This study was approved by the South China Normal University and Guangzhou University Institutional Ethics Review Board (2017–139). Human participants provided written informed consent prior to participation. A total of eighty right-handed (Li et al., 2017; Wasylyshyn et al., 2018) university female students (ages 18–25) who came to the recruitment were enrolled in this study which included 38 PDM and 42 non-PDM controls. PDM and control subjects were matched according to gynecological age. Demographic and clinical information are shown in Table 1. College students with dysmenorrhea (DM) were selected and PDM was further diagnosed in Sun Yat-sen Memorial Hospital, where magnetic resonance imaging (MRI) was performed to ensure there was no macroscopic structural abnormality inside or outside the uterus. Briefly, the diagnostic criteria for PDM were similar to those defined by the American College of Obstetricians & Gynecologists (American College of Obstetricians and Gynecologists, 2006). The following inclusion criteria were used for PDM participants: (1) a menstrual cycle of average 30 days; (2) a history of menstrual pain over more than 12 months; (3) a self-assessed severity of the average menstrual pain of 5 and above over the previous 6 months based on the visual analog scale (VAS, 0 = not at all, 10 = the worst imaginable pain); and (4) a pelvis MRI scan did not show any anatomical pelvic disease. The inclusion criteria for the controls were similar to those for the PDM subjects except that the controls had a self-assessed VAS of 0. Exclusion criteria included pregnancy, organic pelvic disease, alcohol or drug abuse, failure of MRI scans due to metal or pacemaker implants, and formal diagnosis of psychiatric conditions. Urinary luteinizing hormone tests were performed to verify experimentally whether the participants were in their periovulatory phase (i.e., days 12–16 of the menstrual cycle), which is the phase when influence of chronic PDM was usually evaluated (Wei et al., 2016; Liu J. et al., 2017; Liu et al., 2018).

The scheme of the experiment is presented in Figure 2A. After arriving at the lab, participants were asked to complete a detailed consent form. Psychological questionnaires (including the Positive and Negative Affect Scale, PANAS (Watson et al., 1988), and the Basic Needs Questionnaire, BNQ) (Bernstein and Claypool, 2012) were processed to obtain the baseline emotional state. Next, participants with PDM underwent abdominal and pelvic cavity, uterus and accessory MRI scans, and clinicians made the final diagnosis on whether the subjects were PDM patients. Participants who met the inclusion criteria then read the instructions outside the MRI room, and the researcher orally interpreted the instructions to the participants, indicating the duration of the experiment, the requirement to keep the head fixed and to warn about the noise of the machine and related equipment. Then, groups of three participants (including one real participant in the study and two fixed female ‘actors’) engaged in a 10-min group interaction session, after which participants were told to begin a Cyberball game, a paradigm based on a virtual ball-tossing game, where participants believe they are playing with other real participants, although in fact these are computer-generated (Eisenberger et al., 2003). Blood oxygenation level-dependent (BOLD) signal changes were recorded during the Cyberball task.

In the scanner, participants saw an animated ball-tossing game, with an icon representing their own hand at the bottom and the two other players depicted as animated icons in the upper corners. The names of the group members were shown next to each icon and participants could throw the ball to whoever they liked. Pressing “1” delivers the ball to the member in the upper left corner, while pressing “2” throws the ball to the member in the upper right corner. Participants were instructed to throw the ball within 2 s of receiving it. If the time exceeds 2 s, the system will temporarily transfer the ball randomly. The computer players waited 0.5–2.0 s before making a throw to heighten the sense that the participant was actually playing with other individuals. Each participant participated in three rounds of the Cyberball game during three fMRI scans. Each round of the Cyberball program consisted of 60 throws (including participant and computer players’ throws) and lasted 3 min. The three rounds of Cyberball (Figure 2B) were as following, (1) Cyberball observation, implicit social exclusion (ISE), where participants were told that the intranet connection was not effective yet because of technical problems, but that they could watch the other participants playing; (2) Cyberball inclusion (INCL), where participants were told they were connected and played with the other players, participant and other players were equally likely to throw the ball; and (3) Cyberball exclusion, explicit social exclusion (ESE), where participants received three throws and were then excluded from the game (i.e., the other players started playing exclusively together, and the real participant never received the ball again) (Eisenberger et al., 2003).

