Cognitive fatigue and cortical-striatal network in old age

Demographic and behavioral data analysis

The sample characteristics, as well as their correlations with CF in the entire sample are shown in Table 1. There was a significant subgroup (high- vs. low-CF) by block (1 vs. 6) interaction on fatigue score from VAS (F(1, 24) = 48.67, p < 0.001) (Figure 2B). Figure 2C shows a significant CF subgroup (low- vs. high-CF) by task block (1 vs. 6) interaction on IIVRT (F(1, 24) = 4.81, p = 0.038). Consistently, IIVRT analysis showed the task performance became worse in individuals with higher CF, supporting our approach of low- and high-CF group definition. IIVRT was positively correlated with CF scores (r = 0.42, p = 0.004), but not MOCA (r = -0.11, p = 0.49) (Figure 2D). Additionally, there was neither significant interacting effect between task type and block (F(1, 44) = 0.23, p = 0.64) on IIVRT, nor main effect of task type on IIVRT (F(1, 44) = 1.03, p = 0.32).

Figure 2. CF profile for the entire sample. (A) The distribution of CF change (post-pre) induced by cognitive tasks during fMRI scanning. According to different perceived fatigue, two groups were selected: low-CF subgroup (12 subjects with CF change ≤ 0, shown in blue) and high-CF subgroup (14 subjects with CF change ≥ 1, shown in red). In addition, there were 20 subjects outside the high- or low-CF subgroups (shown in grey). (B) The VAS fatigue score before and after performing cognitive tasks for each group, showing significant difference between high- and low-CF subgroup. (C) The IIVRT in Block1 and Block6 for each group showed the performance change with time by CF subgroups. There were significant different trends in IIVRT change between high- and low-CF subgroup. (D) Higher CF was significantly associated with larger IIVRT change (worse cognitive performance). Note: CF, cognitive fatigue; IIVRT, intra-individual variability of reaction time. low-CF subgroup: blue; high-CF subgroup: red; others: gray.

Cortical-striatal network in low- and high-CF subgroups

Examining the interaction effects of CF subgroup (low- vs. high-CF) and task block (1 vs. 6), we identified four significant regions in the left cortical-striatal network: left superior frontal gyrus (SFG), right anterior cingulate cortex (ACC), left posterior cingulate cortex (PCC), and right cuneus, and four significant regions in the right cortical-striatal network: left medial frontal gyrus (MedFG), left middle frontal gyrus (MFG), left MFG2, and left occipital gyrus (OG). Table 2 provides a description on each region. Figure 3 displays the relationship between functional connectivity of each cortical-striatal network across two task blocks by CF subgroup.

Figure 3. The interaction effects of CF subgroup and task block on left and right cortical-striatal network connectivity. For the left cortical-striatal network, there were significant differences of connectivity change in L SFG, R ACC, L PCC and R Cuneus. For the right cortical-striatal network, there were significant differences of connectivity change in L MFG, L MFG2, L MedFG and L OG. During the fatiguing tasks, the connectivity strength of cortical-striatal network increased in low-CF group, while deceased in high-CF group. Note: L, left; R, right; SFG, superior frontal gyrus; ACC, anterior cingulate cortex; PCC, posterior cingulate cortex; MedFG, medial frontal gyrus, MFG, middle frontal gyrus; OG, occipital gyrus. CF, cognitive fatigue.

Correlations between cortical-striatal network and CF

To examine correlations between the cortical-striatal network and CF in the entire sample, partial correlation was applied to each brain region we found above controlling for age, education and whole-brain gray matter volume. Greater CF was significantly related to less increase in functional connectivity between striatum and the four frontal regions identified above, but not posterior regions, for False Discovery Rate (FDR)-corrected p < 0.05 (Figure 4). We also examined correlations between striatal network and CF for Stroop and 1-back tasks separately, and found similar patterns (Supplementary Figure 2).

Figure 4. Correlations between connectivity change and CF in the entire sample. In the cortical-striatal network, the partial correlation analysis showed CF change was significantly correlated with the anterior regions, including L SFG, R ACC, L MFG, LMFG2 and L MedFG. The CF change showed negative correlation trend with the posterior regions but not significant, including L PCC, R Cuneus and L OG. Note: L, left; R, right; SFG, superior frontal gyrus; ACC, anterior cingulate cortex; PCC, posterior cingulate cortex; MedFG, medial frontal gyrus, MFG, middle frontal gyrus; OG, occipital gyrus. p values here are FDR-corrected across all regions. CF, cognitive fatigue; IIVRT, intra-individual variability of reaction time.

