Northwestern University and the University of Texas Southwestern Medical Center Institutional Review Boards approved the protocol. Fifty-seven healthy eumenorrheic females with moderate physical activity and no history of lower limb disorder were enrolled in the study. Women with moderate physical activity were defined as those who exercised less than 7 hours per week and were not participating in competitive-level sports. We had to exclude a total of 33 women: two due to lack of adherence to the protocol, two due to sensitivity to electrical stimulation, four due to the absence of estradiol and/or progesterone fluctuations across the menstrual cycle, 12 withdrew from the study, and 13 decided not to participate in any part of the study after being enrolled. Thus, the results presented here are based on data acquired from 24 healthy eumenorrheic women (age: 27.0 ± 4.4 years; BMI: 24.8 ± 4.8 kg/m²; cycle length: 30.3 ± 3.8 days; means ± SD) involved involved in the study. All participants signed the written informed consent form and were free to participate.

Women who exercised more than 7 h per week or participated in competitive level sports were excluded because of the high rate of undiagnosed menstrual dysfunction in this population (Gibbs et al., 2013). Participants were asked to maintain stable exposure to caffeine, alcohol, and exercise during the 12 hours before each testing session to minimize changes in reflex excitability (Motl and Dishman, 2003; Walton et al., 2003; von Dincklage et al., 2007).

Each participant was tested every other day for one menstrual cycle (about 15 to 20 testing sessions depending on the length of the cycle). The personnel performing the testing were blinded to subject’s menstrual cycle phase. The start date of the testing was assigned randomly. The tesing time was scheduled based on participant’s preference and maintained at the same time of the day throughout the testing sessions to minimize the effect of diurnal fluctuations of hormone levels (Bao et al., 2003; Bungum et al., 2011) and circadian variations of the flexion reflex (Sandrini et al., 1986a).

The hormone concentrations were obtained from blood samples collected by means of venipuncture. The samples were collected at each testing session from antecubital area of the participant’s arm using a 5 mL Vacutainer serum separator tube (BD, Franklin Lakes, New Jersey). Once the tube was filled, the sample was mixed well and allowed to clot. The samples obtained at Northwestern were processed at Northwestern Memorial Hospital Outpatient Laboratory and the samples obtained at UT Southwestern were sent to MedFusion. The hormonal concentrations analyzed were estradiol and progesterone.

The participants were placed in a supine position on a padded examination table. Each participant was instructed to choose a comfortable position for her head and arms. The position was recorded so it could be replicated at subsequent testing sessions. They were instructed not to activate their muscle prior, during, and after the stimulation paradigm.

Bipolar surface electromyogram (sEMG) signals were detected from the tibialis anterior either with pre-amplified electrodes (DE-2.1 Bagnoli single differential, interelectrode distance: 10 mm, contact dimensions 10 × 1 mm) or bipolar Ag/AgCl pre-gelled electrodes (diameter 10 mm, inter electrode distance 20 mm). The electrode type for each subject was consistent throughout the testing sessions.

Prior to electrode placement, the skin was cleaned with alcohol, slightly abraded (Nuprep®, Weaver and Company, CO, United States), and cleaned with water to remove any residue. The locations of the electrodes were recorded and marked with a waterproof or blacklight marker to ensure the same electrode location in each session. The signals were digitized with a Micro1401 Data Acquisition Unit (Cambridge Electronic Design, United Kingdom) at a rate of 2,000 Hz.

Classically, examination of the flexion reflex pathway in humans used two unique paradigms: peripheral nerve electrical stimulation and electrical stimulation to the skin in the sole of the foot (Van Wezel et al., 1997; Knikou, 2007). In this study we incorporated both paradigms, in almost evenly split cohorts, 11 and 13 subjects respectively. Implementation of both paradigms allows for a tertiary analysis to compare the consistency in the flexion reflex response across the two paradigms.

