Neural Basis of Increased Cognitive Control of Impulsivity During the Mid-Luteal Phase Relative to the Late Follicular Phase of the Menstrual Cycle

Zhuang, Jin-Ying; Wang, Jia-Xi; Lei, Qin; Zhang, Weidong; Fan, Mingxia

doi:10.3389/fnhum.2020.568399

ORIGINAL RESEARCH article

Front. Hum. Neurosci., 12 November 2020
Sec. Cognitive Neuroscience
Volume 14 - 2020 | https://doi.org/10.3389/fnhum.2020.568399

Neural Basis of Increased Cognitive Control of Impulsivity During the Mid-Luteal Phase Relative to the Late Follicular Phase of the Menstrual Cycle

Jin-Ying Zhuang^1,2

Jia-Xi Wang^2*

Qin Lei²

Weidong Zhang²

Mingxia Fan³

¹Faculty of Education, East China Normal University, Shanghai, China
²School of Psychology and Cognitive Science, East China Normal University, Shanghai, China
³Department of Physics, Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, China

Hormonal changes across the menstrual cycle have been shown to influence reward-related motivation and impulsive behaviors. Here, with the aim of examining the neural mechanisms underlying cognitive control of impulsivity, we compared event-related monetary delay discounting task behavior and concurrent functional magnetic resonance imaging (fMRI) revealed brain activity as well as resting state (rs)-fMRI activity, between women in the mid-luteal phase (LP) and women in the late follicular phase (FP). The behavioral data were analyzed and related to neural activation data. In the delay discounting task, women in the late FP were more responsive to short-term rewards (i.e., showed a greater discount rate) than women in the mid-LP, while also showing greater activity in the dorsal striatum (DS). Discount rate (transformed k) correlated with functional connectivity between the DS and dorsal lateral prefrontal cortex (dlPFC), consistent with previous findings indicating that DS-dlPFC circuitry may regulate impulsivity. Our rs-fMRI data further showed that the right dlPFC was significantly more active in the mid-LP than in late FP, and this effect was sensitive to absolute and relative estradiol levels during the mid-LP. DS-dlPFC functional connectivity magnitude correlated negatively with psychometric impulsivity scores during the late FP, consistent with our behavioral data and further indicating that relative estradiol levels may play an important role in augmenting cognitive control. These findings provide new insight into the treatment of conditions characterized by hyper-impulsivity, such as obsessive compulsive disorder, Parkinson disease, and attention deficit hyperactivity disorder. In conclusion, our results suggest that cyclical gonadal hormones affect cognitive control of impulsive behavior in a periodic manner, possibility via DS-dlPFC circuitry.

Introduction

It has been posited that observed influences of menstrual cycle phase in women on reward-related impulsivity reflect alterations in brain dopamine function and thus downstream effects of those alterations on dopamine efferent targets in the basal ganglia and frontal cortex (Xiao and Becker, 1994; Jackson et al., 2006). These circuits are critically involved in temporal and reward processing (McClure et al., 2004; Kim et al., 2012); they are organized by gonadal steroid hormones during early development and are modulated by these hormones in adults (Xiao and Becker, 1994; Jackson et al., 2006). When estradiol levels are highest during the late follicular phase (FP), women have been shown to be more responsive to rewarding substances, such as cocaine (Terner and De Wit, 2006), and reward-related activation in the mesolimbic system has been shown to be enhanced in the late FP, compared with that in the mid-luteal phase (LP) when progesterone levels peak (Dreher et al., 2007). Self-reported liking of smoked cocaine is greater during the late FP than the mid-LP (Sofuoglu et al., 1999; Evans and Foltin, 2006). Animal studies have also provided evidence of estrogenic modulation of reward-related impulsivity. For example, female rats exhibit their maximal cocaine self-administration levels shortly after estradiol peaks, and exogenous administration of estradiol enhances the acquisition of cocaine self-administration in ovariectomized female rats (Lynch et al., 2001; Jackson et al., 2006). These findings suggest that the proclivity of female mammals to wait for a higher reward is reduced in the late FP relative to the mid-LP.

