Introduction

Emotional processes are known to interact with executive control (Lindstrom & Bohlin, 2011). One important aspect of executive control is the ability to retain relevant information and update it according to task demands. This is referred to as working memory (WM), and it is prefrontal cortex-dependent (Miyake et al., 2000). There are shared and valence-specific neural correlates of emotion processing such that changes in brain activity in the left prefrontal cortex (PFC) and striatum are related to positive emotion processing, whereas changes in the right PFC and anterior cingulate cortex (ACC) are associated with negative emotion processing (Mak et al., 2009; Grimm et al., 2012; Grosse et al., 2019). However, processing of emotional stimuli in conjunction with a WM task has not been shown to have a consistent impact on brain activity, especially because different types of WM tasks exist. For example, both positive and negative emotional stimuli, compared with neutral stimuli, in a verbal WM task elicit more brain activity in the dorsolateral PFC in healthy men, and this is associated with longer reaction times (Grimm et al., 2012). Comparatively, in a nonverbal WM task that is limited to emotional faces, positive stimuli processing results in more brain activity in the medial PFC and only negative stimuli processing leads to more brain activity in the dorsolateral PFC compared with neutral stimuli in a clinical population (Becerril & Barch, 2010). Few, if any, studies to date have used a WM paradigm, such as an n-back task encompassing a wide array of emotional images. Such a task would be an important contribution to the emotional cognition literature and would provide an advantage over previous work, because individuals encounter a variety of emotional stimuli in their daily lives. Emotional stimuli also are more salient than neutral stimuli, suggesting that WM for emotional stimuli would be neurophysiologically different than WM for neutral images. While some attempts have been made, the majority of these tasks introduce images of emotional faces surrounding the WM stimulus images as distractors, rather than the focus of the task (Bertocci et al., 2012; Tavitian et al., 2014).

One factor shown to modulate the integration of emotion processing and memory is the fluctuation of female sex hormones (Sundstrom & Gingnell, 2014; Catenaccio et al., 2016). This fluctuation occurs throughout the menstrual cycle, which is characterized by changes in estrogens (mostly 17β-estradiol) and progestins (mostly progesterone) due to regulation of the hypothalamic-pituitary-gonadal (HPG) axis. The menstrual cycle can be generally divided into the follicular (low hormone), ovulatory (peak estradiol), and luteal (high progesterone/high estradiol) phases (Fauser & van Heusden, 1997). Menstrual cycle studies show that ovarian steroid hormone modulation results in structural and functional plasticity in the central nervous system. For example, brain regions that commonly exhibit follicular/luteal structural plasticity include the hippocampus, parahippocampal gyrus, fusiform gyrus, ACC, insula, middle frontal gyrus, thalamus, and cerebellum (Catenaccio et al., 2016). Moreover, several studies report changes in brain activity associated with emotion processing and memory across the menstrual cycle. Estradiol demonstrates predominantly anxiolytic properties, whereas progesterone displays mixed anxiolytic and anxiogenic properties due to its metabolites pregnenalone and allopregnenalone (Andrade et al., 2005; Söderpalm et al., 2004; Andréen et al., 2009). The most consistent finding so far appears in the luteal phase when progesterone levels are high, resulting in an increased amygdala response to negative emotional stimuli and an increased ACC response to positive emotional stimuli (Protopescu et al., 2005; Amin et al., 2006; Andreano and Cahill, 2010; Gingnell et al., 2012; Bayer et al., 2014). On the other hand, estradiol levels during the follicular and preovulatory phases result in better emotion recognition accuracy independent of emotional stimuli (Pearson & Lewis, 2005; Derntl et al., 2008a,b) and more brain activity in the ACC and dorsolateral PFC during response inhibition to positive stimuli (Amin et al., 2006). In contrast to emotion processing, the effects of female sex hormones on emotional memory are mixed. While a longitudinal study reports decreased recognition for negative items during the luteal phase (Bayer et al., 2014), cross-sectional studies have demonstrated no difference (Felmingham et al., 2012) or enhanced memory for emotional items during this phase (Ertman et al., 2011). No study to date has examined the influence of female sex hormones on emotional processing in the context of a WM task, which could have implications for daily life (e.g., relationships, mood, and cognitive functioning).

