Brain functional connectivity alterations in patients with anterior cruciate ligament injury

Recent advancements in neuroimaging have illustrated that anterior cruciate ligament (ACL) injuries could impact the central nervous system (CNS), causing neuroplastic changes in the brain beyond the traditionally understood biomechanical consequences. While most of previous functional magnetic resonance imaging (fMRI) studies have focused on localized cortical activity changes post-injury, emerging research has suggested disruptions in functional connectivity across the brain. However, these prior investigations, albeit pioneering, have been constrained by two limitations: a reliance on small-sample participant cohorts, often limited to two to three patients, potentially limiting the generalizability of findings


Introduction
Anterior cruciate ligament (ACL) injuries are prevalent among knee injuries, affecting approximately 85 out of every 100,000 individuals aged 16 to 39 years (Kaeding et al., 2017;Lynch et al., 2015).As a vital structural component for knee stability, the ACL's integrity is crucial for maintaining proper joint mechanics, muscle strength, and overall joint function.Damage to the ACL not only compromises these functions but also accelerates joint degeneration, impairs athletic performance, and increases the risk of developing osteoarthritis (Alkjaer et al., 2002;Chmielewski et al., 2001).The ACL and its surrounding region are equipped with numerous mechanoreceptors, responsible for transmitting essential sensory information to the central nervous system (CNS), facilitating the CNS's ability to generate precise motor responses to external stimuli (Banios et al., 2022;Gilman, 2002;Pitman et al., 1992).This sensory information is then processed within the CNS, enabling precise responses to external stimuli via motor signaling (Riemann and Lephart, 2002).When the ACL is damaged, these mechanoreceptors are disrupted, leading to alterations in the ascending afferent neural pathways and subsequent modifications in CNS functionality.This disruption is evidenced by alternations in somatosensory evoked potentials (SEPs) observed in patients with ACL injury or anterior cruciate ligament reconstruction (ACLR), highlighting the profound neural impact of such injuries (Courtney et al., 2005;Valeriani et al., 1999;Valeriani et al., 1996).
Over the past decade, functional magnetic resonance imaging (fMRI) has provided a unique non-invasive tool for investigating the neural consequences of ACL injuries or reconstruction within the CNS, significantly advancing our understanding of how the brain adapts to injuries and surgical interventions (Baez et al., 2020;Criss et al., 2023;Diekfuss et al., 2020;Diekfuss et al., 2019;Grooms et al., 2017;Grooms et al., 2015;Kapreli et al., 2009;Kim et al., 2023;Schnittjer et al., 2023;Strong et al., 2022).For instance, a pioneering fMRI study demonstrated that individuals with ACL injuries exhibited reduced activation in various sensorimotor cortical regions during a knee extension/flexion task, compared to healthy controls (Kapreli et al., 2009).Using the same approach, Grooms et al. (2017) also observed significant alterations in brain activation within the sensorimotor cortex of patients who had undergone ACLR.Regions typically engaged during knee movement tasks exhibited varied levels of activation, either increasing or decreasing.Their findings collectively suggested the brain's adaptations to altered joint mechanics and proprioceptive feedback caused by the ACL injury and its surgical repair.
However, many previous fMRI studies have focused on localized cortical activity changes post-injury or after surgical reconstruction, primarily assessing specific brain regions.However, this approach may not capture the broader network-level interactions and compensatory mechanisms that occur across the brain's complex network of interconnected regions.The resting-state functional connectivity (rs-FC) analysis, by contrast, evaluates the temporal correlations between spatially remote but functionally connected brain areas, providing a more holistic view of brain function.This method has proven effective in capturing subtle but crucial neural interactions across various cognitive domains (Baniqued et al., 2018;Carter et al., 2010;Demirakca et al., 2016;Gallen et al., 2016;Verstraete et al., 2010;Wu et al., 2009).
Similarly, the rs-FC approach enables an analysis of the collaborative processes among various brain regions as they adjust to the challenges of altered sensory feedback and motor control resulting from ACL injuries or reconstruction.Two seminal studies employing this approach have revealed that alterations in specific brain functional connectivity may predict ACL injury risk, highlighting that pre-injury disruptions in brain connectivity related to movement control may increase injury susceptibility (Diekfuss et al., 2020;Diekfuss et al., 2019).These findings demonstrated that rs-FC analysis has the potential to uncover broader network-level adaptations and compensatory mechanisms that traditional localized fMRI studies may overlook.However, these prior investigations, despite their groundbreaking nature, have faced two constraints.Most have depended on small participant samples, typically involving only two to three patients, which may limit the generalizability of their conclusions.Moreover, these studies often focused exclusively on specific regions of interest, potentially neglecting the complex interactions across the entire brain network.
With a cohort of 15 ACL-injured patients, our study aims to build upon these foundational works by overcoming previous limitations and analyzing the whole-brain rs-FC beyond predetermined ROIs through two network analysis techniques.First, we used Network-Based Statistics (NBS) (Zalesky et al., 2010) to analyze changes at the level of network components, rather than individual connections between brain regions as is typical in traditional FC analyses.This method allows for the comparison of network differences between ACL-injured patients and healthy controls.Second, we conducted a modular connectivity analysis to explore potential alternations alterations in functional integration within modules and communication or coordination between distinct functional modules.This analysis aims to unravel the brain's functional organization post-ACL injuries at the modular level (He et al., 2009).Both analyses are complementary and enable a comprehensive examination of the neural changes following ACL injuries.