Posterior tibial nerve stimulation: The flexion reflex of 11 participants was elicited by delivering a train of eight pulses (2 ms pulse width, 200 Hz) using a Grass Stimulator S48 (Grass Instruments, Quincy, MA) connected to an isolation stimulus unit (SIU5, Grass Instruments) and passed through constant current unit (CCU1, Grass Instruments). The bipolar stimulating electrode (electrode diameter 0.8 cm, inter electrode distance 2 cm) was placed over the posterior tibial nerve located posterior and inferior to the medial malleolus, with the anode positioned posteriorly. Optimal placement of the stimulating electrode was found by moving the stimulating electrode over the nerve until a low-current stimulus elicited the reflex response. To find the reflex threshold, the stimulus intensity was increased manually with 0.2 mA increments until a distinguishable and stable response was clearly seen. Then a maximum of 12 stimuli with an intensity at the reflex threshold were delivered to the subjects with a 30–35 s random inter-stimulus interval until at least four responses were obtained.

Plantar cutaneous afferents stimulation: For the remaining 13 participants, a pair of square gel electrodes (5 cm x 5 cm) were used. The cathode was placed over the area between the first and second metatarsal, while the anode was placed on the arch of the foot. A train of five pulses (1 ms pulse width, 200 Hz) was delivered using a DS7A constant current stimulator (Digitimer Ltd., Hertfordshire, England). The reflex threshold was defined as the smallest intensity that elicited the reflex response. Then a maximum of 20 stimuli with an intensity 10% higher than the reflex threshold was delivered to the participant with a 30–35 s random inter-stimulus interval.

Post-hoc analysis indicated that within subjects, stimulation intensity across visits showed no statistically significant correlation on visits suggesting that there was no effect of habituation nor hormone concentration dependency.

The signals were processed offline using Matlab software (R2019a. The MathWorks Inc., MA, United States). A 4th order zero phase Butterworth bandpass filter (final cut off frequency 20–500 Hz) was applied to the signal to attenuate movement artifacts and other interference. The envelopes of rectified EMG signals extracted using a 4th order zero phase Butterworth low-pass filter with a final cut off frequency at 200 Hz were used to define the onset and offset of the reflex response. The start and end of the reflex response were determined when the envelope of the reflex signal crossed 15% of the peak amplitude measured from the baseline value (Figure 1). The baseline value was defined as the average of the 250 ms EMG envelopes before the onset of the stimulus. The onset and offset of the reflex response were used to calculate root mean squared (RMS) value, duration, and latency, measured from the onset of the stimulus. The average value of each parameter for all identified reflex responses was then calculated and used for further analysis.

The magnitude of the flexion reflex reflex (RMS value), duration, and latency were analyzed at: (1) the follicular phase, where the estradiol concentration increased from low level to its maximum while the progesterone concentrations remained low, and (2) the luteal phase, where the progesterone increased from low level to its maximum along with background estradiol concentration.

Data are presented as mean ± SD. The statistical analysis for this study was performed in the R language (version 4.2.2, R Foundation for Statistical Computing, Vienna, Austria). To evaluate the effect of changes in hormones level on reflex RMS, duration, and latency, a generalized estimating equation (GEE) with log link function and exchangeable correlation structure was employed using geepack R package (Ballinger, 2004; Halekoh et al., 2006). Each of the flexion reflex parameters was simultaneously regressed on estradiol, progesterone, and the estradiol × progesterone interaction. The estradiol concentration in ng/ml was used in GEE to keep the units consistent between estradiol and progesterone. Significance was accepted for p values <0.05.

The estradiol and progesterone profiles obtained from a non-linear mixed effect model with harmonic terms (Albert and Hunsberger, 2005) is shown in Figure 2. The cycle day was normalized to the cycle length and centered such that 0 corresponds to first day of menses, 0.5 corresponds to peak estradiol, 0.75 corresponds to peak progesterone, and 1.0 corresponds to last day of the cycle. Estradiol concentrations at the first day of menses, peri-ovulatory (peak estradiol concentration), and mid-luteal (peak progesterone concentration) time points were 41.7 ± 16.2 pg/mL, 307.23 ± 99.2 pg/mL, and 186.7 ± 60.2 pg/mL, respectively. The corresponding progesterone concentrations at those time points were 1.0 ± 0.5 ng/mL, 0.7 ± 0.3 ng/mL, and 13.6 ± 4.4 ng/mL. These concentrations are consistent within the range of estradiol and progesterone levels reported in the literature (Sehested, 2000; Marsh et al., 2011; Mumford et al., 2012).

