Resting‐state fMRI functional connectome of C9orf72 mutation status

Abstract

Objective

The resting‐state functional connectome has not been extensively investigated in amyotrophic lateral sclerosis (ALS) spectrum disease, in particular in relationship with patients' genetic status.

Methods

Here we studied the network‐to‐network connectivity of 19 ALS patients carrying the C9orf72 hexanucleotide repeat expansion (C9orf72+), 19 ALS patients not affected by C9orf72 mutation (C9orf72−), and 19 ALS‐mimic patients (ALSm) well‐matched for demographic and clinical variables.

Results

When compared with ALSm, we observed greater connectivity of the default mode and frontoparietal networks with the visual network for C9orf72+ patients (P = 0.001). Moreover, the whole‐connectome showed greater node degree (P < 0.001), while sensorimotor cortices resulted isolated in C9orf72+.

Interpretation

Our results suggest a crucial involvement of extra‐motor functions in ALS spectrum disease. In particular, alterations of the visual cortex may have a pathogenic role in C9orf72‐related ALS. The prominent feature of these patients would be increased visual system connectivity with the networks responsible of the functional balance between internal and external attention.

Introduction

Amyotrophic lateral sclerosis (ASL) is a neurodegenerative disorder that affects motor neurons in the brain and spinal cord. ¹ , ² Alongside progressive muscular atrophy and motor symptoms, 15–20% of ALS patients present non‐motor symptoms, behavioral changes, and severe cognitive deficits within the ALS—frontotemporal dementia (FTD) clinical spectrum. Growing evidence has linked ALS with a set of genetic mutations. ³ In particular, the intronic hexanucleotide G₄C₂ expansion in the chromosome 9 open reading frame 72 (C9orf72) is the major genetic cause of ALS and FTD and it is associated with more severe forms of ALS. ⁴ , ⁵ This mutation has been observed in 20–40% of familial and 3–8% of sporadic ALS cases. ⁶ , ⁷ The disease onset is earlier in mutation carriers (C9orf72+), typically characterized by with bulbar signs, cognitive symptoms, and faster disease progression that leads to a shorter survival. ⁸ , ⁹ , ¹⁰ Structural imaging analyses have also revealed widespread extra‐motor pathology in addition to motor cortex damage in C9orf72+ patients. ¹¹ , ¹² , ¹³ Mutation carriers show gray matter volume loss and cortical thinning in the primary motor cortex, extending to prefrontal, occipitotemporal, and parietal areas, ¹¹ , ¹³ , ¹⁴ , ¹⁵ alongside significant thalamic atrophy. ¹¹ , ¹² , ¹⁴ , ¹⁶ Despite this extensive corpus of knowledge, functional network alterations in ALS according to genetic mutations are still insufficiently investigated. Only few resting‐state functional magnetic resonance imaging (rs‐fMRI) studies have been performed on this topic and the abnormal connectivity organization in C9orf72+ ALS patients is still only marginally understood. ¹¹ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ This literature gap is particularly striking when considering that functional connectivity alterations at rest have been suggested to potentially represent the earliest detectable brain abnormality in carriers of ALS‐causing gene mutations. ¹⁷ , ²¹ Previous studies have reported patterns of increased/altered connectivity at rest associated with C9orf72 mutation. When compared with ALS sporadic cases, C9orf72+ patients showed increased connectivity within the visual, anterior default mode, and right dorsal attention networks, ¹¹ , ¹⁹ and diminished connectivity within salience, sensorimotor, posterior default mode, and pulvinar/thalamic‐seeded networks. ¹¹ , ¹⁷ , ¹⁸ , ¹⁹ These connectivity alterations allegedly emerge during development and worsen after disease onset. ¹⁸ , ²² When compared to healthy participants, C9orf72+ patients have been reported to show reduced posterior and subcortical connectivity, ²⁰ and decreased meta‐states changes ²³ (i.e., reduced dynamic changes in brain connectivity patterns). However, the relationship between intrinsic connectivity networks remains unclear and this vast range of alterations does not allow for a systematic understanding of the impact of C9orf72 mutation on the overall brain's functional organization.

