Molecular mechanisms underlying resting-state brain functional correlates of behavioral inhibition

,


Introduction
Behavioral inhibition, a basic facet of executive functioning (Hofmann et al., 2012), refers to the ability to suppress inappropriate, automatic, reflexive, or habitual prepotent responses to produce a controlled goal-directed response (Isoda and Hikosaka, 2011;Masharipov et al., 2022).Behavioral inhibition has strengthened its position as the most well-characterized cognitive domain in clinical neuroscience, given that poor behavioral inhibition is thought to be a characteristic feature of many mental disorders (Gut-Fayand et al., 2001;Nigg, 2001;Nigg et al., 2006;van Rooij et al., 2015).Empirical evidence from cognitive science, neuroimaging, and neuropsychiatry has established the presence of sex differences in behavioral inhibition performance (Knežević, 2018;Sjoberg and Cole, 2018;Weafer and de Wit, 2014), brain activation induced by behavioral inhibition tasks (Garavan et al., 2006;Li et al., 2006;Liu et al., 2012), and prevalence of behavioral inhibition-related diseases (Coffino et al., 2019;Polanczyk et al., 2007;Slade et al., 2016).For example, behavioral data have demonstrated that females outperform males in behavioral inhibition (Knežević, 2018;Sjoberg and Cole, 2018).Neuroimaging studies have yielded mixed results, with some showing greater activation in females in many cortical regions during a Go/No-Go task (Garavan et al., 2006) and others reporting greater activation in males in a wide array of cortical and subcortical areas during a stop signal task (Li et al., 2006).Brain-wide association studies, which correlate individual variations in cognitive phenotypes to individual differences in brain structural or functional measures, have become a dominant approach for linking cognition and brain.Notably, linking human cognition to resting-state brain function is a central question in systems neuroscience.However, there is limited evidence that speaks directly to the question of whether or not there are sex-dependent associations between behavioral inhibition and resting-state brain function and, if so, how they are modulated by the underlying molecular mechanisms (i.e., genetic architecture and neurochemical basis).
Resting-state functional magnetic resonance imaging (fMRI) is a noninvasive neuroimaging technique to measure spontaneous fluctuations in brain activity based on the blood-oxygen-level-dependent (BOLD) signal (Biswal et al., 1995), temporal correlations of which across brain regions are defined as functional connectivity (FC) that typically reflects the intrinsic functional organization of the brain (Fox and Raichle, 2007).Due to reliance on a priori definition of seed regions or restriction to specific networks, hypothesis-driven seed-based FC analysis or independent component analysis may lack a global and independent view, and thus could have yielded biased results.By contrast, functional connectivity density (FCD) mapping, a data-driven graph-theory approach, has been developed to construct and analyze whole-brain FC networks in a voxel-wise manner (Tomasi andVolkow, 2010, 2011a, b).Brain regions with higher FCD show more functional connections to other regions and thus are more critical for information processing across the whole brain.The FCD approach has been successfully applied to examine resting-state brain functional correlates of several aspects of human behavior.For instance, Liu et al. found that working memory performance was associated with FCD in several frontal, parietal, and occipital regions in a sample of healthy young adults (Liu et al., 2017).Likewise, Xu and colleagues demonstrated a positive correlation between neuroticism and FCD in the ventral striatum (Xu et al., 2020).
With the recent introduction of comprehensive, brain-wide gene expression atlases such as the Allen Human Brain Atlas (AHBA) (Hawrylycz et al., 2012;Shen et al., 2012) comes the emergent field of imaging transcriptomics, whose goal is to explore the genetic architecture underlying a variety of neuroimaging phenotypes by identifying genes with spatial profiles of regional expression that track anatomical variations of a certain neuroimaging measurement (Arnatkeviciute et al., 2019;Chen et al., 2022;Fang et al., 2023;Fornito et al., 2019;Li et al., 2023;Liu et al., 2022;Shen et al., 2022;Song et al., 2023;Sun et al., 2023;Xu et al., 2023;Zhang et al., 2022;Zhao et al., 2022a).Rather than directly making inference on the identified genes as well as for the purposes of biological interpretation, gene category enrichment analysis (GCEA) is commonly pursued to investigate functional gene categories that drive the transcriptome-neuroimaging relationships.Nevertheless, classical GCEA is markedly impacted by false-positive biases arising from spatial autocorrelation and gene-gene co-expression.To address this shortfall, a flexible ensemble-based null model tailored to allow for more valid and interpretable inference of GCEA is emerging (Fulcher et al., 2021).In parallel, it is now clear that behavioral inhibition has been closely linked to several neurotransmitter systems including dopamine (Cummins et al., 2012;Robertson et al., 2015), gamma-aminobutyric acid (GABA) (Quetscher et al., 2015;Silveri et al., 2013), and serotonin (Drueke et al., 2013;Macoveanu et al., 2013).Continuing improvements in nuclear imaging techniques and tracers have afforded reliable and accurate quantification of a range of neurotransmitter receptors and transporters, facilitating a detailed characterization of the neurochemical basis underlying brain structure and function (Dukart et al., 2021;Hansen et al., 2022).Combined, the current availability of whole-brain gene expression and neurotransmitter atlases along with the refinement of methodologies for multi-modal datasets could open new avenues to examine the spatial associations between these atlases and neuroimaging phenotypes, which may contribute to a mechanistic understanding of cognition-related neural substrates.
The objective of this study was to investigate the sex-specific neural correlates of behavioral inhibition and their underlying molecular mechanisms.To achieve this goal, we initially computed FCD using resting-state fMRI data to examine their associations with behavioral inhibition ability measured using a Go/No-Go task across a large cohort of 510 healthy young adults.Note that the Go/No-Go task was selected because it is a widely used neuropsychological test to assess behavioral inhibition ability (Chan et al., 2023;Weiss and Luciana, 2022).Then, we examined the spatial relationships of the FCD correlates of behavioral inhibition with gene expression and neurotransmitter atlases to explore their potential genetic architecture and neurochemical basis.A systematic flowchart of the study design is shown in Fig. 1.

