Impact of sex and serum lipids interaction on working memory: A large‐scale brain networks study

Abstract Backgrounds Previous studies have demonstrated that both serum lipid levels and sex are crucial factors associated with individual cognition. However, the impact of sex and serum lipid interaction effects on the brain and cognition remains largely unknown. This study aimed to explore the underlying neural mechanisms among sex, serum lipids, and cognition using large‐scale brain networks. Methods Resting‐state functional MRI data were collected from 157 young healthy adults. Independent component analysis was used to examine large‐scale inter‐ and intra‐network functional connectivity (FCs). Peripheral venous blood samples were collected to measure serum lipid levels. The three‐back task was employed to assess cognition (i.e., working memory). General linear model, correlation, and mediation analyses were conducted to examine the interaction effects of sex and serum lipids on large‐scale brain networks and their relationship with working memory. Results We found that inter‐network connectivity with the executive control network at its core was more susceptible to sex and triglyceride interaction effects. The intra‐network connectivity in the dorsal attention networks (DANs), lateral visual networks, and anterior default mode networks was influenced by the interaction effects of sex and total cholesterol (TC)/low‐density lipoprotein cholesterol. Specifically, correlations between serum lipids and affected brain networks were found to be sex‐specific. In addition, higher intra‐network FC in the right inferior parietal (R‐IPL) of the DAN correlated with a longer three‐back reaction time in females. More importantly, the relationship between serum TC levels and three‐back reaction time was mediated by intra‐network connectivity in the R‐IPL of the DAN. Conclusions Our findings describe the impact of sex and serum lipid interaction effects on large‐scale brain networks, as well as on cognitive function. Our data suggest that sex‐specific usage of serum lipids or brain networks would be beneficial for monitoring and therapy in dyslipidemia‐related cognition decline.


INTRODUCTION
Previous studies have demonstrated that serum lipid levels are associated with individual differences in cognition, such as working memory Xia et al., 2015), executive control, and sustained attention (Gendle et al., 2008). Numerous cross-sectional and longitudinal studies have consistently shown that higher serum total cholesterol (TC) (Gendle et al., 2008;Solomon et al., 2009;Xia et al., 2015), triglycerides (TG) (Parthasarathy et al., 2017), and low-density lipoprotein cholesterol (LDL-C) (Bates et al., 2017;Power et al., 2018) levels are associated with poor cognitive performance, whereas higher serum high-density lipoprotein cholesterol (HDL-C) levels predict the better maintenance of cognitive function (Bates et al., 2017;Reynolds et al., 2010;Sun et al., 2019). In humans, the brain has the secondhighest lipid content after adipose tissue, accounting for 50% of its dry weight (Hornemann, 2021). High lipid content plays a crucial role in the brain, because lipids provide structural integrity and modulate the fluidity of brain neuronal cells. Lipid dysregulation has been associated with the etiology and progression of neurodegeneration and other neurological pathologies (Pfrieger, 2003;Yoon et al., 2022). Therefore, lipids are emerging as important potential targets for the early diagnosis and prognosis of neurological diseases. Recent advances in neuroimaging technologies, especially brain functional connectivity (FC), have facilitated studies examining the relationship between serum lipids and the brain in both normal and disease-affected populations. For example, cholesterol was found to accelerate the impact of age on neural trajectories by disrupting FC in circuits implicated in integrative processes and behavioral control (Spielberg et al., 2017). High serum cholesterol has been associated with the disruption of FC in the salience network in non-demented elderly persons .
Disruption of the resting-state connectivity in the hippocampus and middle frontal gyrus may mediate the relationship between poorly controlled cholesterol, impaired attention, and executive function in type 2 diabetes mellitus (T2DM) patients (Xia et al., 2015).
Similar to serum lipids, biological sex is also an important factor that can impact the cognitive function of humans. Sex-specific differences have been identified not only in advanced neurocognitive processes, such as attention (Bangasser et al., 2019), verbal working, and spatial memory (Chen et al., 2020;Voyer et al., 2021), but also in cognitive development and psychiatric disease (Kaczkurkin et al., 2019). Earlier studies suggested that sex-specific cognitive differences were due to differences in the structure and function of the brain in males and females (de Lacy et al., 2019;Gur & Gur, 2017;Ingalhalikar et al., 2014;Sang et al., 2021;van Eijk et al., 2021). Indeed, the brain network not only exhibits sexual dimorphism but also plays a vital role in mediating sex-related cognitive changes Li et al., 2022;Murray et al., 2021;Tunc et al., 2016;Xu et al., 2020;Zhao et al., 2021;Zhao et al., 2020). In addition, sex-specific differences in serum lipid profiles have been frequently reported in healthy aged individuals and have been associated with worse cognitive performance (Lu et al., 2017). Previous studies have suggested that physiological doses of sex steroid hormones can affect serum lipid and lipoprotein levels in humans (Engelberg & Glass, 1955). However, the interactive role of sex-specific brain network mechanisms and specific lipids on cognitive performance remains unclear.
The human brain is a complex system consisting of multiple distinct and interacting functional networks that subserve different functions (De Luca et al., 2006;Power et al., 2011). Each functional network is composed of several brain regions with a high degree of consistency in signal change over the course of resting-state functional MRI (rs-fMRI). Different networks display diverse activity patterns. These functional networks can be automatically identified by independent component analysis (ICA) of rs-fMRI data, a useful method enabling data-driven, exploratory investigation of temporal correlations among brain regions at rest (Calhoun et al., 2001;van de Ven et al., 2004). This approach has been broadly applied to explore the underlying relationships between cognition and large-scale functional organization (interand intra-network FCs) in normal and abnormal brains Zhang et al., 2021). Inter-network FC reflects the information exchange capability between different regions, whereas intra-network FC reflects information specialization within specific functional networks (Liao et al., 2017). Abnormal integration and separation of information between brain networks have been examined in some psychiatric disorders, which can potentially lead to cognitive changes (Wei et al., 2022;Yang et al., 2016;Zhu et al., 2016). However, as previous studies mainly focused on the relationship between sex or lipids on cognition in specific brain regions or networks, little is known about the roles of inter-and intra-network FCs in the interaction effects of sex and serum lipids on cognition.
Based on these earlier studies, we sought to determine how largescale brain networks mediate the interaction effects of sex and serum lipids on cognitive function by performing an rs-fMRI analysis with ICA and quantifying large-scale inter-and intra-network FCs in 157 healthy adults. Peripheral venous blood samples were collected to measure serum lipid levels. The three-back task was employed to assess working memory. Using these data, we first sought to determine the interand intra-network FCs that were influenced by the interaction effects of sex and serum lipids. Second, we aimed to assess the associations of affected brain networks with working memory in males and females separately. Finally, we attempted to determine the mediative role of these identified FC markers in accounting for the associations between serum lipids and working memory in males and females separately. Based on our findings, we hypothesize that serum lipids affect cognition through large-scale brain networks in a sex-specific manner.