On these grounds, it seems crucial to develop a better understanding of the neurofunctional differences between mutation carriers and non‐mutation carriers. In this study, we used rs‐fMRI to compare a group of ALS patients carrying the C9orf72 hexanucleotide repeat expansion (C9orf72+), with a group of well‐matched patients not affected by the gene mutation (C9orf72−), and a control group of ALS mimics patients ²⁴ (ALSm). Participants were carefully matched for disease severity, as previuously described, ²⁵ to describe the genotype signature controlling for clinical phenotypes. In this study, we used a functional network connectivity ²⁶ (FNC) approach to investigate the impact of C9orf72 mutation on both intra‐ and inter‐network connectivity. Based on previous evidence of alterations within single connectivity networks, we expect C9orf72 mutation to significantly affect the brain's between‐networks organization.

If proven significant, the relationship between inter‐network connectivity and C9orf72 mutations in ALS would allow a broader understanding of C9orf72‐based alterations and potentially represent an early‐detectable biomarker in asymptomatic and presymptomatic carriers.

Materials and Methods

Participants

Thirty‐eight (N = 38) patients diagnosed with ALS (EI Escorial Revised Criteria ²⁷ ), and 19 (N = 19) ALSm patients took part to the study. Comorbidity of severe neurological or psychiatric conditions as well as causes of focal or diffuse brain damage, including lacunae, and extensive cerebrovascular disorders at routine MRI were considered as exclusion criteria. Nineteen ALS patients carried the C9orf72 hexanucleotide repeat expansion (C9orf72+), while 19 ALS patients were not affected by the pathological C9orf72 expansion (C9orf72−). The group of ALSm patients included in this study did not have primary neurodegenerative diseases but rather clinical conditions with features resembling those of ALS: cervical spondylotic myelopathy (N = 3), myasthenia gravis (N = 9), monomelic amyotrophy (Hirayama) (N = 1), and peripheric motor neuropathy (N = 6). All ALSm patients were initially recruited as suspected or possible ALS cases and were later diagnosed as mimics. ALSm, C9orf72+, and C9orf72− participants were carefully matched for age and sex assigned at birth using a one‐by‐one pairing technique. C9orf72+ and C9orf72− patients were additionally matched for ALSFRS‐ total score and disease duration. Number of body regions involved was assessed using King's staging system ²⁸ (KSS). All included patients had no mutation in SOD1, TARDBP, and FUS genes. Demographic and clinical data are reported in Table 1.

Table 1.

Demographic and clinical data.

	C9orf72+ (N = 19)	C9orf72− (N = 19)	ALSm (N = 19)	P a
Male/female ratio	10/9	10/9	10/9	1
Bulbar/spinal onset	13/6	11/8	–	0.68
Cognitive classification	n	n	–	0.36
Normal cognition	7	11	–
ALSbi	1	2	–
ALSci	2	1	–
ALScbi	1	0	–
ALS‐FTD	4	1	–
Unknown	4	4	–
KSS (number of body regions involved)	n	n	–	0.58
KSS 1	9	7	–
KSS 2	5	8	–
KSS 3	4	2	–
KSS 4	1	2	–

	Median (IQR)	Median (IQR)	Median (IQR)	P b
Age (years)	55.50 (51.15–69.70)	59.50 (53.45–66.05)	60.00 (53.00–69.00)	0.99
ALSFRS‐R score	43.00 (38.50–45.00)	43.00 (37.50–45.00)	–	0.95
Disease duration (months)	11.0 (8.00–14.50)	11.0 (9.00–13.50)	–	0.94
Overall survival (years)	2.57 (2.19–3.17)	2.70 (2.13–3.40) (4 alive)	–	0.91
Disease progression (ΔALSFRS‐R)	0.53 (0.28–0.78)	0.56 (0.27–0.91)	–	0.67

KSS, King's staging score system; IQR, interquartile range; ALSFRS‐R, Revised Amyotrophic Lateral Sclerosis Functional Rating Scale.

^a Indicates results for chi‐squared test between C9orf72+, C9orf72−, and ALSm.

^b Indicates results for two‐sample independent t‐tests between C9orf72+ and C9orf72−, or one‐way ANOVA tests between C9orf72+, C9orf72−, and ALSm.