Participants and experimenters were in a double-blind state, namely, participants were told that the purpose of the study was to examine the effect of imagination on mission performance; the experimenter did not know who among the participants was a PDM subject. Immediately following the scanning session, participants completed the PANAS and BNQ questionnaires again. At the end of the experiment, participants were asked whether they believed they were playing the Cyberball game with the other two participants (i.e., the ‘actors’ they met initially). Subjects who failed to initiate social exclusion were excluded according to their answers (for example, if they did not believe that they were playing a real game of pitching, but thought it was a pre-set experimental procedure) and the PANAS and BNQ scales scores before and after the Cyberball game were compared (Eisenberger et al., 2006). In this study, both PDM subjects and controls showed a significantly higher post-BNQ score than pre-BNQ score, indicating that the participants noticed the exclusion and felt excluded (Table 1). Finally, each participant received 70 yuan (RMB) as compensation and was thoroughly debriefed about the purpose of the study.

Magnetic resonance (MR) data were acquired on a 3.0 Tesla clinical scanner (Achieva TX; Philips Healthcare, Best, Netherlands) with an 8-channel head coil in Sun Yat-sen Memorial Hospital, Sun Yat-sen University. We performed resting-state fMRI (rs-fMRI) before task fMRI (T-fMRI) using a T2^∗-weighted fast-field echo-planar imaging (FFE-EPI) sequence (rs-fMRI/T-fMRI TR = 2000 ms/3000 ms, TE = 30 ms, FA = 90°, FOV = 240 mm × 240 mm, acquisition matrix = 64 × 64, thickness = 4.0 mm, 33 transverse slices covering the whole brain; 240/60 volumes were obtained for rs-fMRI/each round of T-fMRI). High-resolution structural images were collected using a T1-weighted 3D FFE sequence (TR = 8.2 ms, TE = 3.7 ms, FA = 8°, FOV = 256 × 256, acquisition matrix = 256 × 256, thickness = 1 mm; 168 sagittal slices covered the whole brain).

Rs-fMRI data were preprocessed using SPM 12¹ and the DPABI v3.1 toolbox² in MATLAB. For each subject, the first 10 functional images were discarded to reach magnetization equilibrium and to allow adaptation to the MR environment. Then, a slice-timing correction was conducted by setting the middle slice (17th) as the reference. Realignment was performed to estimate head motion, and two PDM subjects were excluded due to excessive head motion (translation more than 2 mm, or rotation more than 2°). We also calculated the mean frame-wise displacement based on Jenkinson’s model (FD-Jenkinson), and ensured there was no significant group effect on the FD-Jenkinson (Jenkinson et al., 2002). High-resolution structural images were co-registered into functional images and segmented into white matter, gray matter and cerebrospinal fluid. Then, we spatially normalized the functional images to the individual structural image in standard MNI-152 standard space with a resampled voxel size of 3 mm × 3 mm × 3 mm; one PDM and four healthy subjects were excluded due to bad normalization (such as, no alignment between functional image and MNI-152 template; functional signal loss of normalization map). Furthermore, nuisance covariates (Friston 24 head motion parameters, mean white matter and mean cerebrospinal fluid) were regressed out to reduce the effect of complex noise. Finally, we further conducted spatial smoothing with a Gaussian kernel of 4 mm full-width at half maximum (FWHM) and performed band-pass filtering (0.01–0.1 Hz) to reduce high-frequency physiological noise. Finally, 35 PDM subjects and 38 controls were included in the further rs-fMRI analysis.

In the T-fMRI, 34 PDM subjects and 36 controls completed the three rounds of Cyberball. For each round of this task, the first five volumes were removed to control for interference between rounds. Other T-fMRI preprocessing steps, including slice-timing, realignment, co-registration, normalization and spatial smoothing, were conducted as for rs-fMRI preprocessing. A high-pass filter (cutoff 128 s) was used to remove low-frequency noise. After quality control, nine subjects with excessive head motion or bad normalization were excluded, leaving 30 PDM subjects and 31 healthy controls for further analysis.

