3.1 Demographic information

We found no significant difference of age (p >.5) and sex (p >.08) between HC and ALS group (Table 1). The average age of the ALS patients was 50.6 years old, which is younger than the study reported by Ingre and colleagues (Ingre, Roos, Piehl, Kamel, & Fang, 2015), highly similar with the previous study reporting sporadic ALS in China (Chen et al., 2015). Among the ALS patients, five of them are bulbar-onset and the rest are limb-onset. Since the genetic examination was not free of charge, only five of the ALS patients underwent gene test. We found one ALS patient with SOD1 and one ALS patient with VAPB mutation. We assigned ALS patients into slow and fast progression subgroups based on the median of PR (0.41). Each subgroup contains 19 patients. No significant difference of age (p >.06) and sex (p >.7) between the two subgroups were found. The slow progression group showed less severe symptoms (higher ALSFRS-R score) than the fast progression patients (T(36) = 3.75, p = .001). Disease duration was significantly longer in slow progression patients (T(36) = 2.22, p = .033). There was no group difference of age and sex between the HCs and ALS patients in the slow and fast group (p values >.11, Table S1).

3.2 Decreased ALFF and network FCS in the PSMA

Voxel-wise two-sample T-test on the ALFF maps showed significantly decreased ALFF in the bilateral PSMA (Figure 1a). Further ROI analysis revealed that both the slow and fast progression groups showed significantly decreased ALFF in both the left and right PSMA (5 mm sphere ROI around peak voxels), while the difference of ALFF between the two ALS subgroups was not significant (Figure 1b,c). The whole brain exploratory analysis revealed that only PSMA ALFF showed significant reduction (T <−3.24) while the posterior cingulate cortex showed increased ALFF at an uncorrected threshold (T >3.24). To provide a view of the whole brain for any potential difference, the whole brain p < 0.05 uncorrected result was provided in Figure S1. ANOVA analysis revealed similar distribution of the group differences in the bilateral PSMA at p <.001 uncorrected threshold. However, no voxels survived FDR correction (Figure S2).

Reduced primary sensorimotor area (PSMA) amplitude of low frequency fluctuation (ALFF) in amyotrophic lateral sclerosis (ALS) patients. Panel a: two-sample T-tests showed decreased ALFF in ALS patients (p <.05 FDR corrected in the PSMA). The averaged ALFF extracted from spherical ROIs centered at the left and right peak voxels of reduced ALFF showed significantly reduced ALFF for both fast and right progression ALS patients in both the left and right PSMA, while the two subgroups showed no significant difference (panels b and c). Error bars represent standard error

T-test revealed that there was a significant reduction of network FCS in the ALS patients (T(73) = −2.35, p = .022). Specifically, the network FCS was significantly reduced in the slow progression patients (T(54) = −2.80, p = .007) compared to HCs while no significant difference was found in the fast progression patients (p = .25, Figure 2).

Reduced primary sensorimotor area (PSMA) network functional connectivity strength (FCS) in amyotrophic lateral sclerosis (ALS) patients. Panel a illustrates the parcellated 102 nodes and their connections within the entire PSMA. Panel b shows the significant reduced FCS of the PSMA in slow progression patients, while there was no significant difference between the two ALS groups, nor between the HC and fast progression group. Error bars represent standard error

It should be noted that the network FCS was to measure the functional integration of the whole PSMA. To further investigate whether the part showing significantly reduced ALFF also show impaired FC, we calculated the FC between the sphere ROIs (Figure 1b,c) located at the peak voxel in each hemisphere. Unlike the network FCS which measures the overall connectivity within the PSMA, the ROI FC measures only the connectivity between the area that showed most significant reduction of brain activity. We found that both the fast (T(54) = −2.43, p = .018) and slow (T(54) = −3.10, p = .003) progression groups showed significant reduced ROI FC compared with HCs. However, there was no significant difference between the fast and slow subgroups (p = .49). Whole brain p <.05 uncorrected results of the seed FC maps are provided in Figures S3 and S4. ANOVA model also revealed significant group effect (F(2,72) = 3.46, p = .037).

3.3 Correlation between RS-fMRI metrics and clinical features in the slow and fast progression groups

When taking all ALS patients together, no significant correlations were detected between the RS-fMRI metrics and clinical measurements including ALSFRS-R score (p values ≥.1), disease duration (p values ≥.08, one subject with a 204-month duration was excluded) and UMN score (p values ≥.62). Detailed information was provided in Table S2.

