Presymptomatic grey matter alterations in ALS kindreds: a computational neuroimaging study of asymptomatic C9orf72 and SOD1 mutation carriers

One of the important paradigm shifts in amyotrophic lateral sclerosis (ALS) research is the departure from the concept of “one-drug for all” to the pursuit of precision, genotype-specific pharmacological interventions [1, 2]. The recognition of the fundamental heterogeneity of ALS led to the nuanced characterisation of various ALS genotypes and phenotypes [3‐6]. It is increasingly recognised that symptom manifestation in ALS is preceded by a long presymptomatic phase [7] and degenerative changes may be detected decades before symptom manifestation [8, 9]. The ideal timing of therapeutic intervention should therefore be reconsidered, especially in genetically susceptible cohorts. Antisense oligonucleotide (ASO)-therapies have been approved for the treatment of spinal muscular atrophy and Duchenne muscular dystrophy [10‐12], and also trialled in ALS [1]. Accordingly, the assessment of disease burden prior to symptom manifestation is of pressing practical relevance. Two most commonly studied genotypes in ALS are the GGGGCC hexanucleotide repeat expansions (HRE) in C9orf72 and SOD1. C9orf72 HRE may lead to a spectrum of clinical manifestations spanning from ALS to frontotemporal dementia (FTD) and clinical manifestations are thought to be closely associated with patterns of phosphorylated 43 kDa TAR DNA-binding protein (pTDP-43) burden. Striatal [13], temporal [14], frontal [15], cerebellar [16], and thalamic [13, 16] grey matter alterations have previously been described in asymptomatic C9orf72 cohorts. Presymptomatic orbitofrontal [17], corpus callosum, cingulate, uncinate [8, 13, 18], and corticospinal tract [9, 13] white matter changes have also been consistently detected and PET studies captured frontotemporal, thalamic, and basal ganglia hypometabolism [19]. Three mechanisms have been proposed for C9orf72 HRE-associated pathophysiology; loss of C9orf72 function through haploinsufficiency, toxic gain-of-function due to the generation of aberrant HRE-containing RNA, and toxic gain-of-function through the accumulation of dipeptide repeat proteins translated from hexanucleotide repeat RNA [20]. These mechanisms are thought to trigger a multitude of cellular responses, of which pTDP-43 accumulation may only be one amongst several processes. Accordingly, cerebral involvement outside the neocortex may represent a distinguishing feature of C9orf72 distinct from other genetic variants, such as SOD1 mutations [21]. The presymptomatic imaging signature of SOD1 is thought to be relatively unique; white matter changes in the posterior limb of the internal capsule [22], reduced superior spinal cord NAA/Cr and NAA/Myo ratios [23], and reduced left frontotemporal junction flumazenil binding [24], and have been reported. While structural, metabolic, and function thalamic [8, 13, 16, 18, 25‐30] changes have been previously described in presymptomatic C9orf72 hexanucleotide carriers and longitudinal thalamic changes have also been explored [25, 31, 32], the predilection for specific thalamic nuclei remains poorly characterised despite the unique role of thalamic nuclei in relaying specific sensory, cognitive, and behavioural functions [33‐36]. Presymptomatic amygdalar and hippocampal alterations are also under evaluated despite the selective involvement of these structures in symptomatic mutation carriers [37‐41]. Accordingly, the principal objective of this study is the nuanced characterisation of cortical and subcortical grey matter changes in two cohorts of presymptomatic mutation carriers using a panel of supplementary imaging techniques. Our hypothesis is that presymptomatic hexanucleotide expansion carriers exhibit focal subcortical degeneration with concomitant alterations in their cortical projection areas.

The study was approved by the Ethics Committee of the University of Ulm (reference 68/19), in accordance with the ethical standards of the current version of the revised Helsinki declaration. All participants gave informed consent prior to enrolment. Recruitment strategy and genetic testing have been previously described [17].

The main study groups were matched for age and education and differed in sex ratios (Table 1). We note that all statistical models were corrected for age, sex, and education.