Participates
The present study enrolled 15 patients with ACL injuries, all from sports-related professions, such as athletes and military personnel, who had to cease their training due to knee injuries.To compare their brain function, we enrolled an age-and sex-matched control group of 15 healthy individuals, also from similar professions but with no history of knee injuries.This sample size is consistent with that used in Grooms et al. (2017).All ACL injury diagnoses were confirmed through MRI and validated by an experienced orthopedic surgeon.
Criteria for inclusion required participants to have a chronic (over six months) and unilateral ACL injury with joint instability and planned surgical reconstruction.Additional inclusion criteria were adapted from Kapreli et al. (2009) to suit current clinical standards: 1) a Lysholm knee score below 65 %; 2) a Tegner knee activity rating score below 5; 3) experiencing more than one episode of knee joint giving way between the time of injury and the screening examination (excluding the initial injury); and (4) scoring less than 80 % on the Activities of Daily Living Scale.These criteria were revised to suit current clinical standards.The timed hop test in Kapreli et al. (2009) was replaced with the Tegner activity scale, which is more commonly used in clinical assessments for evaluating patients' activity levels (Frobell et al., 2010;Rezende et al., 2015).Additionally, the global rating question score was substituted with the Lysholm knee scoring system due to its specific design to evaluate knee stability, function, and associated pain more accurately (Marx, 2003).
Following Kapreli et al. (2009), the exclusion criteria for all participants included prior lower limb ligament injuries (except for meniscus injuries without joint motion limitation), reported limb pain in controls, or contraindications to MRI such as metallic implants or pacemakers.Individuals with a history of neurological or cardiovascular diseases, those on medication therapy, or those aged over 40 were also excluded to avoid confounding factors such as osteoarthritis, which could influence knee function recovery.

Ethical approval and informed consent
This study was conducted in accordance with the ethical guidelines of the 1975 Declaration of Helsinki and approval by the institution in which the research was conducted.Prior to data collection, informed consent was obtained from all participants.

