The present study was focused on the evaluation of brain metabolic changes that differentiate ALS patients with PB and PS clinical involvement. The main finding was a relative hypometabolism in motor cortices in PB compared to PS patients, closely overlapping with the somatotopic representation of bulbar functions in the motor homunculus described by Penfield and Boldrey. Both PB and PS showed extensive hypometabolism as compared to HC, involving extramotor regions, in agreement with previous findings from our centre [
6]. Our results are consistent with the hypothesis that ALS starts as a focal process [
2] and its spread gives rise to the heterogeneity of motor phenotype. Some neuroimaging studies supported the concept of disease focality in vivo. An MRI study based on voxel-based morphometry showed that bulbar onset patients were characterized by bilateral focal atrophy in the bulbar segment of the homunculus of motor cortex, compared with patients with spinal onset, who displayed focal cortical changes in the limb segment of their motor strip [
3]. Similarly, the site of onset (bulbar/spinal) resulted to be associated with the thinning of the corresponding part of the primary motor cortex in other MRI studies [
4,
5]. Previous brain 2-[
18F]FDG-PET works comparing bulbar and spinal onset patients showed divergent findings. An investigation from our centre including 32 patients identified large relatively hypometabolic clusters in bulbar onset patients as compared with spinal onset ones in prefrontal and frontal cortices [
7]. A subsequent study based on the enlargement of the previous dataset (
n = 195 ALS subjects) reported a relative hypometabolic cluster in the left motor and premotor cortex in bulbar onset patients as compared to spinal onset ones [
6]. A work performed in a different population did not detect any difference between patients with bulbar and spinal onset [
8]. In this context, a study aiming at testing the diagnostic accuracy of 2-[
18F]FDG-PET in discriminating spinal and bulbar onset ALS patients showed poor performance [
9]. Nevertheless, in this study a number of spinal onset patients had bulbar impairment and vice versa at the time of PET. Focusing on patients without functional evidence of the spread of the disease from spinal to bulbar regions and the other way round, in the present study, we demonstrated that the metabolic clusters including the regions representing bulbar function in the motor strip can discriminate between pure bulbar and pure spinal patients with an error rate < 10%. Namely, PB patients showed a relative hypometabolism as compared to PS ones in such regions. This finding seems in agreement with a study investigating disease spreading in ALS through the assessment of clinical evidence of upper and lower motor neuron signs [
15]. The authors suggested that when the disease had a limb onset, bulbar neurons were more resistant to be involved as compared with other spinal regions. Moreover, neurophysiological data pointed towards a more relevant cortical dysfunction in bulbar onset ALS compared to limb onset disease, as expressed by hyperexcitability, possibly associated with loss of GABAergic inhibitory interneurons [
16]. Notably, a recent MRI study employed VBM and tract-based spatial statistics (TBSS) to evaluate patterns of grey and white matter changes in bulbar and limb onset ALS. The authors suggested that bulbar onset ALS probably originates from the orofacial segments of the primary motor cortex, and that limb onset ALS patients have fewer cerebral structural integrity loss in the early phase of the disease, possibly due to a lower degree of neuronal network integration of the first pathologically affected regions [
17]. The lack of clusters of relative hypometabolism in PS subjects compared to PB ones is challenging to interpret. Similarly to our results, a recent MRI study [
5] found a significant cortical thinning of the bulbar regions of the motor strip in bulbar onset patients compared to spinal onset ones. Conversely, they did not detect any area of reduced thickness in the motor cortex of patients with spinal onset ALS compared to cases with bulbar onset. A possible explanation of these findings might be the prevalence of a dying-back mechanism in PS patients, leading to a relatively higher damage of lower compared to upper motor neurons. Despite some studies supporting this hypothesis [
18], it remains speculative. Further data are necessary to clarify this issue and appropriately define the possible role of brain 2-[
18F]FDG-PET as a biomarker in clinical trials. Taken together, our findings and literature data suggest that bulbar onset ALS is associated with an increased vulnerability of the bulbar regions of the motor cortex. The high discriminant value of the metabolism of the clusters identified in the present study provides some hints for further research. First, it supports the concept of the focal onset of the disease, whose phenotypic heterogeneity could be modulated by other factors. Second, it underlines that neuroimaging techniques can be implemented with other biomarkers to stratify patients in clinical trials towards a precision medicine approach. It is worth noting that the adoption of a high explainability (white box) artificial intelligence approach allowed to derive mechanistically relevant information from the model.
Our study has some limitations. First, to classify patients as PB or PS, we used the ALSFRS-R score, which is a functional scale and cannot disclose smooth clinical alterations detectable with a full neurological examination. Second, we lack a measure of upper motor neuron burden, allowing further analysis to highlight motor cortex dysfunction in PB patients. Third, our data are cross-sectional, while longitudinal data are necessary to place the focality of the onset in the context of the spread of the disease. Fourth, we did not perform partial volume effect correction for cortical atrophy, because we did not have MRI scans of the whole sample. However, previous studies employing voxel-based atrophy correction of resting glucose metabolism showed that metabolic measurements were relatively independent of brain atrophy [
19].
In conclusion, by using cross-sectionally a single biomarker, we found clusters of relative hypometabolism in bilateral motor cortex in PB compared to PS patients, closely overlapping with the somatotopic representation of bulbar functions in the motor homunculus. The metabolism of such regions showed very high capability to discriminate between PB and PS patients. Our data provide in vivo support for the concept of the focality of ALS onset and strengthen the idea that 2-[18F]FDG-PET can play a role as a biomarker for precision medicine oriented clinical trials.