Neural emotion regulation during pregnancy: An fMRI study investigating a transdiagnostic mental health factor in healthy first-time pregnant women

Pregnancy is a physiological and psychosocial transition phase presenting a window of vulnerability for mental health. Anxiety and depression rates increase up to 15% during pregnancy and after giving birth, 10–20% of women suffer from postpartum depression (PPD; Haga et al., 2012; Roos et al., 2011; Sacher et al., 2020). Sex hormone fluctuations have been associated with the onset of mental disorder symptoms and are suggested to play a role in the pathophysiology of pregnancy-related mood disorders (Pinares-Garcia et al., 2018; Rehbein et al., 2020; Sacher et al., 2020; Soares & Zitek, 2008). During pregnancy, there is a drastic surge in sex hormones including estradiol (E2), exceeding any levels females will experience in their entire non-pregnant life (Lauritzen & Klopper, 1983; Rehbein et al., 2020; Soares & Zitek, 2008). In the postpartum period, E2 levels drop sharply, returning to beyond pre-pregnancy levels in only about 5 days (Hendrick et al., 1998; Rehbein et al., 2020; Sacher et al., 2020). By passing the blood–brain barrier due to its lipophilic properties along with widespread expression of E2 receptors, E2 can influence the brain (Soares & Zitek, 2008; Sundström Poromaa & Gingnell, 2014). Considering the incomparable rise and fall of E2 levels during and following pregnancy, neuroimaging studies have confirmed neuronal plasticity and neural network reorganization (for reviews see Barba-Müller et al., 2019; Luders et al., 2022; Sacher et al., 2020). Affected areas are associated with social cognition and emotion processing and overlap with the Theory-of-Mind- and the Default Mode Network (DMN; Chechko et al., 2022; Hoekzema et al., 2017, 2022; Nehls et al., 2024; Rehbein et al., 2022). Recent data on pre- versus post-pregnancy resting-state functional magnetic resonance imaging (fMRI) demonstrated an increase in temporal coherence in the DMN (Hoekzema et al., 2017) which promotes the assumption that neural structural and connectivity changes during pregnancy might serve a behavioral adaptation purpose in the transition to motherhood (Chechko et al., 2022; Hoekzema et al., 2017, 2022; Nehls et al., 2024; Rehbein et al., 2022). Even though only few studies have examined emotional functions during pregnancy, changes are evident (Amin et al., 2006; Pearson et al., 2009; Raz, 2014; Roos et al., 2012). In terms of self-reports, pregnant females indicate higher mood instability and emotional sensitivity, as well as reduced emotion regulation, which were related to sex hormone levels (Li et al., 2020; Pearson et al., 2009; Roos et al., 2011). Further, studies have shown that pregnant females have improved accuracy for encoding threatening facial expressions and an enhanced sensitivity to negative stimuli (Pearson et al., 2009; Roos et al., 2011, 2012). This may be explained by the effect of enhanced hormones, mainly E2, on fronto-amygdala circuits (Roos et al., 2011). In the postpartum phase, a positive correlation of amygdala response to emotional stimuli and PPD symptoms was demonstrated, suggesting enhanced arousal for salient stimuli in mothers with PPD (Wonch et al., 2016). However, research on behavioral and neural emotional functions during pregnancy is still limited, even more so regarding emotion regulation (Rehbein et al., 2020; Sacher et al., 2020).

The ability to regulate emotions is a requirement for intact social interaction and fundamental for our well-being and health (Goldin et al., 2008; McRae et al., 2012; Rammensee et al., 2023). As impairments in emotion regulation are present in many disorders and contribute to the development and/or maintenance of psychopathology, emotion regulation is proposed as a transdiagnostic factor for mental health (Goldin et al., 2008; Harvey et al., 2004; Radke et al., 2018; Rammensee et al., 2023). The most frequently studied strategy to regulate emotions is cognitive reappraisal, which targets the process of cognitive re-evaluation (Goldin et al., 2008; McRae et al., 2012). General and distinct neural networks are assumed to underly emotion regulation and the different strategies, most of them including the bilateral inferior frontal gyrus (IFG) and middle frontal gyrus (MFG; for meta-analyses see Kohn et al., 2014; Morawetz et al., 2017b, 2020). Furthermore, fronto-amygdala coupling seems particularly relevant for successful emotion regulation (Morawetz et al., 2017a). A recent meta-analysis on task-dependent functional connectivity during emotion regulation reported task-modulated coupling between prefrontal regions and the left amygdala, which could be enhanced with regulation success (Berboth & Morawetz, 2021). Here, specifically the left IFG seems to play a key role and a direct link between the left IFG and the left amygdala is suggested during emotion regulation (Berboth & Morawetz, 2021; Morawetz et al., 2016a, 2017b). In contrast, deficits in emotion regulation might be explained by a failure to adequately recruit the neural regulation networks (Radke et al., 2018). Modulating factors of emotion regulation on the behavioral and neural level are sex hormones, especially E2 (Newhouse & Albert, 2015; Rehbein et al., 2021; Wu et al., 2014). Previous studies have shown that emotion regulation success is influenced by E2 levels throughout the female menstrual cycle: Whereas increased regulation effort is required in phases of low E2 (Rehbein et al., 2021; Wu et al., 2014), also reflected in enhanced recruitment of frontal neural resources (Rehbein et al., 2021), females are more successful at downregulating emotional arousal in high E2 phases (Graham et al., 2017). To date, however, little is known about emotion regulation abilities during phases of extremely high E2 levels, such as pregnancy. In pregnant women, inadequate self-reported emotion regulation has been associated with higher hair cortisol levels indicating chronic stress (Conradt et al., 2020). Also, sleep problems, substance use as well as increased rates of depression, anxiety, self-injurious thoughts, and behaviors have been linked to difficulties in emotion regulation (Bradizza et al., 2017; Kalmbach et al., 2020; Lin et al., 2019; Penner & Rutherford, 2022). After birth, the use of regulation strategies was found to predict PPD symptoms, whereby women diagnosed with PPD report less frequent use of adaptive strategies (Coo et al., 2022; Hajal & Paley, 2020; Marques et al., 2018). Most importantly, emotion regulation is suggested to have long-term implications for parental health, caregiving behavior, as well as health and development of the child and is considered a protective factor for psychopathology of mother and child (Lin et al., 2019; Penner & Rutherford, 2022). Nevertheless, no study to date has examined behavioral and neural emotion regulation during the antepartum period, although changes in brain structures related to these functions are evident after pregnancy (Hoekzema et al., 2017).

