Cognitive impairment falling short of dementia occurs in approximately 35% of patients with amyotrophic lateral sclerosis (ALS), and a further 10% develop a frontotemporal dementia (Rippon et al., 2006; Montuschi et al., 2015). Cognitive impairment in non-demented ALS patients characteristically includes executive and language deficits (Beeldman et al., 2016). The revised consensus criteria for the diagnosis of ALS propose a classification scheme based on deficits in specific cognitive domains (Strong et al., 2017): patients are categorized as non-impaired (ALSni), cognitively impaired (ALSci), behaviorally impaired (ALSbi), cognitively and behaviorally impaired but not demented (ALScbi), or as having frontotemporal dementia (ALS-FTD).

Shen et al. (2018) reported group differences in frontotemporal cerebral blood flow between ALSni, ALSci, and ALS-FTD (Shen et al., 2018). Resting-state functional connectivity (RSFC) analyses based on the BOLD signal in fMRI have shown differences between healthy controls (HCs) and ALS patients in several neuronal networks. Schulthess et al. (2016) reported a general expansion (i.e., an increase in RSFC) of ALS RSFC maps in the default mode network (DMN), motor network (MN), and ventral attention network (VAN) over time. Additionally, they reported that the expansions of the MN followed a pattern similar to the TDP-43 distribution in the brain. Current evidence regarding network expansion or contraction is heterogeneous. In the DMN, hyperconnectivity was found in the left precuneus in ALS patients when compared with healthy controls (Agosta et al., 2013; Li et al., 2017). In contrast, hypoconnectivity in the DMN has been reported in the ventral anterior cingulate cortex, the posterior cingulate cortex (Mohammadi et al., 2009; Agosta et al., 2013; Li et al., 2017), the orbitofrontal cortex, and the medial prefrontal cortex (Mohammadi et al., 2009; Schulthess et al., 2016; Li et al., 2017). This is in line with the notion that ALS patients seem to show lower RSFC in the anterior- and higher connectivity in the posterior nodes of the DMN than HC (Li et al., 2017). A further distinction into ALS with bulbar onset and ALS with limb onset by Trojsi et al. (2017) showed an even stronger increase in posterior RSFC in patients with bulbar onset.

Given that ALS is considered a multi-systemic disorder, which includes cognitive and motor impairment, it is of interest to establish a possible association between clinical symptomatology and possibly underlying changes in RSFC. The aim of the present study was to investigate differences between ALSni, ALSci, and ALS-FTD in resting-state functional connectivity. We expected to replicate group differences between ALS patients compared with HC in the DMN, VAN, and MN (Schulthess et al., 2016). We further expected a decreased functional connectivity to correlate with worse cognitive and motor function.

The primary results are summarized in Table 2. The MN and VAN comparisons between HCs and all ALS patients combined into one group were significant at p < 0.001 and remained so after family wise error correction. The same was true for between-patient group difference in the motor network (Table 3), but not in the VAN.

A number of comparisons were only significant at the more liberal P < 0.005 (Table 2). Parametric mapping expects the RSFC values to be normally distributed. If they are non-normally distributed, the effects provided by the full-factorial ANOVA may be less reliable. Simultaneously, unequal group sizes may lead to heterogeneous distributions in the outcome, affecting the ANOVA’s reliability. To reinforce our effects’ reliability, we verified the normality and homogeneity of the RSFC values in each network by group (Figure 1) before reporting the primary analyses in detail (Figures 2, 3).

We applied a seed-based RSFC approach to investigate group differences between HC, ALSni, ALSci, and ALS-FTD patients in the default mode network, motor network, and ventral attention network. When applying conservative approaches to significance, we observed moderate to large differences between HC and patient groups in the motor network. The expected hypoconnectivity across all networks compared with controls was only present at a more liberal significance level. Applying this, we observed numerous differences in RSFC between ALS subgroups, but these did not follow any specific pattern. ALSci patients and ALS-FTD patients exhibited hypoconnectivity of the MN compared with ALSni. In the VAN, ALSci, and ALS-FTD patients showed hypoconnectivity compared with ALSni. None of these between-group effects outside the MN survived the FDR corrections. In the ALS patient group, we correlated the functional connectivity with three clinical parameters: motor function, global cognitive performance, and specifically shifting ability in the executive function. Worse performance on the executive function of shifting was associated with hyperconnectivity of the VAN even after the FDR correction. Conversely, better-preserved motor function was associated with hyperconnectivity of the motor network; this effect also survived the FDR correction. Throughout the recent ALS literature, the application of statistical thresholds and corrections has ranged heterogeneously from very liberal P < 0.05 without corrections to very conservative P < 0.001 with corrections (Mohammadi et al., 2009; Jelsone-Swain et al., 2010; Agosta et al., 2013; Zhou et al., 2013; Li et al., 2018). Consequently, we will discuss our findings at P < 0.001 and P < 0.005. None of our results ran in the opposite direction of our unidirectional hypotheses.

