Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease which is primarily characterized by the degeneration of upper and lower motor neurons [1, 2]. While frontotemporal pathology is often observed [3], sensory neuropathy, dorsal root ganglion pathology and the degeneration of brain structures involved in sensory processing are not recognised as core features of the disease. Nonetheless, sensory involvement has been observed clinically [4, 5, 6, 7], and demonstrated by electrophysiology [4, 8, 9, 10, 11], neuroimaging [9, 12, 13] and neuropathology studies [4, 14, 15, 16, 17]. Sensory dysfunction is thought to be more common in familial ALS than in sporadic ALS [1, 18], and sensory disturbances have also been observed in other motor neuron disease phenotypes [19, 20, 21, 22]. The literature of sensory dysfunction in ALS is conflicting; most studies report negative findings [23, 24, 25], but few studies have evaluated sensory regions specifically [9]. Clinically, subtle paraesthesia [5, 26, 27, 28], numbness [2], pain [29], vestibular symptoms [30], proprioceptive deficits [31], changes in olfaction [32, 33, 34] and taste [35, 36] may be overshadowed by progressive limb weakness and bulbar dysfunction [37]. It is noteworthy, that Riluzole, the first FDA approved for the treatment of ALS has been associated with mild circumoral paresthesia [38] but this is rarely a cause for discontinuation. Sensory changes have also been observed in the larynx on fibre-optic endoscopic evaluation of swallowing with sensory testing (FEEST) [39, 40].

The substrate of sensory symptoms is most commonly studied by electrophysiology or post mortem, and it is seldom specifically evaluated on neuroimaging [41]. Alterations in sensory nerve action potentials (SNAP) [11, 42], somatosensory evoked potentials (SEP) [8, 43, 44, 45, 46], visual evoked potentials (VEP) [47, 48], auditory evoked potentials [8, 48] and sensory action potential amplitudes (SAPA) [49] have been sporadically described. There are inconsistent reports on the impairments of thermal sensory pathways, but study protocols differ considerably [24, 50]. Overt visual deficits are rarely reported by patients, but impaired visuospatial function has been consistently demonstrated in ALS [51, 52]. Microscopically, dorsal root ganglion, dorsal column, thalamus, sensory cortex and sensory nerve degeneration has also been observed [25, 53, 54]. Occipital lobe pathology has been noted in both familial and sporadic ALS patients [55] and occipital pTDP-43 pathology has also been described [56]. On neuroimaging, dorsal column degeneration [9, 57, 58], thalamic pathology [59, 60] and postcentral gyrus [61, 62] atrophy have consistently been reported. Structural [63], diffusion [64], functional [65], susceptibility [66], and MR spectroscopy [67] have all captured some degree of parietal and occipital lobe involvement. Despite these reports, there is a prevailing notion that ALS spares sensory networks [68], and no dedicated cerebral studies have systematically investigated the cortical, subcortical and white matter components of sensory pathways.

The primary objective of this study is the systematic evaluation of cortical, subcortical and white matter structures involved in the processing, relaying and mediation of sensory information. Accordingly, a comprehensive panel of cerebral structures was defined associated with specific somatosensory, visual and auditory processes (Table 1). Some of the selected regions such as the postcentral gyrus, the ventral posterolateral (VPL), ventromedial (VM), lateral geniculate (LGN) and medial geniculate (MGN) nuclei of the thalamus and the medial lemniscus are either exclusively or primarily involved in sensory functions. Other regions included in the panel, such as the insular cortex, supramarginal gyrus, superior temporal gyrus, and the posterior thalamic radiation mediate a combination of sensory and non-sensory functions. We hypothesised that cerebral structures involved in somatosensory processes are not spared in ALS. Direct clinico-radiological correlations were not sought, as peripheral and spinal components of sensory pathways were not evaluated in this study [9, 69].

A total of 287 participants, 160 patients with ALS (‘ALS’) 127 healthy controls (‘HC’) were included in a prospective, single-centre imaging study. All participants provided informed consent as required by the Ethics Approval of this research project (Beaumont Hospital, Dublin, Ireland). Exclusion criteria included a history of traumatic brain injury, cerebrovascular events, neoplastic, paraneoplastic and neuroinflammatory diagnoses. A total of 7 patients were excluded after enrolment because of incidental radiological findings. Patients with ALS were diagnosed according to the revised El Escorial criteria. Key demographic and clinical variables (ALSFRS-r) were recorded for all participants at the time of their scan and a dedicated sensory assessment was performed in a subset of C9orf72 negative patients (n = 45). The cohort of patients who underwent sensory assessment had no cervical or lumbar spondylosis, carpal tunnel syndrome, type I diabetes or pharmacotherapy associated with sensory symptoms. Methods for genetic screening for hexanucleotide repeat expansions in C9orf72 have been described previously [70]. Repeat-primed PCR was used and expansions longer than 30 repeats were considered pathological.