Since static FC assumes that brain connectivity is temporally stationary (Fox and Raichle, 2007), we calculated the correlation coefficient between ACC and HIPP over the whole scan time to reflect the overall connections. The bilateral ACC and HIPP were selected as regions of interest (ROIs) for ROI-based FC based on the Automated Anatomical Labeling atlas (AAL) (Rolls et al., 2015), which is widely used for human brain imaging analysis. The specific MNI locations of the ROIs (ACC and HIPP) are viewed on the ICBM152 human brain surface (Mazziotta et al., 2001; Figure 1A). For each participant, we extracted the time courses of each ROI from preprocessed images. Then, the Pearson’s correlation coefficients (r) between ACC and HIPP were computed. To improve the normality of group analysis, Fisher’s z-transformation was performed to convert the r-value into a z-value.

Dynamic FC between ACC and HIPP was calculated using a sliding time-window approach. We fixed the length of the time-window at 20 TRs (40 s) and the sliding step to 1 TR: 211 FC matrices between ACC and HIPP were obtained for each subject [time points: TP = 230 TRs; length of time-window (L) = 20 TRs; sliding step = 1 TR; dynamic FC matrices: T = TP – L + 1 = 211] (Li et al., 2014). The standard deviation of FC across all 211 slide-window FC matrices was calculated as FC variability, which reflects the discreteness of FC.

Whole-brain general linear model (GLM) analysis was performed using standard hemodynamic response function to identify brain activation during the Cyberball task. Each round of Cyberball (ISE, INCL, ESE) was modeled using a block design. The contrasts of interest – implicit exclusion compared with inclusion (ISE-INCL), explicit exclusion compared with implicit exclusion (ESE-ISE), and explicit exclusion compared with inclusion (ESE-INCL) – were then computed to depict social exclusion resulting from peer rejection. The GLM parameter maps of the contrasts were generated for each participant. Significant differences in brain activation were evaluated by performing a group comparison of the GLM-parameter maps.

Psychophysiological interaction (PPI) analysis (Friston et al., 1997) was conducted to determine which cerebral regions were functionally connected with the ROI for each of the ISE-INCL, ESE-ISE, ESE-INCL contrasts separately. Prior studies suggested that the ACC is activated by social exclusion (Bolling et al., 2011; Masten et al., 2011) and therefore the ACC (ACC.L/ACC.R/combined bilateral ACC) was chosen to initiate the PPI analysis. The deconvolved activity time-series of the left ACC, right ACC, and combined bilateral ACC were extracted and adjusted for effects of interest (ISE-INCL, ESE-ISE, ESE-INCL), and the PPI term was created using the ROI eigenvariate and the specific task contrasts (ISE-INCL, ESE-ISE, ESE-INCL). Finally, a second GLM was performed with a condition-specific regressor probing each contrast (ISE-INCL, ESE-ISE, ESE-INCL) to obtain the PPI parameter map, allowing the identification of ACC-connectivity changes for social exclusion during Cyberball. To further clarify the connectivity between ACC and HIPP under social exclusion conditions, the mean PPI parameter of HIPP was extracted for the group comparison.

The T1-weighted structural images were preprocessed with the VBM toolbox in the SPM12. The GM volume of each subject was evaluated using voxel-based morphology (VBM) analysis. First, the structural image was manually reoriented to MNI space and centered on the anterior commissure to facilitate the following segmentation step. The reoriented image was then segmented into GM and white matter using the unified segmentation approach. Spatial smoothing with 4 mm FWHM was performed on the warped and modulated GM map to improve the spatial resolution. Finally, we restricted our search regions to the ACC and HIPP, rather than the whole brain. To achieve this, we extracted the GM signal of the ACC and HIPP for each subject.

Female Sprague-Dawley (SD) rats were purchased from the Laboratory Animal Center of Southern Medical University (Guangzhou, China) and kept at the School of Life Sciences, South China Normal University, with controlled humidity and temperature, and a 12 h (6:30 AM to 6:30 PM) light–dark cycle. Rats involved in experiments were used according to international and university ethical standards. Food and water were available ad libitum. Animals weighing 250–300 g (120–150 days old) were given on average 7 days to adjust to the new environment prior to the experiments. Pelvic pain model was generated by intraperitoneal (IP) injection of estradiol benzoate and oxytocin (both Ningbo Hormone Inc., China).

