Brainstem pathology in amyotrophic lateral sclerosis and primary lateral sclerosis: A longitudinal neuroimaging study

Highlights

• •Computational neuroimaging captures focal brainstem pathology in motor neuron diseases in contrast to both healthy- and disease controls.
• •ALS patients exhibit progressive medulla oblongata, pontine and mesencephalic volume loss over time.
• •Brainstem atrophy in ALS and PLS is dominated by medulla oblongata volume reductions.
• •Vertex analyses of ALS patients reveal flattening of the medullary pyramids bilaterally.
• •Morphometric analyses in ALS detect density reductions in the mesencephalic crura consistent with corticospinal tract degeneration.

Abstract

Background

Brainstem pathology is a hallmark feature of ALS, yet most imaging studies focus on cortical grey matter alterations and internal capsule white matter pathology. Brainstem imaging in ALS provides a unique opportunity to appraise descending motor tract degeneration and bulbar lower motor neuron involvement.

Methods

A prospective longitudinal imaging study has been undertaken with 100 patients with ALS, 33 patients with PLS, 30 patients with FTD and 100 healthy controls. Volumetric, vertex and morphometric analyses were conducted correcting for demographic factors to characterise disease-specific patterns of brainstem pathology. Using a Bayesian segmentation algorithm, the brainstem was segmented into the medulla, pons and mesencephalon to measure regional volume reductions, shape analyses were performed to ascertain the atrophy profile of each study group and region-of-interest morphometry was used to evaluate focal density alterations.

Results

ALS and PLS patients exhibit considerable brainstem atrophy compared to both disease- and healthy controls. Volume reductions in ALS and PLS are dominated by medulla oblongata pathology, but pontine atrophy can also be detected. In ALS, vertex analyses confirm the flattening of the medullary pyramids bilaterally in comparison to healthy controls and widespread pontine shape deformations in contrast to PLS. The ALS cohort exhibit bilateral density reductions in the mesencephalic crura in contrast to healthy controls, central pontine atrophy compared to disease controls, peri-aqueduct mesencephalic and posterior pontine changes in comparison to PLS patients.

Conclus

ions: Computational brainstem imaging captures the degeneration of both white and grey matter components in ALS. Our longitudinal data indicate progressive brainstem atrophy over time, underlining the biomarker potential of quantitative brainstem measures in ALS. At a time when a multitude of clinical trials are underway worldwide, there is an unprecedented need for accurate biomarkers to monitor disease progression and detect response to therapy. Brainstem imaging is a promising addition to candidate biomarkers of ALS and PLS.