Anterior vs. posterior cortical-striatal network analysis

To examine the relationship between anterior and posterior cortical-striatal network and CF, we calculated the functional connectivity for anterior and posterior cortical-striatal networks separately. There was a significant subgroup (high- vs. low-CF) by block (1 vs. 6) interaction on the functional connectivity for both anterior (F_{(1, 24)} = 16.21, p < 0.001) and posterior network (F_{(1, 24)} = 8.34, p = 0.001) (Figure 5A). Using the entire sample, the CF change was significantly correlated with both change of functional connectivity in the anterior network (r = -0.45, p = 0.002), and posterior network (r = -0.40, p = 0.04). Notably, the correlation between CF change and anterior network was more robust relative to the posterior network (Figure 5B).

Figure 5. Relationships between the anterior and posterior cortical-striatal network functional connectivity and CF. (A) During the cognitively fatiguing tasks, both of the anterior and posterior network showed significant interaction effects between subgroup (high- vs. low-CF) and block (1 vs. 6), that connectivity strength increased in low-CF group while decreased in high-CF group (B) Although CF was negatively associated with change of anterior or posterior connectivity, the negative relationship was more robust in anterior cortical-striatal network. Note: CF, cognitive fatigue; IIVRT, intra-individual variability of reaction time.

Posterior-anterior shift, CF and cognitive function

PAS was calculated to measure the posterior-anterior shift over the course of cognitively fatiguing tasks. Moderation analysis was applied to examine the influence of CF level on the relationship between PAS and cognitive function (Figure 6). CF significantly interacted with PAS in predicting task performance during the fatiguing tasks (R² = 0.29, F(6, 39) = 2.62, p = 0.031); greater PAS was related to better cognitive performance in the low-CF subgroup while worse cognitive performance in the high-CF subgroup. CF also significantly interacted with PAS in predicting executive function (R² = 0.32, F(6, 39) = 3.06, p = 0.015). We also calculated the moderation models for anterior and posterior cortical-striatal network separately. Of note, there was a significant moderating effect of CF change on anterior, but not posterior, network and IIVRT change/executive function (see Supplementary Figure 3).

Figure 6. Moderation analysis of relationship between PAS and cognitive function. There was a significant interacting effect between CF and PAS in predicting task performance, showing greater PAS was related to better cognitive performance (smaller IIVRT) in the low-CF subgroup while worse cognitive performance (larger IIVRT) in the high-CF subgroup. Consistently, CF also significantly interacted with PAS in predicting executive function. Note: CF, cognitive fatigue; IIVRT, intra-individual variability of reaction time. PAS, posterior-anterior shifting.

Participants

Fifty-two older adults with normal cognition, indexed by Montreal Cognitive Assessment ≥ 26 and Rey Auditory Verbal Learning Test delayed recall ≥6, were recruited in the current project. Participants were required to have intact capacity to give consent (indexed by intact score at UCSD Brief Assessment of Capacity to Consent, UBACC), have adequate visual and auditory acuity for testing by self-report, be free from major depression (indexed by Geriatric Depression Scale-15 item score <6), sleep disorder (indexed by Pittsburg Sleep Quality Inventory global sleep quality score < 14), and chronic fatigue (indexed by 20-item Multidimensional Fatigue Inventory general fatigue subscale < 20), be ≥60 years of age, English-speaking, and community-dwelling. Exclusion criteria included (1) potential disease or medication confounded with fatigue symptoms: neurologic or vascular disorders (e.g., multiple sclerosis, stroke, transient ischemic attack, heart attack, or traumatic brain injury within the past five years, severe cerebrovascular disease, Parkinson’s disease); an episode of a diagnosed and active psychiatric disorder (i.e., major depression, anxiety, bipolar disorder) within the past five years; schizophrenia (regardless of the time since the last episode); a clinical diagnosis of mild cognitive impairment or dementia as defined by the most current version of DSM; change in beta-blocker dosage within the past three months; and (2) MRI contraindications (e.g., pacemaker, metallic implant, claustrophobia). The study was approved by the university’s research subject review board (IRB No. RSRB00063603). All the participants were required to understand the experimental procedure, answer UBACC correctly, and sign the informed consent.

