Interaction of Catechol‐O‐methyltransferase Val158Met polymorphism and sex influences association of parietal intrinsic functional connectivity and immediate verbal memory

Abstract Introduction Sex differences modulate catechol‐O‐methyltransferase (COMT) genotype effect at a synaptic dopamine level, which influences brain function as well as cognitive performance. In this study, we investigated how COMT Val158Met polymorphism and sex affect intrinsic functional connectivity and memory. Methods Intrinsic functional networks were extracted using independent component analysis of resting‐state functional magnetic resonance imaging data from 186 healthy young COMT‐genotyped participants. The association of these functional networks and memory function was tested to investigate whether the effect of COMT × sex interaction influences the association of intrinsic functional connectivity and memory performance. Quadratic curve fit estimation was used to examine the relationship between functional connectivity and speculative dopamine level among groups. Results COMT MM/MV carriers, relative to VV carriers, showed increased functional connectivity in left superior parietal lobule and right inferior frontal gyrus. Further, male MM/MV carriers showed significant higher mean functional connectivity in left inferior parietal lobule relative to male VV carriers and female MM/MV carriers, which was associated with worse immediate verbal recall performance. Additionally, the relationship between inferior parietal lobule functional connectivity and speculative dopamine level among groups fits the quadratic curve. Conclusions These findings suggest that the interaction of COMT genotype and sex might regulate synaptic dopaminergic concentrations and influence the association of intrinsic functional connectivity and immediate verbal memory in left inferior parietal lobule.

. Several behavioral and task-fMRI studies also provide supporting evidence to demonstrate the relationship between COMT and memory performance (Bertolino, Rubino, & Sambataro, 2006;Egan, Goldberg, & Kolachana, 2001;Frias et al., 2004;Raz, Dahle, Rodrigue, Kennedy, & Land, 2011), as well as COMT × sex interaction effect on cognitive performance (Holtzer et al., 2010;Kempton et al., 2009;O'Hara et al., 2006). However, the question of whether differences in synaptic dopaminergic concentration that arises from COMT × sex interaction effect influences the association between RSNs and memory performance remains unclear. Still, brain structure has also been reported to be influenced by COMT and sex. Some previous studies have revealed COMT × sex difference in gray matter volume in prefrontal regions (Kates, Antshel, & AbdulSabur, 2006;Tian et al., 2013;Xu, Qin, & Li, 2016a), whereas others have found no relationship between COMT genotype and sex on gray matter volume (Barnes, Isohanni, & Barnett, 2012;Zinkstok, Schmitz, & van Amelsvoort, 2006). Therefore, whether COMT × sex interaction influences the gray matter volume in healthy young people still needs to be explored.
In the current study, we investigated the effects of COMT genotype and sex on RSNs and memory functions in healthy young participants. Within the RSNs, we focused on DMN, ECN, and FPN that are highly associated with cognitive functions and examined their relationship with COMT genotype and memory functions.
Previous studies have demonstrated an "inverted-U-shaped" function for the relationship between dopamine levels and cognitive performance (Cools & D'Esposito, 2011;Dickinson & Elvevag, 2009;Mattay, Goldberg, & Fera, 2003;Vijayraghavan, Wang, Birnbaum, Williams, & Arnsten, 2007), and a similar "inverted-U-shaped" relationship was also found for brain functional connectivity (FC) and participants with varying dopamine concentrations (Cole et al., 2013;Cools & D'Esposito, 2011;Tian et al., 2013). Thus, we hypothesized that the interaction of COMT Val 158 Met polymorphism and sex would modulate the association of FC strengths and memory performance. Specifically, we expected that male MM/MV carriers to have optimal dopamine level in the "inverted-U-shaped" function and the strongest FC strength that associated with higher memory performance.