In contrast, progesterone may play a role in reducing impulsive behavior by favoring more cognitive control. Progesterone is a female gonadal hormone that is often included in hormonal contraceptives and drugs used to maintain pregnancies (Jones et al., 2005). Exogenously delivered progesterone can alleviate stimulant abuse in animals (Anker et al., 2009; Zlebnik et al., 2014) and may support stimulant use cessation in humans (Evans and Foltin, 2006, 2010; Quinones-Jenab and Jenab, 2010; Fox et al., 2013; Carroll and Lynch, 2016; Swalve et al., 2016). Furthermore, progesterone, or its metabolite allopregnanolone, reduce stress and impulsive behavior as measured by the Stroop test in humans (Milivojevic et al., 2016). Neuroimaging studies have demonstrated that increases in progesterone levels from the late FP to the mid-LP result in changes in activity in several prefrontal cortex regions (Dreher et al., 2007; Van Wingen et al., 2008; Ossewaarde et al., 2011; Marečková et al., 2012). Thus, the relatively high progesterone and low estradiol levels during the mid-LP may affect the functioning of brain regions involved in cognitive control in a manner that results in reduced impulsivity.

The main estrogen receptor is highly expressed in the amygdala (Wharton et al., 2012), which sends glutamatergic efferents to the striatum, especially the dorsal striatum (DS), a crucial brain area in the reward system related to impulsivity (Yager et al., 2015). Cummings et al.’s results (2014) suggested that estradiol enhances dopamine release in the DS, but not ventral striatum (VS), in female rats. A recent study showed that women’s responses to drug cues in the DS, specifically in the putamen, were modulated by menstrual phase (Franklin et al., 2019).

A large body of evidence indicates that the dorsolateral prefrontal cortex (dlPFC) plays an important role in cognitive control (Sheline et al., 2010; Cieslik et al., 2013). Moreover, the quality of communication between the DS and dlPFC may modulate cognitive control of impulsivity. A neural circuit linking the right caudate nucleus head and putamen with the dlPFC has been identified as a cognitive loop (Rotge et al., 2008). Commonly, patients with altered frontal-striatal function—such as in obsessive compulsive disorder (Heuvel et al., 2005), Parkinson disease (Williamsgray et al., 2007), and Huntington’s chorea (Watkins et al., 2000)—exhibit executive functioning deficits. Reduced resting state (rs)-functional connectivity between the right caudate and dlPFC has been associated with cognitive control deficits in internet gaming disorder (Yuan et al., 2016). Remarkably, in a functional magnetic resonance imaging (fMRI) study examining rs-functional connectivity in women smokers, Wetherill et al. (2016) found that, compared to data obtained during the LP, when in the FP of their menstrual cycles women had lower rs-functional connectivity between cognitive control areas (dorsal/subgenual anterior cingulate cortex and medial orbitofrontal cortex) and a reward-related region (i.e., the VS) as well as less cognitive control over their smoking behavior.

Humans have a tendency for delay discounting, a phenomenon wherein the value of a delayed rewards are over-discounted relative to that of a reward that can be obtained immediately or sooner (Frederick et al., 2002; Green and Myerson, 2004). Delay discounting has been conceptualized as an index of impulsivity (Staubitz, Lloyd, and Reed). Theoretically, the more a person discounts the value due to its delay, the more impulsive their choice is considered. Patients with conditions that affect brain dopamine systems, namely Parkinson disease and obsessive compulsive disorder, have been shown to exhibit more pronounced delay discounting than healthy and neurotypical people (Guttman and Seeman, 1985; Volkow et al., 1993; Hesse et al., 2005). This intensified preference for selecting of smaller, sooner rewards over larger, later rewards in these populations has been termed a Now bias (Smith et al., 2014).

The aim of the present study was to investigate the neural mechanisms underlying the modulatory effects of menstrual cycle phase on cognitive control of impulsivity with an event-related monetary delay discounting behavioral task and a rs-fMRI study. We chose to compare behavioral and imaging results between the late FP, when there are high estradiol levels with low progesterone levels, and the mid-LP, when both estradiol and progesterone levels are relatively high (Figure 1). We hypothesized that, compared to women in the mid-LP, women in the late FP would show a greater Now bias in the delay discounting task concomitant with higher activity in the DS. Furthermore, we hypothesized that a brain regions related to cognitive control, including the dlPFC, would be more active during the mid-LP, leading to the behavioral acceptance of later, larger rewards. Finally, we hypothesized that DS-dlPFC functional connectivity may play an important role in regulating impulsivity. To test our hypotheses, choices in the delay discounting task (McClure et al., 2004), performed during fMRI scanning, were compared between women in the late FP and women in the mid-LP (task fMRI study). Because fMRI signal fluctuations at rest contain information about functional network architecture (Fox et al., 2005; Fox and Raichle, 2007; Biswal, 2012), we also compared rs-functional connectivity between these two groups (rs-fMRI study) and correlated the connectivity data to hormone level data.