CF assessment

An 18-item visual analogue scale (VAS) was used to assess fatigue before and immediately after MRI scan. The VAS was adapted [33], using a Likert scale ranging from 0 (“not at all") to 10 (“very much”). Participants were instructed to rate their current state based on items related to mood and energy (e.g. tired, keeping your eyes open is difficult, lively, efficient). All “alertness” items were reversely coded and a mean score across all items was developed with higher scores indicating higher fatigue. The Cronbach’s alpha of the two measures were both >.90. CF was computed as the discrepancy between fatigue mean scores after and before MRI scan, where higher scores indicated more fatigue after the tasks. There is no standard cutoff score for defining fatigue vs. non-fatigue after performing cognitive tasks in older adults, especially since the meaning of scores can be varied depending on the type, duration, and intensity of the manipulation task. To better extract the CF related cortical-striatal network, we first selected two subgroups with distinct CF sensitivity, based on the entire sample’s CF score distribution, as the low-CF (CF≤ 0, n = 12) and high-CF (CF ≥ 1, n = 14) group (see Figure 2A). We next used the entire sample’s CF score for further analyses.

Design and CF task protocol

An MRI related study visit occurred during a controlled morning window (8-12am) to avoid a potential diurnal fluctuation in the CF indices. Participants were instructed to eat breakfast but to avoid nicotine, caffeine, and exercise for at least 2 hours before their arrival. Participants rested for 5 minutes upon arrival, and rated pre-task fatigue using an 18-item VAS [33]. The MRI scan started with a T1-weighted sequence, followed by cognitively fatiguing task related BOLD fMRI. Another VAS rating of post-task fatigue was completed immediately after the BOLD fMRI.

To induce fatigue, the participant was required to complete two cognitive tasks in the scanner: Color-word Stroop task and Dual 1-back task (Figure 1). Event-related design was used. In the Stroop task, participants were shown serial colored words on the screen, and asked to judge the color of the word regardless of the meaning of the word. In the Dual 1-back task participants were shown an English letter on the screen and asked to judge whether the current stimulus matched the letter and position of the previous one. For both tasks, a visual stimulus can last up to 4 seconds; participants were required to make response before the stimulus disappeared. A feedback was provided for 1.5 seconds. A 3-second interval was provided between stimuli. For Stroop task, an extra 0.5-second fixation (white dot before the onset of stimulus) was placed to remind the participant of preparing for the appearance of next stimulus. For 1-back task, since the task required a comparison between stimuli, subject did not need a 0.5 second fixation for reminding. Each of the tasks lasted 15 minutes, and the order of the two tasks was randomized across participants.

Behavioral data analysis

In the two cognitively fatiguing tasks, data on the accuracy and reaction time (RT) in response to each stimulus were collected. To assess cognitive performance, we calculated intra-individual variability of reaction time (IIVRT), using standard deviation divided by average reaction time to trials with correct responses for each task. IIVRT is considered a better measurement of cognitive function than accuracy or RT in aging [34], which ensures comparability of performance between two types of task. Before calculating IIVRT, the two cognitive tasks were first sorted based on the task order irrespective of the type. Then we divided each task into 3 blocks (5 minutes each), and calculated the IIVRT within each block. Cognitive performance throughout the protocol was calculated as the IIVRT difference between Block 6 and Block 1, with lower scores indicating better cognitive performance.

Assessment of overall executive abilities

General cognitive performance was assessed using tests reflecting working memory (1-back with calculation as the distraction task) and cognitive control (Set shifting task) that are conceptually similar to the cognitively fatiguing tasks, but format-wise different, from the EXAMINER package [35]. Higher composite score indicated better executive function. This additional outside-MRI measure of general cognitive capacity was included to test whether the relationship between CF and PASA model can be applicable to more generalized circumstances.

