Alterations of large‐scale functional network connectivity in patients with infantile esotropia before and after surgery

Abstract Background Growing evidences have indicated neurodevelopmental disorders in infantile esotropia (IE). However, few studies have analyzed the characteristics of large‐scale functional networks of IE patients or their postoperative network‐level alterations. Methods Here, individuals with IE (n = 32) and healthy subjects (n = 30) accomplished the baseline clinical examinations and resting‐state MRI scans. A total of 17 IE patients also underwent corrective surgeries and completed the longitudinal clinical assessments and resting‐state MRI scans. Linear mixed effects models were applied for cross‐sectional and longitudinal network‐level analyses. Correlation analysis was performed to assess the relationship between longitudinal functional connectivity (FC) alterations and baseline clinical variables. Results In cross‐sectional analyses, network‐level FC were apparently aberrant in IE patients compared to controls. In longitudinal analyses, intra‐ and internetwork connectivity were observed with significant alterations in postoperative IE patients compared to the preoperative counterparts. Longitudinal FC changes are negatively correlated to the age at surgery in IE. Conclusions Obviously, altered network‐level FC benefiting from the corrective surgery serves as the neurobiological substrate of the observed improvement of stereovision, visuomotor coordination, and emotional management in postoperative IE patients. Corrective surgery should be performed as early as possible to obtain more benefits for IE in brain function recovery.


INTRODUCTION
Strabismus is a common ocular disorder that affects 2-6% of the children, resulting in undesirable consequences including impaired visual acuity, loss of stereoscopic depth perception, poor hand-eye coordination, and psychosocial disturbance (Suttle et al., 2011;Torp-Pedersen et al., 2017). Infantile esotropia (IE) is a subtype of concomitant strabismus, which occurs during the first 6 months of life and is distinguished by large-angle nasal-ward deviation of both optic axes, cross-fixation, latent nystagmus, and dissociated vertical deviation (Mehner et al., 2023). Although kinds of theories have been proposed for the etiology of IE, such as impairment in brain regions associated with fusion ability (Yang et al., 2014), the enhanced subcortical ocular motion pathway plasticity (Brodsky, 2019), binocular monocular nasal-temporal optokinetic asymmetry resulting from the maldevelopment of visual cortex (Brodsky, 2018), the underlying neuropathological mechanism remains controversial and urgently needs to be further explored.
In recent years, functional magnetic resonance imaging (fMRI) has been popular in examining the neurobiological changes of patients with strabismus. As a result, abnormalities of brain structure and function in strabismus patients have been demonstrated through varieties of imaging techniques with different data processing protocols.
Specifically, morphometry studies have revealed abnormal gray matter volume (Chan et al., 2004;Su et al., 2022), white matter volume (Ouyang et al., 2017), and cortical thickness (Yin et al., 2021) of some brain regions in patients with strabismus. Altered microstructural properties of white matter implicated in visual pathway have been demonstrated in strabismus patients using voxel-based analysis Yan et al., 2010), tract-based spatial statistics, and automated fiber quantification analysis of diffusion tensor imaging (DTI) (Duan et al., 2015;Li et al., 2018). More specifically, a voxelbased DTI analysis has revealed lower fractional anisotropy in middle occipital gyrus and supramarginal gyrus in strabismus adults (Yan et al., 2010); Higher fractional anisotropy values in inferior fronto-occipital fasciculus and inferior longitudinal fasciculus were apparent in adults with comitant exotropia (Li et al., 2018). An automated fiber quantification study of adults strabismic amblyopia has demonstrated elevated mean diffusivity along several visual-related fasciculi such as vertical occipital fasciculus, optic radiation, and inferior longitudinal fasciculus (Duan et al., 2015). It has been reported that strabismus patients present abnormal functional connectivity (FC) of the primary visual cortex Yan et al., 2019), anomalous interhemispheric FC of visual-related brain subregions (Peng et al., 2021;Zhang et al., 2021), as well as disrupted spontaneous neural activity of regional homogeneity Tan et al., 2022) and amplitude of low frequency fluctuation Hu et al., 2022;Xi et al., 2020) in brain areas related to visual processing. However, few studies have explored the functional alterations of brain in IE. As a rare category of strabismus with congenital fusion deficiency, IE may have uncommon neuroimaging features suggestive of specific neurobiological changes.
Corrective surgery is an essential option in intervention for IE, as alignment of visual axes facilitates the potential for binocularity. Nev-ertheless, it is worth noting that restoring normal ocular alignment do not promise improvement of stereopsis in some IE patients, and the precise mechanism of which remains perplexing (Çerman et al., 2014;Hug, 2015). Strengthened spontaneous brain activity in visual cortex or microstructural alterations in core components of visuospatial network may account for the stereoscopic vision restoration in postoperative patients with comitant exotropia Wu et al., 2022), while no longitudinal studies have investigated the interaction between stereovision recovery and changes in FC in terms of largescale networks in IE patients before and after surgery. In addition, controversies concerning the timing of surgery for IE remain difficult to figure out (Mehner et al., 2023). Our previous cross-sectional research has suggested that, compared to healthy subjects, patients with IE have an aberrant developmental trajectory of FC from birth, exhibiting great potential for correction of compromised neurodevelopment during the early postnatal period, thus, the timely surgical interventions may bring satisfactory benefits for improvements in visual function as well as other visual-related brain functions (Guo, Chen, Liu et al., 2022a). However, longitudinal studies are necessary for better understanding the effects of surgical intervention on functional recovery of IE during the crucial period of growth and development.
In this study, we analyzed the alterations of large-scale functional network connectivity using the cross-sectional sample and what is, as far as we know, the first longitudinal sample of resting-state fMRI data from patients with IE, with the purpose of investigating whether and how the surgical intervention contributes to the restoration of various brain functions as well as exploring the potential neuropathological mechanism of visual-related impairment in IE.