The experimental procedures of this study were approved by the Animal Protection and Use Committee of Guangzhou University and South China Normal University. The chronic PDM rat model was generated by modification of an acute PDM mouse model (Chen et al., 2013; Jesuino et al., 2019). Briefly, estradiol benzoate was injected (IP, two times/week, 4 mg/Kg) for eight consecutive weeks (wks). From the fourth wk through the eighth wk, PDM rat model were injected with oxytocin (IP, one injection/week, 20 IU/Kg/per injection) 24 h after injection of estradiol benzoate. Control rats received injections of an equal volume of estradiol benzoate and saline. Paw withdrawal mechanical threshold (PWMT) (Chaplan et al., 1994; Liu Y. et al., 2017), electrophysiological and biochemical evaluations were conducted in control and PDM rats between 24 and 72 h following the last oxytocin injections.

In vivo dual-site extracellular recordings were conducted as described with a few modifications (Noguchi et al., 2017; Chen et al., 2019). Rats were anesthetized with pentobarbital sodium (IP 80 mg/kg, Sigma, United States) then head-fixed in a stereotaxic apparatus (RWD Life Science, China) with body temperature maintained between 36 and 37°C. When necessary, a supplemental dose of anesthesia was given based on tail reflex. After a midline skin incision was made, two skull holes were drilled above the ACC (2.5 mm anterior to the bregma, 0.4 mm lateral to the midline, 1.7-2.0 mm depth) and the dorsal CA1 subregion of the HIPP (3.6 mm posterior to the bregma, 2.0 mm lateral to the midline, 2.2–2.5 mm depth, 10°) under a stereomicroscope (Sunny Optical Technology, China). Two glass microelectrodes for recording (filled with 0.5 M NaCl, resistance 4–6 MΩ) were slowly inserted until the tips of the electrodes reached the ACC and hippocampal CA1. Each recorded signal was amplified (1,000x) by an electrometer amplifier (Model 3000; A-M Systems, United States) and digitized via a D/A converter (Micro 1401; Cambridge Electronic Design, Ltd., United Kingdom), then sent to data acquisition software (Spike2; Cambridge Electronic Design).

Extracellular recording data were analyzed offline in MATLAB. For processing the local field potential (LFP), a Butterworth low pass filter (300 Hz) was applied to the raw recorded data. Power spectral density was computed using Thomson’s multitaper method for a fast Fourier transformation (FFT) to determine the power for specific frequency bands. Frequency ranges were defined as follows: delta: 1–4 Hz; theta: 4–12 Hz; gamma: 30–100 Hz, of which theta frequency covers a wider range in the rodent (Buzsaki and Draguhn, 2004; Tamura et al., 2017).

We performed synchronization analysis in line with our established methodological protocol (Chen et al., 2019). Simultaneous signals were subjected to cross-correlation estimation to quantitatively evaluate the similarity. The maximal offset was set to ±1 s. After the calculation, spectral coherence between the two LFPs from the ACC and HIPP was analyzed using a FFT number of 2¹², and the values ranged from 0 to 1, meaning non-correlated or completely correlated in the frequency domain.

To further evaluate the synchrony between the oscillations of ACC and HIPP, weighted phase lag index (WPLI) analysis, which is based on the complex conjugate of spectral coherence (Vinck et al., 2011), was conducted to the same dataset. The indices were shown from 0 to 1 as mentioned above.

To access the modulation strength of cross-frequency oscillations, we first derived the instantaneous phase and amplitude from the targeted signals of both areas, then analytically clustered the theta phases binned into 20° intervals with the corresponding gamma amplitude at the same time in the other region. Both directions were analyzed to compare the influence of the theta band on the interregional gamma band.

Acute brain slices containing ACC and HIPP (350 μm) were prepared according to routine procedures (Chen et al., 2017; Luo et al., 2019), from control and PDM rats using a vibratome (VT 1000S, Leica, Germany) in oxygenated ice-cold cutting solution containing (in mM), 119 NaCl, 2.5 KCl, 2.5 CaCl₂,1.3 MgSO₄, 1 NaH₂PO₄, 11 D-glucose, 26.2 NaHCO₃ (pH 7.2-7.4), saturated with 95% O₂/5% CO₂. Slices were kept in artificial cerebrospinal fluid (aCSF) containing (in mM) 140 NaCl, 4.7 KCl, 2.5 CaCl₂, 1.2 MgCl₂, 11 D-glucose, 10 HEPES (pH 7.2-7.4), and gassed with 95% O₂/5% CO₂. Slices were incubated for 1 h at 30–32°C before recording and then transferred to a submerged recording chamber where temperature was held at 32 ± 0.5°C with an automatic temperature controller (TC-324B, Warner Instrument Corporation) with aCSF flow set at 2–3 ml/min.