| Participants
The study was approved by the Medical Research Ethics Committee of Nanjing Drum Tower Hospital. Two hundred young healthy Han Chinese participants were recruited from Nanjing, Jiangsu, China, by advertisements on the Internet. Exclusion of participants was performed via telephonic interview, based on the following criteria: (a) nonright handedness, (b) achromatopsia or hypochromatopsia, (c) smoking, (d) excessive drinking, (e) neurological and psychotic disorders or family history, (f) using drugs recently or drug abuse, and (g) pregnancy. All females were not in their menstrual period at the day of scanning and evaluation.
Written informed consent was obtained from all participants.
Financial compensation was provided after they completed all tests. One hundred and ninety-six participants who met the criteria agreed to participate in the study. Nine participants were excluded due to failure of genotyping and one due to unsuccessful working memory test. Finally, data from 186 participants were

| Genotyping
Genomic deoxyribonucleic acid (DNA) was extracted from white blood cells using a DNA extraction kit (BioTeke, Beijing, China) according to the manufacturer's instructions. COMT

| Episodic memory
The California Verbal Learning Test-Second Edition (CVLT-II) (Delis, 2000), which measures episodic verbal learning and memory abilities, was administered in a quiet room. The test contained two learning lists, A and B. List A comprised of 16 words drawn from four semantic categories, whereas list B comprised of 16 words drawn from two semantic categories of list A and two new semantic categories to create interference. First, participants were presented with list A over 5 learning trials and were asked to recall as many words as they can in any order (free recall) after each trial. Participants were then presented with list B. Free and cued recall of list A were tested immediately after list B (short-delay), and again after 20 min (longdelay). In cued recall, participants were prompted with the semantic category. Finally, participants performed a recognition task in which they indicated whether the presented item was a target word from list A or a distractor (total presented words, 48:16 from list A + 32 from list B and completely new words). Number of correct responses was recorded for the first time of list A (immediate recall), over 5 learning trials of list A (total recall), immediately after list B (shortdelay free and cued recall), and 20 min after list B (long-delay free and cued recall). Recognition discriminability was calculated as d' prime (i.e., numbers of hits minus false alarms).

| Working memory
Participants' working memory capacity was measured using a n-back behavioral task, performed outside scanner on a computer in a quiet room. E-Prime 2.0 software (Psychology Software Tools, Pittsburgh, PA, USA) was used to present the visual stimuli, and participants' button presses and response times were collected. Sixty blocks of 1-back and 3-back tasks were presented. Within each block, 60 letters were presented for 200 ms with an interstimulus time interval of 1,800 ms each. Participants pressed the right arrow button with their middle finger when the letter on the screen was identical to the one that was presented 1 (1-back) or 3 (3-back) letters earlier.
Otherwise, they pressed the left arrow button with their index finger.
They had to respond within a time period of 2,000 ms. Participants practiced the task before actual data collection. Proportion of correct responses and response times for 1-back and 3-back conditions were calculated, and working memory capacity was computed as the absolute value of proportion of correct responses for 3-back minus 1-back condition.

| MRI acquisition
Participants underwent structural and RS-fMRI scanning using the Philips 3T scanner (Ingenia, best, the Netherlands) with a 32-channel phased-array head coil. They were instructed to keep their eyes closed without falling asleep during scanning.
Structural 3D T1-weighted images were collected using a