Additionally, the fluctuating female sex hormones can be suppressed through the use of exogenous hormones in oral contraceptives (OCs). These hormones mimic endogenous estradiol and progesterone, but they inhibit the HPG axis to prevent menstruation and pregnancy (Pletzer & Kerschbaum, 2014). OC use has been linked to altered task-dependent emotion processing and memory; however, the current literature is riddled with some inconsistent findings. For example, OC users display both increased and decreased regional brain activity compared with naturally cycling (NC) women depending on the type of negative emotional stimulus (Petersen & Cahill, 2015; Miedl et al., 2018). Similarly, OC users display enhanced memory for gist, but not for details of an emotional story, compared with NC women (Nielsen et al., 2011), but OC users are no different than NC women during a complex emotion recognition task (Shirazi et al., 2020). Recent findings from our laboratory demonstrate increased brain activity in OC users, compared with NC women, when WM is engaged for negatively, but not positively, arousing images in an emotional n-back picture task (i.e., 2-Back Negative > 1-Back Negative) (Sharma et al., 2020). However, OC users and NC women show no significant differences in brain activity during positive and negative emotion processing of the WM conditions in the same task (i.e., 2-Back Positive > 2-Back Neutral and 2-Back Negative > 2-Back Neutral), suggesting that OC users and NC women share similar neural responses during general emotion processing (Sharma et al., 2020). It remains unclear whether positive and negative emotion processing recruit different brain regions in the context of a visual WM picture task with respect to female sex hormones.

Therefore, the objective of the current study was to examine the role of progesterone and estradiol on positive and negative emotion processing in the context of a picture WM task that is specifically designed for women. We hypothesized that estradiol would show a positive relationship with PFC activation during both positive and negative emotion processing (Amin et al., 2006; Miedl et al., 2018). Furthermore, we hypothesized that progesterone would have a positive relationship with ACC activation during positive emotion processing and the amygdala during negative emotion processing.

Material and Methods

Participants

Women between the ages of 18 and 26 years were recruited through the University of Ottawa’s Integrated System of Participation in Research (ISPR) and in the community. A total of 74 women (mean age 19.6; standard deviation [SD] = 1.96), including both OC users and NC women, participated in the fMRI session. Twenty-six participants were OC users who were using a combined OC formulation at the time of the fMRI session. Forty-eight participants were naturally cycling and were tested during the early follicular phase (days 2 to 6 following the onset of menstruation), preovulatory phase (days 10 to 14), or the luteal phase (days 18 to 24) of the menstrual cycle, which was established via a forward day count. There were 16 NC women in each phase (early follicular, preovulatory, or luteal) at the time of the fMRI session. Because the objective of the current study was to examine the general role of ovarian steroid hormones, OC users and non-OC users (74 participants in total) were combined together as a single group to investigate the association between the brain activation and ovarian steroid hormones, and OC use (1/0) was controlled for as a covariate of no interest. The number of participants was selected based on previous literature (Amin et al., 2006; Andreano & Cahill, 2010). All participants were right-handed, had normal visual acuity, and were screened for histories of psychiatric or medical illnesses. Participants with contraindications to MRI (e.g., nonremovable/implanted metal, back problems, loss of consciousness, pregnancy, or claustrophobia) were excluded. The experimental protocol was approved by the University of Ottawa’s research ethics board and the Royal Ottawa Mental Health Centre’s ethics panel. Informed, written consent was obtained from each participant.

Questionnaires

Participants were asked to complete the Beck Depression Inventory (BDI) before the fMRI session. The BDI was administered to measure individual differences in affect. It contains 21 items with each response being scored on a scale value from 0 to 3 (Beck et al., 1996). The scores on each item of the BDI were summed, and the total score was used in the analyses.

Assessment of ovarian steroid hormones

One milliliter saliva samples were collected before the fMRI experimental procedure. These samples were used in high sensitivity progesterone and 17β-estradiol enzyme-linked immunosorbent assay (ELISA) kits according to the supplier’s instructions (Salimetrics, Kit Lot numbers 1901540 & 1810501, respectively). The sensitivity of the ELISA kits was the following: 19.08 pg/mL ± 4.77 (high control range) and 5.49 pg/mL ± 2.20 (low control range) for 17β-estradiol and 820.13 pg/mL ± 205.03 (high control range) and 47.32 pg/mL ± 18.93 (low control range) for progesterone. To obtain a measure of free estrogenic activity, an estradiol-to-progesterone (E/P) ratio was calculated, because estradiol actions often are counteracted by progesterone due to their opposite effects on various neurotransmitter systems (; Pletzer & Harris, 2019).