Addressing this gap in research, the present study investigated behavioral and neural emotion regulation for the first time in primiparous pregnant females using a standard emotion regulation paradigm for negative emotions during fMRI. Outside the scanner, behavioral responses in a positive emotion regulation paradigm were assessed, together with several self-report measures. To specifically investigate the effect of pregnancy compared with high and low E2 levels, results were compared with nulliparous naturally cycling females with respective levels. Based on previous findings (Li et al., 2020), we hypothesized that pregnant females self-report less application of emotion regulation strategies. As no previous study has examined peripartum emotion regulation beyond self-report, our hypotheses were partly exploratory on the behavioral and neural level. However, referring to evidence of altered emotional functions during pregnancy (Pearson et al., 2009; Raz, 2014; Roos et al., 2011, 2012), we hypothesized reduced regulation success in the pregnant compared with the nulliparous groups. Given results on altered ante- and postpartum neural structure and function (Hoekzema et al., 2017; Raz, 2014; Rehbein et al., 2022; Roos et al., 2012) and the influence of E2 on emotion regulation (Rehbein et al., 2021; Wu et al., 2014), we hypothesized group differences on the neural level, particularly in the amygdala (Sundström Poromaa & Gingnell, 2014) and frontal regions (IFG, MFG; Hoekzema et al., 2017; Rehbein et al., 2021) as well as task- and resting-state functional connectivity (Hoekzema et al., 2022; Wonch et al., 2016). Further, we hypothesized associations between emotion regulation abilities and mood symptoms. Here, we anticipated relations of regulation success with functional activity, specifically positive for frontal regions and negative for the amygdala (Hassa et al., 2021; Morawetz et al., 2016a, 2017a; Rehbein et al., 2021). Additionally, we expected relations of reduced regulation success (Coo et al., 2022; Haga et al., 2012) and altered neural activity (Silverman et al., 2011; Wonch et al., 2016) with increased depression symptoms. As such, the present study sheds light on underlying neural mechanisms involved in emotion regulation during pregnancy, which ultimately has implications to understand specificities of emotion (dys)regulation in the peripartum period.

Forty-seven human females between the age of 19 and 36 years were included in the present study, of which 32 were naturally cycling (NC) and 15 primiparous. All participants were right-handed, did not have any present or past mental, neurological, or endocrine disorders and did not take hormonal contraceptives during the past 6 months or any other medication. NC females were required to have a regular, natural menstrual cycle between 26 and 32 days and no current or past pregnancies. NC participants were equally randomized to the E2 valerate (E2V; N = 16) or placebo group (N = 16), receiving either an E2V or placebo pill, respectively. Pregnant females were in the second trimester of their pregnancy (21st–28th gestational week) and provided a copy of their ultrasound screening to exclude any pregnancy-related complication in the mother or fetus. Participants were recruited via the University Tübingen e-mail provider. Pregnant females were additionally informed about the study in the University Women’s Hospital Tübingen and in surrounding gynecology practices. All participants were offered 100€ for participation. Written informed consent and the data protection agreement were obtained from all participants before inclusion. The study was approved by the Ethics Committee of the Medical Faculty of the University of Tübingen (754/2017/BO1) and data were collected from 08, 2018 till 12, 2021.

The present study was part of a project on pregnancy and the brain. Some of the participants included in the current sample have previously been part of samples in other publications within this project (Rehbein et al., 2021, 2022). The procedure for the study is summarized in Figure 1a and outlined below.

All participants were screened for past or present mental disorders and cognitive abilities were assessed (see Fig. 1a). NC females reported onset of their menstruation and were invited to the laboratory between days 2 and 5 of their menstrual cycle, that is, their early follicular phase. One day before the MRI assessments (day 1), blood was drawn in NC females and participants of the E2V group received 6 mg of estradiol valerate (Progynova® 21; Bayer Weimar GmbH & Co. KG), whereas participants in the placebo group received placebo pills in a double-blind fashion. On day 2, NC participants took a second dose approximately 20–24 h after the first pill intake and 6 h before the fMRI session. This procedure allows the experimental elevation of E2 levels, comparable with the peri-ovulatory phase, while at the same time maintaining low progesterone and testosterone levels (Bayer et al., 2018; Kuhl, 2005). On day 2, all groups were invited to the laboratory, where blood samples were drawn to assess hormone levels and neuroimaging was performed. As the Ethics Committee approved only a 30 min (f)MRI session for pregnant participants, all participants performed a subsequent positive emotion regulation task outside the MRI scanner. Additionally, several self-report measures were assessed (see Fig. 1a). Trait use of emotion regulation strategies was reported on the emotion regulation questionnaire (ERQ; Abler & Kessler, 2009; Gross & John, 2003) and depressive symptoms were assessed with the Edinburgh Postnatal Depression Scale (EPDS; Cox et al., 1987; Herz, 1996) in pregnant females and the Beck Depression Inventory (BDI-2; Beck et al., 1996; Kühner et al., 2007) in NC females. The whole procedure on day 2 took maximum 170 min.