FIGURE 1

Figure 1. Transformed discount rate (k) by menstrual phase. The mean transformed k value was significantly greater during the late FP than during the mid-LP (error bars indicate the standard error of the mean).

Task fMRI Study

Methods

Participants

A cohort of 24 healthy, right-handed, female undergraduates (mean age [± standard deviation (SD)], 22.79 ± 1.44 years; range 20–25) were enrolled. The participants were recruited from a larger, common subject pool with certain inclusion/exclusion criteria, only individuals who were heterosexual, reported having a 28- to 30-day menstrual cycle, and did not take any form of hormones in the previous 3 months were included. Because the criteria for these other studies included heterosexuality as an inclusion criteria, the participants in the current study were exclusively heterosexual. However, it is important to note that this criteria was incidental (and otherwise irrelevant) for the research questions addressed in the current study. The participants were asked to come to the laboratory on two separate occasions (late FP and mid-LP) to complete an intertemporal choice task. The tasks were the same the two testing times, but we did not state this explicitly to the participants.

We used the backward counting method to predict each participant’s next menstrual onset, late FP (14–16 days prior to the predicted menstrual onset), and mid-LP (6–8 days prior to the predicted menstrual onset). This method has been successfully used to predict other effects of theoretical interest (Durante et al., 2011; Zhuang and Wang, 2014). If a woman’s next predicted menstrual onset was 8–14 days away, she was scheduled to complete the mid-LP testing first (N = 10); otherwise, she was scheduled to complete the late FP testing first (N = 14).

All participants had normal or corrected-to-normal vision. No participants reported a history of a psychiatric disorder or current use of a psychoactive medication. The protocol was reviewed and approved by the Ethics Committee of the local University and the study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants, and they were compensated 50 RMB per hour.

Behavioral Task and Procedure

Prior to scanning, the subjects were presented with a 9-rung ladder scale. They were instructed to place themselves on the ladder that ranged from 1 (lowest rung) to 9 (highest rung) based on where they stand compared to others in terms of family economic level and social status level, respectively (Adler et al., 1994).

The intertemporal binary choice paradigm (see Supplementary Material) was performed as described previously (McClure et al., 2004). In each pair of choices, the sooner option always had a lower reward value than the delayed option. The two options were separated by either 2 weeks or 1 month, with wait times ranging from the day of the experiment to 6 weeks later. Thus, the sooner option was sometimes available immediately and sometimes available after a delay (see Figure 2). The reward amount ranged from 31.25 to 250.00 RMB (USD equivalent, $5–40; conversion rate 1:6.25). In each experiment, the participant made 48 binary hypothetical choices, and the order of the choices was randomized within and across the participants.

FIGURE 2

Figure 2. Examples of sooner reward and delayed reward options. During fMRI scanning, in each trial, the participants viewed a choice screen (above), made a self-paced choice response, and subsequently viewed a choice outcome screen (below) for 2 s. The reward options were in Simplified Chinese and read “ www.frontiersin.org ” (now), 2 (2 weeks later), 4 (4 weeks later), 6 (6 weeks later).

The behavioral experiment was conducted entirely inside an fMRI scanner. The participants were instructed to make a series of choices between a smaller, sooner reward (r1 available at delay t1) and a later, larger reward (r2 available at delay t2; where r1 < r2 and t1 < t2). Each participant was instructed to indicate her preference as soon as the choice was displayed by pressing one of two buttons that corresponded with the location of the preferred option with her right hand. The smaller, sooner options were always presented on the left. Decisions were self-paced with a maximum allowed reaction time of 15 s. After each choice, a feedback screen was presented for 2 s to indicate the choice outcome (McClure et al., 2004). All responses were submitted well before the 15-s time limit. Each subject completed two 24-trial runs. Each trial was followed by a jitter interval (1500 ms, 50% of trials; 4000 ms, 25%; 6000 ms, 16.7%; and 12,000 ms, 8.3%). Prior to being presented with the choices, the participants were administered two control questions to familiarize themselves with the nature of the task. The entire experiment took ∼ 8 min. The stimuli were presented via the in vivo Esys system for fMRI (Gainesville, FL, United States).