Memory task

N-back task

Participants were scanned as they engaged in an emotional picture n-back task. The block design task was developed by our laboratory and consisted of stimulus images from the International Affective Picture System (IAPS) database (Lang, 2005). The images were selected based on the 5-cluster solution proposed by Constantinescu et al. (2017). Using a model-based clustering technique, the aforementioned researchers were able to create discernable and discrete groups within the IAPS dataset, classifying the images into five statistically supported clusters based on emotional valence (i.e., neutral, positive, and negative). However, their clusters were created based on combined male and female valence scores from the IAPS dataset. Because our task was to be used on a female population, the combined male and female valence scores of each image in each respective cluster were compared with their female only valence scores. The difference between these two valence scores was taken, and the images were reordered from the smallest to largest difference. The first 60 images from clusters 2, 3, and 4 were selected to be incorporated in the current task, because these images had minimal differences between the combined and female only valence scores. Cluster 2 represented positive emotion (valence = 7.27; arousal = 4.69), whereas cluster 3 represented negative emotion (valence = 2.27; arousal = 5.87). Cluster 4 represented neutral emotion (valence = 5.05; arousal = 3.31) (Constantinescu et al., 2017). Following the stimulus selection, a sorting algorithm was applied to create a balanced block design. E-prime was used to create the final computerized task. In each block, a stimulus image was presented for 1,000 ms with an interval of 500 ms followed by the next stimulus image. In total, there were 16 stimulus images per block with 3 blocks per n-back condition. The 6 main conditions were based on emotional valence and task type (i.e., 1-Back Negative, 1-Back Positive, 1-Back Neutral, 2-Back Negative, 2-Back Positive, 2-Back Neutral). Each individual block consisted of 6 target images and 10 nontarget images (37.5% target images per condition). No responses were required for the nontargets. The order of presentation of each block was counterbalanced. Participants were instructed to “Press for 2-Back” (working memory test condition) when they recognized an image that they saw 2-steps in the sequence. Brain activity was calculated as the average activity of each condition, including both target and nontarget images. Participants viewed the task on a mirror attached to the head-coil reflecting a projection screen, and all images were presented on a black background. Lights were off during task completion (Fig. 1).

Fig. 1
figure 1

Experimental task

Full size image

Performance

During the memory task, average reaction time (RT) was calculated for accurate responses occurring within 900 ms of stimulus presentation. Performance data also were screened for omission errors. A one-way repeated measures ANOVA was used to assess whether there were statistically significant differences in RTs between the 2-Back Positive, 2-Back Negative, and 2-Back Neutral blocks. Post-hoc testing used the Bonferroni correction factor.

fMRI data acquisition

Whole brain echo planar fMRI data were acquired using a Siemens Biograph 3 Tesla Magnetom MR scanner (Siemens, Erlangen, Germany) equipped with a 32-channel head coil at the Royal Ottawa Mental Health Centre’s Brain Imaging Centre. A gradient echo pulse sequence was used (TR/TE 3000/34 ms, FA 90°, FOV 200 × 200 mm2, voxel size 1.6 mm × 1.6 mm × 3 mm, 48 axial slices, slice thickness 3 mm, band-width 2894 Hz). The duration of the functional scan was 11 min and 12 s with a total of 221 functional images.

fMRI data analysis

Functional imaging data were analyzed using Matlab (R2013b) and Statistical Parametric Mapping (SPM12) software. The functional scans were normalized to the standard SPM Montreal Neurological Institute template and spatially smoothed with a 6-mm FWHM Gaussian kernel following SPM12 standard preprocessing procedures (i.e., realignment, normalization, and smoothing). The final images were corrected for field inhomogeneity. Two first-level contrast images were constructed at the individual participant level: (i) 2-Back Positive > 2-Back Neutral; (ii) 2-Back Negative > 2-Back Neutral. The contrasts represent positive and negative emotionality in WM. Contrasts constructed at the single participant level were then entered into second-level multiple regression analyses. Whole-brain multiple regression analyses were conducted to examine the relationship between endogenous progesterone and estradiol levels and brain activity in response to the two contrasts. Individual variability in factors that could affect brain activity were controlled for in the analyses, which included the participants’ total score on the BDI and RTs for the 2-Back Positive or 2-Back Negative blocks. The analyses also accounted for OC use to examine the global effects of endogenous progesterone and estradiol concentration. All reported results were derived from an uncorrected threshold of puncorr < 0.005 at the whole brain level and deemed significant at a threshold of p < 0.05 after family wise error (FWE) correction for multiple comparisons at the cluster level. All reported clusters consisted of ten or more contiguous voxels. To provide detailed statistical information and the multiple regression power, the regions-of-interests (ROI) analysis were employed for the standard multiple regression analysis (please see details in the supplementary information).