Participants
A total of 510 healthy young adults (320 females and 190 males) were recruited from the local universities and community through poster advertisements.All participants met the inclusion criteria of being Chinese Han, right-handed, and within a restricted age range of 18-30 years.Exclusion criteria included neuropsychiatric or severe somatic disorder, a history of head injury with consciousness loss, alcohol or drug abuse, regular smoking, pregnancy, MRI contraindications, and a family history of psychiatric disease among first-degree relatives.The MINI-International Neuropsychiatric Interview and Alcohol Use Disorders Identification Test were used in the process of excluding participants.This study was approved by the ethics committee of The First Affiliated Hospital of Anhui Medical University.Written informed consent was obtained from all participants after they had been given a complete description of the study.Detailed demographic characteristics of the participants are listed in Table 1.

Behavioral inhibition assessment
The Go/No-Go task was performed on a computer to assess the ability of behavioral inhibition (Cai et al., 2021;Kaufman et al., 2003) using E-Prime 2.0.During the task, the letter X or Y was presented at a frequency of 1 Hz on the screen.In "Go" conditions, the current letter is different from the previous one and participants should respond quickly by pressing the button within 900 ms.In "No-Go" conditions (10 % of all trials), the current letter is the same as the previous one and participants cannot press the button; if one presses the button, it would be counted as an error.The Go/No-Go task consisted of a practice test and a formal test.There were 20 trials (15 "Go" trials and 5 "No-Go" trials) in the practice test.If a participant responds correctly in three "No-Go" trials, he or she can shift to the formal test; otherwise, the participant needs to restart the practice test.The formal test was divided into two groups with 210 trials in each group and 30 s break between the two groups.It took about 12 min for the Go/No-Go task.The primary variable of interest is the accuracy in "No-Go" conditions (Acc_No-Go) that reflects behavioral inhibition.It is noteworthy that all participants underwent MRI scanning immediately after completing the Go/No-Go task on the same night.

