Physical activity and frontoparietal network connectivity in traumatic brain injury

Abstract Background Prolonged changes to functional network connectivity as a result of a traumatic brain injury (TBI) may relate to long‐term cognitive complaints reported by TBI survivors. No interventions have proven to be effective at treating long‐term cognitive complaints after TBI but physical activity has been shown to promote cognitive function and modulate functional network connectivity in non‐injured adults. Therefore, the objective of this study was to test if physical activity engagement was associated with functional connectivity of the cognitively relevant frontoparietal control network (FPCN) in adults with a TBI history. Methods In a case‐control study design, resting state function magnetic resonance imaging and physical activity data from a subset of participants (18–81 years old) from the Cambridge Centre for Ageing and Neuroscience (Cam‐CAN) study was analyzed. Fifty‐seven participants reported a prior head injury with loss of consciousness and 57 age and sex matched controls were selected. Seed‐based functional connectivity analyses were performed using seeds in the dorsolateral prefrontal cortex and the inferior parietal lobule, to test for differences in functional connectivity between groups, associations between physical activity and functional connectivity within TBI as well as differential associations between physical activity and functional connectivity between TBI and controls. Results Seed‐based connectivity analyses from the dorsolateral prefrontal cortex showed that those with a history of TBI had decreased positive connectivity between dorsolateral prefrontal cortex and intracalcarine cortex, lingual gyrus, and cerebellum, and increased positive connectivity between dorsolateral prefrontal cortex and cingulate gyrus and frontal pole in the TBI group. Results showed that higher physical activity was positively associated with increased connectivity between the dorsolateral prefrontal cortex and inferior temporal gyrus. Differential associations were observed between groups whereby the strength of the physical activity‐functional connectivity association was different between the inferior parietal lobule and inferior temporal gyrus in TBI compared to controls. Discussion Individuals with a history of TBI show functional connectivity alterations of the FPCN. Moreover, engagement in physical activity is associated with functional network connectivity of the FPCN in those with a TBI. These findings are consistent with the evidence that physical activity affects FPCN connectivity in non‐injured adults; however, this effect presents differently in those with a history of TBI.

nectivity association was different between the inferior parietal lobule and inferior temporal gyrus in TBI compared to controls.
Discussion: Individuals with a history of TBI show functional connectivity alterations of the FPCN.Moreover, engagement in physical activity is associated with functional network connectivity of the FPCN in those with a TBI.These findings are consistent with the evidence that physical activity affects FPCN connectivity in non-injured adults; however, this effect presents differently in those with a history of TBI.

K E Y W O R D S
frontoparietal control network, physical activity, traumatic brain injury