Imaging Data Acquisition

Scanning was conducted on a 3-T Siemens scanner at the fMRI laboratory of East China Normal University in Shanghai. Functional images were acquired with a gradient echo-planar imaging sequence, 2000-ms repetition time (TR), 30-ms echo time (TE), 220-mm field of view (FOV), 3 mm × 3 mm × 3.5 mm voxel size and 32 slices. 112 images per scan were acquired and a total was 224 across the two scans. The first five TRs acquired were discarded to allow for T1 equilibration. Prior to the fMRI scanning, a high-resolution structural image was acquired with a T1-weighted, multiplanar reconstruction sequence (TR = 1900 ms, TE = 3.42 ms, FOV = 256 mm, 1 mm × 1 mm × 1 mm voxel size, and 192 slices).

Behavioral Data Analysis

We employed a calculation method similar to that used by Kim et al. (2012), in which the behavioral data were fitted assuming a hyperbolic discount function. The discounted value function is given by

\begin{matrix} V (r, t) = \frac{r}{1 + k t} \end{matrix}

where r is the reward amount available at delay t, and V is the subjective value of the offer. The discount rate (k) for each subject was estimated by assuming a logistic decision function and maximizing the log-likelihood of the observed choices. Best-fit model parameters were determined in Matlab¹ using a simplex search algorithm with 100 random initial parameter values. Subsequently, the k values were transformed logarithmically to permit parametric statistical analyses. For each participants, we calculated an immediate choice ratio (ICR; the number of sooner choices divided by the total number of choices). Because arcsine-square root transformed ICRs correlate well with the hyperbolic discounting rate variable k (Boettiger et al., 2009; Xu et al., 2009), we subjected the ICRs to arcsine-square root transformation.

Logarithmically transformed k values were compared between the late FP and mid-LP conditions with paired samples t-tests. We also conducted a correlation analysis between the transformed k and ICR values. The paired samples t-test was used to test for an order effect on the transformed k across the two menstrual phase conditions.

Imaging Data Analysis

The fMRI data were analyzed in Statistical Parametric Mapping 8 (SPM8, Wellcome Department of Cognitive Neurology, London, United Kingdom). We performed a slice-timing correction and, subsequently, aligned the data to correct for head movement. Images were smoothed with an 8-mm full-width at the half-maximum Gaussian kernel and then normalized to the Montreal Neurological Institute template and re-sampled at 3 mm × 3 mm × 3 mm resolution.

General linear modeling (GLM) was conducted in SPM8. For the first level GLM analysis, we used an event-related design to estimate neural responses to events of interest. Potentially confounding variables, such as trial-by-trial head movements and choice outcomes (i.e., motor responses), were included in the GLM as regressors of no interest. As suggested by the as soon as possible model (Kable and Glimcher, 2010), we expected participants to overweight the value of the soonest available reward regardless of whether it was offered immediately or after a delay. We pooled all smaller, sooner rewards (ignoring whether they were available immediately or at a delayed time) chosen by the participants in a pair together to form the sooner reward choice condition. The remaining larger, later rewards chosen by the participants were pooled together as the delayed choice condition. Each condition [sooner reward choice in the late FP (SFP), sooner reward choice in the mid-LP (SLP), delayed reward choice in the late FP (DFP), and delayed reward choice in the mid-LP (DLP)] were modeled as reaction times from the decision onset. In the first-level analysis, simple main effects were computed for each participant for each of the above mentioned conditions by applying a ‘1 0’ contrast, where 1 represents one of these conditions, and 0 represents all other possibilities.

For the second (group) level analysis, we conducted random effect modeling (flexible factorial design) to analyze the first-level individual contrast images. The main effect of menstrual phase was calculated by comparing late FP trials versus mid-LP trials. The main effect of delay discounting was obtained by comparing sooner-reward choice trials versus delayed-reward choice trials. The interaction between menstrual phase and delay discounting was calculated to extract brain regions that showed higher or lower sensitivity to the sooner reward outcome over the delayed reward outcome among women in the late FP versus women in the mid-LP.

All data were initially thresholded at a value of p < 0.001 (uncorrected), and the results were reported at a cluster statistical threshold of a family-wise error (FWE)-corrected p < 0.05. Activations were localized with the anatomy toolbox in SPM8 (Eickhoff et al., 2005) using the MRIcron automated anatomical labeling template (Tzourio-Mazoyer et al., 2002).