| RS-fMRI data preprocessing and analyses
RS-fMRI data were preprocessed and analyzed using statistical parametric mapping (SPM12; Wellcome Trust Center for Neuroimaging) and FMRIB Software Library Multivariate Exploratory Linear Optimized Decomposition into Independent Components (FSL-MELODIC; https://fsl.fmrib.ox.ac.uk/fsl/fslwi ki/MELODIC). The quality of images was visualized to exclude scanner artifacts by an experienced radiologist. For each participant, volumes of the RS-fMRI data were realigned to the first volume, unwarped, spatially normalized into Montreal Neurological Institute (MNI) space, resampled into 4 × 4 × 4 mm 3 voxels, and smoothed using a 8mm full width half maximum (FWHM) Gaussian kernel. Framewise displacement (i.e., the averaged spatial displacement between two adjacent volumes) for the RS-fMRI data was calculated for each participant, and those yielded an average framewise displacement of >0.2 mm were excluded in subsequent analyses.
The smoothed normalized RS-fMRI data from all participants were then decomposed into 20 sets of spatiotemporal components using FSL-MELODIC algorithm for independent component analysis (ICA) with a high-pass filter of 125s to remove physiological noise and MCFLIRT motion correction to remove head motion ). Pearson's spatial cross-correlations of the 20 components were calculated using ten well-identified RSNs ) to match the networks identified in our data and kept those components that yielded a significant spatial correlation (Pearson's r > 0.20). We found 16 components that were matched with the ten well-identified RSN templates and 4 components that did not significantly correlate with any of the RSN template and likely due to artefactual effects (Reineberg, Andrews-Hanna, Depue, Friedman, & Banich, 2015).
Thus, 16 components were used to generate subject-specific spatial maps and its associated time series using dual regression (Beckmann, Mackay, Filippini, & Smith, 2009). Specifically, for each participant, group-averaged spatial maps were regressed (as spatial regressors in a multiple regression) into individual participant's 4D space-time dataset. This resulted in a set of subject-specific time series, one per grouplevel spatial map. Next, the time series were regressed (as temporal regressors, again in a multiple regression) into the same 4D dataset to create a set of subject-specific spatial FC maps of each network, one per group-level spatial map (statistical threshold of Z> 2.3).
Finally, to investigate the COMT genotype and sex effects on RSN, we input the subject-specific spatial FC maps of DMN, ECN, left FPN (LFPN), and right FPN (RFPN) into a second-level analysis that modeled 4 RSN networks (DMN, ECN, LFPN, and RFPN) as a within-subject variable and 4 groups (MM/MV males, MM/MV females, VV males, and VV females) as a between-subject variable using SPM. Inferences were made at the second level to allow for random effects analysis and inferences at the population level (Penny & Holmes, 2007). At the random effects level, we tested for main effects of COMT and sex, and COMT genotype × sex interaction differences in each of the spatial maps. Unless otherwise specified, we report activations at family-wise error (FWE) of p < .05 at the cluster level corrected for multiple comparisons across the whole brain (auxiliary threshold of p < .001 uncorrected). The search space for interaction effect was limited to the combination of the template of DMN, ECN LFPN, and RFPN as our search volume of interest (i.e., including 10,064 voxels) . For each participant, mean FC was extracted from the regions-of-interest (ROI) that showed significant effect of COMT genotype × sex and entered into an analysis of covariance to examine the influence of COMT genotype × sex interaction on associations between mean FC and working memory as well as episodic memory measures using IBM Statistical Package for the Social Sciences (SPSS) version 21.0. We also investigated the relationship between mean FC of the ROI and speculative dopamine level among groups by curve fit estimation in SPSS. Considering the "inverted-U-shaped" (i.e., nonlinear) model in dopamine neuromodulation, we used a quadratic model for the curve fit estimation. Taking into account of the effect of COMT genotype and estrogen on the dopamine level, female MM/MV carriers was hypothesized to have the highest dopamine concentration, whereas male VV carriers have the lowest.

| Structural MRI (sMRI) preprocessing and analyses
The structural data were analyzed using SPM12. For each participant, the sMRI data were first segmented into gray matter (GM), white matter (WHM), and cerebrospinal fluid (CSF). The segmented images were then subsequently normalized to the averaged population templates generated from the complete image set using Diffeomorphic Anatomical Registration using Exponentiated Lie algebra (DARTEL) Tools (Ashburner, 2007), spatially normalized into MNI space and spatially smoothed using an 8mm FWHM Gaussian kernel. Total intracranial volume (TIV) for each participant was calculated as the sum of GM, WHM, and CSF volumes.
To investigate whether there are any main effects of COMT or sex, or COMT genotype × sex interaction on structural differences among the four groups, the smoothed GM images of each participant were entered into a full factorial 2-way (genotype and sex) random effects ANOVA. TIV was included in the design matrix to account for intersubject variability due to head size. Unless otherwise specified, we reported significant differences at p < .05 after the cluster level FWE corrected for multiple comparisons across the whole brain (auxillary threshold of p < .001 uncorrected) within a specified search volume. Table 1 shows the mean and standard deviation of age, education years, the scores of California Verbal Learning Test (CVLT), mean accuracy, mean response times, and capacity of n-back tasks of the four groups based on COMT genotype and sex. The distribution of COMT Val 158 Met (8Met/Met (MM), 74Met/Val (MV), 104Val/Val (VV)) genotype was in Hardy-Weinberg equilibrium (p> .05). Given the low frequency of MM homozygotes (4-5 times lower than VV homozygotes), we merged MM and MV participants into one MM/ MV group (Taylor et al., 2007;Xu, Qin, Liu, et al., 2016). Thus, in the final sample, there were 82 participants (27 males, 55 females) in the MM/VV group and 104 participants (31 males, 73 females) in the VV group. There was no main effect of COMT genotype, sex, and interaction of COMT genotype × sex (after correcting for multiple comparisons using false discovery rate ( Benjamini & Hochberg, 1995)) among four groups were found in any of behavioral measures (see Table 1).

