Voxel-based mapping of grey matter volume and glucose metabolism profiles in amyotrophic lateral sclerosis

Participants were 37 patients with ALS (21 men and 16 women; mean age = 61.18 years, standard deviation = 11.11; mean level of education = 10.03 years, standard deviation = 2.96) with either bulbar (n = 10) or spinal (n = 27) onset. Patients with ALS were recruited between November 2009 and June 2014 via the ALS clinic of the neurology department of Caen University Hospital (France). All participants were native French speakers and had a minimum level of education equivalent to a now obsolete school-leaving certificate that was generally taken in 14 years. All the patients met the modified El Escorial criteria for probable or definite ALS [21]. Exclusion criteria were the additional presence of severe and chronic illness, alcohol or drug abuse, traumatic brain injury or an extremely severe communication problem that could seriously compromise the administration of cognitive tests. None of the patients fulfilled the criteria for a diagnosis of FTD according to the core and supportive diagnostic features of FTD detailed in Lund and Manchester’s consensus statement [22]. None of the patients with ALS met the criteria for a possible and/or probable behavioural variant of FTD [23]. Behavioural disorders were explored in 32 patients via the short form of the Neuropsychiatric Inventory [24]. None had behavioural disorders of the frontal type sufficiently severe to meet the criteria for ALS with behavioural impairment. All the patients underwent a neurological assessment that included the ALS Functional Rating Scale Revised (ALS-FRS-R) [25], Norris scale [26] and Medical Research Council (MRC, 1976) Muscle Strength Scale. They were able to speak or to write intelligibly. All the patients gave their written informed consent, and the independent regional ethics committee approved the study. Some of these patients had already been included in a previous study [5]. The ALS patients underwent a neuropsychological assessment covering a range of cognitive functions (for details, see [5]).

For each participant, a high-resolution T1-weighted anatomical image was acquired on a Philips Achieva 3 T scanner using a three-dimensional fast field echo sequence (sagittal; repetition time = 20 ms, echo time = 4.6 ms, flip angle = 10°, 180 slices, slice thickness = 1 mm, field of view = 256 × 256 mm², matrix = 256 × 256).

Each participant underwent a PET examination the day after MRI examination. FDG scans were acquired on a Discovery RX VCT 64 PET-CT device (GE Healthcare) with a resolution of 3.76 × 3.76 × 4.9 mm³ FWHM (axial field of view = 157 mm). Forty-seven planes were obtained with septa out (3D acquisition), with a voxel size of 2.7 × 2.7 × 3.27 mm³. A CT transmission scan was performed for attenuation correction before PET acquisition. Participants were fasted for at least 6 h before scanning. After a 30-min resting period in a quiet and dark environment, ~180 MBq of FDG were intravenously injected as a bolus. A 10-min PET acquisition scan began 50 min after injection. During data acquisition, head motion was continuously monitored with laser beams projected onto ink marks drawn on the forehead, which also served to make any necessary corrections.

Using the VBM5.1 toolbox (http://​dbm.​neuro.​uni-jena.​de/​vbm/​vbm5-for-spm5/​), implemented in SPM5 software (Wellcome Trust Centre for Neuroimaging, London, UK), the raw MRI data were spatially normalized to Montreal Neurological Institute (MNI) space (voxel size = 1 mm³, matrix = 156 × 189 × 157) and segmented into the GM, white matter (WM) and cerebrospinal fluid (CSF). The normalized GM images were modulated by the Jacobian determinants to correct for nonlinear warping only, to obtain maps of local GM volumes corrected for brain size.

FDG-PET data were first corrected for CSF and WM PVEs, using the voxel-by-voxel modified Müller-Gartner method [29, 30], described in detail elsewhere [31], and already used in our laboratory [18‐20]. Using SPM5, the PVE-corrected PET data were then coregistered (rigid-body coregistration) to their respective native MRIs and normalized to MNI space by reapplying the normalization parameters estimated from the VBM protocol described above. The resulting images underwent quantitative scaling using mean GM as a reference.

For the between-group comparison, and in order to blur individual variations in gyral anatomy and increase the signal-to-noise ratio, the registered MRI data were smoothed using an 8 × 8× 8 (x, y, z)-mm Gaussian kernel. PET datasets were also smoothed, using a 12 × 12 × 12-mm Gaussian kernel.

For the direct comparison between GM atrophy and glucose hypometabolism, since the MRI and PET data had different spatial resolutions, differential smoothing was applied to equalize the effective smoothness [32‐34]. A Gaussian kernel of 8 × 8 × 8 (x, y, z) mm was used for the MRI data and 7.1 × 7.1 × 6.4 for the PET data. Finally, images were masked to exclude non-GM voxels from the analysis.

To obtain measurements of atrophy and hypometabolism expressed in the same units so that we could undertake direct comparisons of the different modalities, we computed W score maps for each patient and each imaging modality, using the healthy control group as a reference. W scores are analogous to z scores but are adjusted for specific covariates [35]—age in the present case. The smoothed MRI and PET data were used to create W score maps [(raw value for each patient) − (value expected in the control group for the patient’s age)/(SD for each patient and each modality); see [18] for more details].