INTRODUCTION
Traumatic brain injury (TBI) is a prominent neurological disorder, which results in significant emotional, cognitive, and economical burden (Maas et al., 2022).TBIs result from an impact to the head leading to both contact and inertial forces, which initiate a cascade of secondary injury mechanisms affecting molecular, cellular, and network processes (Naumenko et al., 2023).Changes in functional connectivity, the temporal covariation of the blood-oxygen-level-dependent (BOLD) signal (Sharp et al., 2014), have been reported acutely after injury and are primary driver of patient-reported symptoms and cognitive deficits post-TBI (Morelli et al., 2021;Puig et al., 2020).While some individuals may recover from this dysfunction, others exhibit persistent (over the first year post-injury) alterations in functional connectivity in large-scale networks such as the frontoparietal control network (FPCN; Churchill et al., 2017;DiFabio et al., 2022;Palacios et al., 2017).Yet less is known about how TBI-induced changes in functional connectivity can persist in the years following injury.Moreover, no efficacious interventions exist to treat the negative effects of TBI, either in the acute or chronic phases of injury.Healthy lifestyle habits, however, such as physical activity, may provide a sustainable and longterm approach to maintaining brain health in those with a history of TBI.
Physical exercise and physical activity-based interventions have been shown to improve cognitive function and modulate functional network connectivity in non-injured populations by promoting changes to cellular and molecular mechanisms (Yu et al., 2021).Notably, physical exercise can induce neuroplasticity (de Sousa Fernandes et al., 2020), leading to changes in large-scale cortical networks (Moore et al., 2022), including alterations in functional connectivity within the FPCN in older adults (Voss et al., 2010).Similarly, functional connectivity of the FPCN has been shown to support adherence to physical exercise interventions in older adults (Morris et al., 2022).
Physical activity engagement seems to be increasingly important for those with a prior TBI, with work demonstrating that those with a TBI have a greater association between physical activity and global and cognitive health (Morris et al., 2019 (Morris et al., 2019) in TBI, yet more evidence of its effect on objective brain outcomes is necessary to promote this behavior in those living with a TBI.
Physical activity engagement is thought to be supported by executive functioning (Best et al., 2014;Cheval et al., 2020;Daly et al., 2015;Savikangas et al., 2021), highlighted in the dual processing theory (Brand & Ekkekakis, 2018) and temporal self-regulation theory (Hall & Fong, 2015).Executive functioning includes set-switching, monitoring and updating information, and inhibiting prepotent responses (Miyake et al., 2000), allowing individuals to take top-down control of behaviors and self-regulate (Hofmann et al., 2012).Moreover, executive functioning can aid in suspending default habits (sedentary behavior) and engaging in healthy lifestyle behaviors with effort (physical activity) (Buckley et al., 2014;Hall & Fong, 2015).Evidence shows that the FPCN includes regions identified as supporting cognitive control and decision-making processes, including the dorsolateral prefrontal cortex, anterior cingulate cortex, and inferior parietal lobule (Dosenbach et al., 2008;Seeley et al., 2007;Vincent et al., 2008).Specifically, we focused on the inferior parietal lobule and the dorsolateral prefrontal cortex.The inferior parietal lobule and the dorsolateral prefrontal cortex have distinct, complementary roles that allow the FPCN to support a range of executive functions and attentional processes, ensuring that behavior is goal directed and adaptable.The inferior parietal lobule is responsible for attention, spatial awareness, and coordination of actions (Andersen & Cui, 2009;Culham & Kanwisher, 2001).The dorsolateral prefrontal cortex is responsible for executive functioning, goal-directed movement, and adaptive behaviors (Mansouri et al., 2009;Miller & Cohen, 2001;Tanji & Hoshi, 2008).Moreover, the inferior parietal lobule and dorsolateral prefrontal cortex have both been implicated in TBI (Mayer et al., 2011), specifically with the execution of movement (Gooijers et al., 2016).Therefore, the dorsolateral prefrontal cortex and inferior parietal lobe are key regions to the FPCN This current study first aimed to test differences between FPCN functional connectivity in those with a prior TBI compared to age-and sex-matched controls.Second, the study aimed to test if there was an association between physical activity and FPCN functional connectivity in those with a prior head injury.To determine if this was specific to those with a history of TBI, the study aimed to test if there was a differential relationship between physical activity and FPCN functional connectivity in TBI and matched controls.

Participants
This is a case-control study using a subset of participants from a population-based study of 708 participants recruited as a part of the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study (data used in this work were available at: https://www.mrc-cbu.cam. ac.uk/datasets/camcan/) (Taylor et al., 2017).See Shafto et al. (2014) for more details on the primary study (Shafto et al., 2014).The ethical approval for the study was obtained from the Cambridgeshire 2 Research Ethics Committee.Participants gave written informed consent.Out of the 708 participants, 63 answered "yes" to the question "Have you ever had a serious head injury and been unconscious after it?(Have you ever been knocked out?)."If they answered "yes," participants also provided the age at which their injury occurred.Of those 63 participants, 57 had complete data across all variables used in this analysis.Note that age 57 years (range = 19-81 years old), sex (65% female), and education (60% with a degree) randomly matched controls using the MatchIt package in R were included in the analysis for a total of 114 participants.Full sample characteristics are shown in Table 1.