Forty-five patients with ALS were systematically assessed for clinical signs of sensory dysfunction. Nociception was assessed using a ‘neurotip’ on the face, upper limbs and lower limbs bilaterally. Vibrioception and proprioception were assessed in the upper and lower limbs bilaterally. For nociception, vibrioception and proprioception, a score of ‘2’ indicated normal sensation in the relevant body region, ‘1’ indicated reduced sensation, and ‘0’ indicated sensory loss in the investigated sensory domain. Graphaestesia, the ability to recognise symbols traced on the skin, and stereognosis, the ability to perceive the form of solid objects by touch, were assessed in the upper limbs bilaterally. For graphaestesia and stereognosis, the patient was given a score of ‘1’ for each correct answer, and ‘0’ for each incorrect answer. A composite score of 84 indicated no sensory deficits and lower scores corresponded with more severe sensory deficits.

Additionally, a brief 7-item questionnaire was administered to screen for subjective symptoms pertaining to (1) upper limb paraesthesia, (2) lower limb paraesthesia, (3) cold sensitivity, (4) perioral paraesthesia, (5) tongue numbness, (6) changes in taste and (7) deterioration in sense of smell. Each item was scored 0–2, ‘0’ meaning no symptoms in the relevant sensory domain.

Images were acquired on a 3 Tesla Philips Achieva Magnetic resonance (MR) platform as part of a standardised neuroimaging protocol. T1-weighted (T1w) images were acquired with a 3D Inversion Recovery prepared Spoiled Gradient Recalled echo (IR-SPGR) sequence with the following key parameters: field-of-view (FOV) of 256 $\times$ 256 $\times$ 160 mm, flip angle = 8${}^{\circ }$, spatial resolution of 1 mm${}^3$, SENSE factor = 1.5, TR/TE = 8.5/3.9 ms, TI = 1060 ms. Diffusion tensor images (DTI) data were acquired with a spin-echo echo planar imaging (SE-EPI) pulse sequence with a 32-direction Stejskal-Tanner diffusion encoding scheme; FOV = 245 $\times$ 245 $\times$ 150 mm, 60 slices with no interslice gap, spatial resolution = 2.5 mm${}^3$, TR/TE = 7639/59 ms, SENSE factor = 2.5, b-values = 0, 1100 s/mm${}^2$, dynamic stabilisation and spectral presaturation with inversion recovery (SPIR) fat suppression. Fluid-attenuated inversion recovery (FLAIR) images were acquired with an Inversion Recovery Turbo Spin Echo (IR-TSE) sequence to screen for comorbid inflammatory or vascular FLAIR data were acquired in axial orientation: FOV = 230 $\times$ 183 $\times$ 150 mm, spatial resolution = 0.65 $\times$ 0.87 $\times$ 4 mm, 30 slices with 1 mm gap, TR/TE = 11,000/125 ms, TI = 2800 ms, 120${}^{\circ }$ refocusing pulse, with flow compensation and motion smoothing and a saturation slab covering the neck region. Clinical sequences (T1w and FLAIR) were radiologically reviewed for incidental findings and datasets from all participants underwent quality control review for movement artifacts before computational analyses.

The standard anatomical reconstruction pipeline of the FreeSurfer image analysis suite [71], was implemented, including non-parametric non-uniform intensity normalization, affine registration to the MNI305 atlas, intensity normalization, skull striping, automatic subcortical segmentation, linear volumetric registration, neck removal, tessellation of the grey matter-white matter boundary, surface smoothing, inflation to minimize metric distortion, and automated topology correction [72]. The relevant cortical labels of the Desikan-Killiany atlas were utilized to retrieve average cortical thickness and cortical volumes values from the postcentral gyrus, insula, superior temporal, transverse temporal, supramarginal, and lateral occipital cortices in the left and right cerebral hemispheres separately. The thalamus was parcellated into 25 sub-regions by Bayesian inference using a probabilistic atlas developed based on histological data [73]. The volumes of the following thalamic nuclei were evaluated in the two hemispheres separately: ventral posterolateral (VPL), ventromedial (VM), lateral posterior (LP), lateral geniculate (LGN), medial geniculate (MGN). The amygdala was segmented into 9 sub-regions using a probabilistic atlas developed based on histological [73, 74]. The volumes of the lateral (LN), medial (MN), and cortical nuclei (CN) were estimated in each hemispheres separately.

Raw diffusion tensor data were eddy current corrected, skull stripped, and a tensor model was fitted in FMRIB’s software library to create maps of axial diffusivity (AD), fractional anisotropy (FA), mean diffusivity (MD), and radial diffusivity (RD). FSL’s tract-based statistics (TBSS) module was implemented for non-linear registration, skeletonisation and the creation of a mean FA mask. Diffusivity values were retrieved from the four (AD, FA, MD, RD) concatenated diffusivity files using the medial lemniscus and posterior thalamic radiation labels of the ICBM-DTI-81 white-matter atlas [75].