MRI data acquisition
MRI data were acquired using a 3.0-Tesla MR system (Discovery MR750w, General Electric, Milwaukee, WI, USA) with a 24-channel head coil.During scanning, tight but comfortable foam and earplugs were used to minimize head movement and scanner noise.All participants were instructed to relax, keep their eyes closed but not fall asleep, think of nothing in particular, and move as little as possible.Highresolution 3D T1-weighted structural images were acquired using a brain volume (BRAVO) sequence with the following parameters: repetition time (TR) = 8.5 ms; echo time (TE) = 3.2 ms; inversion time (TI) = 450 ms; flip angle (FA) = 12 • ; field of view (FOV) = 256 mm × 256 mm; matrix size = 256 × 256; slice thickness = 1 mm, no gap; voxel size = 1 mm × 1 mm × 1 mm; 188 sagittal slices; and acquisition time = 296 s.Resting-state BOLD fMRI data were acquired using a gradient-echo single-shot echo planar imaging (GRE-SS-EPI) sequence with the following parameters: TR = 2000 ms; TE = 30 ms; FA = 90 • ; FOV = 220 mm × 220 mm; matrix size = 64 × 64; slice thickness = 3 mm, slice gap = 1 mm; 35 interleaved axial slices; 185 vol; and acquisition time = 370 s.Routine T2-weighted images were also collected to exclude any organic brain abnormality.All MR images were visually inspected to ensure that only images without visible artifacts were included in subsequent analyses.

fMRI data preprocessing
Resting-state BOLD data were preprocessed using Statistical Parametric Mapping (SPM12, http://www.fil.ion.ucl.ac.uk/spm) and Data Processing & Analysis for Brain Imaging (DPABI, http:// rfmri.org/dpabi) (Yan et al., 2016).The first 10 vol for each participant were discarded to allow the signal to reach equilibrium and the participants to adapt to the scanning noise.The remaining volumes were corrected for the acquisition time delay between slices.Then, realignment was performed to correct the motion between time points.Head motion parameters were computed by estimating the translation in each direction and the angular rotation on each axis for each volume.All participants' BOLD data were within the defined motion thresholds (translational or rotational motion parameters less than 2.0 mm or 2.0 • ).We also calculated frame-wise displacement (FD), which indexes the volume-to-volume changes in head position.Several nuisance covariates (the linear drift, the estimated motion parameters based on the Friston-24 model, the spike volumes with FD > 0.5 mm, the white matter signal, and the cerebrospinal fluid signal) were regressed out from the data.The datasets were then band-pass filtered using a frequency range of 0.01 to 0.1 Hz.In the normalization step, individual structural images were firstly co-registered with the mean functional image; then the transformed structural images were segmented and normalized to the Montreal Neurological Institute (MNI) space using the diffeomorphic anatomical registration through the exponentiated Lie algebra (DARTEL) technique (Ashburner, 2007).Finally, each filtered functional volume was spatially normalized to the MNI space using the deformation parameters estimated during the above step and resampled into a 3-mm cubic voxel.

FCD analysis
FCD was computed according to the method described by previous studies (Tomasi and Volkow, 2010;Yang et al., 2020;Zhao et al., 2022b;Zhu et al., 2017Zhu et al., , 2022;;Zhuo et al., 2018Zhuo et al., , 2014)).First, Pearson correlation coefficients were calculated between the BOLD time courses of all pairs of voxels within the whole brain, generating a whole-brain voxel-wise functional connectivity matrix for each participant.Two voxels with a correlation coefficient above a threshold of 0.25 were considered significantly connected.For a given voxel, its FCD was defined as the total number of functional connections with correlation coefficients above 0.25 between that voxel and all other voxels within the whole brain.This calculation was repeated for all voxels within the whole brain, yielding a FCD map per participant.Note that FCD is also referred to as the nodal degree centrality of binary networks in graph theory, and regions with high FCD are considered functional hubs that are highly connected to the rest of the brain.Then, we normalized the FCD value of each voxel by dividing it by the global mean FCD value.The resultant FCD maps were spatially smoothed with a 6 mm full-width at half-maximum Gaussian kernel.

