Effects of gender, digit ratio, and menstrual cycle on intrinsic brain functional connectivity: A whole‐brain, voxel‐wise exploratory study using simultaneous local and global functional connectivity mapping

Abstract Introduction Gender and sex hormones influence brain function, but their effects on functional network organization within the brain are not yet understood. Methods We investigated the influence of gender, prenatal sex hormones (estimated by the 2D:4D digit ratio), and the menstrual cycle on the intrinsic functional network organization of the brain (as measured by 3T resting‐state functional MRI (rs‐fMRI)) using right‐handed, age‐matched university students (100 males and 100 females). The mean (±SD) age was 20.9 ± 1.5 (range: 18–24) years and 20.8 ± 1.3 (range: 18–24) years for males and females, respectively. Using two parameters derived from the normalized alpha centrality analysis (one for local and another for global connectivity strength), we created mean functional connectivity strength maps. Results There was a significant difference between the male mean map and female mean map in the distributions of network properties in almost all cortical regions and the basal ganglia but not in the medial parietal, limbic, and temporal regions and the thalamus. A comparison between the mean map for the low 2D:4D digit ratio group (indicative of high exposure to testosterone during the prenatal period) and that for the high 2D:4D digit ratio group revealed a significant difference in the network properties of the medial parietal region for males and in the temporal region for females. The menstrual cycle affected network organization in the brain, which varied with the 2D:4D digit ratio. Most of these findings were reproduced with our other datasets created with different preprocessing steps. Conclusions The results suggest that differences in gender, prenatal sex hormone exposure, and the menstrual cycle are useful for understanding the normal brain and investigating the mechanisms underlying the variable prevalence and symptoms of neurological and psychiatric diseases.

The data created with a threshold of Z=3.28 also showed significant difference in the distribution of global hubs between males and females (p = 4.95x10 -5 ) ( Figure S6B).

Effects of preprocessing upon the network properties
This data set was used to investigate the effect of preprocessing upon the results, because it is, in theory, impossible to discriminate signals related solely to the brain activity from the functional image data (Bright and Murphy, 2015;Pujol, et al., 2014) and there is no standard preprocessing method for the functional images (Liu, 2016). In particular, global signals could convey brain activity information, and its regression away from the data could cause artificial noise (Liu, et al., 2017), though global signal regression has been shown to be effective in removing nuisance noise related to head motion, cardiac pulsation, and respiration (Power, et al., 2016).
When an adjacency matrix was generated with a threshold of 2.56 using the data from preprocessing 2 (CompCor without global signal head motion regression), the mean number of edges for males (1.76x10 6 ± 0.67 x10 6 ) was not significantly different from that for females (1.86x10 6 ± 0.65 x10 6 ) (p = 0.31, t-test). The distributions of the global hubs and global nodes in the 14 regions were significantly different between male and female groups (Chi-square test, p = 0.009 and 1.13x10 -5 ) ( Figure S7) as seen for the data from 'preprocessing 1' (Fig. 3). In other words, the percentages of global hubs in the three frontal regions, the cingulate, and the insula for males were significantly higher than those for females; observations which were validated by the Kaneoke, Y. 2 permutation test (p<0.05). There were no significant differences in the distributions of local nodes and local hubs in contrast to the data from 'preprocessing 1' which showed significant differences among these distributions.
The effects of digit ratios upon the distributions of node types were shown to be significant for both males ( Figure S8A) and females ( Figure S8B). The permutation test revealed that the percentage of global nodes for high digit ratio males in the sensorimotor region was significantly higher than that for low digit ratio males. For females, the percentage of global nodes in the lateral parietal and occipital regions were significantly different when compared high and low digit ratio subgroups.
For the female low digit ratio group, we found that menstrual phase cause significant effects upon the distributions of global hubs, global nodes, and local nodes.
as revealed by Chi-square tests. Several of these regional differences were further validated by the permutation test ( Figure S9A). Menstrual phase also affected the global node distribution for the high digit ratio group ( Figure S9B). The permutation test revealed that the percentage of global nodes in the medial parietal region for the luteal phase group was significantly higher than that for the follicular phase group (p = 0.034, permutation test).  Figure S2. Gender differences in the network properties for the participants with a handedness score of 100 Gender difference in node type distributions were investigated using data from participants whose handedness scores were 100 (38 males and 57 females). The percentage of each node types in each region are shown as in Fig. 3. The results are similar to those shown in Fig. 3. *p<0.05 and *p<0.01 by permutation test. Chi-square test results are shown by p values (corrected with Bonferroni method) in each graph.

Figure S3. Relationship between Gray matter volume and network properties.
The gray matter (GM) volume (x axis) for each participant and the mean nAC 0 (top) and nAC 1 (bottom) in each region (see Table S1) for each participant were plotted with regression lines. There was no significant relationship between them at all the regions (p>0.05, Pearson's method) for both males and females.
Kaneoke, Y. 8 Figure S4. Relationship between Gray matter volume and network properties.
The gray matter (GM) ratios to total intracranial volume (TIV) (x axis) for each participant and the mean nAC 0 (top) and nAC 1 (bottom) in each region (see Table S1) for each participant were plotted with regression lines. There was no significant relationship between them at all the regions (p>0.05, Pearson's method) for both males and females.
Kaneoke, Y. 9 Figure S5. Relationship between the number of edges and nAC 0 /nAC 1 The number of edges (x axis) for each participant and the mean nAC0 (top) and nAC1 (bottom) in each region (see Table S1) for each participant were plotted with regression lines. See Table 5 for Pearson's correlation coefficients and p values. Gender difference of the node type distributions were investigated using the data from preprocessing 1 and with a threshold of Z = 1.96 (top) and 3.28 (bottom) for an adjacency matrix. The percentage of each node types in each region are shown as in  Gender difference in the node type distributions were investigated using the data from preprocessing 2 with a threshold of Z = 2.58 for an adjacency matrix. The percentage of each node types in each region are shown as in Fig. 3. The results were similar to those for preprocessing 1 and for a threshold of Z = 2.56 (Fig. 5)   The effect of menstrual phase on the network property distributions are investigated using the data for low and high digit ratio groups with preprocessing 2 and a threshold of Z = 2.58 for an adjacency matrix. The percentage of each node types in each region are shown as in Fig. 3