Conflicting findings regarding ALS patients’ divergent DMN connectivity have been reported previously: Agosta et al. (2013) observed a decrease in RSFC of the right orbitofrontal cortex and an increased RSFC of the precuneus in ALSni patients; Mohammadi et al. (2009) described a hypoconnectivity in the anterior and posterior cingulate cortex, medial prefrontal cortices, and inferior parietal cortices. Our findings did not support abnormal connectivity in the precuneus (Agosta et al., 2013), but contrary to the findings of Mohammadi et al. (2009), we found increased connectivity of the cingulate cortices in ALSni patients compared with ALS-FTD patients. Zhou et al. (2013) investigated the motor network connectivity of 12 ALS patients compared with healthy controls and found hypoconnectivity in the primary motor cortex and premotor cortex. Jelsone-Swain et al. (2010) also found hypoconnectivity of the motor network in ALS patients. Our results support such relative decreases between ALSci and HC as well as ALSni. Contrastingly, we observed hyperconnectivity in the left primary motor cortex of the ALS-FTD patients compared with ALSci. Li et al. (2018) applied a frequency—rather than seed-based approach to determine functional connectivity which yielded decreased functional connectivity of the lingual gyrus, postcentral gyrus, right superior temporal gyrus, left middle temporal gyrus, bilateral basal ganglia, and the ventrolateral prefrontal cortices. Our findings partially support those Li et al. (2018) and extend them to include different ALS patient groups. We observed decreased postcentral gyrus connectivity in the motor networks of our ALS patients and hypoconnectivity, not in the right, but in the left mid-temporal lobe of the ALS-FTD patients (Figure 2D). Loewe et al. (2017) reported reduced RSFC in the occipital lobes of ALS patients. Our results support and expand this finding: the reduced connectivity occurred specifically when comparing controls with ALSni and ALSci patients. We did not find changes in the ventrolateral prefrontal cortices, even though increased RSFC of the VAN correlated with impaired executive functions (Figure 3B).

Two theoretical explanations are consistent with our findings: compensatory functional activity and the spectral nature of the ALS-FTD proteinopathies. These explanations are not mutually exclusive.

Roosendaal et al. (2010) hypothesized that subclinical structural damage in early disease is compensated for by functional activity, so clinical impairment only becomes obvious after compensatory mechanisms are exhausted. Our present and previous observations of this sample are consistent with this hypothesis: cortical thinning occurred in the motor cortex, right-sided frontotemporal cortices, and the right inferior precuneus, and was associated with cognitive impairment in a subset of this sample (Schuster et al., 2013, 2014). In a smaller subset, TDP-43 was also associated with cognitive impairment (Prudlo et al., 2016). Heimrath et al. (2014) reported that cognitive deficits were associated with hyperconnectivity in parahippocampal and parietal areas of the DMN in nine ALS patients. Moreover, faster disease progression has been related to motor network hyperconnectivity (Verstraete et al., 2010). The presently reported RSFC patterns in ALS-FTD support the notion of compensatory hyperconnectivity in the premotor cortex of the MN and the putamen and insula lobe. Simultaneously, our ALS-FTD patients had better-preserved motor function and shorter disease duration than the ALSni and ALSci groups. Putaminal lesions have been associated with executive dysfunction, while insular damage has been associated with reduced emotional awareness (Sefcsik et al., 2009; Jones et al., 2010), competencies that are impaired in ALS-FTD. Regarding the overall sample, we observed that better-preserved motor function was associated with hyperconnectivity of the MN, while worse cognitive functions were associated with hyperconnectivity. An association between hyperconnectivity of the MN and motor function has previously been documented using EEG (Dukic et al., 2019). Combined, our findings support the hypothesis that hyperconnectivity occurred in areas that were structurally damaged and with support functions vulnerable to disease-specific impairment.