MRI data acquisition and preprocessing
Imaging data were acquired with a 3T MRI scanner (General Electric, for further motion effect control (volumes with framewise displacement bigger than 0.3 mm and no more than three sequential time points were removed), nuisance covariates regression for global signal, white matter signal, cerebrospinal signal and head motion parameters, band-pass filtering of 0.01-0.1 Hz, spatial smoothing with a 6 mm full width at half maximum Gaussian kernel. Global signal, the averaging signal from all gray matter voxels, contains both the neural and nonneural (like head motion) components. Therefore, whether to remove the global signal is really dependent on the specific question. First, the global signal removal is good to further control the impact of head motion. Then, the focus of this article is to find and delineate the regional alterations associated with disease, for which it is better to remove the global information and enhance the local information in the blood oxygen level-dependent signals. Overall, performing global signal correction is helpful to reduce the potential confounding effects and contributes to the detection of regionally specific FC patterns. Spatial normalization was performed to transform fMRI data from individual space to the University of North Carolina (UNC) brain standard space of 2-year-old (Shi et al., 2011) with a nonlinear registration method, using the Advanced Normalization Tools (ANTs, http://stnava.github. io/ANTs/).

Seed-based FC map
Network-level FC analyses using seed-based FC map were focused on seven formerly proposed intrinsic functional networks, which include visual network (VN), sensorimotor network (

Statistical analysis
A linear mixed effects model (LME) was applied for group-level FC contrasts, in which the age, gender, group, and the interaction items between age and group were taken as fixed effects while the random intercept and slope for group were regarded as random effects.

Demographics and clinical features
At the baseline fMRI scanning, 32 IE patients and 30 healthy participants had similar profiles in age (p = .306) and gender (p = .987) (      Table   S1 for detailed clusters' information).

Longitudinal network-level analyses
The LME was used to compare the functional network connectivity  Table S1 for detailed clusters' information).

F I G U R E 3 Brain regions with significant group-age interaction in functional connectivity of the default mode network (a), occipital pole visual network (b), salience network (c), and dorsal attention network (d).
The functional connectivity of default mode network with left thalamus, occipital pole visual network with right inferior temporal gyrus, left precuneus, left postcentral gyrus, salience network with right inferior parietal lobule, right calcarine fissure and surrounding cortex, dorsal attention network with right precuneus/right superior parietal gyrus, and right inferior temporal gyrus/right fusiform gyrus exhibited significant interaction between group and age in the cross-sectional analysis. Red: Functional connectivity increase with age in infantile esotropia; blue: Functional connectivity decrease with age in infantile esotropia. Cluster-level multiple comparison correction at alpha < .05 with voxel-level p-value < .001 and cluster size > 24.

Correlation of longitudinal FC changes with baseline clinical metrics
Significant negative correlations of the age at surgery with longitudinal FC changes in terms of CEN-DMN (right precuneus; p = 1.15e-02) and V1N-LN (left middle frontal gyrus; p = 2.22e-02) were observed in patients with IE ( Figure 6). There was no evident correlation between longitudinal FC alterations and baseline strabismus degree in IE.

DISCUSSION
Here, we examined alterations of large-scale functional network connectivity in IE patients using cross-sectional samples as well as lon-