Physical activity
Physical activity energy expenditure (physical activity) was gathered as part of a larger self-completed questionnaire.The questions about physical activity were based on the European Prospective Investigation into Cancer Study-Norfolk Physical Activity (Wareham et al., 2002).Individual total physical activity per day (kJ/day/kg) was calculated from self-reported activities into metabolic equivalents, based on the standard definition of 1 MET as 3.5 mL O 2 /min/kg (or 71 J/min/kg) based on the resting metabolic rate (Henry, 2005).Physical activities probed in this questionnaire encompass any physical activity performed at work, at home, during leisure, and commuting typically over the past 12 months.The full questionnaire is found in the supplementary materials.(Behzadi et al., 2007) noise components within white matter and CSF were estimated by computing the average BOLD signal as well as the largest principal components orthogonal to the BOLD average, and motion parameters within each eroded segmentation mask.The mean motion was calculated as the scan-to-scan average framewise displacement (FD) for each participant based on the six motion parameters (Power et al., 2012).No participant was removed from this analysis for excessive head motion of >0.9 mm mean motion.

Statistical analysis
Group-level functional connectivity analyses were performed using ROIs; First, we tested if there were differences in functional connectivity in those with a history of TBI versus matched controls.Second, we tested if there was an association between physical activity and functional connectivity in just the TBI group.Third, we tested if there was a differential association between functional connectivity and physical activity in TBI and matched controls.Then, we used the "interactions" (Long, 2021) package in R to calculate simple slopes of the relationship between functional connectivity and physical activity in each group.
We tested between group differences with an analysis of covariance while controlling for age, sex, education, and years since injury.Model assumptions were checked using Q-Q residual versus fitted plots to assume dichotomy of outcome variables and linearity.We first tested for differences in mean head motion.We then tested for differences in physical activity between TBI and matched controls.

Functional connectivity in TBI versus controls
Results revealed no significant differences in mean head motion (F = .01,p = .91)(Table 1).No participants were removed due to excessive head motion.
To test for differences in the group of individuals who reported having a prior head injury, seed to voxel results from the dorsolateral prefrontal cortex revealed four clusters that had significantly different functional connectivity between those with a history of TBI versus matched controls (Figure 2).All clusters are voxel p < .001uncorrected, cluster p-FWE < .05.Full statistical information and coordinates are shown in Table 2.A cluster within the cingulate gyrus showed increased connectivity in TBI with both the right and left dorsolateral prefrontal cortex (Figure 2a).A cluster spanned the intracalcarine cortex and lingual gyrus and showed decreased connectivity in TBI with both the right and left dorsolateral prefrontal cortex (Figure 2b).A cluster sitting within the frontal pole showed increased connectivity in TBI with both the right and left dorsolateral prefrontal cortex (Figure 2c).A cluster sitting within the cerebellum showed decreased connectivity in TBI with both the right and left dorsolateral prefrontal cortex (Figure 2d).Functional connectivity of the inferior parietal lobule was not significantly different between groups.

The association between functional connectivity and physical activity in TBI
Seed to voxel results from the dorsolateral prefrontal cortex revealed one cluster that was associated with physical activity in TBI (Voxel p < .001uncorrected, Cluster p-FWE < .05)sitting within the inferior temporal gyrus, showing increased connectivity is associated with increased physical activity (Figures 3a and 4).Full statistical information and coordinates are shown in Table 2. Seed to voxel results from the inferior parietal lobule revealed no clusters that were associated with physical activity in TBI.

Physical activity in TBI versus Controls
There were no differences between physical activity and TBI status (F = 1.23, p = .27)(Figure 5a).
Seed to voxel results from the dorsolateral prefrontal cortex from the interaction between physical activity in TBI and physical activity in controls revealed no clusters that had significant differential connectivity between groups.Seed to voxel results from the inferior parietal lobule from the interaction between physical activity in TBI and physical activity in controls revealed one cluster that had significant differential connectivity between groups sitting within the inferior temporal gyrus (Figure 5b).Full statistical information and coordinates are shown in Table 2.The TBI group demonstrated a positive relationship between the right inferior parietal lobe and inferior temporal gyrus and physical activity, while the control group demonstrated a negative relationship between those two regions.The false discovery rate corrected simple slopes demonstrated a significant relationship between the right inferior parietal lobule and the inferior temporal gyrus in the controls (Figure 5d).The interaction effect is less strong in the left inferior parietal lobe (Figure 5e).