ALS-FTD is increasingly recognized as a spectrum of cognitive–behavioral impairments (Murphy et al., 2007), underlain by a spectrum of proteinopathies (Geser et al., 2009; Mackenzie et al., 2010) and driven by an etiologically diverse molecular genetic spectrum (Van Langenhove et al., 2012; Müller et al., 2018). These spectra present alongside a possible morphological continuum of reduced cortical thickness, ranging from ALSni to ALS-FTD (Rohrer et al., 2009, 2010; Schuster et al., 2014; Agosta et al., 2016). Shen et al. (2018) investigated whether the cognitive profiles of ALS showed distinct perfusion patterns and found that hypoperfusion often occurred alongside atrophy. Atrophy and hypoperfusion worsened gradually between ALSni, ALSci, and ALS-FTD patients (Shen et al., 2018), suggesting that hypoperfusion occurs as a continuum. In our study, liberally determined differences in RSFC networks signified hypoperfusion between ALSni and ALSci patients, alongside hypo- as well as hyperperfusion between ALS-FTD and ALSni and ALSci patients. This suggests that RSFC is of a spectral nature: if there were a continuum with ALSni and ALS-FTD at its opposite ends, Figure 4 would contain a discernible pattern (and Figure 1 would contain progressive changes in violin shape). Instead, we see distinctive profiles between groups.

Ours was a prospective study with homogenized operating procedures and identical MRI acquisition protocols at both sites, including phantom tests [following the guidelines of the American College of Radiology (American College of Radiology, 2018)]. Analyses incorporated age, sex, location, and verbal IQ to account for their effects on connectivity and cognition, and we ascertained the normality and homogeneity of our data to document the reliability of our main results in light of our unequal group sizes (Figure 1).

Our study has three major limitations. First, uneven group sizes are undesirable even when normality and homogeneity of variance are present, and even when they reflect the natural dispersion of cognitive impairment, as is the case here. However, the proportion of the ALS-FTD group within the broader ALS cohort in our study (8.2%) would, in fact, reflect the proportional numbers expected among ALS patients in general. Second, we did not include a measure of social cognition, so we could not investigate the previously proposed association of cingulate connectivity and this domain (Trojsi et al., 2017). Third, our findings varied greatly depending on the significance levels. This may be rooted in the very small ALS-FTD group and relates back to the first limitation. Fourth, future work might also include behaviorally impaired profiles of the ALS-FTD spectrum. These challenges leave room for future methodological expansions of the present work.

In conclusion, our findings support distinct rs-fMRI network connectivity in cognitive subgroups of ALS. This encourages the notion that ALS is a spectrum disorder with ALSni, ALSci, and ALS-FTD presenting distinctly, without progression between them. Our observation that worse attentional shifting correlated directly with VAN connectivity reinforces the idea that increased connectivity is a support mechanism.

The datasets presented in this article are not readily available because the MRI data cannot be shared publicly due to ethical constraints. Requests to access these datasets should be directed to JP, johannes.prudlo@med.uni-rostock.de.

The studies involving human participants were reviewed and approved by the Ethics Committee of the University of Rostock, and of the University of Magdeburg. The patients/participants provided their written informed consent to participate in this study.

AGMT and MD developed the methodology, performed the statistical analysis, interpreted the results, wrote the original draft (Methodology, Results, and Discussion). CB conceptualized the study, conducted the statistical analyses, and wrote the original draft (Introduction). EK and JM also conceptualized the study, developed the neuropsychological battery, collected the neuropsychological data, and reviewed the manuscript. JK developed the imaging methodology, collected the imaging data, and reviewed the manuscript. SV also conceptualized the study, collected the neurological data, and reviewed the manuscript. ST also conceptualized the study, performed the statistical analysis, interpreted the results, and reviewed the manuscript. JP also conceptualized the study, collected the neurological data, acquired the grant, and reviewed the manuscript. All authors contributed to the article and approved the submitted version.

ST participated in scientific advisory boards of Roche Pharma AG, Biogen, Grifols SA, Dr Willmar Schwabe GmbH & Co. KG, and MSD, and received lecture fees from Roche Pharma AG and MSD. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.