Region of interest (ROI) analyses

Region of interest analyses were conducted to further refine our hypothesis. ROIs were selected based on our hypotheses and previous studies (Cummings et al., 2014; Franklin et al., 2019). Based on prior relevant work (Pine et al., 2010), the following Montreal Neurological Institute (MNI) coordinates for bilateral ROIs were defined: putamen, MNI coordinates 30 −3 −12 and −24 12 −9; caudate, MNI coordinates 21 24 −3 and −9 12 9; and dlPFC, MNI coordinates 42 18 27 and −45 33 18. We extracted an average beta value for each ROI in each condition (sooner reward choice and delayed reward choice in each menstrual phase) for each participant by selecting a 6-mm sphere around the coordinates using the MarsBaR ROI toolbox 0.44 in SPM8 (Brett et al., 2002). All beta values were submitted to a 2 (delay discounting: sooner vs. delayed reward choice) × 2 (menstrual phase: FP vs. LP) × 6 (the aforementioned ROIs) mixed measures analysis of variance (ANOVA). We also completed a 2 (delay discounting: sooner vs. delayed reward choice) × 2 (menstrual phase: FP vs. LP) mixed measures ANOVA for each ROI. Mean beta values are reported with SDs.

Functional connectivity analysis

To test our neural circuitry hypothesis, we conducted beta series correlation (BSC) analyses of the functional connectivity between the DS (putamen and caudate) and dlPFC within each condition (SFP, SLP, DFP, and DLP). First, based on the above-defined ROIs, each trial was modeled as a separate event of interest (Rissman et al., 2004) and the beta series associated with each trial type within each ROI were extracted and sorted by study condition. Pair-wise BSC analyses were performed for the DS and the dlPFC. After calculating the correlation between activities in each ROI pair individually for each subject and for each condition across the time series, the correlation values obtained were subjected to Fisher transformation prior to being subjected to ANOVAs designed to detect which correlations varied across delay discounting choice and menstrual phase.

Correlations between DS-dlPFC functional connectivity and discount rate

We conducted correlation analyses between discount rate (transformed k) and functional connectivity (as indexed via BSCs) within each condition to investigate the effect of DS-dlPFC circuitry on impulsivity.

Results

Demographics

A total of 8 participants were excluded from further analysis, including 3 for excessive head motion (>2-mm displacement in the x, y, or z dimension, or >2° angular shift) during scanning, 2 because tracking of their next menstruation revealed that we did not have an accurate menstrual date, and 3 whose behavioral data did not conform to the model from the log-likelihood of observed choices. The 16 remaining participants (mean age 22.44 ± 1.31 years; range, 20–24) included 8 participants who were initially scanned during their late FP, and 8 who were initially scanned during their mid-LP. The mean scores of the 16 participants in the final analyses had mean family economic and social status level scores of 5.81 ± 0.40 and 5.44 ± 0.51, respectively. There was very high agreement among the participants on subjective socioeconomic status [intra-class correlation (1,15) = 0.788], which suggests that behavioral effects observed in the experiment cannot be attributed to a confounding effect of subjective value variations.

Behavior

A paired samples t-test did not show an effect of testing order (i.e., which phase women were in during first test) on transformed k values [t₂,₁₄ = −0.97, p = 0.35, d = −0.52]. Meanwhile, a paired samples t-test showed a significant difference in transformed k values between the late FP and mid-LP (t₁,₁₅ = −2.14, p = 0.049 < 0.05, d = 0.43), with a significantly greater mean discount rate being observed in the late FP (−1.32 ± 0.49) than in the mid-LP (−1.51 ± 0.46) (Figure 1). Transformed k and ICR values were highly correlated within each menstrual phase (FP: r = 0.94, p < 0.001; LP: r = 0.96, p < 0.001) (see Table 1).

TABLE 1

Table 1. Means and correlations of k and ICR values within menstrual phase datasets.

Whole Brain Analysis

We observed main effects of menstrual phase on activity in several visual areas, including greater activation during the mid-LP than during the late FP in the bilateral lingual gyrus, bilateral calcarine gyrus, left middle occipital gyrus, and left inferior occipital gyrus. No other meaningful brain areas were found in the opposite contrast. With respect to delay discounting, brain regions that were preferentially activated by the prospect of a sooner reward choice over a delayed reward choice (sooner > delayed) included the right putamen, the right thalamus, and the right supplementary motor area. In the reverse contrast (delayed > sooner), greater activation was identified in the left postcentral gyrus and left superior parietal lobule. Analysis of interactions between menstrual phase and delay discounting revealed more active regions in the left putamen, bilateral caudate, bilateral visual areas [Brodmann area (BA)17 and 18], left hippocampus, and left insula in the (late FP – mid-LP) - (sooner – delayed) comparison; there was no regions that were significantly more active in the reverse comparison (Table 2).