The study was carried out in accordance with the Declaration of Helsinki; all clinical data were collected and analyzed under the local ethics committee approval (Azienda Ospedaliero‐Universitaria Città della Salute e della Scienza di Torino, SC Neurologia 1U, Turin, Italy). All participants gave their written consent.

MRI acquisition

Magnetic resonance imaging acquisition was performed with a 1.5 Tesla General Electric Signa HD‐XT scanner (General Electric, Milwaukee, WI, USA) equipped with an eight channels 8HRBrain head coil, at the unit of Neuroradiology Unit, CTO Hospital, AOU Città della Salute e della Scienza di Torino, Italy. For each participant a T1‐weighted anatomical image was acquired with the following parameters: repetition time [TR] = 11.85 ms, echo time [TE] = 4.9 ms, flip angle [FA] = 12°, field of view [FOV] = 240 mm, number of axial slices = 120, voxel size = 0.47 × 0.47 × 1 mm³, and whole‐brain coverage. rs‐fMRI scans were acquired with a fast speed Echo Planar Imaging sequence with the following parameters TE = 50 ms; TR = 2250 ms; FA = 90°; number of volumes = 585; FOV = 210; 25 axial slices per volume; slice thickness = 4; interslice gap = 1 mm; voxel size = 3.28 × 3.28 × 5 mm³; phase encoding direction = anterior/posterior; the cerebellum was only partially included in the volume coverage. Participants were asked to keep their eyes closed during the rs‐fMRI acquisition.

Statistical analyses

Demographical analyses

Analyses were performed with SciPy ²⁹ running on Python3. Parametric tests were performed for all demographic and clinical variables, except for those not normally distributed (according to the Kolmogorov–Smirnov test). Differences in age (years) between all three groups were tested with Kruskal–Wallis ANOVA. Group differences in sex assigned at birth distribution were tested with chi‐squared statistics. Differences in clinical variables (i.e., bulbar vs. spinal onset, ALSFRS‐R score, disease duration in months, survival in years, disease progression, and number of body regions involved) were tested between C9orf72+ and C9orf72− with independent sample t‐tests and Mann–Whitney U tests. All results were considered as significant at P ≤ 0.05 and corrected for multiple comparisons.

Preprocessing

T1‐weighted anatomical images were segmented with FreeSurfer (v6.0.0, http://surfer.nmr.mgh.harvard.edu/). ³⁰ The estimated total intracranial volume (TIV), representing the sum of total gray matter, white matter (WM), and cerebrospinal fluid volume (CSF), was extracted for each participant using the same tool. Kolmogorov–Smirnov test was performed to assess the normal distribution of each variable. Group differences in TIV were computed using a one‐way ANOVA to determine the overall atrophy degree in patients' subgroups.

The CONN toolbox ³¹ (version 20.b; MATLAB 9.14) implemented in SPM12 was used to perform neuroimaging data preprocessing and resting‐state functional connectivity analyses. CONN's “default_MNI” preprocessing pipeline was used. The following processing steps were performed: realignment to the first functional image and unwarping, slice‐timing correction (slice order = interleaved, bottom‐up); outlier identification with the artifact rejection toolbox (ART) (ART thresholds: framewise displacement >1.1 mm; global BOLD signal changes >5 SD); direct segmentation in gray matter, WM, and CSF tissue classes; registration and normalization of functional and anatomical data into standard Montreal Neurological Institute (MNI) space ³² ; functional data resampling to 2 mm isotropic voxels, and anatomical data resampling to 1 mm isotropic voxels; smoothing with a full‐width at half‐maximum 6 × 6 × 6 mm³ Gaussian kernel. Data denoising was performed with the anatomical component‐based noise correction (aCompCor) to extract noise signal from white matter and CSF components on a voxel‐by‐voxel level. ³³ , ³⁴ For each participant, realignment parameters, outlier volumes scrubbing parameters, physiological noise components obtained from mean WM and CSF maps, and “effect of rest” (initial magnetization transient effects) were entered as nuisance covariates in the first‐level analysis. ³⁵ Subsequently, a 0.008‐ to 0.1‐Hz bandpass filter was applied to the time‐series to remove low‐frequency drifts and high‐frequency noise.

The quality of structural and functional images was assessed and visually inspected using both the quality assessment (QA) tool implemented in CONN's default, and the BIDS‐compliant software MRIQC. ³⁶ No participant was excluded because of excessive motion, noise, or other MRI artifacts.