1. Introduction

Motor neuron diseases (MND) encompass a range of clinically heterogeneous conditions which typically affect young adults, are relentlessly progressive and have no effective disease-modifying therapies (Chipika et al., 2019). Some of the barriers to the development of effective pharmacological interventions include the considerable clinical heterogeneity of the condition, lack of validated biomarkers and late recruitment into pharmacological trials (Mitsumoto et al., 2014; Schuster et al., 2015; Hardiman et al., 2016). The majority of imaging studies in MND focus on cortical grey matter atrophy (Omer et al., 2017; Bede et al., 2013a), corticospinal tract degeneration (Schuster et al., 2016a), corpus callosum pathology (Filippini et al., 2010; Bede et al., 2015) and spinal cord degeneration (Bede et al., 2012; El Mendili et al., 2019; Lebouteux et al., 2014). With the recognition of extra-motor deficits in MND (Burke, 2017; Elamin et al., 2017; Burke et al., 2016; Christidi et al., 2018; Finegan et al., 2019a) the frontotemporal (Nasseroleslami et al., 2019; Iyer, 2017), basal ganglia (Feron et al., 2018; Christidi et al., 2019) and cerebellar (Abidi et al., 2019) profiles of MNDs have also been gradually characterised (Bede et al., 2018a), but there remains a striking lack of dedicated brainstem studies. Brainstem pathology is a hallmark feature of the disease and has been associated with the condition since its earliest descriptions (Clarke and Jackson, 1867). In one of the first pathologically supported reports, Lockhart Clarke eloquently describes progressive bulbar involvement in a patient with ALS: “September 1865; Her voice changed; she did not pronounce words as usual…Her deglutition now became difficult… The tongue is protruded badly; about a quarter of an inch beyond the lower teeth. It is atrophied on each side, and in folds, reminding one of cerebral convolutions. Her talking is nearly unintelligible; to me it is generally quite unintelligible, but the nurse can usually manage to make out what the patient says.” This moving description from over 150 years ago illustrates the bulbar symptoms in ALS which continue to affect patients today (Yunusova et al., 2019). Following his clinical description, Clarke comments on post mortem brainstem changes; “The medulla oblongata was below the average size. It was not, however, softer than usual, nor was it anywhere damaged by disintegration of tissue; but many of its nerve-cells, particularly about the floor of the fourth ventricle and calamus scriptorius, were in different stages of degeneration.” Despite historical descriptions of brainstem atrophy, it remains surprisingly understudied in vivo, in spite of considerable advances in ALS neuroimaging (Bede and Hardiman, 2014). Moreover, brainstem pathology is regarded as ‘stage 1’ of a recently proposed four-stage pathological staging system based on pathological TDP-43 burden patterns (Brettschneider et al., 2013). Furthermore, brainstem pathology is not unique to ALS, it is a unifying feature of most motor neuron diseases (ALS, PLS, SBMA) affecting the descending pyramidal tracts, brainstem nuclei or both, depending on the phenotype (Brettschneider et al., 2013; Querin et al., 2018a; Li et al., 1998; Finegan et al., 2019b; Geser et al., 2011). So, while brainstem degeneration is a pathognomonic feature of most MNDs, it is seldom evaluated specifically in dedicated imaging studies. This is a missed opportunity at a time when wet and dry biomarkers are urgently sought to inform clinical trial outcomes (Devos et al., 2019; Blasco et al., 2018). Accordingly, our main objective is the systematic characterisation of brainstem involvement in ALS and PLS, using both healthy- and disease controls in a prospective longitudinal study design. Based on the extensive post mortem literature of the disease (Brettschneider et al., 2013; Geser et al., 2011), we hypothesise that brainstem pathology can not only be detected in vivo, but it exhibits progressive changes over time.

2. Methods

2.1. Ethics statement

The study protocol, recruitment, and data management procedures were approved by our institutional ethics committee (Beaumont Hospital, Dublin, Ireland), in accordance with the 1964 Helsinki declaration and its later amendments. All participants provided informed consent and data was handled in accordance with the relevant EU regulations (GDPR).

2.2. Participants

A total of 100 ALS patients, 33 patients with PLS, 30 patients with FTD, and 100 healthy controls were included. Exclusion criteria included prior traumatic brain injury, prior cerebrovascular events, space-occupying intracranial findings, and inability to undergo MRI scanning due to implanted metallic devices (baclofen pumps, pacemakers), claustrophobia, or orthopnoea. Participating ALS patients had ‘probable’ or ‘definite’ ALS according to the El Escorial criteria (Brooks et al., 2000), PLS patients were diagnosed based on the Gordon criteria (Gordon et al., 2006) and FTD patients diagnosed in accordance with the Rascovsky criteria (Rascovsky et al., 2011). ALS patients were scanned twice with a follow-up interval of 4 months. Healthy controls had no known neurological or psychiatric conditions, previous head injuries or established vascular risk factors and no family history of neurodegenerative conditions.

2.3. Neuroimaging methods

T₁-weighted images were acquired on a 3 Tesla Philips Achieva system using an 8-channel receive-only head coil using a 3D Inversion Recovery prepared Spoiled Gradient Recalled echo (IR-SPGR) sequence with the following settings; spatial resolution: 1 mm³, field-of-view (FOV): 256 × 256 × 160 mm, TR/TE = 8.5/3.9 ms, TI = =1060 ms, flip angle = 8°, SENSE factor = 1.5, acquisition time: 7 min 30 s. A multimodal approach was undertaken to comprehensively characterise brainstem degeneration in MND. First, total intracranial volumes (TIV) were estimated and a model-based approach utilised to estimate the volume of the entire brainstem. Subsequently, a Bayesian segmentation algorithm was used to estimate the volumes of the medulla oblongata, pons, and midbrain. Brainstem atrophy patterns in ALS were evaluated with reference to healthy controls, disease controls and PLS patients using vertex analyses. Finally, region of interest morphometry analyses were carried out to evaluate focal brain stem alterations in the study groups.