Correlation between behavioral inhibition and FCD
A voxel-wise approach was used to examine the correlations between Acc_No-Go and FCD across subjects in females and males, separately.We used multiple regression model in the SPM12 to identify any voxels in the FCD images that showed significant correlations with Acc_No-Go while controlling for potential confounding factors including age, education, and FD.The statistical analysis yielded a t map, representing correlations between Acc_No-Go and FCD.For the voxel-based analysis, multiple comparison correction was performed using the cluster-level Fig. 1.A flowchart of the study design.We initially computed FCD to reflect whole-brain functional connectivity profile using resting-state fMRI and calculated Acc_No-Go to measure behavioral inhibition ability using a Go/No-Go task in a large cohort of 510 healthy young adults, followed by cross-subject correlation between FCD and Acc_No-Go to identify neural correlates of behavioral inhibition.Then, we examined the spatial relationships of neural correlates of behavioral inhibition with gene expression and neurotransmitter atlases to explore their potential genetic architecture and neurochemical basis.Abbreviations: Acc_No-Go, accuracy in "No-Go" conditions; FCD, functional connectivity density; fMRI, functional magnetic resonance imaging.The data are presented as mean ± SD.Abbreviations: Acc_No-Go, accuracy in "No-Go" conditions; FD, frame-wise displacement; RT_Go, mean reaction time of correct responses in "Go" conditions; SD, standard deviation.
family-wise error (FWE) method, resulting in a cluster defining threshold of P = 0.001 and a corrected cluster significance of P < 0.05.

Brain gene expression data processing
Brain gene expression data were obtained from the AHBA dataset (http://www.brain-map.org)(Hawrylycz et al., 2015(Hawrylycz et al., , 2012)).This dataset was derived from six human postmortem donors (Table S1 in the Supplementary materials).The expression of > 20,000 genes at 3702 spatially distinct brain tissue samples was measured using custom 64 K Agilent microarrays.A newly proposed pipeline was employed to process gene expression data (Arnatkeviciute et al., 2019).Specifically, the probe-to-gene annotations were first updated based on the latest available information from the National Center for Biotechnology Information using the Re-Annotator package (Arloth et al., 2015).With intensity-based filtering, we excluded probes that did not exceed the background noise in at least 50 % of samples across all donors.Since multiple probes were used to measure the expression level of a single gene, we further used the RNA-seq data as a reference to select probes.After excluding genes that do not overlap between RNA-seq and microarray datasets, we computed the correlations between microarray and RNA-seq expression measures for the remaining genes.After excluding probes with low correlations (r < 0.2), a representative probe was selected for a gene based on the highest correlation with the RNA-seq data.Here, only the tissue samples in the left cerebral cortex were included.For one, all six donors had expression data in the left hemisphere, whereas only two donors had samples in the right hemisphere.For another, the inclusion of subcortical samples might introduce potential biases, given the substantial divergence in gene expression between cortical and subcortical regions (Hawrylycz et al., 2012).To account for potential between-sample differences and donor-specific effects in gene expression, we conducted both within-sample cross-gene and within-gene cross-sample normalizations with the scaled robust sigmoid normalization method.Differential stability (DS) is a measure of consistent regional variation across donor brains.Earlier work has reported that genes with high DS demonstrate more consistent spatial expression patterns between donors and are enriched for brain-related biological functions (Hawrylycz et al., 2015).Since gene expression conservation across subjects is a prerequisite for the transcriptome-neuroimaging spatial correlations, only genes with relatively more conserved expression patterns were selected for analysis.To achieve this goal, we ranked the genes by their DS values and chose the 50 % of the highest DS genes for the main analysis.After these processing procedures, we obtained normalized expression data of 5013 genes for 1280 tissue samples.To ensure reliability, we further focused our analyses on the tissue samples within a cerebral cortical gray matter mask derived from the Human Brainnetome Atlas (Fan et al., 2016), yielding a final sample × gene matrix of 623 × 5013.