ROI‐to‐ROI resting‐state functional connectivity

A region of interest‐to‐region of interest (ROI‐to‐ROI) approach was adopted for the first‐level analyses. For each participant, a connectivity matrix including 26 ROIs of the following rs‐fMRI networks, as implemented in CONN's default atlas, was computed. Default mode network (medial prefrontal cortex, left lateral parietal region, right lateral parietal region, posterior cingulate cortex), sensorimotor network (left lateral region, right lateral region, superior region), visual network (medial region, occipital region, left lateral region, right lateral region), salience network (anterior cingulate cortex, left anterior insula, right anterior insula, left rostral prefrontal cortex, right rostral prefrontal cortex, left supramarginal gyrus, right supramarginal gyrus), dorsal attention network (left frontal eye field, right frontal eye field, left intraparietal sulcus, right intraparietal sulcus), and frontoparietal network (left lateral prefrontal cortex, right lateral prefrontal cortex, left posterior parietal cortex, right posterior parietal cortex). For each ROI of each rs‐fMRI network, Pearson's correlation coefficients were computed between the signal time‐series of that ROI and all the other 25 ROIs. Correlation coefficients were then converted to Z scores with the R‐to‐Z Fisher transform and a 26 × 26 whole‐connectome matrix was generated for each subject. Connectome matrices were entered into the second‐level general linear model to compare the functional connectivity network profiles of patients' subgroups with directional contrasts (i.e., C9orf72+ vs. ALSm, C9orf72− vs. ALSm, and C9orf72+ vs. C9orf72−). Age, sex assigned at birth, and TIV were included as nuisance covariates. ³⁷ , ³⁸ , ³⁹ The functional network connectivity (FNC) method implemented in CONN was adopted to perform cluster‐level inferences on the rs‐fMRI networks of ROIs. ²⁶ , ³⁵ Briefly, FNC first identifies sets of related ROIs with a data‐driven hierarchical clustering procedure. The entire set of connections between all pairs of ROIs is then analyzed using a multivariate parametric general linear model analysis. To select only highly significant pairs of sets of connections the cluster‐level P‐false‐discovery rate (FDR)‐corrected for multiple comparisons was set at <0.05. To identify only the most robust individual connections of the significant sets the connection‐level post hoc uncorrected P was set at <0.05. Cluster‐level and connection‐level significance thresholds were kept at CONN 20b's default settings.

Correlational analyses were performed to determine the impact of disease duration and ALS progression on brain connectivity. For each between‐group second level significant comparison, the mean connectivity of the resulting matrix was computed for C9orf72+ and C9orf72− and correlated with disease duration and King's staging system scores. Moreover, to assess whether variations in network connectivity could be observed across different KSS stages, comparisons between C9orf72+ and C9orf72− groups were also performed restricting the analysis to stages with at least five participants per group. Finally, as both male sex assigned at birth and spinal onset have been described as factors negatively impacting prognosis in C9orf72+ patients, ⁴⁰ additional comparisons were performed (i.e., male C9orf72+ patients vs. female C9orf72+ patients; bulbar onset C9orf72+ patients vs. spinal onset C9orf72+ patients).

Graph theory metrics

In graph theory, the brain's functional organization can be modeled as a complex system of interacting elements (graph) with nodes (ROIs) connected by edges (connections). ⁴¹ , ⁴² Each participant‐specific 26 × 26 ROI‐to‐ROI connectivity matrix was thresholded at a cost value of k = 0.3 as small‐world properties were observed at this threshold (as recommended by CONN's guidelines ³¹ ). Accordingly with CONN's default options, single subjects' negative functional connectivity values were disregarded in these analyses. The following graph theory metrics as implemented in CONN were extracted from each participant's whole 26 × 26 ROI‐to‐ROI connectome: node degree (i.e., the number of connections of each node), global efficiency at a node (i.e., the average of the inverse minimum number of edges between each node and all other nodes in the network, representing a measure of the node's centrality within a network), local efficiency at a node (i.e., the extent of the integration of a sub‐graph consisting of only the nodes surrounding a certain node, representing an index of local integration or coherence), and betweenness centrality (i.e., an alternative measure of node centrality representing the proportion of times that a node is part of a shortest path between any two pairs of nodes within a graph). ⁴² Graph theory metrics were compared between each group with CONN with the same covariates used in the previous for ROI‐to‐ROI connectivity analyses. All results were corrected with a P < 0.05 FDR correction for multiple comparisons.