TABLE 2

Table 2. Brain regions whose activity is altered in relation to menstrual phase or delay discounting.

ROI Analyses

A 2 (delay discounting) × 2 (menstrual phase) × 6 (ROI) mixed measures ANOVA revealed a significant main effect of ROI (F₅,₉₀ = 9.49, p < 0.001, $η_{p}^{2}$ = 0.35) and a significant interaction of delay discounting and menstrual phase (F₁,₉₀ = 8.97, p = 0.004, $η_{p}^{2}$ = 0.09) (Figure 3; non-significant effect data are reported in the Supplementary Material). Regarding the ROI effect, post hoc analysis indicated that the beta value for the left dlPFC was significantly greater than the values obtained for all other ROIs (p_s ≤ 0.005), while the beta value for the right dlPFC was significantly greater than the values obtained for the right putamen (p = 0.003) and right caudate (p = 0.010). Simple effects analysis of the delay discounting × menstrual phase interaction showed that when choosing the sooner rewards, beta values differed significantly between the late FP and mid-LP (F₁,₉₀ = 6.61, p = 0.010, $η_{p}^{2}$ = 0.07), with a greater mean beta value being observed in the late FP (0.98 ± 0.17) than in the mid-LP (0.58 ± 0.17). Specifically, for the left putamen, we found a significant main effect of menstrual phase (F₁,₁₅ = 5.1, p = 0.040, $η_{p}^{2}$ = 0.25), wherein women in the late FP (0.99 ± 1.22) had higher mean beta value than in the mid-LP (0.30 ± 0.77). There was a significant delay discounting × menstrual phase interaction for the right putamen (F₁,₁₅ = 5.02, p = 0.040, $η_{p}^{2}$ = 0.25). Simple effect analysis showed that the mean beta value was significantly higher when women chose sooner rewards in the late FP (0.02 ± 1.12) than when they did so in the mid-LP (−0.76 ± 0.72; F₁,₁₅ = 7.96, p = 0.010, $η_{p}^{2}$ = 0.35). No other effects were significant (Figure 3).

FIGURE 3

Figure 3. Mean beta values for ROIs by condition. (A) The ROIs of the left putamen [–24, 12, –9], right putamen [30, –3, –12], left caudate [–9, 12, 9], right caudate [21, 24, –3], left dlPFC [–45, 33, 18], and right dlPFC [42, 18, 27]. (B) Activation of the left putamen was significantly higher during the late FP than during the mid-LP. Moreover, when choosing sooner rewards, the mean beta values of the right putamen and the right caudate were much greater during the late FP than during the mid-LP. A significant interaction was observed for the right dlPFC; simple effects analysis showed a tendency for right dlPFC activation to be higher in the DLP condition than in the DFP, SLP, and SFP conditions. ^∗The effect is significant at the 0.05 level.

Similarly, we found a significant delay discounting × menstrual phase interaction for the right caudate (F₁,₁₅ = 6.86, p = 0.020, $η_{p}^{2}$ = 0.31). Simple effect analysis showed that the mean beta value was significantly higher when women chose the sooner rewards in the late FP (0.27 ± 0.77) than when they did so in the mid-LP (−0.19 ± 0.80; F₁,₁₅ = 8.65, p = 0.010, $η_{p}^{2}$ = 0.37). We did not find any significant effects for the left caudate (Figure 3).

For the right dlPFC, there was a significant delay discounting × menstrual phase interaction (F₁,₁₅ = 5.90, p = 0.030, $η_{p}^{2}$ = 0.28), but no significant effects were observed in the simple effects analysis. However, we did observe a tendency showing that the mean activation level in the DLP condition (1.67 ± 2.86) was higher than in the other conditions (SLP, 1.02 ± 2.28; SFP, 1.41 ± 1.91; DFP, 1.38 ± 2.34). No significant main effects (delay discounting, F₁,₁₅ = 1.32, p = 0.270, $η_{p}^{2}$ = 0.81; menstrual phase F₁,₁₅ = 0.53, p = 0.480, $η_{p}^{2}$ = 0.34) and no significant delay discounting × menstrual phase interaction (F₁,₁₅ = 0.33, p = 0.570, $η_{p}^{2}$ = 0.02) were observed for the left dlPFC (Figure 3).