DISCUSSION
Here we show that the functional connectivity between the dorsolateral prefrontal cortex and four clusters across the cortex exhibited differences in those with a TBI history compared to those with no TBI history.A general pattern of hypoconnectivity of long-range cortico-cortical connections and hyperconnectivity of short-range cortico-cortical connections was seen, where two posterior clusters (intracalcarine cortex and cerebellum) showed decreased connectivity in the TBI group, while the two anterior clusters (cingulate gyrus and frontal pole) showed increases in connectivity in the TBI group.Moreover, seed to voxel results revealed an association between physical activity and functional connectivity between the dorsolateral prefrontal cortex and the inferior temporal gyrus in those with a prior TBI, with greater physical activity associated with greater connectivity.
The associations between physical activity and functional connectiv-ity between the inferior parietal lobule and a distinct cluster within the inferior temporal gyrus was different in those with TBI compared to matched controls, with a positive relationship between functional connectivity and physical activity in TBI and a negative relationship in controls.
Overall, participants with a prior TBI showed hyperconnectivity in short-range connections and hypoconnectivity in long-range connections compared to age-and sex-matched controls.These functional connectivity shifts may result from changes in firings of neurons, blood flow, oxidative metabolism, or a combination of these factors (Fox & Raichle, 2007;Mayer et al., 2011).Mechanistic processes that would affect the BOLD signal occur in TBI, such as reductions in cerebral perfusion (Soustiel & Sviri, 2007), increased reactivity of smooth muscle in the walls of vessels (Ueda et al., 2006), metabolic failure (Soustiel & Sviri, 2007), and decreases in the size of capillaries (Obenaus et al., 2017).These results are also supported by prior work showing FPCN connectivity differences between TBI and controls early after injury (DiFabio et al., 2022;Palacios et al., 2017).Interestingly, differences in functional connectivity in this study were seen in a cohort of participants regardless of age with varying times since injury, suggesting that these results may reflect a common change in the functional connectivity of the FPCN following a TBI with loss of consciousness.
The main aim of the study was to assess if engagement in physical activity was associated with functional connectivity of the FPCN in those with a TBI history.If so, physical activity interventions may then be promoted as a potential intervention for individuals living with the consequences of a TBI.We focused on the FPCN given the overlap of this network's involvement in TBI pathophysiology (Amir et al., 2021;Han et al., 2016) and physical exercise (Morris et al., 2022;Voss et al., 2010).No differences in physical activity were seen between groups, consistent with prior population-based cohort studies (Morris et al., 2019), however, functional connectivity results demonstrate that ( 1) physical activity engagement is associated with the functional connectivity of the FPCN and (2) there are differences in how physical activity is associated with functional connectivity of the FPCN in those with a history and TBI compared to those without.A similar pattern of results has been reported previously, whereby despite a lack of difference in the proportion of individuals engaging in physical activity, physical activity was significantly and differentially associated with perceived cognitive health in those with a TBI history compared to those without (Morris et al., 2019).It is likely then that lifestyle interventions that aim to improve brain health in those with a TBI should be designed and tested within this specific population as differing results from healthy populations are possible.Notwithstanding, our results demonstrate that physical activity is associated with network-level brain health in those with a history of TBI.
Our findings complement previous animal model literature in the field, adding to the understanding of how physical exercise (organized physical activity with the goal of improving physical fitness) can improve cognitive and brain functioning following a TBI (Griesbach et al., 2007;Karelina et al., 2021;Madathil & Saatman, 2015).Animal models have shown that the benefits of physical exercise after TBI are multifaceted with molecular, cellular, and network effects being shown (Griesbach et al., 2004(Griesbach et al., , 2007;;Sharma et al., 2020).Synaptic dysfunction is common after TBI and a driver of network changes after injury (Wolf & Koch, 2016).Our between group differences in functional connectivity of the FPCN show that a common signature of TBI exists in those with a history of TBI, which may give rise to the differential physical activity/functional connectivity relationship we see in this cohort (Wen et al., 2017).The within-network functional connectivity between the dorsolateral prefrontal cortex and inferior temporal gyrus may be a potential mechanism through which physical activity can improve TBI-related outcomes, albeit, interventional studies are needed to test this hypothesis.Indeed, prior work has demonstrated that physical activity is associated with the inferior temporal gyrus in non-injured adults (Jacini et al., 2009;Raffin et al., 2023).Consequently, this may imply that physical activity interventions could be a way through which long-term sustainable lifestyle interventions can promote brain health in those with a TBI.
There are several limitations to this study that need to be considered when interpreting the results.This study is purely cross-sectional, and ity definition is another limitation to consider, as this is a questionnaire based on recall and not objective.However, the questionnaire is extensive and addresses various types of physical activity.Further, biological sex is important to the recovery of TBI.Specifically, hormonal changes during menopause are associated with changes in functional connectivity (He et al., 2021) but are not addressed in this study.Furthermore, the age range of included participants is large.Age is included as a covariate to attempt to account for age-related effects with reported results indicating an age-independent effect of TBI of functional connectivity of the FPCN and the physical activity-functional connectivity association.Additionally, this study only looks at physical activity and not other healthy lifestyle behaviors, such as sleep, diet, mental health, substance use, and others, which often cluster together (Ai et al., 2023).
Future work should look to examine if there is a combined effect of broader lifestyle changes on functional connectivity within a TBI population.