As the adoption of matrix node thresholds for performing graph analyses can potentially hide small effects, additional analyses were performed with K cost threshold set at 0.1; 0.3 (our default); 0.5; 0.7; 0.9, respectively indicating that the analyses were performed only with the top 10%, 30%, 50%, 70%, and 90% of the connectome edges.

Results

Demographical analyses

No significant difference in age between the three groups was found (H = 0.03, P = 0.98). No significant difference in ALSFRS‐R at the time of MRI (T = 0.06, P = 0.95) and disease duration from the onset (U = 169.5, P = 0.76) was found between C9orf72+ and C9orf72− patients. When comparing TIV among each group, no significant difference emerged (F = 0.55, P = 0.58).

ROI‐to‐ROI resting‐state functional connectivity

The comparison between C9orf72− and ALSm patients did not reveal significant differences. Similarly, no significant difference in network connectivity between C9orf72+ and C9orf72− patients was found.

C9orf72+ patients when compared to ALSm revealed significant differences in the connectivity of one set of networks (F(2, 50) = 11.14, P‐FDR‐corrected for multiple comparisons = 0.001, η ² ≈ 0.92; 95% CI = 0.85–0.96). In particular, C9orf72+ patients relative to ALSm exhibited increased connectivity between the default mode network (bilateral lateral parietal components, medial prefrontal cortex, posterior cingulate cortex) and the visual network (medial, occipital, and bilateral lateral components), and between the frontoparietal network (bilateral posterior parietal cortex, bilateral prefrontal cortex) and the visual network (medial, occipital, and bilateral lateral components). Results are reported in Table 2; and Fig. 1A.

Table 2.

ROI‐to‐ROI resting‐state functional connectivity results.

Cluster	F‐value	P‐FDR
Whole network	11.41	<0.001

Region 1	MNI xyz coordinates	Region 2	MNI xyz coordinates	T‐value	P‐unc
Visual lateral (L)	−37, −79, 10	Default mode LP (L)	−39, −77, 33	4.01	<0.001
Visual lateral (R)	38, −72, 13	Frontoparietal LPFC (R)	41, 38, 30	3.82	<0.001
Visual lateral (R)	38, −72, 13	Default mode LP (R)	47, −67, 29	3.78	<0.001
Visual lateral (L)	−37, −79, 10	Default mode LP (R)	47, −67, 29	3.66	<0.001
Visual lateral (R)	38, −72, 13	Frontoparietal PPC (L)	−46, −58, 49	3.44	0.001
Visual lateral (R)	38, −72, 13	Default mode LP (L)	−39, −77, 33	3.43	0.001
Visual lateral (R)	38, −72, 13	Default mode PCC	1, −61, 38	3.28	0.002
Visual lateral (R)	38, −72, 13	Frontoparietal LPFC (L)	−43, 33, 28	3.24	0.002
Visual lateral (L)	−37, −79, 10	Frontoparietal LPFC (R)	41, 38, 30	3.14	0.003
Visual occipital	0, −93, −4	Default mode LP (R)	47, −67, 29	3.17	0.003
Visual occipital	0, −93, ‐4	Default mode LP (L)	−39, −77, 33	3.17	0.003
Visual lateral (L)	−37, −79, 10	Frontoparietal PPC (R)	52, −52, 45	2.79	0.007
Visual lateral (L)	−37, −79, 10	Frontoparietal LPFC (L)	−43, 33, 28	2.75	0.008
Visual lateral (L)	−37, −79, 10	Default mode PCC	1, −61, 38	2.7	0.009
Visual medial	2, −79, 12	Default mode LP (R)	47, −67, 29	3.38	0.001
Visual medial	2, −79, 12	Default mode LP (L)	−39, −77, 33	3.13	0.003
Visual lateral (R)	38, −72, 13	Frontoparietal PPC (R)	52, −52, 45	2.6	0.012
Visual lateral (L)	−37, −79, 10	Frontoparietal PPC (L)	−46, −58, 49	2.46	0.017
Visual lateral (L)	−37, −79, 10	Default mode MPFC	1, 55, −3	2.23	0.03
Visual medial	2, −79, 12	Frontoparietal LPFC (R)	41, 38, 30	2.6	0.012
Visual medial	2, −79, 12	Frontoparietal LPFC (L)	−43, 33, 28	2.18	0.034
Visual medial	2, −79, 12	Default mode MPFC	1, 55, −3	2.15	0.036
Visual occipital	0, −93, −4	Frontoparietal PPC (R)	52, −52, 45	2.03	0.047