Participants
The study included 157 healthy young adults who were recruited by advertisement. All participants fulfilled the following inclusion crite-

Blood sampling and serum lipid measurement
Blood samples were collected within 1 day before or after MRI examination. After an overnight fasting period, peripheral venous blood samples (2 mL) were collected from all participants the following morning. Samples were centrifuged to separate the serum at 3000 rpm for 10 min at room temperature, and lipid profile analysis was carried out immediately on the fresh serum. Serum TC, TG, and HDL-C levels were measured in an automated clinical analyzer (Roche cobas 8000). Serum LDL-C levels were estimated using the Friedewald equation.

Working memory assessment
The letter three-back task was performed on a computer to assess working memory using E-Prime 2.0 software (http://www.pstnet.com/ eprime.cfm) (Owen et al., 2005). During the task, each participant viewed a series of sequentially presented letters, with each letter stimulus presented for 200 ms and an interval of 1800 ms between stimuli.
Participants were instructed to press a button on the right with their middle finger if the letter appearing on the screen was identical to the one displayed three letters earlier. Otherwise, they pressed a button on the left with their index finger. The task consisted of 60 trials. Prior to the formal test, participants received verbal instructions and underwent a practice test. Working memory performance was measured using two metrics: accuracy and the mean reaction time of correct responses. An increase in accuracy and reduction in mean reaction time signified a better working memory.