The emotion regulation paradigm was adapted from previous studies (Morawetz et al., 2016b, 2017a; Rehbein et al., 2021). During the negative emotion regulation paradigm (see Fig. 1b), participants were exposed to 24 highly negative pictures (International Affective Picture System; Lang & Bradley, 2007) and followed two different instructions: (1) For downregulation, participants were instructed to decrease their emotional response toward the picture, either by changing the perspective on or meaning of the picture, for example, by increasing their personal distance toward the picture, but not by thinking of something positive. (2) In the view instruction, participants were asked to experience arising emotions toward the picture without changing them. In an event-related design, each picture was shown twice for the duration of 8 sec; once preceded by the instruction (2 sec) for downregulation (downward-pointing arrow) and once for view (equal sign), with the two conditions presented in randomized order. Instructions and pictures were followed by a jittered fixation cross (2–6 sec). After each picture, participants rated their emotional state on a continuous visual rating scale (very negative to very positive). The positive emotion regulation paradigm outside the MRI scanner is described in the Supplementary Material.

Hormonal levels of E2, progesterone, and testosterone were analyzed at the central laboratory of the University Hospital Tübingen. Then, 7.5 ml of blood was drawn in serum tubes and analyzed using enzyme-linked immunoassays (ELISA). Sensitivity and range of measurement for E2, progesterone, and testosterone were E2: 43.60–11,010 pmol/l; progesterone: 0.67–190.80 nmol/l; testosterone: 0.24–52.05 nmol/l. Due to problems in drawing blood and technical problems, four blood samples (one placebo, two E2V, one pregnant) were missing.

IBM SPSS Statistics (version 27.0) was used for all statistical analyses, if not specified otherwise. To assess group differences in sample characteristics, univariate analysis of variances (ANOVAs) were applied with group (placebo, E2V, pregnant) as factor. In case of significant group differences, Bonferroni-adjusted post-hoc tests were performed. If the assumption of a normal distribution was not met, nonparametric Kruskal–Wallis H tests were applied. For statistically significant results, follow-up Mann–Whitney U tests were calculated between the individual pairs of groups. For all analyses, the α criterion level was set to p ≤ .05. Effect sizes in the ANOVAs are reported in ηp² (0.01 ηp² = small, 0.06 ηp² = moderate, 0.14 ηp² = large). As this study includes a small but first sample of pregnant females during fMRI measurement, statistical tendencies are reported up to p ≤ .10 for brain activity results. In case data were missing for a participant, it was not replaced but the participant excluded from the respective analysis. Sensitivity analyses were conducted using G*Power (version 3.1; Faul et al., 2007) to identify the minimum detectable effect sizes at 80% power (β = .80) and α = .05, tailored to each analysis.

Emotional state ratings acquired per picture and instruction were summarized in mean values for the view and regulation condition for each participant. To compare between groups and conditions, mixed between-within-subjects ANOVAs were performed with mean state ratings as a dependent interval variable, group (placebo, E2V, pregnant) as between-subjects factor, and regulation (regulation, view) as within-subject factor. To assess emotion regulation success, we subtracted mean emotional state ratings for view from mean ratings for the regulation condition for each participant (Morawetz et al., 2017a). As emotion regulation success for negative pictures was a variable of interest for our regression analyses, we compared between groups by performing a univariate ANOVA with group (placebo, E2V, pregnant) as factor and regulation success as dependent variable. For the negative emotional state ratings, data were missing for three females, one in each group.

Data were acquired on a 3T Siemens PRISMA scanner at the University Hospital Tübingen. Brain structure was measured with a standard magnetization-prepared rapid gradient-echo sequence (MPRAGE; TR = 2.3 sec, TE = 4.16 msec, slice-thickness = 1 mm, voxel size = 1 × 1 × 1 mm, flip-angle 9°, distancing factor 50%, GRAPPA acceleration factor, sagittal orientation). The fMRI sequences (task, resting-state) consisted of standard echo-planar imaging (EPI) protocols (32 interleaved slices, TR = 2 sec, TE = 32 msec, voxel-size 3.4 x 3.4 x 3.4 mm, flip-angle 76° transversal orientation, anterior-posterior commissure orientation, 64-channel head coil). Participants were provided earplugs to reduce the sound intensity. To further ensure that the sound level was kept at a minimum for the unborn child, the volume of the MRI sequence was measured before scanning, reaching approximately 90 dB. Besides, the unborn child in the womb is protected against noise disturbance through the amniotic fluids. Data from one female in the placebo group had to be excluded due to excessive head movement and technical problems and one pregnant participant terminated MRI measurements after the anatomical scan.