Imaging data acquisition

Imaging data were collected at the Rochester Center for Brain Imaging using a 3T Siemens TrioTIM scanner (Erlangen, Germany). The fMRI scan began with a MPRAGE scan (TR/TE = 2530/3.44 ms, TI = 1100 ms, FA = 7, matrix = 256×256, resolution 1×1×1mm, slice thickness = 1 mm, 192 slices). The task-related fMRI data were collected using a gradient echo-planar imaging (EPI) sequence (TR/TE = 2500ms/30ms, FA = 90, slice thickness = 4 mm, matrix = 64 × 64, 4×4 mm in-plane resolution, 42 axial slices, volumes = 360), lasting 15 minutes for each cognitively fatiguing task.

Gray matter volume

Voxel-based morphometry (VBM) analysis was performed to generate a whole-brain gray matter map using SPM8. First, the structural images were segmented into gray matter, white matter and cerebrospinal fluid. Second, a gray matter template was generated through an iteratively nonlinear registration using DARTEL, a toolbox with a fast diffeomorphic registration algorithm [36]. The generated gray matter template was used for normalizing functional images to MNI space. Third, the gray matter mask was applied to extract global gray matter volume from individual’s gray matter image. Since brain atrophy is a crucial factor of neurodegeneration, the whole-brain gray matter volume was used as a covariate.

Functional imaging data preprocessing

The fMRI data were preprocessed using the Data Processing Assistant for Resting-State fMRI (DPARSFA) [37] based on SPM8 (http://www.fil.ion.ucl.ac.uk/spm/). For each participant, the first 6 volumes per task were excluded to obtain steady-state tissue magnetization. The remaining imaging data were corrected for slice timing and head-motion, co-registered to their own structural images, and normalized to the Montreal Neurological Institute (MNI) standard space. All processed fMRI images were resampled to 3×3×3 mm, and smoothed using a Gaussian kernel (FWHM 6 mm). Six subjects were excluded from the formal analysis due to low quality coregistration and head motion greater than 2 mm or 2 degrees. In addition, given the commonality of head motion in old age and its confounding effect [38], we also examined the relationship between head motion and CF, which showed no significant correlations.

Event-related functional connectivity analysis

We applied beta-series correlation analysis, which allows the investigation of functional connectivity in event-related fMRI data [39,40]. To examine the functional change over time, the two fMRI tasks were sorted based on the order irrespective of the type. Then we divided the series of beta-values across the 30-minute protocol into 6 blocks (5 minutes each), the same time segmentation as calculating block-based IIVRT described above. A general linear model (GLM) was applied to estimate the evoked activity within a block; Block-based beta-values related to a given experimental condition was therefore generated for each voxel. To generate the cortical-striatal network, the bilateral striatum were selected as seeds based on the Automated anatomical labeling (AAL) template [41]. The individual’s cortical-striatal networks were generated using beta-series correlation between a seed and a candidate voxel in the BASCO toolbox [40]. We applied repeated-measures ANOVA to examine the interaction effects of CF subgroup (low vs. high) by task block (block1 vs. block6) on the block-based beta-series correlations between the seeds (striatum at each hemisphere) and candidate voxels (whole brain except the striatum being used as the seed), to identify the functional connectivity with significant changes over blocks that differed between CF subgroups. Statistical F maps were generated with threshold at p < 0.05 by GRF correction (individual p < 0.005, cluster p < 0.05).

PASA model

To examine the PASA model in CF related cortical-striatal network, the averaged strength of functional connectivity was calculated, based on beta-series correlations, across regions for anterior and posterior brain respectively, regardless of hemisphere. Posterior-anterior shift (PAS) score was defined as

PAS score = Block6(Anterior – Posterior) – Block1(Anterior – Posterior)

to estimate the shifting degree from posterior to anterior regions with higher scores indicating greater reliance on anterior regions at later time points.

Other statistical analysis

All other statistical analyses were conducted using SPSS 22.0 (IBM Corporation, Armonk, NY). Non-parametric group comparison was conducted for the comparison of high- vs. low-CF subgroups. We also used repeated measures ANOVA to examine the interaction effect between tasks (pre- vs. post-) and groups (high- vs. low-CF) on CF, as well as the interaction effect between task blocks (1 vs. 6) and groups (high- vs. low-CF) on IIVRT. Partial correlation was applied to examine relationships between CF, cortical-striatal networks, and task performance, controlling for age, education and whole-brain gray matter volume. Moderation analysis was used to test whether CF would moderate the relationship between PAS score and cognitive performance using macro PROCESS (http://www.processmacro.org), controlling for age, education and whole-brain gray matter volume.