Behavioral and physiological stress responses
The measured behavioral and physiological data indicated that the experimental paradigm successfully elicited a mild to moderate level of psychological stress in both male and female subjects. The main effect of experimental condition was significant for perceived stress [F(5, 26) = 17.47, P < 0.001], perceived anxiety [F(5, 26) = 19.55, P < 0.001] and heart rate [F(4, 27) = 41.76, P < 0.001], which were immediately elevated in response to the stress tasks, as well as for salivary cortisol [F(5, 26) = 3.22, P = 0.021], which showed a delayed response to the high stress task (). The main effect of gender was not significant for perceived stress/anxiety, heart rate or salivary cortisol measures. However, the interaction of experimental condition and gender was significant for perceived stress [F(5, 26) = 5.52, P = 0.001]. Post hoc analyses indicated that males reported a greater acute response in perceived stress from the low to high stress task [F(1, 30) = 4.39, P = 0.045] compared to females. This effect was not observed for perceived anxiety, although self ratings of stress and anxiety were correlated (R = 0.76, P < 0.001). Despite a higher level of task difficulty [F(1, 30) = 7.20, P = 0.012] and effort required [F(1, 30) = 4.93, P = 0.034] reported by females during the stress tasks (Supplementary Figure 1), men and women performed equally well for the serial subtraction task. There was no significant difference between the two sexes in the recorded number of errors made (male: mean ± s.e.m. = 5.7 ± 1.0, female: 6.2 ± 1.3, Z = 0.17, P = 0.87) and completed subtractions before committing an error (male: 19.2 ± 3.4, female: 15.3 ± 4.8, Z = 1.35, P = 0.18).
Neural pathways associated with perceived stress in men and women
The neural correlates of subjects’ own experience of stress were probed using voxel-wise linear regression analyses of the perfusion fMRI data with perceived stress. First, acute stress responses during the performance of stress tasks were identified by correlating changes in regional CBF and perceived stress from the low to high stress task (High–Low stress task
). Second, lasting stress effects after task completion were identified by correlating baseline CBF variations (Baseline 2–1
) with changes in perceived stress from the low to high stress task. We first examined the right and left prefrontal cortex based on main findings from our prior study (Wang et al
Performing the two regression analyses in each gender revealed that, in the male group, CBF in the right prefrontal cortex (RPFC) was elevated both during the performance of stress tasks and at baseline after task completion in subjects experiencing stress. However, no significant correlation between RPFC activation and perceived stress was observed in the female group either during tasks or at baseline (A and B). This gender difference in RPFC activation was especially evident during the performance of stress tasks, wherein the RPFC CBF showed the lowest correlation with subjective stress ratings in the female group (R
= 0.24, P
= 0.36, A). We further observed that, in the male group, CBF in the left orbitofrontal cortex (LOrF) inferior frontal cortex (IFC) was suppressed both during the performance of stress tasks and at baseline after task completion in subjects experiencing stress (A and B). For females, the association of CBF reduction in LOrF/IFC and perceived stress was only significant during the performance of stress tasks (A). These results suggest that the stress response in men is primarily characterized by RPFC activation accompanied by LOrF/IFC inhibition, a robust response that persists beyond the stress task period. In contrast, women only showed transient suppression of the LOrF/IFC during the performance of stress tasks.
Brain regions with significant findings in regression analyses of CBF data, and in comparison between male and female stress responses
Fig. 2 Axial sections of regression analysis results performed in the male and female group respectively, showing consistent RPFC activation in all the three analyses performed in males. These analyses use the CBF change during stress tasks (high stress – (more ...)
Fig. 3 Axial sections of regression analysis results performed in the male and female group, respectively, showing consistent deactivation of LOrF/IFC in all the three analyses performed in males. The analyses are the same as shown in . Scatter plots (more ...)
We then examined the limbic system along with closely interconnected brain regions including hippocampus, insula and cingulate cortex. During the performance of stress tasks, CBF increases in the left insula/putamen (LIn/Pu), right insula (RIn) and bilateral ventral striatum (LSt and RSt), including caudate and globus pallidus, were correlated with subjective stress ratings only in the female group. In contrast, the male group did not exhibit any stress related brain activation in the limbic regions during stress tasks (A). After completion of stress tasks, persistent activation in the ACC, posterior cingulate cortex (PCC) and RIn were associated with heightened stress level during tasks in the female group. Particularly, in dorsal ACC, the correlation coefficient of CBF with perceived stress was significantly higher in females than males (P = 0.02). In the male group, persistent CBF elevation was observed only in the RIn in stressed subjects (B). These results indicate that the female stress response is primarily associated with limbic activation of the ventral striatum, putamen and insula during stress tasks, and the ACC and PCC persisting beyond the task period.
Fig. 4 Axial sections of regression analysis results performed in the male and female group, respectively, showing limbic and cingulate activation only in females. The analyses are the same as shown in . Scatter plots of corresponding CBF changes in (more ...)
Since women experienced increased cognitive demand relative to men during stress tasks, there exists the concern that our observation may reflect gender differences in performing arithmetic tasks rather than stress reactivity. We therefore repeated the above regression analyses, while including subjective ratings of effort/difficulty as a covariate in conjunction with perceived stress. Behaviorally, subjective ratings of stress and effort/difficulty were not correlated (R = 0.07, P = 0.80). Including effort/difficulty as a covariate along with stress in regression analyses of CBF data did not affect the reported results on gender differences in brain activation associated with perceived stress (Supplementary Figure 2).