Results

Ovarian steroid hormones and BDI

OC users and NC women were hormonally similar since no significant differences were observed in endogenous progesterone and estradiol concentration between the two subgroups (p = 0.100 and p = 0.702, respectively). The mean progesterone concentration was 59.42 (±58.42) pg/mL, and the mean estradiol concentration was 0.616 (±0.326) pg/mL in all participants (see Table 1 for a breakdown of hormone concentrations per subgroup). Similarly, there were no significant group differences in the total BDI score between OC users and NC women. The mean total BDI score was 7 (±6) in all participants. To add, there were no statistically significant correlations between the hormone levels and total BDI score (p > 0.05).

Table 1 Hormone concentrations
Full size table

N-back task performance

The mean RT for the 2-Back Positive block was 607.5 (±75.5) ms with a mean of 16 (±2) correct responses and 2 omission errors in all women. During the 2-Back Negative block, the mean RT was 606.2 (±69.7) ms with a mean of 15 (±2) correct answers and 2 omission errors. Finally, the mean RT for the 2-Back Neutral block was 629.8 (±74.0) ms with a mean of 15 (±2) correct responses and 2 omission errors. A main effect of emotion was identified in the repeated measures ANOVA (p < 0.001, np2 = 0.130, power = 0.988). Post-hoc testing revealed that all participants displayed a significantly slower RT for the 2-Back Neutral block compared with the 2-Back Positive and 2-Back Negative blocks (MD = 23.4, SEM = 5.4, p < 0.001; MD = 23.7, SEM = 6.2, p = 0.001, respectively). There were no statistically significant correlations between the hormone levels, RTs during the 2-Back Positive and 2-Back Negative blocks, and correct answers during the aforementioned blocks (p > 0.05).

Positive emotionality (2-Back Positive > 2-Back Neutral)

Results from the regression of the 74 participants showed bilateral brain activity in the frontal regions of the brain and insula as well as in the right lingual gyrus, right calcarine, left caudate, left putamen, and left inferior occipital gyrus during the positive emotionality contrast (Fig. 2A). Progesterone was positively associated with brain activity in the right insula, right postcentral gyrus, right mid-cingulum, and right supplementary motor area (SMA) (Fig. 2B). Conversely, estradiol was positively related to brain activity in the right superior medial frontal gyrus (SFG), right precentral gyrus, and right superior parietal lobule (Fig. 2C). While no significant relationship was observed between the E/P ratio and regional brain activity at the whole brain uncorrected threshold of puncorr < 0.005, estradiol was positively associated with brain activity in the bilateral superior parietal lobule and the left precuneus. This association was identified from a regression conducted at an uncorrected threshold of puncorr < 0.05 at the whole brain level and deemed significant at a threshold of p < 0.05 after FWE correction for multiple comparisons at the cluster level (Table 2). No significant negative associations between hormone concentration and brain activity were identified for progesterone, estradiol, or the E/P ratio.

Fig. 2
figure 2

(A) Activation of reward-related brain areas, including the caudate [-10 10 8] and putamen [-16 4 16], from the regression analysis of all participants during positive emotion processing (2-Back Positive > 2-Back Neutral). (B) Progesterone positively correlated with brain activity in the SMA [12 -4 50]. (C) Estradiol positively correlated with brain activity in the SFG [4 50 28]. All regions are represented on coronal, sagittal, and transverse sections (p < 0.005 uncorrected and p < 0.05 after FWE correction, k ≥ 20, T value scale presented on the right)

Full size image
Table 2 Brain activation during 2-Back Positive > 2-Back Neutral
Full size table

Negative emotionality (2-Back Negative > 2-Back Neutral)