MRI data acquisition
MRI scans were obtained using a 3.0 Tesla MR system (Discovery MR750w, General Electric, Milwaukee, WI, USA) with a 24-channel head coil. Earplugs were used to reduce scanner noise, and tight but comfortable foam padding was used to minimize head motion.
High-resolution 3D T1-weighted structural images were acquired by employing a brain volume (BRAVO) sequence with the following All images were visually inspected to ensure that only images without visible artifacts (e.g., ghosting artifacts arising from subject movement and pulsating large arteries, metal artifacts, susceptibility artifacts, and blooming artifacts) were included in subsequent analyses. None of the participants were excluded for visually inspected imaging artifacts.

fMRI data preprocessing
Resting-state BOLD data were preprocessed using Statistical Parametric Mapping software (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 volumes 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 (i.e., translational or rotational motion parameters less than 2 mm or 2 • ). We also calculated frame-wise displacement (FD), which indexes the volume-to-volume changes in head position. In the normalization step, individual structural images were first co-registered with the mean functional image; then the transformed structural images were segmented and normalized to the Montreal Neurological Institute (MNI) space using a high-level nonlinear warping algorithm, that is, the diffeomorphic anatomical registration through the exponentiated Lie algebra (DARTEL) technique (Ashburner, 2007). Finally, each filtered functional volume was spatially normalized to MNI space using the deformation parameters estimated during the above step and resampled into a 3-mm cubic voxel. After spatial normalization, all data sets were smoothed with a Gaussian kernel of 6 × 6 × 6 mm 3 full-width at half maximum.

Statistical analysis
The statistical descriptive analyses of demographic, serum lipid levels, and behavioral data were conducted using the SPSS 23.0 software package (SPSS, Chicago, IL, USA). We adopted a multistage approach inter-or intra-network FCs were transformed into Fisher's Z scores and then compared between males and females (Nostro et al., 2017).
Second, in cases of significant interaction effects, we further examined the association between affected brain networks and working memory using partial correlations adjusted for age, FD, and educational level in males and females separately. Finally, to determine whether the relationship between serum lipids and working memory was mediated by identified FC in a sex-specific manner, mediation analysis ( Figure S2) was performed using PROCESS macro (Honey et al., 2009). A detailed description of the mediation models can be found in the Supporting Information section.

The sex and serum lipids interaction effect on inter-network FC and its relationship with working memory
A significant sex and TG interaction effect on inter-network FC was observed (p < .05, FDR corrected; Figure 1). Specifically, we found Next, we examined the effect of TGs on the inter-network FC significantly affected by the sex and TG interaction effect in males and females separately (Figure 1 and Table S1). Partial correlation analysis showed that TG was negatively correlated with inter-network connectivity between ECN and dSMN (partial correlation coefficient Furthermore, the relationship between TG-related inter-network connectivity and working memory in females was not significant.

The sex and serum lipids interaction effect on intra-network FC and its relationship with working memory
Voxel-wise intra-network FC analyses demonstrated that the connectivity within multiple functional networks was influenced by the interaction effect of sex and serum lipids (p < .05, cluster-level FDR corrected; Figure 2). Specifically, the intra-network FC in the right inferior parietal lobe (R-IPL, cluster size = 31 voxels; peak MNI coordinates: x/y/z = 36/−51/48; peak t = −4.24) of the DAN and right middle temporal gyrus (R-MTG, cluster size = 25 voxels; peak MNI coordinates: x/y/z = 48/−69/9; peak t = 4.02) of the lVN were influenced by the interaction effect of sex and TC (Figure 2A,B). The intranetwork FC in the bilateral medial superior frontal gyrus (B-SFGmed, cluster size = 53 voxels; peak MNI coordinates: x/y/z = 3/42/21; peak t = 4.41) of the aDMN was influenced by the interaction effect of sex and LDL-C ( Figure 2C).
Next, we examined the relationship between serum lipids (TC and LDL-C), and the intra-network FCs significantly affected by the sex and serum lipids interaction effect in males and females separately.
As shown in Figure 2 and Table S2,

Sensitivity analysis
To test the possible effect of education level on our results, we included education level as an additional nuisance covariate in the analyses of the sex and serum lipids interaction effect on inter-and intra-network FCs. As shown in Tables S3 and S4, our main results were preserved after additional adjustments for education level (p < .05). Furthermore, there are significant differences in serum TG levels between male and female subjects, which may influence the reliability of the sex and TG interaction effect. To ensure the robustness of our findings, we selected 70 male and 70 female participants from the total sample, ensuring that their TG levels were matched (t = 0.487, p = .627). We subsequently reevaluated the interaction effect of TG and sex on inter-network FC, and our results remained consistent (Table S5). Moreover, our analysis has confirmed that the current sample size is sufficient to yield statistically significant results, as depicted in Figure S3.