Data were preprocessed using SPM12. Preprocessing consisted of slice time and distortion correction, realignment and unwarping, segmentation, co-registration, normalization, and smoothing (6 mm FWHM). Participants with head movement of more than 2 mm were excluded from the analyses. The first-level analysis included the regressors downregulation and view, the emotional state rating period, instruction period, plus six movement parameters, and time derivatives for each participant. At the second level, a full factorial analysis with the factors group (placebo, E2V, pregnant) and regulation (downregulation, view) was calculated, including age as covariate. Whole-brain results were corrected for multiple comparisons with cluster-wise correction (p < .001). Brain areas were labeled with the Anatomy toolbox as available in SPM12. Data were visualized with SPM12.

Region of interest (ROI) analyses. Based on previous findings (Morawetz et al., 2017b; Rehbein et al., 2021) and our a-priori hypotheses, selected ROIs related to emotion regulation were bilateral IFG (MNI(x,y,z): -46, 26, -8 / 50, 30, -8) and bilateral MFG (MNI(x,y,z): -45, 20, 35 / 42, 48, -2). Additionally, due to the impact of E2 and involvement in emotion regulation, bilateral amygdala were included (MNI(x,y,z): -24, -3, -21 / 24, -6, -15). Beta-values were extracted with MarsBarR (Brett et al., 2002; 10 mm sphere around the reported MNI peaks). The effects of group (placebo, E2V, pregnant) and regulation (downregulation, view) were analyzed in a mixed between-within ANOVA. Age was included as covariate as pregnant females were significantly older than nulliparous groups (Table 1). To correct for multiple testing related to laterality, the Bonferroni-corrected significance level was set to p < .025.

Psychophysiological interaction (PPI) analysis. To investigate changes in functional connectivity of brain regions (physiological component) during the downregulation and view condition (psychological component), we conducted ROI-to-ROI PPI analyses (Friston et al., 1997) using the CONN Toolbox (Whitfield-Gabrieli & Nieto-Castanon, 2012). Following Morawetz et al. (2017a), we performed a PPI analysis of (1) the left amygdala with the left IFG. Additionally, we explored connectivity of (2) the left amygdala with the left MFG. Time series from these ROIs were determined, individual connectivity parameters for each regulation condition extracted and entered as dependent variables in separate univariate analysis of covariance (ANCOVA) with group (placebo, E2V, pregnant) as factor and covariates age and regulation success (Berboth & Morawetz, 2021; Morawetz et al., 2017a). To correct for multiple testing, the Bonferroni-corrected significance level was set to p < .025. Due to missing data, 14 females in the placebo, 15 in the E2V, and 14 in the pregnant group were included in the PPI analyses.

Examining the same task-related connectivities at rest, ROI-to-ROI resting-state connectivity parameters of (1) the left amygdala with the left IFG as well as (2) the left amygdala with the left MFG were extracted using the CONN Toolbox (Whitfield-Gabrieli & Nieto-Castanon, 2012). The effect of group (placebo, E2V, pregnant) on resting-state connectivity parameters was analyzed in an ANCOVA with age and regulation success score as covariates. To correct for multiple testing, the Bonferroni-corrected significance level was set to p < .025. Furthermore, we explored resting-state connectivity in (3) an emotion downregulation network consisting of previously reported ROIs (Morawetz et al., 2017b, 2022; for detailed information see Supplementary Material). For the resting-state analysis, data were missing for one pregnant participant.

First, separate linear regression analyses were conducted with brain activity during downregulation as predictor (left/right IFG; left/right MFG; left/right amygdala) of regulation success scores in each group, respectively. This allows a straightforward interpretation of group-specific patterns without obscuring meaningful patterns within the groups. To correct for multiple testing related to laterality, the Bonferroni-corrected significance level was set to p < .025. Second, in each group, we conducted regression analyses of negative emotion regulation success as predictor of depression scores (EPDS for pregnant, BDI-2 for nonpregnant groups). Third, individual regression analyses in each group were conducted with (1) bilateral amygdala activity during downregulation and (2) functional connectivity (left amygdala with left IFG; left amygdala with left MFG during downregulation) as predictors of depression scores. To correct for multiple testing, the Bonferroni-corrected significance level was set to p < .025.

Detailed sample characteristics are presented in Table 1. Mean age between the groups was significantly different (H = 20.53, p < .001) with pregnant participants being older than nonpregnant participants. In terms of cognitive abilities, groups differed in processing speed (TMT-A: H = 8.92, p = .012), with pregnant females being slower than the placebo (p = .017) and E2V group (p = .004). However, groups showed no difference on the delta of TMT-B–A (F(2,43) = .88, p = .423), which is used to remove bias due to psychomotor functioning or visual sequencing (73). In terms of self-report data, the placebo and E2V group differed in alexithymia scores (TAS-20: F(2,46) = 3.33, p = .045, part-η² = .137; placebo vs. E2V group: p = .041) but no difference was found in comparison with the pregnant group (p’s > .478). Also, pregnant females reported reduced subjective sleep quality compared with females in the placebo group (F(2,40) = 3.36, p = .045; part-η² = .144, placebo vs. pregnant: p = .083), but groups did not differ on the total sleep quality index (F(2,34) = .91, p = .413). In terms of hormones, pregnant females had higher levels in all assessed sex hormones compared with both nonpregnant groups (p’s < .031). The two nonpregnant groups only differed in E2 levels (p < .001) but not in progesterone and testosterone (p’s > .451). In terms of trait emotion regulation use (ERQ), groups differed in cognitive reappraisal (F(2,43) = 4.03, p = .025, part-η² = .164), with pregnant females reporting lower ratings than the placebo (p = .022) but not the E2V group (p = .194). E2V and placebo groups did not differ in their cognitive reappraisal ratings (p = .978; see Fig. 2). No group differences appeared for emotional suppression (F(2,44) = 1.51, p = .233). For all other sample characteristics, no differences emerged (p’s > .054).