Pyramidal neurons were identified by their morphology, typically characterized by a triangular-shaped soma, in brain slices (Ramaswamy and Markram, 2015). To record miniature excitatory and inhibitory postsynaptic currents (mEPSCs and mIPSCs) from pyramidal neurons of the ACC and HIPP, voltage was held at –60 and 0 mV, respectively. To block fast sodium channel activity and thus action potential, 1 μM TTX was added to the aCSF (Chen et al., 2017). The pipette was filled with the following internal solution (mM): 100 mM Cs-gluconate, 5 mM CsCl, 10 mM HEPES, 2 mM MgCl₂, 1 mM CaCl₂, 11 mM BAPTA, 4 mM ATP, and 0.4 mM GTP (pH 7.3, adjusted with KOH) at an osmolality of 280–290 mOsm. Data were collected with a MultiClamp 700 B amplifier (Axon Instruments) and filtered during acquisition with a low pass filter set at 2 kHz using pCLAMP10 software (Molecular Devices, United States). The data were analyzed offline using Mini Analysis Program (Synaptosoft Inc., United States).

Rat brains were dissected and ACC and HIPP tissues were removed on ice as previously described (Chen et al., 2017). Tissues were then homogenized in SDS buffer (50 mM Tris pH 7.5, 150 mM NaCl, 5 mM EDTA pH 8.0, 1% SDS). Cellular debris was removed by centrifugation at 4°C (14,000 rpm for 10 min) and the supernatant was collected for analysis. Tissue lysates were subjected to SDS-PAGE, and transferred to nitrocellulose membranes. The membranes were blocked with 5% non-fat dry milk and incubated with specific primary antibodies GluR1 (Abcam, ab31232; dilution 1:1000), GluR2 (Abcam, ab206293; dilution 1:2000), GluR4 (Abcam, ab119995; dilution 1:4000), NMDAR1 (Abcam, ab109182; dilution 1:4000) or NMDAR2B (Abcam, ab65783; dilution 1:4000) overnight at 4°C. GAPDH (Beyotime, AF0006; dilution 1:5000) or anti-β-actin antibody (Sigma, United States; dilution 1:5000) was used as a loading control. After three washes with TBST, HRP-labeled secondary antibody (CWS, China) was added at room temperature for 1 h using 5% milk in TBST followed by three additional washes with TBST. The Immobilon ECL western system (Millipore, United States) was then used to visualize the bands, which were quantified and analyzed with Gel-Pro Analysis software (Media Cybernetics, United States).

Normal distribution was tested in demographics and psychological data. Except for the BNQ score, all data were not distributed normally. Thus, Mann–Whitney tests were used to detect differences in age, age of onset of menstruation, menstrual duration, menstrual cycle, and PANAS between the PDM subjects and healthy controls. Wilcoxon test was used to detect difference in PANAS between pre- and post-tests. Parametric tests (including two-sample t-tests and paired-sample t-tests) were performed to determine differences in BNQ between and within group, respectively. Two-sample t-tests were also performed to detect group differences in the GLM-parameter map and PPI-parameter map, to identify abnormal regional brain activity and abnormal regions functionally connected with the ACC. Multiple comparison corrections were conducted using an AlphaSim correction (both voxel-wise threshold and cluster threshold were set as p < 0.05) and the Gaussian Random Field correction (voxel-wise threshold: p < 0.01; cluster threshold: p < 0.05) separately. A non-parametric permutation test was conducted to identify between-group differences in the static FC, FC variability, GM volume and ACC-HIPP PPI parameters. In the calculation, we took the menstrual cycle as covariate and regressed it out if there was a significant difference in the menstrual cycle between PDM subjects and controls. For LFP analysis, the data were first tested for normal distribution. None of the datasets were distributed normally, therefore a non-parametric test was used for two-group comparisons. Student’s t-test was used in two-group comparisons of western blotting results. For comparisons of multiple groups, one-way ANOVA or two-way ANOVA with post hoc tests were used. Data are shown as mean ± SEM unless otherwise stated. Statistical significance threshold was set at p < 0.05.