DISCUSSION
Our study is the first to investigate the interaction effects of sex and serum lipids on cognitive function using large-scale brain networks. Our analysis demonstrated that sex had an effect on the relationship between serum lipid levels and inter-network FC in the ECN-dSMN, ECN-mVN, ECN-AN, and ECN-DAN and intra-network FC in the R-IPL, R-MTG, and B-SFGmed. We found that intranetwork connectivity in the R-IPL significantly mediated the relationship between serum TC levels and the 3-back reaction time in females, but not in males. Thus, our findings suggest that sex may be an important factor that needs to be considered when examining the relationship between serum lipid levels, brain, and cognition.

F I G U R E 3 Scatter plots showing the correlation between intra-network functional connectivity, serum total cholesterol (TC) levels, and three-back reaction time in females (A, B) and males (D, E). (C, F) Mediation analyses between serum TC (X) and three-back reaction time (Y) with
intra-network functional connectivity in the right inferior parietal (R-IPL) of the DAN as the mediator (M). Path coefficients with p values ( * p < .05 and * * p < .01, respectively). DAN, dorsal attention network; IPL, inferior parietal lobe; R, right.
Extensive studies have revealed that serum lipids, as well as sex dimorphism, are involved in the onset and progression of different neuropsychiatric disorders, such as depression , first-episode schizophrenia (Gjerde et al., 2021), cerebral small vessel disease (Yin et al., 2018), and Parkinson's disease (Seyfried et al., 2018). However, the neurobiological bases of the sex-specific different contributions of lipids fractions on neuropsychiatric disorders are still unclear, probably due to the few studies available to date which have carried out sex-stratified analysis. Although few studies have examined the brain network mechanisms through which blood lipids and sex affect behavioral performance including cognition, some efforts to determine the relationship between serum lipids and brain FC have been made. For example, stronger connectivity in the DMN and lower connectivity in the SN have been observed in nondemented elderly with high serum TC levels . Poorly controlled cholesterol has been shown to impair FC in the hippocampus and middle frontal gyrus in T2DM patients (Xia et al., 2015). In addition, TGs have been shown to affect the inter-network FC between SN and vSMN, which in turn impacts cognition . However, these earlier studies did not examine sex-specific effects, which may impact the correlation between brain network connectivity and serum lipids.
In the current study, we assessed the interaction effects of sex and serum lipid levels on brain networks using the ICA method, which facilitates a more thorough characterization of the whole-brain functional connectome. We found that the functional exchange and integration between executive control and sensorimotor systems (i.e., the internetwork FC in the ECN-dSMN, ECN-mVN, ECN-AN, and ECN-DAN) were preferentially associated with serum TG levels, and that these relationships displayed sex dimorphism-that is, negative correlation in females, and no correlation in males. The ECN is the core network among all the affected inter-network functional connections and is primarily anchored in the prefrontal and lateral parietal regions and is thought to be associated with working memory and attention processes Song et al., 2013). Filippi et al. (2013) examined the effects of sex on resting-state functional networks and found that women displayed stronger connectivity in the frontotemporal regions and within attention and memory-related networks. In addition, Banks suggested that elevated TG levels may compromise the blood-brain barrier transport of insulin and other hormones (Banks, 2012), thereby exerting a pro-inflammatory effect, which may negatively impact cognitive performance. Knopp et al. (2006) reported that high TG levels exerted a worse cardiovascular outcome in women than men. Thus, it is possible that in cognitively impaired patients, hypertriglyceridemia exerts a more detrimental effect in women than in men. These findings are consistent with our study, which demonstrated that inter-network connectivity with the ECN at its core is more susceptible to serum TG levels in females, which may further influence cognitive performance.
In short, such findings may help explain our results, which found that the effect of TG on ECN-related inter-network FC was sex-dependent.
Our results, together with those of previous studies (Gjerde et al., 2021;Seyfried et al., 2018;Xu et al., 2021;Yin et al., 2018), suggest that alterations in inter-network FC may be a potential network mechanism of lipid-involved neuropsychiatric disorders that exhibit sex dimorphism.
Dense intra-network connections increase local clustering and thus facilitate information specialization within a specific functional network. Thus, studies examining intra-network temporal coherence may further our understanding of functional specialization (Liao et al., 2017). Previous voxel-wise analyses of spatial maps revealed no significant correlations between serum lipid levels and intra-network connectivity within any of the functional networks, when sex was considered to be a nuisance variable . However, the effect of sex on the relationship between serum lipid levels and variations in neuroimaging phenotypes should not be ignored. In the current study, we found that sex and serum lipid interaction effects affected the intra-network FC in the DAN, lVN, and aDMN, further suggesting that sex is a key factor that needs to be considered when examining the relationship between serum lipids and brain networks. Previous studies have also shown that the relationship