Assumptions of normality were examined using the Kolmogorov-Smirnov test. Skewness and kurtosis were assessed separately for each study group. Since all variables followed a normal distribution, parametric statistics were applied. The age and ALSFRS-r profile of the study groups were compared multivariate analysis of variance (MANOVA). Sex, symptom onset and handedness proportions in the study groups were contrasted using Pearson’s chi-squared test. White matter variables and cortical thickness measures were contrasted using multivariate analysis of covariance (MANCOVA) with age and sex and covariates. Volumetric profiles (cortical, thalamic nuclei, amygdalar nuclei) of the study groups were also appraised in MANCOVAs with the following covariates, age, sex and total intracranial volume (TIV). Bonferroni corrections were used to account for multiple comparisons and reduce Type I error. Output statistics were summarised in comprehensive tables to report estimated marginal means (EMM), standard error (SE), p-values and post hoc pairwise comparisons. To further illustrate focal pathology, radar plots were generated for the two patients groups where EMMs were expressed as the percentage of reference (HC) EMM in the given anatomical region.

The study groups were matched for age, sex and handedness. Patient groups were matched for site of onset and functional disability (Table 2). Forty-five patients were specifically evaluated for the presence of sensory signs and symptoms, and all of these patients tested negative for the C9orf72 gene.

On clinical examination, vibrioception was the most commonly affected sensory domain detected in 80.0% of the patients, subtle nociceptive deficits were identified in 57.8%, and proprioceptive deficits were ascertained in 28.9% of patients. Graphaesthesia was abnormal in 48.9% and stereognosis was abnormal in 28.9% of the 45 patients who had a dedicated sensory assessment at the time of their scan. Subtle facial sensory deficits were identified in 11.1% of patients. On the sensory symptom questionnaire, 68.9% reported at least one sensory symptom, and only 31.1% patients reported having no sensory symptoms at all. 33.3% had paraesthesia in the hands/fingers, 20% had paraesthesia in the feet/toes and 24.4% of patients reported cold sensitivity in hands/fingers. Tingling around the lips was reported in 8.9% and tongue numbness in 11.1%. 24.4% reported changes in taste and 17.8% of patients reported some change in smell (Table 3).

C9- ALS patients exhibited cortical volume reductions in the insula bilaterally, and the left supramarginal and right superior temporal gyrus compared to healthy controls (Table 4). Each cortical region associated with sensory processing showed atrophic change in C9+ ALS bilaterally in comparison to both healthy controls and to C9- ALS. C9- ALS patients showed cortical thickness reductions in all evaluated regions except the lateral occipital region bilaterally and the right postcentral cortex. Cortical thickness changes were observed in all regions bilaterally in C9+ ALS in contrast to healthy controls, and in all regions except the transverse temporal bilaterally and the right insula when compared to C9- ALS (Fig. 1).

Fig. 1.

Cortical thickness profile in C9orf72 positive ALS patients (C9+ ALS), C9orf72 negative ALS patients (C9- ALS), healthy controls (HC), based on estimated marginal means adjusted for age and sex. Error bars represent 95% confidence intervals. Left denoted by (L) and right denoted by (R). Significant inter-group differences corrected for multiple comparisons are highlighted with asterisks: * p value between 0.01 and 0.05, ** p value between 0.001 and 0.01, *** p value 0.001.

In C9- ALS, volumes of all thalamic nuclei involved in sensory processing were reduced except the lateral geniculate nucleus (LGN) bilaterally and the left lateral posterior nucleus (Table 5). Sensory nuclei volumes were also reduced in C9+ ALS in comparison to healthy controls bilaterally except in the left medial geniculate nucleus. In the amygdalae of C9- ALS patients, volumetric reduction was observed in the right lateral and right cortical nuclei in comparison to healthy controls. Volumes reductions were also observed in the amygdalae of C9+ ALS patients in comparison to healthy controls in the lateral and cortical nuclei bilaterally and the left medial nucleus. When C9+ ALS patients are contrasted with C9- ALS patients, the volume of the left lateral nucleus was observed to be significantly reduced.

Significant inter-group AD and MD differences were observed in the right medial lemniscus and post hoc comparisons revealed AD differences between C9+ ALS and C9- ALS patients (p = 0.038). A trend of FA reduction was observed in C9- ALS compared to healthy controls in the left medial lemniscus (p = 0.054) (Table 6). FA reductions and increased RD were observed in the bilateral posterior thalamic radiation in both C9+ and C9- ALS in comparison to healthy controls. Increased posterior thalamic radiation MD was also detected in C9- ALS in contrast to healthy controls in the right hemisphere (Fig. 2).

Fig. 2.

Radar plots of integrity measures expressed as % of estimated marginal means of controls (green) in regions of interest associated with sensory processing. * indicates significant intergroup differences (p value 0.05).