For the rs-fMRI and structural MRI, no significant differences were found for the age, age of onset of menstruation or menstrual duration between the PDM subjects and controls, while a longer menstrual cycle was found in the PDM subjects. For the T-fMRI, there were no significant between-group differences for age, age of onset of menstruation, menstrual duration, menstrual cycle, PANAS, and BNQ. Both groups showed no significant difference in PANAS between pre- and post-test values, and a significantly higher post-BNQ score than pre-BNQ score. The specific values are presented in Table 1.

To examine the synchronization of blood oxygenation level dependent (BOLD) signals between the ACC and HIPP, we first evaluated FC, utilizing rs-fMRI, for both static and dynamic FC in 35 PDM women with 5.14 ± 2.13 years dysmenorrhea and 38 age-matched controls (Table 1). Static FC increased significantly in PDM subjects between left/right ACC and left/right HIPP (ACC.L-HIPP.L, ACC.L-HIPP.R, ACC.R-HIPP.L, ACC.R-HIPP.R) compared to controls (p < 0.05; Figure 1B), indicating that experiencing 5 years of PDM changes brain FC between the ACC and HIPP. Given that the human brain is a complex and interactive system that dynamically processes information flow and that changes in FC over time are not revealed by static FC evaluations (Hutchison et al., 2013), we next examined the variability in FC, i.e., changes in the connections between the ACC and HIPP, using a sliding time window (see section “Materials and Methods” for details). The dynamic FC calculations revealed significantly increased FC variability between the left ACC and left HIPP (ACC.L-HIPP.L) as well as between the right ACC and left HIPP (ACC.R-HIPP.L) in the PDM cohort (p < 0.05; Figure 1C). Thus, both static and dynamic FC suggest an overall increase in communication between the ACC and HIPP in PDM females.

Given that being socially integrated is a primary human need (Adolphs, 2010), and that emotional/mental stimuli can lead to altered activation of brain areas, such as the ACC (Singer et al., 2004), we next asked whether the experience of PDM, a type of physical pain, would change the brain activation and FC when dealing with negative and positive emotional stimulations, i.e., social exclusion and inclusion conditions. To do so, we used the Cyberball task, a paradigm based on a virtual ball-tossing game, including three scenarios of implicit social exclusion (ISE), inclusion (INCL), and explicit social exclusion (ESE) (Williams et al., 2000; Eisenberger et al., 2003).

We first evaluated the condition-related differences in regional brain activity as measured by T-fMRI utilizing the general linear model (GLM), a method used to evaluate differences in activation under various conditions by subtracting one condition from another, i.e., in this case, ISE-INCL, ESE-ISE, and ESE-INCL (Friston et al., 1997). In the GLM analysis, PDM females showed decreased activation of the right Crus II of the cerebellar hemisphere (CERCRU2) in the ESE-ISE contrast compared to controls (Figure 2C and Table 2). No significant difference was found in the ISE-INCL and ESE-INCL contrast between PDM subjects and controls (Table 2).

We next conducted psychophysiological interaction (PPI) analysis (Friston et al., 1997) to determine which brain regions were functionally connected with the ACC. In the ISE-INCL contrast, the FC between ACC.L-brainstem and ACC-brainstem (including the right parahippocampal region, PHIPP.R) was higher in PDM females than in controls (Figure 3A and Table 3). In the ESE-INCL condition, PDM females demonstrated increased connectivity between ACC.L-right thalamus (THA.R)/PHIPP.L/right inferior frontal gyrus, triangular part (IFGtriang.R), ACC.R-HIPP.R/HIPP.L/IFGtriang.R/right superior frontal gyrus (SFG.R), and ACC-HIPP.R/HIPP.L/right middle frontal gyrus (MFG.R) (Figure 3B and Table 3). Identical FC between regions connected with ACC.L/ACC.R/ACC was observed in PDM subjects and controls in the ESE-ISE contrast. Together, the results indicate increased connections between the ACC and the above brain areas in response to Cyberball-induced social exclusion in PDM females.