| Influence of COMT Val 158 Met and sex on RSNs
Significant increased FC for MM/MV carriers relative to VV carriers was found in left superior parietal lobule and right inferior frontal gyrus in LFPN and not in other networks (see Table 2, Figure 2). Also, significant increased FC for males relative to females was found in each of the four networks (see Figure S1, Table S1). Furthermore, we found a significant COMT genotype × sex interaction on mean FC of left inferior parietal lobule (IPL) in LFPN spatial map within the search volume of interest (see Table 2, Figure 3a). Post hoc comparisons demonstrated increased mean FC for MM/MV group relative to VV group only in males (Z = 4.77, Figure 3b) and increased mean FC for males relative to females in COMT MM/MV group (Z = 7.03, Figure 3b). No COMT genotype × sex interaction on mean FC was found in DMN, ECN, and RFPN spatial maps. Also, we did not find any significant main effects on mean FC for VV relative to MM/MV group or for females relative to males in each of the four networks.
Curve fit estimation in SPSS showed that the quadratic (inverted-Ushaped) model was found to be significantly fit (R-square = 0.048, p = .011, see Figure S2).

| Influence of COMT Val 158 Met and sex on gray matter volume
At the group level, a 2-way (COMT genotype and sex) random effects ANOVA comparing GM volumes did not reveal any significant main effects of COMT or sex, nor effect of COMT genotype × sex interaction.

| Examining the influence of COMT × sex interaction on associations between mean FC of LFPN and working and episodic memory measures
We examined whether the significant COMT genotype × sex interaction of mean FC of left IPL could be explained by differences in memory function. Table 3