Results from the regression of the 74 participants showed bilateral brain activity in the SFG, SMA, insula, and hippocampus, as well as in the right inferior and medial occipital gyri, right inferior temporal gyrus, right cingulum, left mid-temporal gyrus, and left amygdala during the negative emotionality contrast (Fig. 3A). Progesterone was positively related to brain activity in the right rolandic operculum, right putamen, right insula, left pallidum, and left thalamus (Fig. 3B). While no significant positive relationship was observed between estradiol or the E/P ratio and regional brain activity at the whole brain uncorrected threshold of puncorr < 0.005, estradiol was positively related to brain activity in the right postcentral gyrus, left ACC, and left SFG. This regression was conducted at an uncorrected threshold of puncorr < 0.05 at the whole brain level and deemed significant at a threshold of p < 0.05 after FWE correction for multiple comparisons at the cluster level (Fig. 3C). In contrast, the E/P ratio was negatively associated with brain activity in the right parahippocampal gyrus, right inferior temporal gyrus, right lingual gyrus, left calcarine, and left precuneus at an uncorrected threshold of puncorr < 0.05 at the whole brain level and deemed significant at a threshold of p < 0.05 after FWE correction for multiple comparisons at the cluster level (Table 3). No significant negative associations between hormone concentration and brain activity were identified for progesterone and estradiol, and no significant positive association was identified between the E/P ratio and brain activity.

Fig. 3
figure 3

(A) Activation of corticolimbic brain areas, including the hippocampus [-20 -30 -4 and 28 -8 -18] and amygdala [-26 -34 0], from the regression analysis of all participants during negative emotion processing (2-Back Negative > 2-Back Neutral). (B) Progesterone positively correlated with brain activity in the thalamus [−16 −16 10]. (C) Estradiol positively correlated with brain activity in the ACC [−4 28 18]. All regions are represented on coronal, sagittal, and transverse sections (p < 0.005 or p < 0.05 uncorrected and p < 0.05 after FWE correction, k ≥ 20, T value scale presented on the right)

Full size image
Table 3 Brain activation during 2-Back Negative > 2-Back Neutral
Full size table

Discussion

In the current study, we used fMRI to investigate the role of ovarian steroid hormones on brain activation during positive and negative emotion processing in a WM task. This task consisted of images from the IAPS database that were grouped by emotional valence (i.e., neutral, positive, and negative) and were specific to female emotion processing. We identified the neural correlates of positive and negative emotion processing during WM and the regions specific to ovarian steroid hormone regulation. Participants showed brain activity in reward-related areas during positive emotion processing while activation in corticolimbic areas was more prominent during negative emotion processing. Participants also reacted quicker during positive and negative emotion processing than to neutral stimuli. Similarly, during processing for both types of emotion, progesterone appeared to modulate more posterior regions of the brain, whereas estradiol was more related to activation in anterior brain regions.

The RT results from this study are consistent with the emotional enhancement hypothesis, which predicts that emotional stimuli attract additional neural resources relative to neutral stimuli, thereby making it easier to encode and store information in WM capacity (Droit-Volet, 2016). While studies at a behavioural level depict mixed findings for this hypothesis (Crowell & Schmeichel, 2016; Garrison & Schmeichel, 2019), there is evidence for it at a neural level. More specifically, processing affective versus neutral material during WM tasks frequently recruits the PFC, amygdala, and temporo-occipital cortex in healthy participants (Schweizer et al., 2019). Brain activity predominantly in the right PFC has been associated with improvements in WM performance in women (Joseph et al., 2012).

During the current n-back task, participants showed bilateral frontal and insular activity during both positive and negative emotion processing, including additional regions that were either left or right lateralized. The two current leading hypotheses regarding emotion processing are the “right hemisphere hypothesis” (RHH) (Levy et al., 1983; Prete et al., 2015) and the “valence hypothesis” (VH) (Davidson et al., 1987; Prete et al., 2015). According to the RHH, the right hemisphere is superior to the left hemisphere in the processing of both positive and negative emotions. However, according to the VH, the right hemisphere specializes in negative emotion processing and the left hemisphere specializes in positive emotion processing. The current literature suggests that the two hypotheses are not mutually exclusive, resulting in a “modified valence hypothesis” (MVH). In this model, hemispheric superiority depends on prefrontal specialization (in which the left PFC specializes in positive emotion processing and the right PFC specializes in negative emotion processing), with posterior areas showing right hemispheric superiority during both positive and negative emotion processing (Killgore & Yurgelun-Todd, 2007). Our results are somewhat consistent with the MVH, which could be attributed to the design of our n-back task. For example, participants predominantly activated right-lateralized posterior regions, such as the right occipital and temporal gyri during negative emotion processing. While we observed left-lateralized activity in subregions of the frontal cortex during positive emotion processing, activation of right-lateralized frontal subregions may reflect the nonverbal nature of the n-back task, which has been previously identified (Smith and Jonides, 1999; Rottschy et al., 2012). Furthermore, the visual stimuli in the task were not limited to specific subgroups of positive or negative emotions (e.g., sexually arousing, fear, disgust, etc.).

