To the best of our knowledge, this is the first study investigating sex-related RSFC differences in individual neural connections and the whole-brain network in awake rats. Our data indicate that individual connections exhibit significant sex differences, with males showing higher cortico-striatal connectivity, and females displaying stronger hypothalamus-related connectivity. To further examine the relative role of specific regions in RSFC sex differences, RSFC patterns of individual brain regions were used to predict sex with a SVM classification method. The striatum, hypothalamus and retrosplenial cortex exhibited the highest predicting power, again demonstrating regional heterogeneity in RSFC sex differences. At the global network level, females showed stronger within-cortical and within-subcortical segregation, whereas males displayed more prominent cortico-subcortical interactions. These data provide a comprehensive framework of sex differences in brain-wide connectivity, and offer a reference for studies aiming to reveal sex-related RSFC differences in animal models of brain disorders.
A prominent region that displays stronger connectivity in females is the hypothalamus. This result well agrees with the finding of sex differences in brain structure (Spring et al., 2007), suggesting that the hypothalamus is different in shape and size between male and female mice. Indeed, the hypothalamus is arguably the most well-documented region exhibiting sex-dimorphism, known to have abundant steroid receptors, and govern mating and stress response (Squire et al., 2012; Yang et al., 2013). In our study, females show stronger connectivity between the hypothalamus and pituitary, consistent with the literature report of sex differences in the hypothalamic-pituitary-adrenal (HPA) axis (Heck & Handa, 2019). As the central system that regulates the stress response, the HPA axis shows sex differences in the neuropeptide regulation (Heck & Handa, 2019). For example, both basal stress hormone and HPA stimulation-evoked stress hormone release are higher in females (Heck & Handa, 2019; Kokras et al., 2019). In addition to the hypothalamus, female rats show stronger connectivity between two olfactory regions—the piriform cortex and olfactory tubercle, and this connection is tightly associated with the pheromone-related psychosexual function (Cherry & Baum, 2020).
In males, a characteristic region showing sex differences in RSFC is the CPu, which is also in line with the previous finding of sex differences in the shape of CPu in mice (Spring et al., 2007). In addition, our data reveal stronger RSFC between the CPu and polymodal association cortex, hypothalamus, and the sensory-motor cortex in male rats, consistent with the report of sex differences in the striatal-cortical and striatal-subcortical circuits (Lei et al., 2016). In a similar vein, the activity of the putamen is positively correlated with the concentration of steroids hormone (Dreher et al., 2007), and males tend to show increased putamen activation than females in reward-related and risk-seeking behaviors (Lighthall et al., 2012). Furthermore, in our data male rats display stronger connectivity between the lateral hypothalamus and amygdala. This circuit is known to be involved in stress response, especially rodents' avoidance behavior in the context of danger (Dopfel et al., 2019; Weera et al., 2021; Liang et al., 2014). Stronger connectivity in this circuit in males may help elucidate the mechanism underlying differential reactions to potential dangers between males and females.
Based on the finding that multiple brain regions exhibit sex differences in their RSFC patterns, we then tested the possibility to predict sex based on the connectivity patterns of individual regions. Our analysis is different from previous studies on RSFC-based sex classification in humans, which are mostly based on whole-brain connectivity patterns (Stephen M Smith et al., 2013; Zhang et al., 2018). The whole-brain network based method has two drawbacks: First, the high dimensionality of whole-brain connectivity matrices may make the method suffer from the problem of overfitting. Second, results from whole-brain network based classification are more difficult to interpret, as the contributions of specific regions are less straightforward. A recent study used SVM to classify sex in humans based on a single region's connectome and obtained a stable sex prediction rate both within samples and between samples, with the mean prediction rate of 68.7 (Weis et al., 2020). This result is consistent with our study using a similar method (mean prediction rate ± std = 70.2% ± 4.9%). These data together demonstrate that, even with a relatively small sample size, most brain regions can achieve a significantly higher classification rate than a random guess, indicating sex-dependent effects on these regions. Interestingly, separate regions exhibit pronounced differences in the prediction accuracy, indicating spatially heterogeneous sex differences in the brain. For instance, the olfactory-related and cortical-striatum systems showed higher predication power than other regions, agreeing with the results from human studies (Cherry & Baum, 2020; Lei et al., 2016). It has to be pointed out that the ultimate goal of testing the predicting power of individual regions is not necessarily to predict sex, but to examine whether separate brain regions contain more sex-specific information that can be reflected by their functional connectivity patterns. RSFC-based sex prediction in rats may be trivial by itself, but the difference in the prediction power among brain regions may reflect the underlying sex specific neurobiology, which can be significant.
Sex hormone changes during the estrus phase may also affect RSFC. It was reported that fluctuations of estradiol and progesterone can induce behavior differences in rodents, such as the stress level in a restraining condition (Shansky et al., 2006), and these differences can be reflected in RSFC. In the present study, when comparing the RSFC between the proestrus/estrus and non-estrus groups, we did not observe a profound sex hormone effect in RSFC, consistent with meta-analysis studies demonstrating that data from female rats are not more variable than male rats (Becker et al., 2016; Prendergast et al., 2014). Another possibility is that the effects of short-lived hormonal fluctuations along the menstrual cycle cannot be reflected by the RSFC measure (Van Den Heuvel & Sporns, 2011). However, given the limited sample size in individual estrus phase subgroups, this result should be interpreted with caution.
The significance of the present study can be appreciated from two aspects. First, it provides important insight into understanding the sex-related neurobiology. Compared to human studies, the environmental and genetic background in lab animals are relatively uniform, making it more likely to identify sex-related differences with a relatively small sample size. Furthermore, sex-related differences in humans are not only determined by biological factors but can be compounded by environmental and sociocultural factors (Becker et al., 2017). Second, the knowledge gained in studying sex differences in rodents may further be used to inform human studies. A number of studies have shown conserved sex differences between rodents and humans. For instance, studies suggest addiction-like behavior in both humans and rodents shares similar sex-related differences (Becker et al., 2017; Carroll & Lynch, 2016). Thus, understanding sex differences in preclinical models of brain disorders may lend critical information to future human research, and can also reduce sex-related bias at the preclinical stage.
There are several potential limitations in present study. First, 14 females and 21 males were used, making the sample size slightly unbalanced. However, we believe this unbalanced sample size issue can be largely mitigated by the linear mixed model applied. Second, there is lack of behavioral tests in animals, making it less likely to unambiguously interpret the RSFC sex differences.
In conclusion, in the present study we discovered the sex-related RSFC differences at both the regional and systems-organization level. This study lays the foundation for future studies aiming to reveal sex-related RSFC differences in different animal models of brain disorders and/or studies of age-related sex differences in RSFC (e.g. during development or aging) (Ma et al., 2018a; Ma et al., 2021).