While differences in overall subcortical volume differences did not reach significance (Table 2), mediodorsal-paratenial-reuniens in the left thalamus and pulvinar-limitans atrophy in the right thalamus was detected in asymptomatic C9orf72 hexanucleotide repeat expansion carriers compared to gene-negative family members (Table 3.). Furthermore, higher sensory nuclei volumes were identified in C9orf72 hexanucleotide repeat expansion carriers compared to both gene-negative controls and SOD1 mutation carriers in both thalami. Effect sizes are illustrated in Fig. 1. Differences in amygdalar nuclei and hippocampal subfield volumes did not reach significance. Relevant output statistics, univariate p value, and effect sizes are summarised in supplementary tables 1–2.

Vertex analyses identified shape deformations in C9orf72 hexanucleotide repeat expansion carriers compared to gene-negative controls in the anterior, superior, and posterior surface of both thalami as well as the lateral aspect of the left hippocampus (Fig. 2). Vertex analyses in SOD1 carriers did not identify shape deformation compared to gene-negative controls. ROI morphometric analyses revealed bi-thalamic and left pulvinar signal reductions in C9orf72 hexanucleotide repeat expansion carriers compared to gene-negative controls (Fig. 3). In our voxelwise analyses, the contrasts between SOD1 carriers and gene-negative controls did not reach significance.

Our computational image analyses capture thalamic and hippocampal alterations in a cohort asymptomatic C9orf72 hexanucleotide repeat expansion carriers without associated neocortical grey matter atrophy. No cortical or subcortical grey matter pathology was observed in our presymptomatic SOD1 group.

Methodologically, our study benefits from a multiparametric approach, where data were interrogated in multiple pipelines run in different image analyses suites resulting in good anatomical concordance. The results of our post-segmentation volumetric analyses and our ROI-based morphometric subcortical analyses are relatively concordant in identifying bilateral mediodorsal thalamic atrophy (Table 3, Fig. 3.). Our results indicate that as opposed to global thalamic degeneration, selective thalamic involvement characterises the asymptomatic phase of C9orf72. The shape deformation identified on vertex analyses confirms focal thalamic involvement, but the nature of these analyses is that surface-projected changes are captured instead of the intra-thalamic changes described by morphometric and post-segmentation pipelines. Resting-state fMRI studies have consistently described widespread connectivity alterations in C9orf72 mutation carriers including networks relayed through the thalamus [13, 55]. Network integrity alterations were also detected using chronnectomic approaches [56] and thalamic hypometabolism has also been consistently identified by PET studies [57, 58]. MR spectroscopy [59] identified reduced putaminal NAA/Cr, Glu/Cr and Glu/NAA ratios, and reduced thalamic Glu/NAA in asymptomatic C9orf72 mutation carriers [19]. Based on structural data, thalamic atrophy [8, 13, 16, 18, 25‐28] has been previously described in presymptomatic C9orf72 hexanucleotide carriers as well as disease burden in other subcortical grey matter structures, such as the caudate [14, 60], putamen [14], and striatum, but the predilection for specific nuclei or subregions have not been comprehensively analysed. In our study, the two main thalamic regions identified by both our volumetric and morphometric analysis streams are the mediodorsal and pulvinar regions. The physiological role of cortico-basal networks relayed though specific nuclei are fairly well established [33, 35, 61]. Thalamic atrophy has been highlighted in most genetic variants of FTD [35, 62], but medial pulvinar degeneration is thought to be relatively unique to C9orf72 [62, 63]. The involvement of specific groups of nuclei at a presymptomatic phase is relativity novel as previous studies primarily focused on global thalamus pathology [18] or only described focal changes based on voxelwise analyses. In a particularly elegant recent study, spectral clustering, a graph-based partitioning technique was implemented which revealed posteromedial thalamic changes in asymptomatic C9orf72 carriers. [64] The early degeneration of mediodorsal and pulvinar regions in our C9orf72 cohort may herald future dysfunction in associated cognitive domains, such as limbic, executive, and associative processes.