For negative emotional state ratings, a significant main effect of regulation was found (F(1,41) = 22.07, p < .001, part-η² = .350), with reduced negative emotional state ratings after downregulation compared with view. The main effect of group was not significant (F(2,41) = 1.14, p = .331, part-η² = .053) and no significant interaction was revealed (F(2,41) = 1.44, p = .248, part-η² = .066). Likewise, groups did not differ significantly in regulation success (F(2,41) = 1.44, p = .248, part-η² = .066). See Table 1 and Figure 2 for details. Results for positive emotion regulation are reported in the Supplementary Material and Table 1.

Whole-brain analysis. In a whole-brain approach, the 3 x 2 full factorial analysis showed a significant main effect of regulation, with an increased activation of the left MFG during downregulation (MNI (x,y,z) = -41, 7, 49; p_FWE < .001, k = 310; see Supplementary Table S1), while no significantly stronger activation during view was revealed (see Fig. 3a).

Region-of-interest analyses. In line with whole-brain results, we observed a tendency for a main effect of regulation for the left MFG (F(1,40) = 3.33, p = .075, part-η² = .077; see Fig. 3b), with higher brain activity during downregulation than view (p = .003). No significant group effect (F(2,40) = .31, p = .736, part-η² = .014) nor interaction emerged (F(2,40) = 1.57, p = .220, part-η² = .073). For all other frontal ROIs (right MFG, bilateral IFG), no significant effects occurred (p’s > .116). For the left amygdala, a main effect of group emerged (F(2,40) = 4.71, p = .015, part-η² = .191; see Fig. 3b), but no regulation effect (F(1,40) = .39, p = .538, part-η² = .010) nor interaction (F(2,40) = .02, p = .978, part-η² = .001). In terms of the group effect, E2V showed significantly higher activity than the pregnant group (p = .019), but not compared with the placebo group (p = .103). There was no difference between the pregnant and placebo groups (p = .597). For the right amygdala, a tendency of a group effect corrected at Bonferroni level emerged (F(2,40) = 3.67, p = .034, part-η² = .155; see Fig. 3b). No regulation effect (F(1,40) = .39, p = .538, part-η² = .010), nor interaction (F(2,40) = .29, p = .750, part-η² = .014) occurred. Post-hoc comparisons revealed a trend for higher activity in the E2V than the placebo (p = .057) but not compared with the pregnant group (p = .123). Activity in the placebo was not different to the pregnant group (p = 1.00).

PPI analysis. Controlling for age and regulation success (Morawetz et al., 2017a, 2020), connectivity analysis of (1) the left amygdala with the left IFG revealed no significant differences during downregulation (F(2,38) = .47, p = .626, part-η² = .028) and view (F(2,38) = .10, p = .907, part-η² = .011). Likewise, groups did not differ in functional connectivity of (2) the left amygdala with the left MFG during downregulation (F(2,38) = 1.47, p = .243, part-η² = .072) and view (F(2,38) = .73, p = .487, part-η² = .012).

No significant group differences emerged for (1) the left amygdala with the left IFG (F(2,41) = .58, p = .566, part-η² = .030), nor for (2) the left amygdala with the left MFG (F(2,41) = .65, p = .526, part-η² = .053), when controlling for age and regulation success. Resting-state connectivity within the emotion downregulation network also did not differ between groups (all p_FDR > .357; detailed results provided in the Supplementary Material and Supplementary Table S2).

Brain activity and regulation success. In pregnant females, the regression of left amygdala activity during downregulation on regulation success was significant (R² = .36, beta = -.60, p = .023), with higher activity associated with reduced regulation success (see Fig. 4a). Brain activity during downregulation for all other ROIs was not associated with regulation success in the pregnant group (p’s > .140). In both nonpregnant groups, no significant associations of brain activity during downregulation and regulation success emerged (p’s > .100).

Regulation success and depression scores. In the pregnant and both nonpregnant groups, no significant associations of regulation success on depression scores occurred (p’s > .121).

Brain activity and depression scores. In pregnant females, the regression with a Bonferroni-corrected significance level showed borderline significance on depression scores (EPDS) for both the left amygdala (R² = .37, beta = .61, p = .036) and the right amygdala during downregulation (R² = .38, beta = .61, p = .034). The correlation between bilateral amygdala reactivity and EPDS scores was found significant (left: r = .61, p = .018; right: r = .61, p = .017; see Fig. 4b). Further, connectivity parameters of the left IFG with left amygdala showed a tendency for a positive correlation with EPDS scores (r = .439, p = .077), but connectivity of the left MFG with left amygdala showed no correlation with EPDS scores (r = .242, p = .224). In the nonpregnant groups, no significant associations of amygdala activity or functional connectivity during downregulation on depression scores (BDI-2) emerged (p’s > .107). Detailed results are provided in Table 2.