Contrary to the MVH, participants showed brain activity in left dorsal and ventral striatal regions, such as the putamen and caudate. These regions have previously been associated with emotional response inhibition (Amin et al., 2006), reward processing, and positive affect, including the processing of positive verbal stimuli (Hamann & Mao, 2002; Li et al., 2018). However, it was unique to identify these brain regions during emotion processing in the context of WM. Additionally, amygdala activation has been identified during positive emotion processing (Hamann & Mao, 2002; Li et al., 2018). While this activation was not seen in the current study, we observed left-lateralized amygdala activity during negative emotion processing. This is consistent with the lateralization effect identified in women in other studies (Stevens & Hamann, 2012), which could relate to their preference in local/decomposed cognitive strategies (Pletzer, 2014). The amygdala, hippocampus, insula, and ACC form a hypothesized corticolimbic emotion processing network (Davidson et al., 2000) with the amygdala and insula activated by bottom-up emotional processes, and the ACC involved in top-down regulation (Bush et al., 2000; Shin & Liberzon, 2010). Previous publications report amygdala reactivity in response to emotional words and faces throughout the menstrual cycle (Protopopescu et al., 2008; Gingnell et al., 2012). Comparatively, women using OCs show greater or blunted amygdala activity relative to NC women (Marečkovà et al., 2014; Petersen & Cahill, 2015). Our study identifies a global, left-lateralized amygdala response irrespective of menstrual cycle phase and OC use, which has been thought to be mediated by endogenous progesterone (Andreano & Cahill, 2010).

While we did not observe any relationship between progesterone levels and amygdala activity during negative emotion processing, progesterone levels correlated with brain activity in the thalamus and insula. The thalamus is densely localized with the classic nuclear progesterone receptor (Brinton et al. 2008). This region also projects to the orbitofrontal cortex (OFC) and ACC during emotion processing similar to the amygdala and parahippocampal gyrus (Catani et al. 2013; Catenaccio et al., 2016). The rostral ACC within the ventral medial PFC governs amygdala activity when processing threat-related stimuli (Rauch et al., 2006). Thus, it is possible that the relationship between progesterone and amygdala activity was not detected during negative emotion processing due to possible top-down amygdala inhibition (Rauch et al., 2006). However, amygdala activity may be mediated indirectly via progesterone’s bottom-up effects on connected regions, such as the thalamus and insula. The thalamus serves as an integrative center that relays information from sensory organs or subcortical structures to the cortex. It also is activated by tasks that require multiple cognitive functions (Hwang et al., 2017). Similarly, the insula is activated during emotional tasks with cognitive demands (Phan et al., 2002). Both regions show a strong functional connectivity to the central nucleus of the amygdala at rest (Gorka et al., 2018). Consistent with the bottom-up modulation of integrative regions by progesterone during negative emotion processing, we identified a positive relationship between progesterone levels and sensory areas during the positive emotion processing contrast. Progesterone levels were associated with brain activity in the postcentral gyrus and SMA, which are regions known to be sensitive to sensory/premotor processing and acute progesterone administration (van Wingen et al., 2008; Luo et al., 2015). Thus, during emotion processing, progesterone levels may predominantly activate subcortical brain activation and a heightened attention to relevant emotional cues. This is in contrast to what was observed for estradiol.