CONCLUSIONS
Our results highlight the role physical activity may play in brain health following a TBI with loss of consciousness.Strategies to increase vol-untary engagement in physical activity specifically in individuals living with a TBI are necessary, and beyond the well-known general health benefits of physical activity, these strategies may also impact brain health in this population.

F
Regions of Interest for seed-based correlations (SBC) analysis in frontal parietal control network, calculated by overlaying Schaefer atlas overtop Harvard Oxford Atlas anatomical space.(a) Inferior parietal lobe and (b) dorsolateral prefrontal cortex.to study top-down, control-type behaviors such as engagement in physical activity.

Structural
T1w and resting state functional magnetic resonance images were obtained with a MAGNETOM Trio, A Tim Systems 3T Siemens scanner with a 32-channel head coil.An 8 min and 40 s resting state scan was acquired with a total of 261 volumes.T1-weight structural data were acquired by a Magnetization Prepared Rapid Gradient Echo Imaging (MPRAGE) sequence.The parameters were as follows: Repetition time (TR) = 2250 ms, echo time (TE) = 2.99 ms, flip angle = 9 • , field of view (FOV) = 256 × 240 × 192, voxel size = 1 × 1 × 1 mm, GRAP acceleration factor = 2.The resting-state functional images were collected by an echo-planar imaging sequence.The parameters were as follows: TR = 1970 ms, TE = 30 ms, flip angle = 78 • , FOV = 192 × 192, voxel size = 3 × 3 × 4.44, volumes = 261, slices = 32, order = descending.A full protocol with more details is provided in prior Cam-CAN publications(Shafto et al., 2014;Taylor et al., 2017), with the initial preprocessing performed by the Cam-CAN team.Anatomical data seed-based correlations (SBC) within a General Linear Model framework (GLM) using the CONN toolbox(Nieto-Castanon, 2020).All other statistical analyses were performed using R (Version 4.3.1)and Rstudio (Version 2023.12.0).Seed-based connectivity maps (SBC) were estimated characterizing the patterns of functional connectivity with two regions of interest (ROI) (bilateral dorsolateral prefrontal cortex and bilateral inferior parietal lobule; Figure1).ROIs were calculated by overlaying the Shaefer 100 functional atlas overtop the Harvard Oxford Atlas anatomical space(Morris et al., 2022).The average time series from each seed ROI was extracted.Then, Pearson's correlation coefficients were computed between the averaged time series within each seed and the time series in all other voxels of the whole brain and converted to normalized z-scores using Fisher transformation prior to performing second-level analyses.For all analyses, we used a GLM F-test with bilateral seeds jointly entered at the same time and controlling for age, sex, education, mean head motion, and years since injury.Bilateral seeds were entered jointly in the same F-test, a default pipeline in CONN, to reduce the number of comparisons, evaluating any effect among either side.A height-level statistical threshold of p < .001,cluster threshold of p < .05family-wise error (FWE) corrected, and k > 50 were used to determine significant clusters.Three separate F-tests were conducted for each pair of bilateral