We further evaluated the FC of the ACC-HIPP pathway using PPI analysis to reveal whether and how this specific pathway differs between PDM subjects and controls. We found that in the ISE-INCL contrast, PDM females showed significantly increased FC in the ACC.L-HIPP.L and ACC.R-HIPP.L circuits (Figure 3C). Furthermore, the combined ACC analysis, i.e., not separating left and right ACC, also showed significantly increased ACC-HIPP FC in the ISE-INCL contrast in PDM subjects (Figure 3C1). In the ESE-ISE contrast, overall ACC-HIPP FC did not differ significantly in PDM women (Figure 3D1), although FC of the ACC.L-HIPP.L and ACC.R-HIPP.L connections was significantly lower (Figure 3D). The results suggest that exclusion, whether explicit or implicit, triggered similar FC in the ACC-HIPP pathway in PDM and control females. In the ESE-INCL contrast, PDM subjects also showed significantly higher FC for all four ACC-HIPP connections (Figure 3E) as well as the combined analysis (Figure 3E1).

Prolonged nociceptive input to the CNS has been shown to induce functional and structural alterations throughout the nervous system (Denk and McMahon, 2017). Having shown that significant differences occurred in the static and dynamic FC of PDM females, we investigated the GM volume of the ACC and HIPP using a voxel-based morphometry (VBM) approach. However, we found no significant differences between PDM subjects and controls (Figure 1D). Thus, the above abnormal connections between ACC and HIPP are not due to GM volume changes.

Given the ethical constraints of human studies, cellular and molecular studies using rodent models are beginning to be used to provide insights into the mechanisms that give rise to chronic pain (Zhuo, 2014). However, this has not been tested directly in PDM due to the lack of a chronic PDM rodent model. Therefore, we generated a chronic pelvic pain rat model that mimics dysmenorrhea-like pain experience based on an existing acute rodent PDM model (Chen et al., 2013; Jesuino et al., 2019), which was characterized by evaluating the levels of prostaglandin F2α and prostaglandin E2, endometrial thickness, and uterine artery blood flow velocity, etc., features that are similar to those found in human PDM (Yang et al., 2015). In this study, after 4 weeks of intraperitoneal (IP) estradiol benzoate injections, which promotes enhanced sensitivity of the uterus to oxytocin (Chen et al., 2013), oxytocin (PDM) or saline (Control) was IP-injected 24 h after estradiol benzoate and repeated for five consecutive weeks (Figure 4A). The rats given estradiol and oxytocin injections exhibited pain induced writhing, indicating abdominal/visceral pain (Supplementary Material 1), i.e., dysmenorrhea-like behavior, due to uterine contraction (Sun et al., 2002). The number of writhing events within a 30 min time window was evaluated (Figure 4B). Average writhing latency was less than 10 s after each oxytocin injection (Figure 4C). In agreement with findings suggesting that women with PDM have elevated pain reactivity (Iacovides et al., 2013), rat with pelvic pain showed a significantly decreased paw withdrawal mechanical threshold (PWMT) (Figure 4D), indicating that these rats are hypersensitive to pain. We next conducted electrophysiological and biochemical experiments in model and control rats to uncover molecular and cellular alterations induced by chronic pelvic pain.

Neuronal oscillatory activity, which is the neural basis of MRI and is preferentially sensitive to BOLD (Logothetis et al., 2001), is fundamental for the entrainment of precise temporal relationships between neuronal responses involved in cognition, perception and emotion (Mathalon and Sohal, 2015). To further evaluate whether the rat model is indeed representative of PDM, we examined whether abnormal neural oscillations occur in pelvic pain model rats by simultaneous dual-site LFP recording in the ACC and dorsal HIPP (Figure 5A). We found that the model rat demonstrated significantly enhanced theta power in the ACC (Figure 5B) and significantly enhanced gamma power in the HIPP (Figure 5C).

Functional coupling of oscillatory activities between pain processing and affective brain areas underlies the suffering associated with chronic pain (e.g., cognitional and emotional alterations), but this cannot easily be studied in PDM women. We thus examined electrical connectivity between the ACC and HIPP using cross-correlation analysis of LFPs (Engel et al., 2001). Model and control rats demonstrated similar correlation values in both ACC-leading and HIPP-leading directions (Figure 5D), suggesting the existence of similar bidirectional communication between the ACC and HIPP.