The results above indicated that the right DS was more active during the late FP than during the mid-LP when sooner rewards were chosen (i.e., SFP > SLP for right DS). In contrast, the dlPFC showed a tendency to be more active during the mid-LP than during the late FP when delayed rewards were chosen (i.e., DLP > DFP for the dlPFC).

Functional Connectivity Across Menstrual Phases

An ANOVA revealed a significant main effect of menstrual phase on the functional connectivity between the left putamen and left dlPFC (F₁,₁₅ = 7.40, p = 0.020, $η_{p}^{2}$ = 0.33). The magnitude of this connectivity was significantly stronger in the mid-LP (0.22 ± 0.32) than in the late FP (0.00 ± 0.42). There were no other significant ROI functional connectivity effects.

Correlations Between DS-dlPFC Functional Connectivity and Discount Rate Across the Menstrual Phases

In the SFP condition, functional connectivity between the right caudate and the bilateral dlPFC correlated inversely with discount rate (right caudate-right dlPFC: r = −0.64, p = 0.010; right caudate-left dlPFC: r = −0.64, p = 0.010) (Figure 4). Conversely, in the SLP condition, functional connectivity between the right dlPFC and the right putamen correlated directly with discount rate (r = 0.50, p = 0.050).

FIGURE 4

Figure 4. Correlations between DS-dlPFC functional connectivity and discount rate across menstrual phases. (A) ROIs: right putamen [30, –3, –12], right caudate [21, 24, –3], right dlPFC [42, 18, 27] and left dlPFC [–45, 33, 18]. (B) Discount rate correlated positively with right putamen-right dlPFC functional connectivity in the SLP condition. Discount rate correlated negatively with right caudate-bilateral dlPFC functional connectivity in the SFP condition.

Discussion

Using the delay discounting task, which assesses intertemporal choice similar to the task used by McClure et al. (2004), we demonstrated that delay discounting behavior in women was affected by menstrual phase, such that the discount rate was significantly greater in the late FP than in the mid-LP, and this augmented discount rate was associated with enhanced activity in the DS. Specifically, greater activation was observed in the left putamen during the late FP than during the mid-LP. Moreover, women showed greater activation in the right DS (putamen and caudate) when choosing sooner rewards in the late FP than when do so in the mid-LP (i.e., SFP > SLP), indicating that the right DS is more responsive to immediate rewards during the late FP than during the mid-LP. These results are consistent with the prior findings showing that a heightened drug cue-responsivity in women is associated with enhanced activity in the DS, especially in the putamen (Cummings et al., 2014; Franklin et al., 2019). Indeed, behavioral impulsivity has been linked to dopamine levels in the putamen. For example, in adult men, higher trait impulsivity as measured by Barratt Impulsiveness Scale (BIS) scores correlated negatively with dopamine transporter availability in the putamen (Costa et al., 2013). Additionally, a lower dopamine synthesis capacity in the putamen, indexed by 6-[¹⁸ F]fluoro-L-m-tyrosine signal, was shown to be predictive of an elevated Now bias and a reduced willingness to accept low-interest rate delayed rewards (Smith et al., 2016).

The VS/nucleus accumbens have been strongly implicated in the seeking the sooner rewards in intertemporal choice paradigms (McClure et al., 2004; Kim et al., 2012), with nucleus accumbens activation in particular being related to reward magnitude sensitivity (Ballard and Knutson, 2009). Our not finding an effect of menstrual phase on these areas suggests that VS reward sensitivity may not be modulated by menstrual phase. Indeed, intertemporal choice in humans has been shown to vary with region-specific dopamine processing, with regionally distinct associations with sensitivity to delay (the putamen) and reward magnitude (the VS) (Ballard and Knutson, 2009; Smith et al., 2016). Studies have linked dopamine availability in the putamen with time perception, a cognitive process thought to contribute to discounting of delayed rewards (Wittmann and Paulus, 2008; Takahashi, 2011; Smith et al., 2016). For example, Parkinson disease patients, who have deficits in putamen dopamine signaling, show selective impairments in time duration comparison (Dormal et al., 2012). A pharmaco-fMRI study related the relationship between dopamine depletion and time perception specifically to activity in the putamen (Coull et al., 2012). Our finding the putamen was more active in the SFP condition than in the SLP condition suggests that naturally cycled ovarian steroid hormones may modulate an aspect of time perception in intertemporal choice, perhaps through steroid hormone modulation of dopamine levels. However, proving such a modulatory effect would require dopamine level data. Future research should compare dopamine levels across menstrual phases to help elucidate the influence of steroid hormones on intertemporal choices.