MRI scanning and image preprocessing
FMRI Data were obtained on a 3.0-Tesla scanner (SIGNA, GE Healthcare) equipped with a 48-channel head coil.To preserve objectivity, the scanning technician was unaware of the subjects' group assignments.Subjects were stabilized with padding to minimize head movement, and earplugs were provided to mitigate scanner noise.Highresolution anatomical images were acquired using a 3D MPRAGE T1weighted sequence (TR = 7.808 ms, TE = 3.3 ms, slice thickness = 0.8 mm, flip angle = 8 • , matrix size = 276 × 276, field of view (FOV) = 220 mm, and isotropic voxel size of 0.8 mm 3 ).An echo-planar imaging (EPI) sequence was used to acquire resting-state functional images (TR = 1200 ms, TE = 25 ms, FOV = 240 × 240 mm, flip angle = 84 • , matrix size = 120 × 120, slice thickness = 2.0 mm, number of slices = 18, and isotropic voxel size of 2 mm 3 ).During the resting-state fMRI scan, subjects were instructed to relax, avoid specific mental tasks, remain still, and keep their eyes open and directed upward to facilitate accurate resting-state brain activity measurement.The resting-state fMRI scan lasted 288 s.
Resting-state fMRI data were preprocessed using Graph Theoretical Network Analysis software version 2.0.0 (GRETNA) (Wang et al., 2015), implemented in the Statistical Parametric Mapping 12 (SPM12) Y. Guan et al. (Wellcome Department of Imaging Neuroscience, UCL, London, UK), within MATLAB R2021a environment (MathWorks Inc, Natick, MA, USA).The preprocessing procedure included: discarding the first five images to allow for signal equilibration, performing slice timing correction and head motion correction, normalizing the images to the standard Montreal Neurological Institute (MNI) space with a resampled voxel size of 3 mm 3 (Fox et al., 2005), removing nuisance signals from cerebrospinal fluid and white matter, applying Friston 24-parameter head motion correction, conducting signal linear detrending, and implementing a temporal band-pass filter of 0.01 -0.08 Hz (Fox et al., 2005;Greicius et al., 2003).Participants with excessive head motion, defined as more than 2 mm displacement or 2 • rotation, were excluded from further analysis.

Functional connectivity analysis 2.4.1. Functional network construction
Network nodes were defined by segmenting the cortical surface into 160 regions of interest (ROIs, i.e., nodes) based on the Dosenbach 160 atlas (Dosenbach et al., 2010).The mean signal time course at each node was extracted as a measure of spontaneous regional activity.Then, a Pearson's correlation coefficient (r-value) was calculated between each pair of nodal time courses and normalized using Fisher's r-to-Z transformation as a standardized measure of node-to-node FC.Finally, a 160 × 160 FC matrix was obtained from each participant (details in supplement).

Network-based statistics
The NBS tool was used to identify significant differences in pairwise associations between ACL-injured patients and healthy controls, adopting a whole-brain approach (Zalesky et al., 2010).To compare the functional connectivity matrices of the two groups, an independent samples t-test with 1000 permutations was conducted.Multiple comparisons were corrected for each FC edge using the NBS correction, with p-values set at 0.005 at the edge level and 0.05 at the component level, respectively.The size of an effect was evaluated using its extent rather than its intensity.

Modular connectivity analysis
We used the module information from the Dosenbach 160 atlas to examine differences in intra-modular and inter-modular connectivity between ACL-injured patients and healthy controls, as shown in Fig. 2A.The Dosenbach 160 atlas includes 6 modules, namely default-mode network (DMN), frontal-parietal network (FPN), cingulo-opercular network (CON), sensorimotor network (SMN), visual network (VN) and cerebellum (Cere).The functional connectivity matrix was first thresholded at varying levels of sparsity, ranging from 0.03 to 0.26 in increments of 0.01 (details in supplement).At each sparsity level, within the six modules defined by the Dosenbach 160 atlas, we calculated: 1) the sum of the strengths of retained edges within each module; 2) the sum of the strengths of retained edges between different modules.To quantify overall connectivity patterns within and between modules across all sparsity levels, we computed the area under the curve (AUC) for the number of retained edges.Intra-modular connectivity reflects the functional cohesion within a module, while inter-modular connectivity indicates the level of integration among different modules.