| D ISCUSS I ON
This study investigated the influence of COMT genotype and sex on intrinsic FC and memory performance in healthy young adults.  Note: Results are expressed as mean (SD). Variables were calculated by a 2-way (genotype and sex) analysis of variance among four groups based on COMT genotype and sex to identify COMT × sex interaction effect. There is no significant difference among groups in terms of age and education years. CVLT and n-back tests did not show significant difference among groups after false discovery rate correction.  (Chen, Lipska, & Halim, 2004;Lotta et al., 1995). Further, neuropharmacological studies demonstrate that levodopa, which is used as a replacement for dopamine, optimizes the intrinsic FC in FPN (Cole, Oei, & Soeter, 2012;Simioni, Dagher, & Fellows, 2017).
Collectively, these studies, together with our findings, suggest that higher intrinsic FC in MM/MV carriers is related to increased synaptic dopaminergic concentrations. Furthermore, human estrogen levels could also inhibit COMT gene transcription and down-regulate COMT enzyme activity (Jiang et al., 2003;Xie et al., 1999) that affect intrinsic brain FC. Indeed, studies that identified sex differences on RSNs demonstrate higher FC for males relative to females (Biswal, Mennes, & Zuo, 2010;Filippi et al., 2013). Conversely, other studies show higher FC for females relative to males (Allen, Erhardt, & Damaraju, 2011;Bluhm, Osuch, & Lanius, 2008).
Interestingly, our findings reveal significant sex differences in mean FC of left IPL that is associated with CVLT first trial performance. Specifically, we found that male MM/MV carriers have increased functional coupling in LFPN that are associated with impaired performance in immediate verbal recall. IPL activity is correlated with many cognitive performance such as immediate recall score (Huo, Li, Wang, Zheng, & Li, 2018), episodic memory performance (He, Carmichael, & Fletcher, 2012), social cognition, and language (Bzdok et al., 2016). It is connected with other frontal and parietal regions and is considered as the node in FPN (Markett, Reuter, & Montag, 2014). The FPN encompasses lateral prefrontal cortex, anterior cingulate, inferior parietal lobules, and lateral cerebellum (Markett et al., 2014;Vincent, Kahn, Snyder, Raichle, & Buckner, 2008) and has been associated with working memory and executive control (Mohr, Wolfensteller, & Betzel, 2016;Sheffield, Repovs, & Harms, 2015). It is demonstrated that brain dopamine level strengthens FC in FPN, which positively correlates with working memory performance ( Simioni et al., 2017). In addition, previous meta-analysis has demonstrated strong correlation between  Collectively, these studies provide supporting evidence that synaptic dopamine concentration, which is influenced by COMT genotype and sex, plays an important role in the association between FPN intrinsic activities and immediate memory performance. Although CVLT is considered as a sensitive measure of episodic memory function, the meaning of different scores might vary among each other. Specifically, the immediate recall of first trial is related more to working memory, not episodic memory as compared to other CVLT scores (Wolk & Dickerson, 2011). In our study, different from our initial hypothesis, immediate verbal recall of CVLT first trial is negatively associated with spontaneous activity of fronto-parietal working memory network in male MM/MV carriers, which suggests that the immediate verbal memory function is linked with enhanced inter-regional desynchronization in resting state (Fjell, Sneve, & Grydeland, 2015). Analogous electrophysiological findings provide supporting evidence that demonstrate that strong desynchronization is associated with better verbal memory performance (Klimesch, Doppelmayr, Schimke, & Ripper, 1997) and thus is essential for large-scale integration of human brain. RS-fMRI results may provide similar information to understand the structure of human brain networks ( Laufs, Krakow, & Sterzer, 2003).
The "inverted-U-shaped" model suggests that an optimal level of dopamine is the most efficient, while suboptimal and supraoptimal dopamine level will impair brain activity and cognitive performance.
In support to this, our findings demonstrate that mean FC of IPL in female MM/MV carriers is decreased relative to male MM/MV carriers, although the dopamine level of female MM/MV carriers should be higher considering Met and estrogen both down-regulate COMT enzyme activity.
In accord to our findings, studies with schizophrenic patients provide additional supporting evidence, since variations of the COMT genotype have been implicated to have a strong link with memory performance of schizophrenia (Egan et al., 2001;Harrison & Weinberger, 2005;Twamley, Hua, & Burton, 2014). Deficits in immediate verbal memory are frequently reported in patients suffering from schizophrenia (Gold, Rehkemper, & Binks, 2000;Toulopoulou, Rabe-Hesketh, King, Murray, & Morris, 2003). In addition, abnormalities in FPN are typically associated with schizophrenia (Xiang, Xu, & Wang, 2019), and task-fMRI studies have demonstrated immediate memory modulation of reduced fronto-parietal effective connectivity in schizophrenia (Nielsen et al., 2017).
We found no significant effects in GM volume within the search volume of interest, thus suggesting that the effects of COMT genotype and sex in healthy young participants manifest only at the functional level. Despite this, other studies have found mixed results: Some brain regions have either larger or smaller GM volume for Val relative to Met carriers (Cerasa et al., 2008;Taylor et al., 2007) and also a COMT × sex interaction effect on prefrontal GM volume ( Xu, Qin, & Li, 2016 memory in healthy young adults (Barnett, Scoriels, & Munafò, 2008;Wardle, de Wit, Penton-Voak, Lewis, & Munafo, 2013). A possible explanation is that the n-back task paradigm that we used in the current study is not sensitive enough to detect any significant group difference in young healthy participants as the performance of n-back task is at ceiling (Buckert, Kudielka, Reuter, & Fiebach, 2012;Kane, Conway, Miura, & Colflesh, 2007). In addition, our sample size is comparatively small relative to other relevant genetic studies (Egan et al., 2001). Furthermore, although COMT enzyme activity and synaptic dopamine concentration in human brain are closely related to COMT genotype, we did not measure dopamine concentration in human brains directly, thus making the dopamine level hypothesis speculative. We also could not exclude confounding variable of fluid intelligence in our current study, as previous studies have found that COMT genotype influences on the neural circuitry supporting fluid intelligence (Bishop et al., 2008;Green, Kraemer, DeYoung, Fossella, & Gray, 2012 Also, we did not record exact menstrual date of females while collecting the data. Future studies using larger sample size with a more direct manipulation of dopamine levels in human brain and greater task difficulty to measure working memory capacity and fluid intelligence will be required. The difference between eyes-open versus.
closed and the influence of menstrual date on behavior would also be interesting to explore in the future.
In conclusion, our study demonstrates that the COMT Val 158 Met polymorphism affects mean FC in parietal and frontal regions.
Particularly, male MM/MV carriers, who are hypothesized to have higher dopaminergic concentrations, show higher FC in left IPL which is related to impaired immediate verbal recall performance, thus providing supporting evidence for the pathophysiology of schizophrenia.

CO N FLI C T O F I NTE R E S T
The authors report no biomedical financial interests or potential conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Research data are not shared.