact or cross fibres with an orthogonal orientation, whereas TBSS is able to detect non-contiguous clusters of voxels within a tract. TBSS is at a disadvantage for evaluating diffusion measures other than fractional anisotropy values. TBSS analyses a thinned fractional anisotropy skeleton that consists of a core of voxels having the highest fractional anisotropy (Smith et al., 2006), thus excluding the full cross-section of tracts that are more variable between individuals. Thus, TBSS will be relatively insensitive to changes in mean diffusivity that occur in voxels outside the core with the highest fractional anisotropy. Fibre tracking and TBSS provide complimentary methods for evaluating white matter, and the limitations of each method should be taken into account for interpretation of differing results.