For the first time, behavioral and neural emotion regulation were investigated in healthy first-time pregnant females in the second trimester and compared with nulliparous naturally cycling (NC) females in the early follicular phase, who either received E2 valerate to specifically increase E2 levels (E2V group) or a placebo to maintain low E2 levels (placebo group). While pregnant females reported lower use of cognitive reappraisal in daily life, emotion regulation performance and prefrontal activity did not differ between groups. However, group differences emerged in amygdala reactivity. In pregnant females, amygdala activity during downregulation predicted reduced regulation success and related to increased self-reported depression symptoms. These findings contribute several insights into emotional and neural functions during pregnancy, as discussed further.

Contrary to our expectations, pregnant females in the second trimester showed no difference in whole-brain fMRI activity compared with nonpregnant females. While considerable changes in brain structure and function related to pregnancy have been reported, most of this evidence in humans stems from pre- versus post-pregnancy comparisons or cross-sectional studies in pregnant versus nonpregnant women (Barba-Müller et al., 2019; Sacher et al., 2020). Up until now, the only functional neuroimaging studies during antepartum used scalp-recorded near-infrared spectroscopy (NIRS; Roos et al., 2012) and electro-encephalography (EEG; Raz, 2014). While NIRS showed different prefrontal activation in response to emotional stimuli between trimesters of pregnancy, no differences compared with nonpregnant controls were reported (Roos et al., 2012). Using EEG, distinctions in neural recruitment during emotional processing were recorded for third trimester pregnant compared with nonpregnant women (Raz, 2014). Little is known about when and how reported neural changes occur in human pregnancy and if the second trimester could be a phase in which functional differences are not yet apparent using fMRI. For example, whereas MRI from pre-pregnancy to postpartum revealed reduced gray matter volume in anterior and posterior midline structures, bilateral temporal- and prefrontal cortex, including IFG, compared with nulliparous women (Hoekzema et al., 2017, 2022), MRI in primiparous women in the second trimester showed smaller gray matter volume merely in the left putamen (Rehbein et al., 2022). Research in female rodents supports that processes underlying pregnancy-related neural changes are especially prominent during late pregnancy and throughout the postpartum period, among others due to the involved steroid hormones (Ammari et al., 2023; Servin-Barthet et al., 2023). Therefore, it may not be until the third trimester of pregnancy, when hormone levels have increased even more, or until postpartum, when hormone levels dropped sharply, that functional differences can be observed in whole-brain fMRI.

Yet, ROI analyses indicated hypothesized group differences in the left amygdala as well as borderline significant differences in the right amygdala. Here the E2V group showed higher activity, regardless of regulation condition, compared with the pregnant in the left amygdala and a trend for the placebo group in the right amygdala. As illustrated in Figure 3, the observed activity pattern in the amygdala shows an inverted-U-shaped response in relation to E2 levels. This reflects an activation pattern with the highest values in the medium range and low levels at the extremes. This was already observed in NC women, for which activity in the medial posterior hippocampus was reduced under low and supraphysiological E2 levels, whereas it was enhanced for physiologically high E2 levels (Bayer et al., 2018). Characteristics of different estrogen receptors and their expression ratios are proposed to underlie these E2-dependent neuronal patterns (Bayer et al., 2018; Österlund et al., 2000; Soares & Zitek, 2008). Thus, our data suggest that the inverted-U-shaped E2 dose–response relation might be applied to the amygdala region in response to emotional stimuli. However, E2 is not the only ovarian hormone involved in pregnancy with an influence on amygdala function and it should be carefully examined whether the observed differences can solely be attributed to E2-specific effects. For example, progesterone and testosterone are suggested to modulate amygdala activity in response to emotional stimuli in NC women and also increase and fluctuate during peripartum (Radke et al., 2015; Sacher et al., 2020; Soares & Zitek, 2008; van Wingen et al., 2009, 2011). Although it is not yet clear how the dynamic and complex hormonal milieu during pregnancy influence brain function, research in rodents shows that several pregnancy-related hormones have an influence on neural excitability (Ammari et al., 2023), including the amygdala (Servin-Barthet et al., 2023). Eventually these hormonal changes facilitate the development of maternal behavior (Ammari et al., 2023; Servin-Barthet et al., 2023), which was supported in human mothers, particularly in response to infant cues (Barrett et al., 2012; Kim et al., 2010). Apart from maternal functions, the amygdala is generally involved in socioemotional processing as well as in identifying and appropriately responding to salient environmental stimuli (Adolphs et al., 2002; Cunningham & Brosch, 2012; Pessoa, 2010). Hence, as part of an “maternal caregiving circuit” (Numan, 2020), it is apparent and adaptive that pregnancy-related plasticity processes affect the amygdala and thus influence emotional functions—all of which promotes maternal behavior (Servin-Barthet et al., 2023). As the present study provides the first data on amygdala function in human pregnancy, associations with regulation success and mood symptoms were specifically assessed.