Significance threshold is set at P‐false‐discovery rate (FDR)‐corrected <0.05 (cluster level) and at P‐uncorrected <0.05 (connection level). L, left; R, right.

Figure 1.

(A) Results from the C9orf72+ > ALSm contrast, in the ROI‐2‐ROI analyses. Results are shown at connection‐P‐uncorrected <0.05, cluster‐P‐FDR‐corrected for multiple comparisons <0.05. L = left; R = right. Red color indicates increased ROI‐to‐ROI connectivity. (B) Resting‐state functional connectivity main effects at the whole‐connectome level. Results are shown at connection‐P‐uncorrected <0.05, cluster‐P‐FDR‐corrected for multiple comparisons <0.05. Red color indicates positive ROI‐to‐ROI connectivity, blue color indicates negative ROI‐to‐ROI connectivity.

No significant correlations between the mean connectivity of significant networks and disease duration, and KSS score were identified. No significant difference was observed when grouping patients by KSS stages. No significant difference associated with male sex assigned at birth or bulbar onset were observed in the C9orf72+ group.

Graph theory metrics

When comparing C9orf72+ with ALSm, graph theory analyses showed a significant enhancement in the node degree of the whole connectome in the former (i.e., the whole system of interconnected networks; T(51) = 4.65, P‐unc < 0.001, d = 0.65, 95% CI = 0.22 to 1.08). Conversely, the bilateral lateral nodes of sensorimotor network showed decreased node degree (left node: T(51) = −3.78, P‐FDR = 0.011, d = −0.53, 95% CI = −1.06 to 0; right node: T(51) = −3.19, P‐FDR = 0.032; d = −0.45, 95% CI = −1.08 to 0.18) and decreased global efficiency at node (left node: T(51) = −3.78, P‐FDR = .009, d = −0.53, 95% CI = −1.06, −0; right node: T(51) = −3.58, P‐FDR = 0.009; d = −0.5, 95% CI = −1.06 to 0.06) in C9orf72+ patients.

No significant difference was found when considering local efficiency and betweenness centrality, and when comparing C9orf72+ with C9orf72− and C9orf72− with ALSm patients. Whole‐connectome main effects for each group are reported in Figure 1B.

Analyses performed with K cost threshold values ranging from 0.1 to 0.9 are reported in the Table S1 section.

Discussion

The relationship between brain functional connectivity at rest and the gene mutation status in ALS is not fully understood. This lack of consistency may depend, at least in part, on adequate patients‐controls matching and methodological choices in data processing. In particular, most of the aforementioned studies adopted an independent component analysis‐based approach to explore connectivity changes within—and not between—isolated networks. Due to the brain's complex and interconnected nature, neurological disorders are known to disrupt communication between multiple interdependent large‐scale functional connectivity networks. ⁴¹ , ⁴³ , ⁴⁴ , ⁴⁵ In this study, we tested inter‐ and intra‐network connectivity between C9orf72 mutation carriers, noncarriers, and ALS‐mimic patients. Two main results were obtained when comparing C9orf72+ with ALSm. Functional Network Connectivity analyses revealed enhanced connectivity of the visual network with both the default mode network and the frontoparietal network. Moreover, graph theoretical metrics revealed increased number of connections within the whole connectome in the face of functional segregation of sensorimotor bilateral nodes.