As the two datasets had different original spatial resolutions, the MRI data were coregistered and resliced into PET space to obtain images with the same dimensions.

GM anatomical localization was carried out using AAL automated labelling software implemented in SPM5 [36].

To determine the pattern of GM atrophy in the ALS group as well as glucose hypometabolism and hypermetabolism, compared to the control group, we used a two-sample t test in SPM5. This yielded three maps of statistically significant GM atrophy and metabolic abnormalities in patients relative to controls. To address the issue of multiple comparisons, for each analysis, the significant cluster size was determined using a Monte Carlo simulation program (AlphaSim) as previously done in our laboratory [37, 38], with a cumulative proportion criterion of less than 0.05 [39]. For the GM volume analysis, this equated to a cluster volume (=k) of 475 voxels and for the metabolism analyses a k = 113 voxels.

Individual GM MRI W score maps and PET W scores maps were entered in a one-paired t test analysis with one group (patients with ALS) and two images per participant (i.e. MRI and PET W score maps), using SPM5. We put years of education and Mattis (total score) as confounding covariates in this analysis. Both contrasts were assessed (W-PET > W-MRI and W-MRI > W-PET) to generate statistical maps reflecting predominant structural abnormalities or glucose hypometabolism. To address the issue of multiple comparisons, significant cluster size was determined using a Monte Carlo simulation program (AlphaSim), with a cumulative proportion criterion of less than 0.05 [39]. This equated to a cluster volume of 31 voxels.

Atrophy was greater than hypometabolism in two main regions: the temporal lobe (anterior, lateral and medial) and the calcarine sulcus. Based on previous studies carried out in Alzheimer’s disease, frontotemporal dementia or chronic alcoholism [18, 20, 63, 64], it is generally thought that loss of brain cells other than neurons might result in atrophy without associated hypometabolism. In ALS, an increasing number of studies have suggested that astrocytosis and/or microglial activation could be involved in the pathophysiology of the disease [65, 66].

The finding of a normal, i.e. higher than expected, level of glucose metabolism in atrophic cerebral zones may also suggest that compensatory mechanisms are at work in these structures, helping to maintain a moderately high metabolic level relative to structural alterations—a hypothesis previously developed for Alzheimer’s disease [18]. In this disease, according to the authors, the presence of abnormally phosphorylated tau proteins that aggregates to form neurofibrillary tangles may be one of the processes underlying severe atrophy and moderate hypometabolism in the hippocampus. In ALS, TDP-43 inclusions are located in neurons and astrocytes of ALS patients, not only in the motor regions but also in the temporal lobe [67]. Following the same reasoning, the presence of TDP-43 could lead to a massive neuronal loss, whereas moderate hypometabolism could be explained by a compensation of the remaining neurons.

We found that hypometabolism was more severe than atrophy in the left superior medial frontal cortex. This result suggests that ALS is characterized by genuine functional alterations (metabolic, chemical or molecular) on top of the neuronal loss, heightening the functional consequences of local GM atrophy. Supporting the notion of metabolic alterations in ALS, a study used the benzodiazepine GABA_A marker ¹¹C-flumazenil to study brain dysfunction in 17 patients [58]. Flumazenil is an antagonist at the benzodiazepine subunit of the GABA_A receptor. These authors found reduced binding of ¹¹C-flumazenil in the dorsomedial prefrontal cortex notably. This may reflect the downregulation of postsynaptic GABA_A receptor expression [58]. Given that glucose metabolism reflects synaptic activity [68], the detrimental effects of ALS on neurotransmission systems could explain why metabolic dysfunction precedes GM atrophy (e.g. within the frontal lobe). Greater hypometabolism than GM atrophy has also been observed in the behavioural variant of FTD [20], suggesting that this could be a remote effect of GM atrophy on metabolism, or a diaschisis. Diaschisis refers to a change in the metabolic activity of neurons that are anatomically or functionally connected to a damaged area. The medial prefrontal cortex is connected to limbic structures such as the medial temporal lobe and putamen [69]. The neuronal loss reflected by GM atrophy within the temporal lobe and putamen (see above) may remotely affect the metabolism of the medial prefrontal cortex, though possibly only for a limited period of time, as disconnected neurons eventually die, giving rise to different patterns of brain volume and metabolic impairment.

The study has some limitations. The major one is the existence of two groups of healthy subjects. Indeed, as our protocol did not include the neuroimaging examinations for healthy subjects, we used the scans of healthy subjects of another protocol of our laboratory. Then, the threshold that we used in this article is relatively liberal even if it has been employed in several MRI [45, 49, 70‐75] or PET studies [14, 16] in ALS. However, with a more stringent threshold, we could have missed some interesting findings. Finally, this study should be replicated in a bigger group of ALS patients.