Transcriptome-neuroimaging spatial correlation and gene category enrichment analysis
We adopted transcriptome-neuroimaging spatial correlation and the newly developed ensemble-based GCEA to investigate the genetic architecture underlying neural correlates of behavioral inhibition.Specifically, we drew a spherical region of interest (radius = 3 mm) centered at the MNI coordinate of a given brain tissue sample and extracted the average t-value of voxels within the sphere from the t map for the correlations between Acc_No-Go and FCD.Then, Pearson correlation between gene expression and t-values across tissue samples was calculated in a gene-wise manner, yielding 5013 spatial correlation coefficients (henceforth referred to as gene scores).In accordance with the Fulcher et al. study (Fulcher et al., 2021), we conducted neuroimaging-spatial ensemble-based GCEA for these gene scores in the following way.First, updated GO term hierarchy and annotation files were obtained from the GO (http://geneontology.org/) on 11th July 2022.Second, direct gene-to-category annotations were performed for the 5013 AHBA genes, and we restricted our analyses to GO categories with 10-200 annotations.Third, the gene scores were agglomerated at the level of GO categories as a mean score of genes annotated to each GO category.Fourth, 10,000 surrogate maps with spatial autocorrelation matching the t map were generated using BrainSMASH package (https://github.com/murraylab/brainsmash),based on the spatial-lag model (Burt et al., 2020).Null distributions (i.e., neuroimaging-spatial ensemble-based null model) of mean gene scores for each GO category were generated through spatial correlations between gene expression and the 10,000 spatial autocorrelation-preserving surrogate maps.Finally, statistical significance of a GO category was assessed by comparing the GO category score derived from the real data to the neuroimaging-spatial ensemble-based null.The significance threshold was set at two-sided P < 0.05 (i.e., higher or lower than the null).

Correlation with neurotransmitters
JuSpace (https://github.com/juryxy/JuSpace) is a useful tool allowing for spatial correlation analyses between cross-modal neuroimaging data (Dukart et al., 2021).To determine the neurochemical basis underlying neural correlates of behavioral inhibition, we adopted JuSpace to investigate the spatial correlations of the t map with nuclear imaging derived measures covering various neurotransmitter systems including dopamine, serotonin, glutamate, GABA, acetylcholine, opioid, cannabinoid, noradrenaline, and fluorodopa (Table S2 in the Supplementary materials).Specifically, Pearson correlation coefficients between the t map and these neurotransmitter maps were calculated across 210 cerebral cortical regions derived from the Human Brainnetome Atlas (Fan et al., 2016) while adjusting for spatial autocorrelation and partial volume with the gray matter probability map.Exact P values were computed using spatial permutation-based null maps with 5000 permutations.Correlations were considered significant at P < 0.05.

Validation analysis
Considering that reaction time might influence task performance, we examined the correlations between Acc_No-Go and FCD in the significant clusters additionally adjusting for mean reaction time of correct responses in "Go" conditions.During the FCD calculation, we used the correlation coefficient threshold of 0.25 to eliminate weak correlations possibly arising from noise signals.To test the effects of different threshold selections, we re-calculated FCD using two other thresholds (0.2 and 0.3) and then repeated the correlation analyses between FCD and Acc_No-Go.