In this study, the visual system resulted more strongly connected with anterior and posterior regions of the default mode network and the frontoparietal network in C9orf72+ when compared to ALSm. Early pathological changes within the occipital cortex accompanied by local functional and structural abnormalities have been reported in C9orf72+ patients. ⁴⁶ , ⁴⁷ , ⁴⁸ Agosta and colleagues ¹¹ showed enhanced visual functional connectivity in C9orf72+ patients in the presence of occipital cortical thinning and decreased inferior frontooccipital fasciculus microstructural integrity. Moreover, Premi and colleagues ²³ recently reported a C9orf72 mutation‐driven association between pathological brain connectivity and processing speed in a visual search task (TMT‐A) that might support a negative impact of gene mutation on visuospatial functions.

On the other hand, alterations in both default mode and frontoparietal networks have been reported in mutation carriers. The default mode network consists of a group of interconnected brain regions whose activity increases while at rest and during “inactive” states (e.g., self‐referential processing and mind wandering), and decreases when the brain is engaged in attention‐demanding and goal‐directed tasks. ⁴⁹ , ⁵⁰ Conversely, the frontoparietal network is part of the “central executive network” whose activity is typically associated with directing attention to perform intentional goal‐driven behavior and exert adaptive cognitive control. ⁵¹ , ⁵² In C9orf72+ patients with respect to sporadic cases and healthy participants, both increased and decreased connectivity within the default mode network have been observed. ¹⁷ , ¹⁸ , ¹⁹ Comparing presymptomatic C9orf72+ participants with healthy participants, longitudinal trajectories of intra‐network homogeneity of the frontoparietal network showed increased connectivity as aging. ²² Moreover, according to the well‐known “triple network model,” the activity of the default mode network and the frontoparietal network is typically found to be connected and mediated by the salience network, as attention naturally switches from internal to external awareness states and vice‐versa. ⁵³ Alteration of this specific functional organization of information transfer would be impactful on executive functions sustained by the main hubs of these large‐scale networks, such as prefrontal, parietal, and cingulate cortices. ⁵⁴ , ⁵⁵ , ⁵⁶ This interpretation is in line with neuropsychological evidence of greater impairment in executive functions, and visual and verbal memory in C9orf72 mutation carriers when compared with C9orf72− patients. ⁹ The altered visual network connectivity with the default mode and frontoparietal networks in C9orf72+ patients is an original finding of this study that needs to be further explored. A potential explanatory interpretation may involve the causal relationship between thalamic degeneration in C9orf72+ patients and dysfunctions in brain connectivity at rest. Thalamic alterations have been recurrently reported in mutation carriers, and have been considered a robust morphological signature of C9orf72 mutation. ¹⁴ , ¹⁶ , ⁴⁷ , ⁵⁷ , ⁵⁸ In particular, pulvinar nuclei within the thalamus are a fundamental point of convergence in visual pathways and are involved in visual cognition and visuospatial attention processes. Moreover, pulvinar nuclei are functionally and anatomically connected with the visual network and have topographically organized visual field maps. ⁵⁹ , ⁶⁰ Lee and colleagues ¹⁷ , ¹⁸ observed medial pulvinar degeneration associated with disrupted salience network connectivity in C9orf72+ presymptomatic patients when compared with healthy controls (see also Premi and colleagues ²³ ). Our findings may suggest that the visual network in C9orf72+ patients, potentially “disanchored” from atrophic pulvinar nuclei, loses its functional specificity and becomes more strongly connected with internal and external attention networks that are typically mediated by the salience network in healthy controls. ⁵³

Considering graph theoretical metrics, our results revealed that C9orf72+ patients are characterized by a connectome with a larger number of connections. This finding is in line with previous accounts of increased resting‐state connectivity observed in C9orf72 mutation carriers ¹¹ , ¹⁹ and it is most probably mirrored at the structural connectivity level. ⁶¹ Hyperconnected systems (i.e., enhanced functional connectivity in terms of the number or the strength of connections) have been observed in several neurological and psychiatric disorders, either as a compensatory response to neurological disruption or following the loss of local inhibition. ²¹ , ⁶² , ⁶³ , ⁶⁴ According to graph theory, network hyperconnectivity generally indicates high‐cost information transfer, with redundant connections that are metabolically expensive to maintain and associated with increased vulnerability to pathologies. ⁴² , ⁶⁴ , ⁶⁵ In addition to whole‐connectome increased number of connections, C9orf72+ patients showed isolated sensorimotor cortices when compared with the ALSm group. This finding is in line with the known alterations of the sensorimotor cortex integrity and connectivity observed in ALS patients, irrespective of the mutation status. ¹¹ , ²² , ⁶⁶ Our results suggest that functional connectivity at rest in C9orf72+ patients is costly and inefficient, with abnormalities in high‐cost components such as network hubs and long‐distance connections. ⁶⁵ Within this system, sensorimotor processes would be inadequately integrated. This pathological organization allegedly originates from combined neurodevelopmental factors and neurodegenerative processes that would lead to dysfunctional motor and extra‐motor behavior in mutation carriers.