2.4. TIV and total brainstem volumes

Total intracranial volume (TIV) was calculated for each study participant to be used as a covariate for subsequent volumetric comparisons. The FMRIB Software Library (FSL) was used for TIV estimations. Each subject's skull-stripped brain was linearly aligned to the standard MNI152 brain image, the inverse of the determinant of the affine registration matrix was calculated and multiplied by the size of the template. FSL-FLIRT was used for registration to template (Jenkinson and Smith, 2001), and tissue type segmentation was performed using FSL-FAST (Zhang et al., 2001). For the estimation of total brain stem volume the model-based approach of FSL-FIRST was implemented. As described previously (Bede et al., 2018b), FSL-FIRST (Patenaude et al., 2011) uses a two-stage affine registration approach to register raw T1 data sets to the Montreal Neurological Institute 152 (MNI152) standard space first and then a model-based method is implemented for the segmentation of subcortical structures.

2.5. Brainstem segmentation

A Bayesian segmentation algorithm was used for the segmentation of the brainstem into the medulla oblongata, pons and midbrain. The protocol relies on a probabilistic atlas of the brainstem and its neighbouring anatomical structures generated based on 49 scans (Iglesias et al., 2015). The FreeSurfer image analysis suite (Fischl, 2012) was used for the pre-processing of T1-wieghted data which included the removal of non-brain tissue, segmentation of the subcortical white matter and deep grey matter structures, intensity normalization, tessellation of the grey matter-white matter boundary, and automated topology correction. Subsequent to pre-processing and brainstem segmentation, raw volumetric values were retrieved for each study participant at each timepoint for statistical interpretation with the appropriate covariates.

2.6. Vertex analyses

In order to characterise anatomical patterns of degenerative change in the brainstem beyond overall volume reductions, shape analyses were performed between the relevant study groups. FMRIB's subcortical segmentation and registration tool FIRST (Patenaude et al., 2011) was used for vertex analyses to generate quantitative, surface-projected information about focal brainstem involvement. Vertex locations of individual participants were projected on the surface of an average template shape as scalar values, positive value being outside the surface and negative values inside. Permutation based non-parametric statistics were used for group comparisons, design matrices included age, gender and education (Winkler et al., 2014), probability maps were corrected for multiple comparisons, and a statistical threshold of p < .05 FWE was considered significant.

2.7. Region of interest morphometry

In order to detect focal density alterations within the brain stem, region-of-interest morphometry was performed to complement the vertex analyses. The FMRIB Software Library (FSL) was used for brain extraction and tissue-type segmentation, following which grey-matter partial volume images were aligned to the MNI152 standard space using affine registration. Subsequently, a study specific template was created, to which the grey matter images from each subject were non-linearly coregistered. A voxelwise generalized linear model and permutation-based non-parametric testing was used to highlight density alterations in a merged brainstem mask accounting for age, gender and education (Winkler et al., 2014; Nichols and Holmes, 2002). The labels of the Harvard-Oxford probabilistic structural atlas (Desikan et al., 2006) and the labels of the Talairach probabilistic (Lancaster et al., 2000) atlas were used to generate a merged brainstem mask incorporating the medulla oblongata, pons and midbrain.

2.8. Genetics

Of the 100 ALS patients, C9orf72 repeat expansion status was determined in 97 patients (Byrne et al., 2012). DNA samples from patients were tested for the presence of the pathogenic GGGGCC hexanucleotide repeat expansion in C9orf72 by repeat-primed PCR using the Applied Biosystems (Foster City, CA, USA) 3130xl Genetic Analyser and visualised using GeneMapper version 4.0 as described previously (Byrne et al., 2012; Bede et al., 2013b). Patients carrying more than 30 hexanucleotide repeats were considered positive for the expansion. Whole-genome sequence data were available for 44 ALS patients (Project Min, 2018) and targeted DNA sequence data for a further 27 ALS patients (Kenna et al., 2013). Whole exome sequence data were available for 29 of the 33 PLS patients (Finegan et al., 2019a). Putative variants were defined as protein altering variants in the exons and splice sites of 33 genes linked to ALS on the ALS online database (Abel et al., 2013) and 70 genes linked to HSP (Klebe et al., 2015). C9orf72 hexanucleotide repeat expansion status in the PLS cohort was also determined using repeat-primed polymerase chain reaction (PCR).