Neural correlates of behavioral inhibition in females
Our voxel-wise analysis revealed a significant negative correlation between Acc_No-Go and FCD in the left superior parietal lobule (SPL) (cluster size = 64 voxels, peak MNI coordinate x/y/z = − 24/− 54/63, peak t = − 4.617, partial correlation coefficient [pr] = − 0.263, P = 1.93 × 10 − 6 ) in females (P < 0.05, cluster-level FWE corrected) (Fig. 2).However, no significant correlation between Acc_No-Go and FCD was observed in males.For completeness, we conducted correlation analysis between Acc_No-Go and FCD in the whole sample of 510 participants and also found no significant correlation.As such, the subsequent analyses were carried out for the FCD correlates of behavioral inhibition in females solely.

Gene categories associated with neural correlates of behavioral inhibition
A combination of transcriptome-neuroimaging spatial correlation   and the ensemble-based GCEA showed that the FCD correlates of behavioral inhibition were spatially associated with gene expression of GO categories predominantly implicating essential components of the cerebral cortex (glial cell, neuron, axon, dendrite, and synapse) and ion channel activity (spatially-constrained permutation-based P < 0.05).Specifically, the FCD correlates of behavioral inhibition were positively associated with regulation of microglial cell activation, and negatively associated with neuron projection fasciculation, motor neuron axon guidance, neuron projection terminus, axonal fasciculation, axon terminus, regulation of dendrite extension, synaptic vesicle exocytosis, vesicle-mediated transport in synapse, GABA-ergic synapse, integral component of synaptic membrane, regulation of calcium ion transmembrane transporter activity, and sodium channel activity (Fig. 3 and Supplementary file 1).

Validation analysis
After additionally adjusting for the mean reaction time, the negative correlation between Acc_No-Go and FCD in the left SPL in females remained significant (pr = − 0.234, P = 2.70 × 10 − 5 ).When FCD was recalculated using two other correlation coefficient thresholds (0.2 and Fig. 4. Spatial correlations between neurotransmitters and neural correlates of behavioral inhibition.The outermost ring displays the names and maps of 27 neurotransmitter receptors/transporters.The second circle displays the distribution of neurotransmitter values across 210 cerebral cortical regions derived from the Human Brainnetome Atlas.The third circle displays cross-region Pearson correlation coefficients between these neurotransmitter maps and the FCD correlates of behavioral inhibition; the light orange background represents positive correlations and the light blue background represents negative correlations.The innermost ring displays the permutation-based statistical significance of the spatial correlations, i.e., -log 10 (P); * P < 0.05.The t map for the correlations between Acc_No-Go and FCD lies in the center.Abbreviations: 5-HT, 5-hydroxytryptamine; Acc_No-Go, accuracy in "No-Go" conditions; CB1, cannabinoid type 1; D, dopamine; DAT, dopamine transporter; FCD, functional connectivity density; FDOPA, Fluorodopa; GABAa, gamma-aminobutyric acid a; mGluR5, metabotropic glutamate type 5; MOR, mu opioid receptor; NAT, noradrenaline transporter; SERT, serotonin transporter; VAChT, vesicular acetylcholine transporter.
S. Cui et al.