Statistical analysis
Statistical analyses were performed using MATLAB and GraphPad Prism version 8.0 (GraphPad Software, La Jolla, CA, USA).The spatial distribution of brain regions was visualized using BrainNet Viewer (Xia et al., 2013).To compare demographic parameters and quality of life measures between groups, we employed independent samples t-tests.One-sample t-tests were used to evaluate differences between the knee status scores of ACL-injured patients and the standard scores of healthy participants (i.e., 100 points).Differences in functional connectivity between the two groups were analyzed using independent samples ttests, with corrections for multiple comparisons implemented via NBS (refer to Section 2. .Additionally, to investigate differences in modular connectivity between the two groups, independent samples t-tests with Bonferroni corrections were applied.

Demographics, clinical and quality of life information
Four healthy participants were excluded from further analysis due to excessive head movements.As shown in Table 1, among the remaining participants, statistical analysis revealed no significant differences between the two groups in terms of age (t 24 = 0.709, p = 0.485, Cohen's d = 0.281), educational level (t 24 = 0.703, p = 0.489, Cohen's d = 0.100), and BMI (t 24 = 0.134, p = 0.894, Cohen's d = 0.054).However, the ACLinjured group showed significantly lower Lysholm scores than the healthy controls (t 14 = 5.860, p < 0.0001, Cohen's d = -1.513),indicating poorer knee function.The ACL-injured patients also showed lower physical health scores on the SF36 scale compared to the controls (t 24 = 2.397, p = 0.025, Cohen's d = 0.952).There was no significant difference in mental health scores between two groups (t 24 = 0.310, p = 0.759, Cohen's d = 0.123).

Dysconnected subnetwork
Nonparametric NBS analysis revealed significant alterations in the functional connectivity of ACL-injured patients compared to healthy controls.As shown in Fig. 1A, these patients exhibited reduced connectivity in a widespread dysconnected subnetwork consisting of 73 regions and 83 connections (all ts > 3.1, all ps < 0.05, NBS corrected) (details see Supplemental Table 1).To better visualize and understand the distribution of these connections, we categorized each module of the Dosenbach 160 atlas within the identified subnetwork, as belonging to six distinct functional modules: DMN, FPN, CON, SMN, VN and Cere, illustrated in Fig. 1B.The majority of connections that were dysconnected in ACL-injured patients involved links between cerebellar regions and widespread cortical regions (occipital, parietal, and frontal cortices), as well as fronto-parietal and parietal-occipital links.
Following Kim et al. (2019), we further identified hub regions within the subnetwork, defined as having degrees larger than the 2SD away from the mean.These hub regions, from which the greatest number of Education level was classified according to the number of years of education completed.The SF-36, a self-administered survey, evaluates health-related quality of life, scoring each domain on a scale from 0 to 100, where higher scores represent better health status.An asterisk (*) indicates statistically significant differences between the two groups.
dysconnections originate, represented the influential brain regions compared with other regions in the network.These hub regions included the Med Cerebellum (label 147 in Dosenbach 160 atlas, degree = 9), Lat Cerebellum (labels 143, degree = 7), Post Occipital (labels 137 and 140, degree = 7 each), and Mid Insula (label 74, degree = 6).Detailed information about the regions that constitute this subnetwork and their degrees is provided in Supplemental Table 2.We did not identify any subnetwork with increased connectivity in the ACL-injured patients.

Decrease in inter-modular and intra-modular connectivity
It should be noted that in the NBS analysis, significance is not attributed to individual dysconnections or hyperconnections; significance can only be declared for the entire subnetwork, whether disconnected or hyperconnected, as a whole (Zalesky et al., 2010).Therefore, to further understand the contribution of specific connections to the disruption of functional connectivity in ACL-injured patients, we carried out a modular connectivity analysis to explore alterations in functional organization at the module level, including functional integration both within and between distinct modules.Significant changes in both interand intra-modular connectivity were observed in ACL-injured patients compared to healthy controls.Specifically, there was a significant decrease in inter-modular connectivity between the SMN and the Cere modules in the ACL-injured group (t 24 = 2.201, p = 0.038; Cohen's d = 0.874) (Fig. 2B).In addition, these ACL-injured patients also showed reduced intra-modular connectivity within the DMN module (t 24 = 2.148, p = 0.042; Cohen's d = 0.853) (Fig. 2B).