2.9. Statistics

Analyses of covariance (ANCOVA) were performed to evaluate intergroup differences in brainstem volumes. Assumptions of normality, linearity and homogeneity of variances were verified. Brainstem volumes were included as dependent variables, and study group allocation as the categorical independent variable. TIV, age, education, and gender were included as covariates (Bede et al., 2013c). Summary tables were generated to present the estimated marginal means of brainstem volumes, standard error, between-group ANCOVA significance and p-values for Bonferroni-corrected post hoc testing. Estimated marginal means of volumes were plotted with confidence intervals to illustrate the volumetric profile of each study group. To demonstrate the divergent subcortical profiles of ALS, PLS and FTD percentage volume reductions were calculated with reference to the estimated marginal means of healthy controls and plotted on a radar (spider) chart. Differences in longitudinal change between C9orf72 hexanucleotide carriers and those without the repeat expansions were evaluated with mixed effect models.

3. Results

Study participants were matched for age, gender, handedness and education. Despite the comparable demographic profile of the study groups, each statistical model of our imaging analyses included age, gender, education and total intracranial volumes as covariates Table 1.

Table 1.

The clinical and demographic profile of the study cohorts.

	ALS T1 n = 100	ALS T2 n = 100	PLS n = 33	FTD n = 30	HC n = 100	p value
Age	59.83 (11.155)	60.21 (11.235)	60.48 (10.488)	63.97 (14.571)	59.65 (11.876)	0.484
Gender (male)	62 (62%)	62 (62%)	19 (58%)	13 (43%)	54 (54%)	0.332
Education (years)	13.45 (3.176)	13.45 (3.176)	12.88 (3.380)	12.80 (3.210)	12.47 (3.383)	0.171
Handedness (right)	90 (90%)	90 (90%)	29 (88%)	27 (90%)	94 (94%)	.784
ALSFRS-r	36.62 (7.470)	35.40 (7.335)	34.36 (5.337)	N/a	N/a	0.23

3.1. Volumetric analyses

The volumetric analyses of brainstem regions revealed considerable multisegmental brainstem atrophy in both ALS and PLS with medullar predominance. ALS patients at their second timepoint exhibited statistically significant volume reductions in both the medulla and pons compared to healthy controls. Total brainstem volume reductions in ALS did not reach statistical significance with reference to controls, suggestive of selective medullar and pontine involvement. In the PLS cohort, brainstem atrophy was detectable in all three regions and medullar atrophy was not only apparent in comparison to healthy-, but also with reference to disease controls. PLS patients also had significantly lower medullar, pontine and mesencephalic volumes compared to ALS patients. Across the two time points ALS patients exhibited progressive volume loss in all three brainstem segments but longitudinal changes did not reach statistical significance. ALS patients showed a trend of medullar atrophy compared to healthy controls at their first time point which has reached significance by their second timepoint. Similarly, while pontine differences were not significant between healthy controls and ALS patients at the first time point, due to progressive pontine volume loss, ALS patients at the their second time point exhibited statistically significant pontine changes (Table 2 and Fig. 1).

Table 2.

Brainstem volumes in healthy controls (HC), patients with frontotemporal dementia (FTD), patients with amyotrophic lateral sclerosis (ALS) and primary lateral sclerosis (PLS). Estimated marginal means and standard error are adjusted for age, gender, education and total intracranial volume (TIV) (Covariates appearing in the model are evaluated at the following values: Age = 59.59, Gender = 1.43, Edu = 13.63, TIV = 1435355.28) Pairwise post hoc comparisons are Bonferroni corrected. Significant intergroup differences are flagged with asterisks ‘*’. Statistical trends are flagged with superscript ‘t’. ALS_T1: ALS patients first time point, ALS_T2: ALS patients second time point, FTD: patients with frontotemporal dementia, HC: healthy controls, PLS: patients with primary lateral sclerosis.