We then performed correlation analyses for the slow and fast progression groups separately. There were no significant correlations of the ALSFRS-R score with either the bilateral ALFF or the network FCS (p values >.2, Figure 3a,b) in the slow progression patients. But in the fast progression patients, we found significant positive correlations of the ALSFRS-R score with both the bilateral ALFF (R = .62, p = .005) and the network FCS (R = .49, p = .034) (Figure 3c,d). No significant correlations were found between the ROI FC and the clinical measurements in any ALS groups (p values >.053).

Correlations of PSMA amplitude of low frequency fluctuation (ALFF) and network functional connectivity strength (FCS) with ALFRS-R. Correlations of ALSFRS-R score with ALFF and network FCS in the slow progression group (Panels A and B) and fast progression group (Panels C and D). *: p <.05. ALSFRS-R: ALS Functional Rating Scale-Revised

Since there was a significant correlation between the PSMA local activity (ALFF) and network FCS in the ALS group (R = .7, p <.001, Figure S5) but not in the healthy group (R = .26, p = .13), we further performed partial correlation analyses between the clinical features and RS-fMRI metrics in the slow and fast progression groups separately. In the fast progression ALS group, while regressing out the ALFF or network FCS, the network FCS or ALFF had no longer significant correlation with the ASLFRS-R (Figure 4c,d). But in the slow progression ALS group, regressing out the network FCS yielded no significant correlation between ALFF and ALSFRS-R (p = .19, Figure 4a), however, regressing out the ALFF yielded a significant negative correlation between the network FCS and ALSFRS-R (R = −.54, p = .026. Figure 4b). Partial correlation analyses for the slow progression patients were based on 18 ALS patients since one patient (disease duration = 48, ALSFRS-R = 33) was detected as significant outlier (partial regression model's Cook's distance = 1.25).

Partial correlations of PSMA amplitude of low frequency fluctuation (ALFF) and network functional connectivity strength (FCS) with ALFRS-R. Partial correlation analysis of ALSFRS-R with ALFF (network FCS as covariate of noninterest. Panels a and c) or with the network FCS (ALFF as covariate of noninterest. Panels b and d) in the slow progression group (Panels a and b) and fast progression group (Panels b and d). Cov: regressing out the covariate

Despite its reduction in the slow progression group compared to HCs, the network FCS may play a compensatory role as revealed by the negative correlation with ALSFRS-R when the ALFF effect was controlled. An interesting question is that which component of the network FCS in the whole PSMA contributes more to this potential compensatory effect. We thus calculated the FCS between the left and right hemisphere (namely inter-hemisphere FCS) and within the left and right hemisphere (average of the left and right PSMA FCS, namely intra-hemisphere FCS). Both the inter-hemisphere FCS (T(73) = −2.20, p = .031) and intra-hemisphere FCS (T(73) = −2.34, p = .022) were reduced in all ALS groups. Specifically, inter- (T(54) = −2.52, p = .015) and intra-hemisphere (T(54) = −2.87, p = .006) FCS in the slow progression patients were reduced while the comparisons between the two subgroups as well as for fast progression patients versus HCs were not significant (p values >.19, Figure 5a,b). While regressing out the ALFF in the slow progression group, the inter-hemisphere FCS showed relatively more significant negative correlation with the ALSFRS-R (R = −.58, p = .014) (Figure 5c) than the intra-hemisphere FCS (R = −.45, p = .071) (Figure 5d). The same outlier was excluded in the partial correlation analyses as we mentioned in the previous paragraph.

Inter- and intra-hemisphere network functional connectivity strength (FCS). Inter- (panel a) and intra-hemisphere (panel b) FCS in the amyotrophic lateral sclerosis (ALS) patients. Panel c shows, in the slow progression group, the ALSFRS-R score showed significant negative correlation with the inter-hemisphere FCS (p = .014) but not significant with intra-hemisphere FCS was as shown in panel D (p = .071) when amplitude of low frequency fluctuation (ALFF) was controlled as covariate of noninterest to remove the shared variance with network FCS

We further decomposed the network FCS into FCS within/outside the area that showed reduced ALFF and replicated the partial correlation analyses in the slow progression group. Both the FCS within and outside the area showing reduced ALFF were correlated negatively with the ALSFRS-R, yet the correlation between within-FCS and the ALSFRS-R was more significant (R = −.61, p = .004) than that of the outside-FCS (R = −.50, p = .035).