Discussion
It has been suggested that ACL injuries can lead to disturbances in neuro-muscular control, impacting central processes in the brain and potentially causing motor response changes and limb dysfunction (Courtney et al., 2005;Kapreli and Athanasopoulos, 2006).Building on this understanding, our study revealed significant changes in brain connectivity associated with ACL injuries, enhancing our understanding of the broader neurological consequences of such injuries.Using the NBS method, we identified widespread reductions in connectivity that spanned multiple brain regions in ACL-injured patients compared to healthy controls, suggesting network-wide dysfunctions following ACL injuries, rather than localized disruptions as previously suggested (Diekfuss et al., 2020;Diekfuss et al., 2019).Furthermore, our modular connectivity analysis revealed pronounced reductions in inter-modular connectivity between the SMN and the Cere, as well as decreased intra-modular connectivity within the DMN following ACL injuries, indicating that ACL injuries may impair the coordination and communication between critical cognitive and motor control.Taken together, our findings highlight the extensive impact of ACL injuries on neural networks, affecting not only the primary motor control areas but also extending to cognitive networks involved in motor coordination, and higher-order brain functions, which might be vital for understanding the challenges of recovery and adaptation to physical impairments postinjury.
Our study contributes significant new insights to the existing body of research on ACL injuries and their neural correlates.Previous fMRI studies have primarily focused on task-based assessments to identify localized changes in brain activity post-ACL injury, observing increased activation in motor-related areas, suggesting adaptive compensations to maintain motor function after the loss of sensory inputs from the injured knee.For instance, Grooms et al. (2017) reported alterations in the activation of the sensorimotor cortex and cerebellum during motor tasks, which they attributed to neuroplastic adaptations in response to altered joint mechanics and disrupted proprioceptive feedback.These findings highlighted the brain's capacity for neuroplasticity following physical injuries but were limited by their focus on specific regions during controlled tasks.In contrast, our study extended these findings by employing network analysis approach and resting-state fMRI to examine broader, network-wide disruptions in neural connectivity following ACL injuries, which was not commonly used in earlier ACL research.Our findings of widespread reductions in connectivity, involving multiple brain regions, aligned with the notion of a broad neural impact of ACL injuries (any refs from the literature of ACL research, no matter whether it is an fMRI or behavioral study).However, unlike previous studies which documented increased localized activity, our approach highlights a network-wide dysfunction, providing a more comprehensive view of the brain's adaptation to ACL injuries.
Moreover, while studies such as that by Diekfuss and his colleagues (2019, 2020) focused on the predictive potential of brain functional connectivity before ACL injuries, suggesting that certain neural patterns could predispose individuals to injuries.Our application of modular connectivity analysis further differentiated our work by providing insights into the functional organization of the brain post-ACL injury, offering insights into the extensive network alterations that occurred.We discovered significant decreases in both intra-modular and intermodular connectivity, prominently affecting the sensorimotor and cerebellar networks, as well as the DMN.It has been well-documented that the SMN is crucial for integrating sensory information to produce motor responses, while the cerebellum plays a fundamental role in finetuning motor actions, balance, and coordination (Buckner, 2013;Enzinger et al., 2008;Hsu et al., 2023;Kelly and Strick, 2003;Koziol et al., 2014;Liu et al., 2024;Sanes and Donoghue, 2000;Solodkin et al., 2004).These networks are not only crucial for basic motor functions but are also involved in higher-order cognitive processes that support complex motor planning and execution (Buckner, 2013;Hsu et al., 2023;Koziol et al., 2012).Therefore, based on these understandings, we inferred that the neural impact of ACL injuries extended beyond mere motor control disruptions to involve cognitive networks associated with higher-order brain functions and motor coordination.By integrating both NBS and modular connectivity analyses, we provided a more detailed understanding of how ACL injuries affect not just isolated brain regions but interconnected neural networks, thereby laying a foundation for future research aimed at developing targeted interventions that consider both motor and cognitive components to improve neurological and functional recovery in ACL-injured patients.Previous studies have demonstrated that ACL injuries cause changes not only in brain regions associated with motor control, but also in other brain regions involved in motor processes (Diekfuss et al., 2020;Grooms et al., 2015).Our study echoed the existing theories regarding the neural impact of ACL injuries but went further to elucidate the extensive and complex nature of brain adaptations following such injuries.