Discussion
This study examined the sex-dependent neural correlates of behavioral inhibition and their underlying molecular mechanisms.Our data showed a significant negative correlation between behavioral inhibition and FCD of the left SPL in females but not males.Further spatial correlation analyses demonstrated that the identified neural correlates of behavioral inhibition were associated with gene expression of GO categories predominantly implicating essential components of the cerebral cortex and ion channel activity, as well as were linked to the serotonergic system.Our findings may provide a critical context for understanding how biological sex might contribute to differences in behavioral inhibition and its related disease risk.
The current observation of a significant cross-subject correlation between behavioral inhibition performance and FCD in the SPL emphasizes the crucial role of the parietal cortex in this cognitive domain, which is in line with previous literature.For instance, Kolodny and colleagues highlighted the contribution of the parietal cortex to inhibitory control using a Go/No-Go task and fMRI (Kolodny et al., 2017).It is generally accepted that the posterior parietal cortex is not only responsible for space, attention, and action-related processing, but also engaged in adaptive visual processing (Xu, 2018).As a hub region of the dorsal frontoparietal attention network, the SPL is involved in spatial attention control via its connections with frontal regions (Szczepanski et al., 2010(Szczepanski et al., , 2013)).Koenigs et al. reported that the SPL is critically important for the manipulation of information in working memory (Koenigs et al., 2009).Moreover, a brain parcellation study using multi-modal neuroimaging approaches identified several subregions in the SPL, with anterior subregions primarily linked to action processes and visually guided visuomotor functions, whereas posterior subregions mainly associated with visual perception, spatial cognition, reasoning, working memory and attention (Wang et al., 2015).It is noteworthy that behavioral inhibition has played a key role in the above-described complex cognitive processes pertinent to the parietal cortex.Remarkably, the finding that greater FCD in the SPL was related to worse behavioral inhibition ability seems counterintuitive at first sight and invites the speculation that excessive engagement of the SPL in global information processing may interfere with normal inhibitory control.Complementing and extending prior research, we found this behavior-neuroimaging association in females rather than males.The reasons for the observed sexual dimorphism are likely to be multi-factorial and might be of interest in the future.
What molecular mechanisms might underlie the observed associations between behavioral inhibition and FCD? Transcriptomeneuroimaging spatial correlation and the ensemble-based GCEA revealed that the FCD correlates of behavioral inhibition were spatially associated with gene expression of GO categories predominantly implicating essential components of the cerebral cortex (glial cell, neuron, axon, dendrite, and synapse) and ion channel activity.Microglial cells are responsible for regulating immune response in the brain, involving antigen presentation, debris phagocytosis, and cytokine production (Gertig and Hanisch, 2014).In addition, microglia shape neuronal networks through multiple potential mechanisms such as the elimination of redundant immature neurons, the regulation of synapse formation and pruning, the modification of synaptic connections, and the secretion of neuromodulatory factors (Salter and Beggs, 2014).FCD reflects a brain region's functional connectivity profile with all other regions across the whole brain.It is largely known that normal communication between brain regions relies on the integrity of neuron, axon, dendrite, and synapse.Indeed, information exchange between neurons primarily takes place at synapses of axons and dendrites (Harris et al., 2012;Pereda, 2014).Crucially, although GABA has long been associated primarily with the mediation of synaptic inhibition, it can also act as a trophic factor during nervous system development to influence events such as proliferation, migration, differentiation, synapse maturation and cell death (Owens and Kriegstein, 2002).From a behavioral viewpoint, GABA deficits have been associated with inhibitory control impairments in aging (Hermans et al., 2018) as well as several clinical conditions (Hines et al., 2013;Murley et al., 2020;Prevot and Sibille, 2021).With respect to ion channels, mounting evidence has supported their pivotal role in brain development and normal brain function (Catterall, 2011;Dolphin and Lee, 2020;Hull and Isom, 2018;Moody and Bosma, 2005), with dysfunction of ion channels contributing to various brain diseases (Eijkelkamp et al., 2012;Kim, 2014;Kumar et al., 2016;Pietrobon, 2002).
Spatial correlation analyses demonstrated a significant link between neural correlates of behavioral inhibition and the serotonergic system (5-HT1b and 5-HT2a).Serotonin is a highly evolutionary conserved monoamine neurotransmitter that modulates multiple psychophysiological functions (Beliveau et al., 2017;King et al., 2008;Magalhaes et al., 2010;Young and Leyton, 2002).The intimate relationship between behavioral inhibition and serotonin has been well documented (Pattij and Schoffelmeer, 2015).For example, impulsive and aggressive behaviors can be modulated by serotonergic signaling, specifically through the 5-HT1b receptor (Nautiyal et al., 2015).In addition, agonists of the 5-HT1b receptor have been reported to exacerbate obsessive-compulsive disorder symptoms (Shanahan et al., 2011).Moreover, previous data suggest that the dysfunctional cortical systems underlying response inhibition deficits in behavioral variant frontotemporal dementia could be partially restored by increasing serotonergic neurotransmission (Hughes et al., 2015).Clinical studies have shown altered levels of the 5-HT1b receptors in behavioral inhibition-associated mental disorders like pathological gambling and alcohol dependence (Hu et al., 2010;Potenza et al., 2013).Likewise, individual differences in impulsive action have been proved to relate to variation in the cortical serotonin 5-HT2a receptor system (Fink et al., 2015).Empirical evidence from genetics has corroborated the involvement of 5-HT2a gene polymorphism in the pathophysiology of schizophrenia, depression, and pathological gambling (Gu et al., 2013;Wilson et al., 2013;Zhao et al., 2014).
Several caveats should be considered when interpreting the current findings.First, Acc_No-Go represents one of the most frequently used neurocognitive measures to assess behavioral inhibition.Our preliminary results should be verified with use of other neuropsychological tools.Second, the correlational nature of the analyses limits our ability to make assumptions about causality.Third, our analyses were conducted in a cohort of healthy young adults, which might restrict generalizability to the general population.Therefore, these findings need replication in different samples.Fourth, FCD was calculated on the basis of our resting-state functional MRI data, whereas the brain maps of gene expression and neurotransmitters were derived from publicly available datasets.Our spatial correlation analyses ignored such variability across individuals, which should be captured and examined in future investigations.Fifth, the imbalance in the number of female and male participants may bias our results.That said, the non-significant neuroimaging-behavior correlation in males may be partially due to limited statistical power from the relatively small sample size.In the future, more male participants will be recruited to validate our findings.Sixth, our neuroimaging results were derived from females while the gene expression data almost from the male donors, which may influence our interpretation.Finally, we used an absolute correlation threshold of 0.25 to calculate FCD.Given that a proportional threshold can preserve a uniform density of connections across individuals prior to calculating FCD, this thresholding method may reduce bias and should be adopted in future studies.