We next evaluated ACC-HIPP connectivity in the frequency domain (Figure 5E) by coherence analysis (Chen et al., 2021), and observed significantly increased coherence in the gamma range in model rats (Figure 5F). To better understand the connectivity in more precise phase ranges, we next used WPLI analysis (Vinck et al., 2011) which can reduce the contingency caused by the bidirectional connection. The result showed a significantly increased gamma-specific WPLI in model rats (Figures 5G,H), which was in agreement with the coherence analysis. Given that resting state FC indicated by BOLD output reflects the contributions of low frequency LFP signals and their dynamic changes (Shi et al., 2019), and that gamma band modulations co-localize with BOLD (Lachaux et al., 2007; Scheeringa et al., 2011), the enhanced synchronization observed between the ACC and HIPP in the gamma range reflects the increased FC shown by rs-fMRI in the ACC-HIPP pathway of PDM females.

The above results uncovered, in addition to increased ACC-HIPP connectivity, altered theta and gamma oscillations in the ACC and HIPP, respectively. The ACC is a key cortical region for pain perception (Vogt, 2005; Zhuo, 2008) and increased theta oscillation may indicate an alteration in how the ACC modulates HIPP activity. To confirm this, we next determined the strength of cross-frequency PAC between ipsilateral ACC and HIPP (Figure 5I). Despite an increase in the power of ACC theta oscillations, the modulating effect of ACC theta on HIPP gamma oscillations decreased significantly. At the same time, the modulatory effect by HIPP-theta on ACC-gamma oscillations remained unchanged (Figures 5J,K). Thus, chronic pelvic pain results in a reduction of ACC theta-HIPP gamma coupling, suggesting that pelvic pain, such as PDM may change top-down ACC informational input into the HIPP.

Brain oscillations emerge as a consequence of local interactions between excitatory and inhibitory synaptic activities (Verret et al., 2012). Thus, the abnormal oscillations observed in the present study may implicate altered synaptic activities, because theta and gamma oscillatory activities relate to synaptic plasticity, in addition to network synchronization and memory formation (Tesche and Karhu, 2000; Buzsaki and Draguhn, 2004). We therefore examined whether or not abnormal excitatory and inhibitory synaptic activity in the ACC and HIPP underlie the aberrant oscillatory activity. Whole-cell patch-clamp recordings were conducted in brain slices containing either ACC or HIPP (Figure 6A). In voltage clamp mode and in the presence of TTX to block fast sodium channels and thus action potentials, the amplitude and frequency of mEPSCs and mIPSCs in the pyramidal neurons of the ACC and HIPP were measured. A significantly increased mEPSC amplitude was observed in pyramidal neurons in both the ACC and HIPP of model rats (Figures 6B,C,F,G), in agreement with the reported hyperactivity of ACC (Zhang et al., 2017; Zhou et al., 2018) and HIPP (Cardoso-Cruz et al., 2013) neurons during chronic pain. In contrast, mEPSC frequency in both ACC (Figure 6C) and HIPP (Figure 6G), as well as mIPSC amplitude and frequency (Figures 6D,H), was identical. These results suggest that increased excitation, which in turn leads to changes in the excitation/inhibition ratio, may contribute to abnormal oscillatory activities (Verret et al., 2012) in the ACC and HIPP.

Postsynaptic excitatory responses are mediated mainly by two groups of glutamatergic receptors, AMPA and NMDA receptors (AMPARs, NMDARs) (Dai et al., 2019). The observed enhancement of mEPSC amplitude could be accounted for by increased levels of AMPARs and/or NMDARs. We next performed western blotting experiments with anti-AMPAR- and anti-NMDAR-subunit antibodies on ACC and HIPP lysates of model and control rats, which revealed significantly increased levels of both the AMPAR subunit, GluR1, and the NMDAR subunit, NR1, in the ACC (Figure 6E) and the HIPP (Figure 6I) of model rats. Together with the data showing increased mEPSC amplitude, these results suggest that upregulated levels of excitatory receptors may underlie hyperactivity of the ACC and HIPP in chronic PDM female.