A striking finding of this study is that, within the pregnant group, left amygdala activity significantly predicts negative emotion regulation success. Specifically, pregnant females with increased amygdala activity were less successful in regulating their emotional state. This is in line with evidence of reduced amygdala activity when emotions have been regulated successfully in men and nonpregnant women (Morawetz et al., 2017a; Rehbein et al., 2021), as well as with evidence of enhanced amygdala activation when individuals with impaired regulation abilities engage in cognitive reappraisal (Hassa et al., 2021). Hence, the same neural mechanism underlying successful emotion regulation seems to apply to our sample of pregnant females. At the same time, our results suggest that the association of amygdala activity and regulation success could be more pronounced for pregnant women, possibly due to extreme E2 levels: while the amygdala x regulation success association was significant in the pregnant group, only a tendency emerged in the E2V group, and no significant relation occurred in the placebo group. Specifically, heightened amygdala activity in pregnant females could indicate increased sensitivity or susceptibility (Wonch et al., 2016), which, in relation to reduced regulation success, proposes a risk marker for mental ill-health. Amygdala hyperresponsiveness is associated with lower resilience, that is, reduced capacity to tolerate stress, heightened vulnerability for depression, and has been proposed as a transdiagnostic factor for psychopathologies (Barbour et al., 2020). For example, increased amygdala activity for negative stimuli is seen in nonpregnant patients with depression as compared with healthy controls (Drevets et al., 2008; Hamilton et al., 2012; Sheline et al., 2001; Suslow et al., 2010; Victor et al., 2010). Evidence on neural correlates of peripartum depression is still scarce and inconclusive; however, data in our pregnant sample indicated a positive correlation of bilateral amygdala activity during downregulation with depression scores (EPDS). This was supported by results of the regression analysis, where amygdala activity predicted depressive scores, for which borderline significance was reached, underscoring heightened amygdala activity as a vulnerability marker for mental health. Besides, in our pregnant group a trend for a positive correlation between functional connectivity of the left IFG with the left amygdala during downregulation and EPDS scores emerged. Although we cannot draw conclusions on effective connectivity, research supports that frontal regions are recruited to reduce amygdala activation during successful cognitive reappraisal (Morawetz et al., 2017a). Thereby, our findings could reflect that women who need to exert more neural regulatory effort to reduce amygdala activity and regulate their emotions successfully also have higher depression scores. The relation of regulation success and depressive scores was not found significant in our group of healthy pregnant females. Yet, an association of emotion regulation abilities with depression scores has been reported for females diagnosed with PPD before (Coo et al., 2022; Haga et al., 2012). Presumably, the healthy status of our pregnant sample might account for these null findings, at least at the time point of their second trimester. Higher mood instability has been reported from the third trimester onward until the early postpartum (Li et al., 2020), so that changes in depression scores, regulation success, as well as their relation in healthy pregnant women might only be observed at a later stage. Nevertheless, the present study is the first to show that amygdala activity in response to emotional stimuli is associated with early mood symptoms in the second trimester of pregnancy. Reduced regulation success is a risk marker for peripartum depression and anxiety (Coo et al., 2022; Haga et al., 2012; Hajal & Paley, 2020; Radke et al., 2018), positioning heightened amygdala activity during pregnancy as a key neural vulnerability correlate for peripartum depression. To further substantiate these associations, future research should assess amygdala response ideally before, during, and after pregnancy, alongside monitoring symptoms of depression in both mothers with and without peripartum depression.

Comparing the groups on emotion regulation performance, pregnant females were just as successful as nonpregnant females in regulating their emotional state when confronted with emotional stimuli and asked to apply cognitive reappraisal. While this confirms and extends results in nulliparous women with low versus high E2 levels, who showed no difference in behavioral emotion regulation success (Rehbein et al., 2021; Wu et al., 2014), this contradicts reports of altered emotional functions among pregnant women (Pearson et al., 2009; Raz, 2014; Roos et al., 2011, 2012). Inconsistent findings could be explained by the time point of our assessment, which was conducted in the second trimester of pregnancy. Previous studies, however, reported increased mood instability, greater emotional sensitivity, and altered evaluation of emotional stimuli during the third trimester (Li et al., 2020; Pearson et al., 2009; Raz, 2014). Pregnancy and motherhood mark transition phases which are accompanied by several psychosocial challenges, including diverse and varying internal experiences and societal expectations (Sacher et al., 2020). Over the course of pregnancy, these challenges can change and ultimately influence emotion regulation abilities and the use of (mal)adaptive regulation strategies (Hayes et al., 2006; Sacher et al., 2020). Consequently, emotion regulation should be investigated in different peripartum stages to draw conclusions about potential changes in regulation abilities and the use of regulation strategies throughout pregnancy. The absence of behavioral differences in the present sample was mirrored on the neural level, where no group differences were found for functional activity related to emotion regulation in whole-brain- and frontal ROI analyses. This was contrary to our hypothesis based on pregnancy-related plasticity in brain regions associated with emotion regulation (Hoekzema et al., 2017, 2022). Instead, in accordance with the emotion regulation literature, all participants had increased activity in the left MFG during the downregulation of negative emotions (Goldin et al., 2008; Kohn et al., 2014; Morawetz et al., 2017b), indicating the same neural mechanism of emotion regulation being recruited during pregnancy. Notably, however, for the pregnant group, we observed pronounced activation differences in the left MFG during the downregulation compared with the view condition. While this observation must be interpreted carefully, the activation differences during cognitive reappraisal could indicate higher regulation effort exerted by pregnant females. This has been suggested for nonpregnant women, in which increased frontal activation during emotion regulation, that is, neural effort, was found in women with low compared with high E2 levels (Rehbein et al., 2021). Replication using fMRI in a larger sample of antepartum women might reveal whether increased left MFG activation during downregulation can be verified. Further, we did not find any group differences in functional connectivity during downregulation, nor in resting-state connectivity. Hence, any neural plasticity in regions and circuits involved in emotion regulation (Hoekzema et al., 2017, 2022; Sacher et al., 2020) did not seem to affect neural downregulation of negative emotions in healthy pregnant females during the second trimester. However, parity is proposed to modulate both neural activity and connectivity (Wonch et al., 2016), indicating that differences might be observed in multiparous as compared with primiparous and nulliparous women. Insufficient difficulty of the functional paradigm could also explain lack of group differences, that is, recording ceiling effects instead of detecting differences in regulation abilities. Yet, the present paradigm is an established fMRI paradigm and was used in previous studies (Morawetz et al., 2016a, 2016b, 2017a).