Although we observed a significant menstrual phase × delay discounting interaction influence on right dlPFC activity, simple effect analysis revealed only a trend toward higher activation in the DLP condition relative the other conditions. Post hoc analysis of ROIs indicated that beta values were significantly higher for the dlPFC than for the DS. Because the striatum and prefrontal cortex are intermodulated via frontostriatal networks (Yuan et al., 2016), impulsivity variance across menstrual phases may be determined, at least in part, by relative activity levels between the DS and dlPFC. Although the activity in the dlPFC was stable across menstrual phases, heightened DS activity during sooner reward selection in women late FP may lead to relatively less potent top-down cognitive control. Conversely, lesser DS activity during the mid-LP may enable more potent cognitive control. Thus, our results suggest that the activities of the DS and dlPFC relative to each other may determine the level of cognitive control over impulsivity, such that there is stronger cognitive control during the mid-LP than during the late FP.

The present results also support the notion that DS-dlPFC circuitry is involved in the regulation of impulsivity. Specifically, when participants chose sooner rewards during the late FP, right caudate-dlPFC functional connectivity correlated negatively with discount rate, indicating a linear relationship between the reduced right caudate-dlPFC connectivity and increased impulsivity (indexed by discount rate) during the late FP, consistent with rs-functional connectivity findings in participants with internet gaming disorder (Yuan et al., 2016) and cigarette-dependent women (Wetherill et al., 2016). Conversely, in the SLP condition, right putamen-right dlPFC functional connectivity correlated directly with discount rate. The hierarchical reinforcement learning theory (Holroyd and Yeung, 2012) has suggested the dlPFC and motor structures in the DS (mainly the putamen) execute options chosen by other brain regions. If so, the aforementioned correlation may reflect an effect of execution function.

Anatomically, the DS includes the dorsal regions of the putamen and caudate nucleus (Porter et al., 2014). Although these two subregions both receive nigrostriatal dopaminergic projections (Fallon and Moore, 1978) and are involved in motivated behavior via the prefrontal cortex, functionally, the caudate is important for reward-related cognition whereas the putamen is more involved in motor behaviors (Baskerville and Douglas, 2010). Thus, the opposing directionality found in our results may reflect different functions of the DS-dlPFC circuitry.

Our results suggest that DS-dlPFC functional connectivity may modulate impulsivity in intertemporal choices, with opposing directionality and differential involvement of brain regions depending upon menstrual phase.

This study had notable limitations. First, although we hypothesized that, brain regions related to cognitive control would be more active during the mid LP, we did not find a significant main effect of menstrual phase on the activation of brain regions related to cognitive control. This negative finding could be related to inherent characteristics of cognitive involvement in the task (McClure et al., 2004), which could confound the effect of menstrual phase. Second, menstrual phase was not confirmed by any biological tests such as hormonal assays, potentially reducing phase designation accuracy. Previously studies have found that estimates based on the backward counting method, as used here, are correct 80–90% of the time (Carroll, 2018). Indeed, several participants were excluded after follow-up revealed inaccurate estimates. Third, the delay discounting task is hypothetical. Although previous studies have shown no significant differences behaviorally or neurologically between hypothetical and real reward outcomes (Johnson and Bickel, 2002; Madden et al., 2004; Lagorio and Madden, 2005; Bickel et al., 2009; Kim et al., 2012), it is possible that an extraordinarily impulsive person might much more easily forego a tantalizing immediate reward for a delayed larger reward in a hypothetical situation than in a real situation (Reynolds et al., 2006; Rosati et al., 2007; Jimura et al., 2009; Smits et al., 2013; Carroll, 2018). Fourth, the sample size of the present study was small, which weakens the strength of our results. Finally, that the smaller, sooner options were always presented on the left during scanning would influence our results on hemispheric asymmetry.

Use of the bilateral dlPFC as a seed can be a valuable tool for exploring the function of cognitive control networks in rs-fMRI studies (Cieslik et al., 2013; Hwang et al., 2015; Tao et al., 2017). Thus, to further probe whether there is increased cognitive control functioning during the mid-LP, relative to the late FP, and the involvement of DS-dlPFC circuitry in impulsivity regulation, we conducted a rs-fMRI experiment in the following accompanying study with a inclusion of hormonal assay confirmed menstrual phases.