F I G U R E 2
Resting state seed-based connectivity (rsFC) in the dorsolateral prefrontal cortex (dlpfc) between traumatic brain injury (TBI) and no TBI.Violin plots are demonstrating the connectivity strength between dorsolateral prefrontal cortex and four resulting clusters in both groups and the transparent threshold map of connectivity correlations (r = −.1-.1) for this analysis with the significant clusters outlined in black.(a) Cluster within the intracalcarine cortex with lower connectivity in the TBI group.(b) Cluster spanning the cerebellum with lower connectivity in the TBI group.(c) Cluster within the frontal pole higher connectivity in the TBI group.(d) Cluster within cingulate gyrus with higher connectivity in the TBI group.F I G U R E 3 Plot showing effect sizes of each bilateral seed of the dorsolateral prefrontal cortex (DLPFC) and its associated cluster.Negative effect sizes indicate hypoconnectivity and positive effect sizes indicate hyperconnectivity.F I G U R E 4 Seed-based connectivity with associations between physical activity and functional connectivity in traumatic brain injury (TBI) group.(a,b) Higher physical activity (PAEE) was significantly and positively associated with functional connectivity between both the L and R dorsolateral prefrontal cortex (DLPFC) and a cluster in the R inferior temporal gyrus (ITG).(c) Unthresheld map of connectivity correlations.Significant cluster outlined in black.(d) Effect size plot demonstrating the effect sizes of each bilateral seed.

E 5
No traumatic brain injury (TBI) and TBI differences in physical activity (a) violin plot showing physical activity in each group.(b) Seed-based connectivity results from the L and R inferior parietal lobule (IPL) showing the differential relationship between physical activity (PAEE) and TBI and controls in a cluster in the R inferior temporal gyrus (ITG).(c) Thresholded (r = .01)map of connectivity correlation strength with statistically significant cluster outlined in black.(d) interaction plot depicting the differential association between physical activity and functional connectivity in TBI (purple) compared to controls (blue).(e) Effect size plot demonstrating the effect sizes of each bilateral seed for each group.
interventional studies are needed to definitively assess the mechanistic effects of physical activity on functional connectivity in TBI populations.The lack of a clinical characterization of TBI means the severity of injury in these participants is unknown and the self-report nature of the diagnosis is open to reporting error.Nevertheless, the addition of reporting a loss of consciousness and the time since injury increases confidence in the accurate reporting of the condition.Physical activ- Participant demographics.
(Geerligs et al., 2017)Nieto-Castanon, 2012) white matter, and cerebral spinal fluid (CSF) tissue classes, and warped into a population-based template through DARTEL procedure(Taylor et al., 2017).Functional images were unwarped using field maps, motion corrected, slice timing corrected, co-registered to T1 images, and normalized into MNI space.Resting-state images were further preprocessed using CONN to identify and remove scanner and physiological artifacts(Whitfield- Gabrieli & Nieto-Castanon, 2012).Functional data were smoothed with 6-mm kernel and denoised using a standard denoising pipeline (Nieto-Castanon, 2020), including wavelet despiking, regression of potential confounding effects characterized by white matter, CSF, six motion parameters, linear trends, followed by bandpass frequency filtering [0.01, Inf](Geerligs et al., 2017)of the BOLD time series.Com-pCor Summary of functional connectivity results from all models.
TA B L E 2