Structure	Study group	Estimated marginal means(adjusted for age, gender, education and TIV)	Standard error	ANCOVA Sig. (p)	Post-hoc comparisons (Bonferroni-corrected)
Medulla				<0.001	ALS T1 vs HCp = .061t ALS T1 vs PLSp = .002* ALS T2 vs HCp < .001* ALS T2 vs PLSp= = .057 t FTD vs PLSp = .001* HC vs PLSp < .001*
	HC	4808.641	4714.924
	ALS T1	4623.070	47.012
	ALS T2	4530.937	47.023
	FTD	4747.986	86.610
	PLS	4268.433	81.904
Pons				<0.001	ALS T1 vs PLSp = .003* ALS T2 vs HCp = .017* ALS T2 vs PLSp = .014* HC vs PLSp < .001*
	HC	15410.591	159.537
	ALS T1	14829.294	157.394
	ALS T2	14699.411	157.433
	FTD	14670.462	289.969
	PLS	13682.707	274.212
Midbrain				0.004	ALS T1 vs PLSp = .073 t HC vs PLSp = .02*
	HC	6173.757	49.524
	ALS T1	6071.860	48.859
	ALS T2	6002.527	48.871
	FTD	5994.261	90.014
	PLS	5807.431	85.122
Whole Brainstem				< 0.001	ALS T1 vs PLSp < .001* ALS T2 vs PLSp = .002* FTD vs PLSp = .058 t HC vs PLSp < .001*
	HC	22694.377	202.067
	ALS T1	22326.919	199.353
	ALS T2	22038.915	199.402
	FTD	21956.655	367.272
	PLS	20556.064	347.314

Fig. 1.

The volumetric profile of the brainstem in ALS, PLS, disease and healthy controls. Corners: Estimate marginal means of the medulla, pons, midbrain are plotted for each study group adjusted for age, gender, education and total intracranial volumes. Centre: Percentage volumetric change is plotted with reference to healthy controls. ALS_T1: ALS patients first time point, ALS_T2: ALS patients second time point, FTD: patients with frontotemporal dementia, HC: healthy controls, PLS: patients with primary lateral sclerosis.

3.2. Vertex analyses

Focal shape deformations were detected at the second timepoint of ALS patients compared to both disease- and healthy controls following the appropriate statistical corrections. Considerable symmetric medullary pyramid atrophy was detected in comparison to healthy controls. In contrast to disease controls, an asymmetric pattern of anterolateral medullary atrophy was detected which centred on the right pyramid but incorporated the right olivary nuclei. In contrast to PLS, ALS patients exhibited widespread anterior and posterior pontine shape deformations including atrophy of the floor of the fourth ventricle.

3.3. Brainstem morphometry

Morphometric analyses revealed density reductions in the ALS cohort compared to healthy controls in the mesencephalic crura. The ALS FTD contrast highlighted anterior medullar and central pontine atrophy at the second timepoint of ALS patients compared to disease controls. Compared to PLS patients, the ALS cohort exhibited peri-aqueduct mesencephalic, central pontine and posterior pontine density reductions. No statistically significant morphometric changes were detected between the two timepoints following the appropriate corrections in the ALS cohort.

3.4. Genetics

Eleven ALS patients and none of the PLS patients carried the GGGGCC hexanucleotide repeat expansion in C9orf72. ALS patients tested negative for other genes previously implicated in ALS. No PLS patients carried ALS or HSP associated mutations. In the ALS cohort, no brainstem differences were identified between hexanucleotide repeat carriers and those without the C9orf72 mutation at baseline or on follow-up. Moreover, patients with repeat expansions have not shown increased volume loss over time compared to C9orf72 negative ALS patients.

4. Discussion

Our study provides evidence of focal and progressive brainstem pathology in ALS and highlights that brainstem alterations can be characterised in vivo using routine MRI sequences. While volumetric analyses provide an overall indication of atrophy, the characterisation of shape deformations and focal density reduction enables the identification of the most vulnerable brainstem regions.

Our data indicate that total brainstem volume reductions are primarily driven by medulla oblongata pathology, but pontine changes also contribute to brainstem atrophy. PLS which is an upper motor neuron disorder (Finegan et al., 2019b) exhibits marked medullar, pontine and mesencephalic atrophy compared to ALS, suggesting that volume reductions are primarily caused by upper motor neuron degeneration. Furthermore, the medullar atrophy observed in PLS was not only confirmed with reference to healthy controls but also in contrast to disease controls. It is also noteworthy that using only a 4-month follow-up interval progressive volume loss can be detected in all three brainstem segments. Our morphometric analyses revealed focal density reductions in ALS compared to healthy controls in the bilateral mesencephalic crura where the descending corticospinal and corticonuclear tracts run. This further supports the notion the UMN degeneration contributes to brainstem volume reductions. This is echoed by the ALS FTD contrast which revealed focal changes in the medullar pyramids. The ALS PLS morphometric comparison is particularly interesting as both conditions are associated with widespread UMN degeneration. This contrast revealed widespread peri-aqueduct mesencephalic, central and posterior pontine grey matter changes.