Conjunction of stress networks in men and women
Conjunction analyses were carried out to detect the overlap between the neural networks associated with perceived stress in the male and female group. At the significance level of P < 0.005, there was no overlapping brain activation between the two gender groups. At P < 0.01, the conjunction analysis only revealed one common brain region during the performance of stress tasks—CBF reduction in the LOrF/IFC, as well as one common region after completion of stress tasks—persistent CBF elevation in the RIn/Pu (Supplementary Figure 3A). When the significance level was relaxed to uncorrected P < 0.05, the conjunction analysis started to reveal more overlapping brain regions between the male and female group, including common activation of LSt, left insula/superior temporal cortex (LIn/STC) and right globus pallidus/thalamus (RGP/Th) as well as LOrF deactivation during the performance of stress tasks. The conjunction analysis also revealed common persistent activation of RPFC, RIn and ACC at baseline after task completion in both genders (Supplementary Figure 3B). Nevertheless, the acute RPFC response was still uniquely associated with male subjects even with P < 0.05.
Neural pathways associated with salivary cortisol in men and women
The earlier results were based on regression of brain responses with subjective stress experience, which may differ between men and women. For instance, it has been reported that females may have a lower threshold for perceived stress compared to males since puberty (Hampel and Petermann, 2006
). We therefore performed a third regression analysis to detect associations between baseline CBF variations (Baseline 2–1
) and AUC measures of salivary cortisol—a physiological index of overall stress elevation caused by undergoing the experimental stress paradigm. Again, in the male subjects, we found that baseline CBF increase in the RPFC and CBF reduction in the LOrF/IFC were correlated with AUC measures of salivary cortisol (C and C). In contrast, significant cortisol related CBF increases were observed in the dorsal ACC (dACC) and left thalamus (LTh) only in the female but not the male group (C and C). Females also showed cortisol related CBF reduction in the left IFC (LIFC), but at a much weaker significance level compared to the LOrF/IFC suppression observed in males (C). A subsequent conjunction analysis with a lowered threshold of uncorrected P
< 0.05 revealed only one common cortisol related brain activation in dACC in both males and females (no significant overlap at P
< 0.01). These additional analyses relying on a physiological parameter—salivary cortisol—are consistent with our findings based on behavioral assessments of stress.
Neural pathways associated with perceived anxiety in men and women
Although we demonstrated that differences in subjective feelings of effort/difficulty did not contribute to the observed gender-specific brain activation pattern under stress, there remains the concern that females may feel more threatened by the arithmetic task than males, thereby eliciting greater limbic activation. In terms of behavioral measures of anxiety, neither gender nor the interaction of gender and experimental condition showed a significant effect. Regression analyses of CBF data with perceived anxiety revealed primary limbic activation in both genders. During the performance of stress tasks, anxiety was associated with CBF elevation in the left amygdala/insula/putamen and RIn/Pu in both male and female groups. Perceived anxiety also elicited CBF increase in LSt and RSt only in female subjects, as well as CBF reduction in left IFC only in male subjects (Supplementary Figure 4). During the baseline conditions, men showed persistent LIn/Pu activation, whereas women showed persistent activation in RIn/Pu, dACC and LSt that were associated with changes in reported anxiety level during stress tasks. In these analyses, targeting neural pathways mediating subjective feelings of anxiety, prefrontal activity was largely missing in the male group. This observation suggests that the gender-specific central stress response in our study cannot be (solely) attributed to potential differences in negative emotions such as perceived anxiety.
Comparison of average stress responses between men and women
To address whether the average brain activation pattern under stress differs between men and women, we compared the mean acute (High–Low stress task) and persistent (Baseline 2–1) CBF responses to stress between the male and female group, using a regression analysis including gender as the independent variable (A) (i.e. unpaired t-test between male and female groups). During the stress tasks, men showed predominantly greater CBF augmentations than women in the right hemisphere including the RPFC and right parietal cortex/angular gyrus (RPC/AG), whereas women only showed greater activation in PCC compared to men. The greater acute RPFC activation in the male group was the most significant finding (peak Z = 3.96). This activation survived the small volume corrected threshold (P = 0.04) using the right frontal lobe as the search volume. When perceived stress was also included as a covariate along with gender in the regression analysis, this gender effect in the RPFC was still significant (Supplementary Figure 5). Based on estimation of effect size using analysis of variance (ANOVA), gender and perceived stress accounted for 51.5% (P < 0.001) and 49.8% (P = 0.015) of the total variance of RPFC CBF changes during stress tasks, respectively.
Fig. 5 Three-dimensional rendering of the results comparing the mean acute (High–Low stress task) and lasting (Baseline 2–1) CBF responses between the male and female groups. The greater right-sided male activation during tasks (A) and greater (more ...)
In contrast to the acute stress responses, during the post- vs pre-stress baseline conditions, women showed much greater CBF elevations than men primarily in the left hemisphere, including the LOrF, left insula (LIn), dorsal ACC and left parietal cortex/supramarginal gyrus (LPC/SMG) (B), whereas men only showed greater activation in the right thalamus compared to women. Taken together, the group comparison results and the regression analyses carried out independently in the male and female groups suggest that the RPFC activation provides a unique biomarker of the acute stress response in men. In contrast, females show greater persistent activation of the dorsal ACC and LIn, and less suppression of the LOrF after task completion compared to men.
Classification of stress responses in men and women
We further employed a SVM based linear classification approach to differentiate the female and male stress response. As shown in C, the CBF changes in the RPFC from the low to high stress task provided a relatively clean separation between the male and female groups, which yielded an accuracy of 93.8% (two errors in 32 subjects) for SVM classification based on just a single ROI of the RPFC. We then sequentially included corresponding CBF changes in stress-related brain regions demonstrating gender differences into the SVM classification, including the LOrF, dorsal ACC and LIn. We were able to achieve a perfect (100%) separation of the male and female group, when all 4 ROIs were included (D). The RPFC was the most important factor in the SVM classifier, with a weighting factor of 51.5%.