Notably, pregnant females reported the lowest tendency of using cognitive reappraisal in their daily life. Previous evidence on higher mood instability in pregnant women has also been based on self-reports (Li et al., 2020). Presumably, rather than failing to regulate emotions, pregnant women could perceive reduced emotion regulation abilities or experience more fluctuating emotions to regulate. Possibly, they might also rely on other strategies than cognitive reappraisal in daily life, such as interpersonal regulation strategies (Coo et al., 2022; Li et al., 2020). Since the present self-report questionnaire only asked about cognitive reappraisal and expressive suppression, future studies should assess the use of various strategies.

Our data have some limitations that may influence data interpretation and raise ideas for future research. Despite assessing human primiparous females in the MRI scanner, which poses some challenges, we acknowledge the small sample size that should be taken into consideration when interpretating the data. While our study reported large effect sizes across self-report, behavioral, and functional activity measures, the sensitivity analyses indicated that the study lacked sufficient power to reliably detect moderate-to-large effects (e.g., part-η² < .145 for main effects of group, part-η² < .08 for interaction effects). For instance, the observed interaction effect of group and regulation in the left MFG (part-η² = .073) and the main effect of group in functional connectivity during downregulation (left amygdala with left MFG, part-η² = .072) highlight possible effects. Thus, the lack of statistical significance may stem from limited power rather than the absence of an underlying effect. These findings emphasize the importance of conducting future studies with larger sample sizes to further explore these relationships. Our cross-sectional study only included healthy primiparous females during the second trimester of pregnancy. To understand pregnancy-related changes in emotional functions and neuronal plasticity, future research should conduct repeated-measures designs in different stages of pregnancy and the postpartum period. Cross-sectional designs comparing primiparous and multiparous mothers would improve our understanding of the potential modulating role of parity on the brain (Wonch et al., 2016). Besides replication in healthy samples, it is of utmost importance to assess clinical samples of ante- and postpartum women regarding their behavioral and neural emotion regulation to characterize and confirm the outlined risk associations.

The results of the present study provide pioneering insights into functional brain activation during human pregnancy, highlighting a risk association between amygdala function, regulation success, and depression scores in pregnant females, with considerable potential for advancing clinical applications. Up to 80% of pregnant women with high anxiety or depressive symptoms go unidentified and do not receive treatment, which has long-term consequences for mother and child (Barbour et al., 2020). Recognizing the peripartum period as a unique opportunity to identify women at risk for pregnancy-related mental health problems (Barba-Müller et al., 2019), this study suggest that vulnerability markers and risk factors can be detected through antepartum self-reports, behavior, and brain function. Screening emotion regulation abilities and mood symptoms via self-report measures during and after pregnancy provides a practical early detection approach, especially in settings where brain imaging is not feasible (Haga et al., 2012; Sacher et al., 2020). Promoting adaptive emotion regulation for those at risk may serve as a preventive intervention for PPD (Haga et al., 2012; Sacher et al., 2020). As a resilience factor for mental health, understanding and fostering adaptive regulation strategies like cognitive reappraisal benefits both maternal well-being and child development, including parent–child relations (Coo et al., 2022; Lin et al., 2019; Penner & Rutherford, 2022). Nonetheless, further research is needed to unravel the neural mechanisms of emotion regulation during the peripartum period, deepening our understanding of the brain–maternal mental health connection and guiding targeted interventions to improve outcomes for mothers and children.

The data and code used in the present study can be requested by sending a research proposal to the principal investigator (PI) Prof. Dr. Birgit Derntl (e-mail: [email protected]). The research proposal is 1-page maximum. Please provide a clear and informative title for your proposed research. Please briefly describe the overall rationale for your study and summarize the specific aims/hypotheses that you will test with the specific data elements you are requesting. A data sharing agreement is subsequently drawn up.

F.W.: Formal analysis; Investigation; Visualization; Writing—original draft. L.K.: Supervision; Methodology; Validation; Writing—review & editing. E.R.: Data curation; Writing—review & editing. C.M.: Methodology; Validation; Writing—review & editing. I.S.P.: Conceptualization; Funding acquisition; Methodology; Writing—review & editing. A.S.: Supervision; Writing—review & editing. B.D.: Conceptualization; Funding acquisition; Methodology; Project administration; Resources; Supervision; Validation; Writing—review & editing.

This study was funded by the Center for Integrative Neuroscience, Tübingen, Germany (CIN; EXC307, MRTG Pregnancy & the Brain) and the International Research Training Group “Women’s Mental Health Across the Reproductive Years” (DFG, IRTG2804). We acknowledge support from the Open Access Publication Fund of the University of Tübingen.

The authors declare no competing interests.

Supplementary material for this article is available with the online version here: https://doi.org/10.1162/imag_a_00529.