Our results should be considered in light of two main limitations. First, it was not possible to collect imaging data from healthy controls due to the hospital setting in which the data were acquired. Comparing our cohorts of patients with a control group of healthy individuals would have allowed a broader understanding of the neurofunctional impact of C9orf72 mutation in ALS. Nonetheless, the inclusion of ALSm patients did allow us to identify those features that are unique to ALS and the C9orf72 mutation in particular, irrespective of the common symptomatic manifestations with symptomatic mimics. In addition, the group of ALSm patients considered in this study did not have primary neurodegenerative diseases, which could have influenced the comparison. Moreover, the heterogeneous group of clinical conditions within the control group might have contributed to significant variability in ALSm connectivity. This limitation is again derived by the inherent difficulties in obtaining closely matched healthy controls within a clinical setting. However, such variability could have partially reduced possible between‐group differences. The effect size and confidence intervals of the ROI‐to‐ROI network connecitivity comparison indicate a large between‐group effect, with a narrow confidence interval, thus suggesting robust effect only marginally influenced by group variability. Nonetheless, future studies could better address this issue directly by enrolling healthy control subjects. As a second limitation, our analyses did not reveal significant differences between C9orf72+ and C9orf72− patients. The two samples were accurately matched for age, sex assigned at birth, and disease burden. Of note, the mean disease duration in both cohorts was inferior to 1 year. Group differences between mutation and non‐mutation carriers may appear later, as dysfunctional connectivity changes over time with dissimilar disease trajectories. ⁸ , ¹⁸ , ²² , ⁶⁷ On these grounds, the two groups might have been too similar for large‐scale network connectivity differences to emerge significantly at this early stage of pathology. A longitudinal study would be better suited to explore mutation‐based differences in the impact of disease trajectories on functional connectivity.

Taken together, our results provide an original overview of the impact of the genetic status on large‐scale connectivity networks in ALS, paving the way for future research on gene‐based connectomics and gene‐targeted therapies. C9orf72 mutation is associated with a neuroimaging phenotype caused by neurodevelopmental and neurodegenerative processes that induce pathological hyperconnectivity between extra‐motor networks in the face of sensorimotor network isolation.

Conclusions

In the present study, we used network neuroscience as a methodological framework to investigate how C9orf72 hexanucleotide expansion in ALS patients affects the brain's functional connectivity at rest. Our findings highlight the unique role played by C9orf72 mutation in the pathological neurofunctional profile of ALS patients. In particular, mutation carriers showed abnormal visual network connectivity with internal and external attention networks and a broadly hyperconnected connectome with isolated sensorimotor nodes.

Author Contributions

Dr Nigri had full access to all of the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis.

Conflict of Interest

Adriano Chiò serves on scientific advisory boards for Mitsubishi Tanabe, Biogen, Roche, Denali Pharma, Cytokinetics, Lilly, Zambon Biotech, and Amylyx Pharmaceuticals and has received a research grant from Biogen. No conflicts of interest are present for the other authors.

Supporting information

Table S1. Graph theoretical results (C9orf72+>ALSm), at different K cost thresholds.

Funding Statement

This work was funded by Italian Ministry of Health grants GR‐2019‐12371291, RF‐2016‐02362405, and RRC; Ministero della Salute ; Ricerca Sanitaria Finalizzata; Giovani ricercatori; European Commission's Health Seventh Framework Programme grant 259867; Italian Ministry of Education, University and Research; Progetti di Ricerca di Rilevante Interesse Nazionale, PRIN grant 2017SNW5MB; Joint Programme – Neurodegenerative Disease Research ; Department of Excellence; ‘Rita Levi Montalcini’ Department of Neuroscience; University of Torino, Italy.

Data Availability Statement

Data will be available upon request by interested researchers.