Conclusion
We found the neural correlates of behavioral inhibition in females that were potentially modulated by specific genetic architecture and neurochemical basis.Our findings may not only yield important insights into the molecular mechanisms underlying the female-specific neural substrates of behavioral inhibition, but also provide a critical framework for understanding how biological sex might contribute to variation in behavioral inhibition and its related disease risk.

S
.Cui et al.

Fig. 2 .
Fig. 2. Neural correlates of behavioral inhibition in females.Left panel: Acc_No-Go was negatively correlated with FCD in the left SPL.Right panel: a scatter plot of the correlation between Acc_No-Go and FCD.Abbreviations: Acc_No-Go, accuracy in "No-Go" conditions; FCD, functional connectivity density; L, left; pr, partial correlation coefficient; SPL, superior parietal lobule.

Fig. 3 .
Fig.3.Gene categories associated with neural correlates of behavioral inhibition.A combination of transcriptome-neuroimaging spatial correlation and the ensemble-based GCEA showed that the FCD correlates of behavioral inhibition were spatially associated with gene expression of GO categories predominantly implicating essential components of the cerebral cortex (glial cell, neuron, axon, dendrite, and synapse) and ion channel activity.The y-axis represents GO category and the x-axis denotes GO category score.The color denotes the spatially-constrained permutation-based statistical significance of the spatial correlations, i.e., -log 10 (P).The triangle represents positive association and the circle represents negative association.Abbreviations: BP, biological process; CC, cellular component; FCD, functional connectivity density; GCEA, gene category enrichment analysis; GO, gene ontology; MF, molecular function.

Table 1
Demographic and behavioral characteristics of the participants.