It is noteworthy that the patterns of grey matter changes identified in this asymptomatic cohort, are also relatively consistent with the “post-symptomatic” signature associated of the genotype. Selective thalamic atrophy in symptomatic C9orf72 hexanucleotide expansion carriers is evidenced by a multitude of large, prospective neuroimaging studies including both ALS and FTD phenotypes [35, 65]. Grey matter involvement in C9orf72 carriers outside the neocortex has a predilection for the thalamus, [66‐70] although hippocampal, putaminal, striatal, caudate, cerebellar, and nucleus accumbens changes are also commonly reported [14, 40, 71, 72]. Even though grey matter disease burden is thought to be more pronounced in C9orf72-associated ALS [71, 73], considerable subcortical grey matter pathology is also readily identified in C9orf72-negative ALS and PLS [40, 47, 74‐77]. Basal ganglia changes have also been consistently described in sporadic patients and linked to neuropsychological and extra-pyramidal deficits [78‐80]. Similar to our findings in this cohort of presymptomatic patients, a study of symptomatic C9orf72 patients with ALS fulfilling the EL Escorial criteria [69] also identified the selective degeneration of thalamic nuclei in C9orf72-positive ALS preferentially involving the mediodorsal–parateniual–reuniens group of nuclei, which play a central role in executive processes. Interestingly, we have detected higher sensory nuclei volumes bilaterally in C9orf72 hexanucleotide carriers compared to both “gene-negative” ALS kindreds and asymptomatic SOD1 mutation carriers. While somatosensory impairment is not classically associated with the core clinical features of ALS, in symptomatic cohorts, imaging studies have consistently confirmed the involvement of somatosensory structures [34, 70]. The marked differences between SOD1 and C9orf72 showcase the notable heterogeneity of ALS, and given the relatively young age profile of participants, the higher volumes detected hexanucleotide repeat carriers may support the role of neurodevelopmental factors [17, 81, 82]. Our vertex analyses also highlight hippocampal atrophy in hexanucleotide carriers which is also consistent with observations from symptomatic patient groups [14, 40, 71] and in line with the reports of early memory impairment in clinical subgroups of ALS [83]. Also, these subjects exhibit nucleus reuniens involvement of a key hub of thalamic afferents to the hippocampus [84]. Thus, C9orf72 carriers display concomitant thalamic and hippocampal alterations which are probably functionally interlinked. However, given that our hippocampal findings are unilateral, further validation is needed by larger studies.

The practical relevance of multimodal presymptomatic studies stems from the prospect of developing accurate predictive models to foretell the approximate time of phenoconversion and the likely clinical phenotype. This would enable precision care planning for individual subjects and optimised timing for clinical trial inclusion. The utility of machine learning (ML) has already been demonstrated in a variety of prognostic and diagnostic applications in ALS [85‐87]. Imaging-based ML models in ALS increasingly include subcortical measures [88‐91] in addition to cortical grey matter and cerebral white matter metrics [92‐95]. Feature importance analyses and cluster analyses consistently confirmed the discriminatory potential of subcortical indices and integrity metrics of networks relayed through subcortical nuclei [96, 97]. In asymptomatic C9orf72 HRE carriers, emerging spinal cord imaging techniques may be particularly useful in delineating incipient ALS from FTD [9, 98]. MRI-based predictive models have already been successfully trialled in symptomatic ALS patients [92, 99] and similar strategies could be adopted in asymptomatic cohorts to foretell likely phenoconversion.

Our results highlight the fundamental heterogeneity of genetic ALS from its earliest stages, by demonstrating the strikingly divergent disease-burden patterns between C9orf72 and SOD1 carriers. Despite the insights generated by a cross-sectional study, ultimately, large, multi-timepoint longitudinal studies are required to track mutation carriers from a very young age until fulfilling diagnostic criteria and beyond. The lack of longitudinal analyses is the biggest limitation of this cross-sectional study. While our descriptive statistics demonstrate focal grey matter changes, they only offer a mere snapshot in time. Characterising the evolution of grey matter alterations from birth to phenoconversion through several timepoints would reveal the full biological trajectory of these processes [100]. Importantly, the subjects of this study have not yet met relevant diagnostic criteria, and therefore, we cannot be sure whether individual hexanucleotide expansion carriers will develop a syndrome primarily consistent with ALS or FTD, although the index cases (affected, symptomatic family members) of the current study all had ALS. Finally, we acknowledge the limitations of the raw data and wish that additional spectroscopic or resting-state fMRI data would be at our disposal to comprehensively interrogate metabolic and connectomic alterations.

C9orf72 hexanucleotide repeat expansions are associated with presymptomatic thalamus and hippocampus alterations which may precede detectable neocortical involvement. The identified radiological changes may be mediated by a multitude of C9orf72-associated pathophysiological processes which take place decades before symptom manifestation.