between serum lipid levels and cognition exhibited sex specificity. For example, in a cross-sectional study of aging women, serum high-density lipoprotein was found to be associated with better verbal learning and memory performance (Bates et al., 2017). Another 16-year longitudinal study suggested that higher HDL-C and lower TG levels predicted better maintenance of cognitive functions (including verbal ability and perceptual speed) in women than men (Reynolds et al., 2010). Interestingly, in the present study, we found sex-specific differences in the association between lipid fractions and the intra-network FC. In particular, a positive association between intra-network FC in the R-IPL of DAN and working memory was found only in women. Similarly, intra-network FC in the DAN mediated the relationship between serum TG levels and working memory, but this was only observed in women. In contrast, an inverse association between lipid fractions and intra-network FC was found in men. Recent studies have reported that activation or impairment of the DAN is associated with short-term memory (Majerus et al., 2012) and memory coding (Mallas et al., 2021). Our study extends these findings by showing that the density of FC in the DAN is more susceptible to serum TC in females, and that changes in the degree of information specialization within DAN are related to working memory. Furthermore, the correlation between the intra-network FC in the R-MTG of the lVN with TC and the correlation between the intranetwork FC in the B-SFGmed of the aDMN with LDL-C were also found to have sex-specific differences. The lVN is mainly implicated in the processing of visual information, whereas the DMN is thought to support internally oriented cognitive processes, including recalling autobiographical and episodic memories, envisioning the future, and making social inferences (Raichle et al., 2001). Our results may provide new perspectives on the treatment of lVN-and DMN-related functional disorders, through sex-specific serum lipid interventions.
Furthermore, different strategies should be used to adjust serum lipid levels or lipid-related brain networks in the treatment of psychiatric disorders associated with sex differences such as autism, which is more prevalent in men (Lai et al., 2015), and depression, which is more prevalent in women (Ferrari et al., 2013).
Here, we found that both the inter-and intra-network FCs were influenced by the serum lipids and sex interaction effect and further affected cognition in a sex-specific manner. This may be because estrogen and progesterone fluctuations throughout the menstrual cycle are not only related to brain networks such as transient reorganization and topological properties of functional brain networks (Liparoti et al., 2021;Mueller et al., 2021), but also to serum lipids. For example, estrogen has been associated with reduced LDL-C levels and increased HDL-C and TG levels (Engelberg & Glass, 1955;Gambrell & Teran, 1991;Rezvani et al., 2014). Another possible explanation is that higher levels of testosterone, and its stable concentration across the life span in males may be responsible for sex-specific differences in the associations between serum lipids and large-scale brain networks (Liu et al., 2015).
There are several limitations in the present study. First, our crosssectional design does not allow for inference to be drawn regarding causality. Nevertheless, due to our relatively large sample size, our findings can provide a useful springboard for further research. Second, as our study sample was selected from a group of educated young adults, our findings may not be generalizable to older populations with mental illness. Third, diet has a diverse effect on serum lipids (Li & Shi, 2017). However, as all participants enrolled in the current study fasted overnight prior to the collection of blood samples, diet should not significantly impact our results. Fourth, the lack of data on serum sex hormones is also an important constraint as, for example, estrogen and progesterone fluctuations over the menstrual cycle are associated with the brain network and serum lipid levels. However, it may be difficult to fully understand the effects of sex hormones in a clinical setting due to the complex regulatory mechanisms that are involved. Finally, in the current study, cognitive function was measured using a three-back task test, which only assesses working memory. Future studies need to utilize a more comprehensive measure involving multiple cognitive domains.
To our knowledge, this is the first study that has specifically examined the interaction effects of sex and serum lipids on the relationship between large-scale brain networks and cognition in healthy adults.
Our findings suggested that interactions between sex and serum lipids may affect cognition through large-scale inter-and intra-network FCs.
Our data may be of clinical relevance because they indicate that sex should be taken into account when using serum lipid levels or related-brain networks to monitor treatment response during cognition decline. In addition, our results may facilitate the development of novel therapeutic strategies that are more tailored to the sex of the patient, thus increasing the likelihood of successful treatment.

AUTHOR CONTRIBUTIONS
Shujun Zhang and Yongqiang Yu conceptualized and designed the study. Shujun Zhang was responsible for conducting the analyses, preparing the first draft of the manuscript, and preparing the manuscript for submission. Yongqiang Yu was responsible for obtain-ing funding for the study, supervising the analyses, and editing drafts of the manuscript. All authors contributed to and approved the final manuscript.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy or ethical restrictions.