Our vertex analyses also confirmed the symmetric involvement of the medullary pyramids in ALS. In contrast to PLS, ALS patients also exhibited widespread posterior pontine shape deformations including atrophy of the floor of the fourth ventricle as historically described by Clarke in 1867 (Clarke and Jackson, 1867). The widespread pontine shape deformations detected in ALS in contrast to PLS raises important questions about central brainstem involvement in ALS. As vertex analyses are essentially surface-projected representations of atrophy, the circumferential, anterior-posterior-lateral deformation detected in ALS compared to PLS (Fig. 2(C)) is likely to be driven by central volume reductions i.e. grey matter degeneration. This would not only be consistent with the differential grey matter (LMN / CNN) involvement in ALS compared to PLS clinically, but also consistent with our morphometric findings, where central pontine and mesencephalic density reductions were also detected in ALS compared to PLS (Fig. 3(C)). The integrative interpretation of the shape and morphometric alterations in ALS compared to PLS, which is a pure UMN condition, suggest that brainstem grey matter degeneration can be detected in vivo. In summary, while overall volume reductions are more prominent in PLS (Fig. 1) vertex and voxelwise analyses capture central brainstem pathology in ALS (Figs. 2(C) and 3(C)).

Fig. 2.

Anatomical patterns of brainstem pathology based on vertex analyses correcting for age, gender, education and total intracranial volumes. Shape deformation in ALS are shown compared to healthy controls (A), disease controls (B) and PLS patients (C). 3D brainstem mesh is presented in blue (A), green (B) and pink (C). Patterns of atrophy is illustrated in orange with respect to healthy controls (A), in red with reference to disease controls (B) and blue in comparison to PLS patients (C). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3.

Morphometric changes in the brainstem A, Morphometric brainstem alterations in ALS patients compared to healthy controls at p < .05 TFCE FWE corrected for age, gender, education and TIV. B, Morphometric brainstem alterations in ALS patients compared to disease controls (FTD) at p < .05 TFCE FWE corrected for age, gender, education and TIV. C, Morphometric brainstem alterations in ALS patients compared to PLS patients at p < .05 TFCE FWE corrected for age, gender, education and TIV. Statistical maps are presented in MNI space, MNI coordinates are provided under each view, and the brainstem mask is presented as a white underlay. Radiological convention is used; Lt – Left, Rt – Right.

Surprisingly few human imaging studies have specifically evaluated brain stem structures in ALS, but some whole brains studies have commented on infratentorial changes. Diffusion tensor imaging studies have been inconsistent in detecting pyramidal tract degeneration at medullar or pontine levels. Some studies readily detected inferior brain stem diffusivity changes (Schuster et al., 2016a; Bede et al., 2015; Iwata et al., 2008) while others only captured mesencephalic pyramidal tract alterations (Yin et al., 2004). Brainstem FA reductions based on pooled multisite data have been linked to stage 2 of a recently proposed pathological staging system (Muller et al., 2016). The findings of functional studies in ALS are even more conflicting (Proudfoot et al., 2018; Bede, 2017). Several PET studies have shown increased pontine metabolism (Cistaro et al., 2012) which was interpreted as reactive astrocytosis (Johansson et al., 2007). Hypermetabolism in the brainstem has been linked to bulbar onset (Cistaro et al., 2012) and C9orf72 hexanucleotide repeats (Cistaro et al., 2014; Floeter and Gendron, 2018). Some spectroscopy studies however have detected decreased Naa/Cr ratios in the brainstem (Pioro et al., 1999; Bradley et al., 1999) while other did not (Sivak et al., 2010; Kalra, 2019). Animal imaging studies enable the histological interpretation of radiological changes (Evans et al., 2012; Bede, 2019) and high-field SOD1 transgenic mouse models showed T2 weighted hyperintensities in the nucleus ambiguus, facial nucleus, trigeminal motor nucleus, rostroventrolateral reticular nucleus, lateral paragigantocellular nucleus and the substantia nigra with associated neuronal loss of histological evaluation (Zang et al., 2004). Our morphometric and vertex analyses captured central pontine and inferior mesencephalic changes in ALS compared to PLS. Few imaging studies have evaluated lower motor neuron pathology to date in ALS, and without exception, these were advanced, quantitative spinal cord imaging studies, where the accurate segmentation of white and grey matter components is possible (Querin et al., 2018b,2019a).