Our study, while providing novel insights into the neurological consequences of ACL injuries, encompassed several limitations that may impact the generalizability and interpretation of our findings.First, our research adopted the resting-state fMRI approach.Though previous studies have established the relevance of spontaneous resting-state functional connectivity to brain function (Biswal et al., 1995;Power et al., 2014;Smith et al., 2009), resting-state fMRI may not fully represent the dynamics of brain activity during specific tasks.Future research using a task-based functional connectivity approach (Greene et al., 2018), which allows for a direct comparison of neural activation patterns during specific motor tasks, is thus needed to validate our findings.Second, our analysis was conducted with unequal sample sizes, with fewer subjects in the control group compared to the ACL-injured group.This imbalance can reduce statistical power and increase the risk of Type I errors, potentially skewing the interpretation of our Fig. 2. Alterations in inter-module and intra-module connections.(A) Distribution of brain regions within six functional modules, mapped according to the Dosenbach 160 atlas.(B) a comparative analysis of connectivity between and within these modules in ACL patients versus healthy controls.There were significant differences in connectivity between the sensorimotor network (SMN) and the cerebellum (Cere), as well as within the default-mode network (DMN) connections between the two groups.An asterisk (*) denotes significantly reduced connectivity in ACL patients compared to healthy controls.
results.Future studies should aim for equal or proportionate sample sizes to replicate our findings.Third, the participants in the ACL-injured group were non-copers.This selection criterion limited the generalizability of our results to the broader ACL-injured population, which includes both copers and non-copers.Future studies should consider including a more diverse participant pool to ensure that our findings are applicable to all individuals with ACL injuries.Last, our study did not collect critical measures such as details on rehabilitation protocols, functional outcomes (e.g., muscle strength, hop distance), and the timing of injury post-occurrence.These measures, typically used as covariates in neuroimaging research, accounted for variations in recovery and physical abilities (Criss et al., 2023;Grooms et al., 2015;Kapreli et al., 2009;Scheurer et al., 2020).The absence of this information might have influenced the observed brain connectivity patterns.Acknowledging these limitations, our study should be viewed as a pilot investigation that emphasized the need for further systematic research with carefully controlled variables to elucidate the complex interplay between ACL injuries and brain function.

Conclusion
Through the use of NBS and modular connectivity analyses, our study has demonstrated that ACL injuries result in extensive neural network dysfunctions that may affect both motor control areas and relevant cognitive networks.Our findings emphasized the profound impact of ACL injuries on the brain's functional organization, extending beyond mere motor dysfunctions to involve essential cognitive processes.

Reporting statement
This study was conducted in accordance with the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) statement.

Fig. 1 .
Fig. 1.Subnetwork identified through NBS analysis in ACL-Injured patients.(A) Regions within this subnetwork are organized according to the Dosenbach 160 atlas and divided into six modules.Gray lines represent functional connections linking regions across different modules, while colored lines indicate connectivity within the same module.Regions are positioned according to their center of mass.(B) The identified subnetwork consists of 73 regions and 83 connections.DMN, defaultmode network; FPN, frontal-parietal network; CON, cingulo-opercular network; SMN, sensorimotor network; VN, visual network; Cere, cerebellum.

Table 1
Demographics and Clinical Characteristics of Participants.