Compared with progesterone, estradiol levels were positively related to brain activity in the right SFG during positive emotion processing and in the left SFG and ACC during negative emotion processing, which is inconsistent with the VH. While the association between estradiol and brain activity during negative emotion processing was identified at a liberal whole brain threshold, it was important to examine, because previous literature has not yet shown consistent findings in the context of emotion processing during WM. However, in other types of WM tasks, such as verbal processing, estradiol has been associated with more brain activity in the left PFC (Berent-Spillson et al., 2015). Estrogen receptors are widely expressed in several brain regions, including the PFC, amygdala, and hippocampus (Osterlund et al., 1999). Previous research also routinely shows increased activation during cognitive processing in frontal and cingulate cortical regions of the WM circuitry with increasing estradiol levels (Dumas et al., 2010; Joffe et al., 2006; Shaywitz et al., 1999). Our findings indicate that estradiol modulation has a lateralization effect (right vs. left) during emotion processing. Similar to our findings, Miedl et al. (2018) also reported a positive correlation between estradiol and ACC activity during traumatic versus neutral film viewing in NC women. Other researchers propose that estradiol may influence top-down regulation of emotion processing (Zeidan et al., 2011), which is reflected in the current study and suggests that estradiol acts on regions with connections to both cognitive and limbic influences. For example, women in the early and late follicular phase show more brain activity in the SFG during a figure comparison task compared with women in the mid-luteal phase of the menstrual cycle (Weis et al., 2008). Women with high estradiol levels also show more brain activity in the ACC along with other brain regions during fear conditioning and extinction learning relative to men (Hwang et al., 2015). Similarly, in the context of menopausal hormone replacement therapy, estradiol treatment increases brain activity during affective processing, verbal WM, and figural memory tasks (Comasco et al., 2014). Thus, it appears that estradiol predominantly activates top-down brain activation during emotion processing in the context of a WM task. The underlying mechanisms of estradiol’s boost on brain function and cognitive functioning include its upregulation of the serotonergic, dopaminergic, and cholinergic neurotransmitter systems (Comasco et al., 2014; Barth et al., 2015). While it is unclear whether estradiol levels correlate with these neurotransmitters to regulate top-down brain activation in the current study’s participants, our findings add newer evidence to the existing literature through our novel n-back task.

While our results from the E/P ratio do not necessarily reflect this top-down regulation, activation of the precuneus shows opposing associations with the ratio during positive and negative emotion processing. The precuneus responds to a wide variety of cognitive processes, including visual processing and recollection and memory (Kraft et al., 2015). Taken together, our results add newer evidence to existing studies on the modulatory roles of ovarian steroid hormones on emotion processing during WM.

However, there are some limitations in the study. First, all scans were based on a single visit. While this was necessary to avoid ceiling effects on task performance, a longitudinal design would provide additional evidence that brain activation during emotion processing fluctuates as a function of hormone concentration. Second, the relationship between estradiol or the E/P ratio and brain activity during the emotion processing contrasts were based on lenient whole brain thresholds compared with the more stringent thresholds used for progesterone. However, in light of the large number of covariates included in the analysis, this thresholding could be considered justified. Additionally, the brain regions shown to have a relationship with the hormone levels are not all consistent with the regions recruited during the affective contrasts; thus, our findings cannot establish causation of hormonal regulation on the brain and other mechanisms could be underlying this inconsistency. Moreover, using the forward day count to establish menstrual cycle position and not accounting for the timing of OC intake in relation to the fMRI session results in less reliable hormone concentrations compared to using other methods for hormone analysis and hormone measures (e.g., liquid chromatography/mass spectrometry, blood samples) (Blue et al., 2018; Hampson, 2020). Finally, the current study’s sample size was small, which should be improved in follow-up studies. Future studies could recruit women at one specific phase of the menstrual cycle and during a specific testing period to allow for more control over hormone levels, such as circadian rhythms in hormones (Herrera et al., 2019). Furthermore, the incorporation of radioligand positron emission tomography could be included to provide essential insight into how hormonal states may directly influence performance on an n-back task through the regulation of various neurotransmitter systems (Lewis et al., 2019)

Conclusions

Despite these limitations, this is an under-researched area that lacks consistent findings. The current study shows that female sex hormones modulate brain activity during positive and negative emotion processing in the context of WM. Because our analyses controlled for OC use, our findings suggest that there are global neural responses across women, in general, during emotion processing tasks. Progesterone appears to predominantly modulate bottom-up regulation by activating posterior, sensory-processing regions during both positive and negative emotion processing. On the other hand, estradiol appears to modulate top-down regulation by activating frontal regions during emotion processing. In summary, a novel n-back picture task was designed that provides insights into the neurophysiology of emotion processing during WM in a large sample of young women and that can be used in women’s cognition/emotion research. Our findings also underline that women's hormonal status should be considered in future cognition research on healthy adults as well as when men and women are grouped together in clinical settings.