Cross-sectional network-level FC alterations in IE
Majority of the studies on resting-state FC, in the context of strabismus, laid importance on V1-based FC analyses, with similar findings of deficient interactions between visual cortex and brain regions pertaining to stereopsis or eye movement Liu et al., 2022;Yan et al., 2019;Zhu et al., 2018). Attaching importance to intra-and internetwork connectivity, instead of regional or global FC, suggests an advance on preceding studies, particularly in view of prominent evidence of disrupted between-network FC in patients with IE (Guo, Chen, Liu et al., 2022a). The VN-SMN internetwork FC is disrupted in IE patients, of which the defects occur in the early postnatal period and persist across the infancy and early childhood, showing a characteristic developmental trajectory with age. The interaction between VN and SMN is implicated in visuomotor coordination (Burgos et al., 2018;Hou et al., 2022), thus, impairment in VN-SMN internetwork FC could account for the observed hand-eye coordination disorder in IE patients. As a higher-order functional network, SN is considered to be implicated in integrating external and internal information, attaching salience weighting to different events, regulating the response of other networks to external stimulus or internal physiological signal, coordinating the correlation between functional networks (e.g., DAN-DMN) (Gao et al., 2013;Gao et al., 2017;Seeley et al., 2007). The FC between SN and subcortical structure was reported to be capable of predicting the visuospatial working memory performance and cognitive function (Alcauter et al., 2014). Furthermore, the FC between SN and limbic regions plays a critical role in the regulation of emotion, reward, and motivation processing (Seeley et al., 2007). Taken together, impaired network-level FC between SN and those networks involved in visual attention (e.g., DAN), cognition and emotion (e.g., DMN), coupled with abnormal FC within SN and disrupted interaction between DAN and DMN, indicate the disorder of SN serving as "regulators," which may contribute to the disturbance of visual function, neurocognitive development, and emotional management in IE patients. The FC between DMN and thalamus (i.e., LN) emerges during the infancy and involves self-concept construction (Philippi et al., 2012). Anomalous DMN-thalamus FC further provides neurobiological evidence for the potential cognitive processing impairment in patients with IE.
Our results indicate that, in IE, different network-level FC are not equally interfered in developmental trajectories. More interestingly, internetwork FC are more prone to be interfered than intranetwork ones. One possible explanation is that primitive functions based on single brain network mature very early after birth to guarantee survival, whereas complicated and high-order functions relying on internetwork interactions require prolonged environment-related tuning and, thus, are vulnerable to abnormal visual experience (Gao et al., 2015).

Longitudinal network-level FC changes in IE
The VN is responsible for broadly categorized visual information processing such as the visual content information and the visuospatial information (Yamamoto et al., 2012). Here, the obviously altered functional network connectivity within VN might play a pivotal role in the improvement of binocularity in postoperative IE patients. Recently, a longitudinal resting-state MRI study also demonstrated significant functional changes in core regions of VN in strabismus patients with binocularity recovery following corrective surgery . In this study, the VN-DMN internetwork FC was decreased in the pre-operative IE patients but significantly increased in the postoperative counterparts. VN-DMN internetwork FC was reported to be involved in social visual engagement behaviors, the impairment of which contributes to ignoring some specific kinds of input information from environment to brain and results in social communication difficulties (Lombardo et al., 2019

Limitations
Several limitations should be considered when illustrating the findings. First of all, due to the young age of the IE patients enrolled in this study, postoperative stereovision evaluation was only performed in those capable of subjectively cooperating in the Titmus stereogram examination. As a result, available quantified stereopsis data were not sufficient in the correlation analysis with network-level FC alterations. Besides, our research with a relatively small sample size and future work with a larger number of subjects in both cross-sectional and longitudinal sample, coupled with multiple followup time points, are necessary to note at which time the process of stereovision improvement begins following the corrective operation.
Furthermore, such further researches should also acquire estimates of cognitive function based on assessment scales to better delineate the interaction between network-level connectivity alteration and cognition function restoration in IE. Also of note, different depths of sleep may exert effect on the properties of network or functional network connectivity. However, it is not easy to carry out overt monitoring of sleep status employing electrophysiological methods in this population . Future studies would be per-formed to try to compare the network connectivity under different sleep stages. By generating masks based on the healthy controls alone, we aimed to capture the functional network patterns specific to the healthy population, producing typical masks of functional networks for group-level comparison, however, this may introduce group-specific confounding effects in this study.

CONCLUSIONS
In conclusion, patients with IE exhibit significantly impaired FC between or within large-scale functional networks, which are implicated in cognitive, emotional, visual, and sensorimotor processing. Alterations of intra-or internetwork FC of those networks in postoperative IE patients render new insights into the neurobiological process underlying the observed functional improvements in terms of stereoscopic depth vision, hand-eye coordination, and emotional regulation. Moreover, the internetwork FC is more prone to be disturbed under the influence of abnormal visual input in IE. More importantly, the corrective surgery for IE patients should be performed as early as possible to help them to obtain more benefits in visual-or emotionalrelated brain functions. On the basis of our results, future explorations with preclinical purposes, such as predicting surgical benefits or discovering biomarkers for predicting stereopsis restoration for IE, would be conducted to attain more meaningful and practicable achievements for clinical practice.

AUTHOR CONTRIBUTIONS
YP, NL, and DH designed the study. JG was in charge of acquiring imaging data and drafting the manuscript. WL, LH, YL, and HK contributed to recruiting participates and collecting clinical data. YC was responsible for data analysis. All authors approved the final manuscript.