The benefit of using standard T1-weighted sequences and morphometric analyses is that high resolution 3D T1W sequences are typically already part of clinical protocols and are readily available on any imaging platform irrespective of manufacturer or field strength. Morphometric brain stem measures could therefore be easily integrated in clinical trial protocols and data acquisition could be harmonised across multiple centres (Muller et al., 2016; Turner et al., 2012). The brainstem measures derived from T1-weighted data are therefore relatively attractive candidate biomarkers as they can be acquired non-invasively without special sequences, the acquisition time is typically short (<10 min), the measures discriminate patients from disease controls, and the derived metrics exhibit progressive change over time.

The identification of imaging measures which readily discriminate ALS patients from healthy and disease controls is hugely relevant for emerging machine learning applications (Bede et al., 2018a) and survival prediction (Schuster et al., 2017). ALS is associated with strikingly focal anatomical vulnerability profiles (Bede et al., 2016) which enable disease-specific feature selection for classification algorithms (Grollemund et al., 2019; Schuster et al., 2016b). Classification models enable the categorisation of blinded datasets into diagnostic, phenotypic or prognostic categories and the incorporation of the discriminatory brain regions such as the brainstem is likely to increase the accuracy of such models further (Querin et al., 2018b; Bede et al., 2017).

Longitudinal imaging in ALS is typically limited by high attrition rates (Chipika et al., 2019) and relatively few large presymptomatic studies have been published to date (Schuster et al., 2015; Eisen et al., 2014; Querin et al., 2019b). Despite the challenges of longitudinal imaging, a number of multi-timepoint studies have now been published (Menke et al., 2018; Bede and Hardiman, 2018) which suggest early white matter degeneration with a ceiling effect and progressive grey mater degeneration. From a biomarker point of view, the ability to track progressive changes in the later stages of the disease is essential; as indices exhibiting early ceiling effects are not suitable for monitoring.

Despite a number of pioneering studies, relatively little is known of the imaging profile of PLS and existing studies suffer from considerable sample size limitations. Furthermore, most PLS studies report overlapping imaging patterns with ALS (Müller et al., 2018; Van Weehaeghe et al., 2016), highlighting motor cortex (Butman and Floeter, 2007; Schuster et al., 2013), pyramidal tract (Müller et al., 2018; Iwata et al., 2011), and corpus callosum pathology (Unrath et al., 2010; Agosta et al., 2014) and very few PLS studies highlight distinctive imaging features (Finegan et al., 2019a, b; Clark et al., 2018). Our data indicate that brainstem atrophy is particularly marked in PLS and the differences between the ALS and PLS cohort are statistically significant. The description of features which discriminate PLS and ALS may have implications for the categorisation of UMN predominant patients and ‘suspected PLS’ patients who don't fulfil current diagnostic criteria (Gordon et al., 2006,2009).

The presented study has several limitations. No post-mortem data are available to validate our imaging findings which would be of particular interest in the PLS where patterns of pTDP-43 burden are poorly characterised. The longitudinal arm of the study is a two-time point design which precludes commenting on linear versus exponential longitudinal changes. While we intentionally used T1-weighted data alone in the various analysis streams, novel diffusion sequences such as HARDI, Q-ball imaging or NODDI (Barritt et al., 2018; Broad et al., 2018) may have been particularly useful in characterising crossing fibre integrity in the brainstem. Notwithstanding these limitations, our study provides evidence that standard T1-weighted data can be used to detect global and focal brainstem pathology in motor neuron diseases in contrast to both healthy and disease controls.

5. Conclusions

Progressive brainstem pathology can be detected in ALS in vivo and brainstem measures may have a biomarker role in diagnostic, monitoring and prognostic applications.

Declaration of Competing Interest

Peter Bede is a member of the UK Motor Neuron Disease Association (MNDA) Research Advisory Panel, the steering committee of Neuroimaging Society in ALS (NiSALS) and the medical patron of the Irish Motor Neuron Disease Association (IMNDA). These affiliations had no impact on the design, analyses, and presentation of this work.