A typical flexion reflex response is shown in Figure 3. The average stimulation intensity used in this study was 7.2 ± 3.5 mA. The stimulation intensity applied on the posterior tibial nerve stimulation ranged from 2.5 to 15.0 mA and the stimulation intensity applied on the plantar cutaneous afferents ranged from 3.0 to 20.9 mA across participants. Participants reported no pain during and following stimulation. Post-hoc analysis on the stimulation intensity indicated no significant difference between the two groups (p > 0.05). The average response rate of posterior tibial nerve stimulation and plantar cutaneous afferents stimulation were 78% (8 responses out of 10 stimuli) and 47% (9 responses out of 20 stimuli) respectively. A non-parametric Kurskal-Wallis test showed that reflex latency and duration at menses did not differ between the two stimulation paradigms: posterior tibial nerve and plantar cutaneous afferents stimulation (Table 1, p > 0.05), suggesting that both paradigms activated an equivalent flexion reflex pathway.

A GEE model on data collected during the follicular phase showed that estradiol concentration was not associated with reflex latency, duration, and RMS magnitude (Table 2). In the luteal phase, reflex duration and RMS magnitude were not associated with fluctuation of either estradiol or progesterone. As indicated in Table 2, in the luteal phase, the GEE revealed a main effect of estradiol (b = 0.228, p = 0.038) on reflex latency. Our data suggests that increases in estradiol in the luteal phase predicts longer reflex latency better than chance. Increases in progesterone make this association slightly less pronounced (b = −0.017, p = 0.081).

The latencies of the reflex responses observed in this study were longer than 60 ms, the natural separation time between the RII and RIII responses (Andersen, 2007). Hence, the observed reflex response can be categorized as the RIII response mediated by Aδ fibers, although the stimulus intensities delivered to the participants were below the pain threshold. The Aδ fiber afferents mainly terminate in laminae I and II, where estrogen receptors (ERα, ERβ, and G protein-coupled receptor 30) are present (Shughrue et al., 1997; VanderHorst et al., 2005; Takanami et al., 2010). In an animal model, estrogen was shown to modulate nociceptive transmission (Li et al., 2009). Specifically, ERα and ERβ were shown to inhibit and enhance nociceptive transmission, respectively (Coulombe et al., 2011). The invariant flexion reflex response during the follicular phase observed in this study may likely be due to the activation of both estrogen receptors, resulting in zero total effect of estradiol.

In the luteal phase, we observed a significant association between estradiol concentration and the latency of the reflex. Although the magnitude of the association is modest, it is tempting to speculate that with the increase in estradiol concentrations in the luteal phase the processing of flexion reflex may be involved to a more complex network when compared to the neuronal circuit at the follicular phase, prior to projection into the motor neuron (Abraira and Ginty, 2013). The mechanism of estradiol’s effect on the latency of the flexion reflex in the presence of background progesterone warrants further examination.

Given the lack of association between estradiol and the latency of the flexion reflex during the follicular phase, one could argue that the observed association between estradiol and latency during the luteal phase is likely facilitated by the presence of progesterone. Histological evidence from animal models indicates that progesterone receptors are present in the lamina IX of the spinal cord (Labombarda et al., 2000), where motor neurons are located (Osseward and Pfaff, 2019), likely modulating changes in the flexion reflex arc. The isolated neuronal effect of progesterone has been explored, for the most part, in cortical structures, specifically hippocampal neuron (Woolley and McEwen, 1993; Weiland and Orchinik, 1995; Foy et al., 2010), with a primary effect in modulating the synaptic plasticity and regulation of GABA_A receptors. In the spinal cord, results from the rodent models were restricted to the neuroinflammatory effect of progesterone and its role in neurodegeneration after trauma (Thomas et al., 1999; Gonzalez Deniselle et al., 2002). Specifically, progesterone has been shown to be neuroprotective through brain derived neurotrophic factor (BDNF) mRNA upregulation (González et al., 2004; De Nicola et al., 2006), downregulation of neuronal Na, K-ATPase regulatory subunits (Labombarda et al., 2002), and promotion of myelin formation (Labombarda et al., 2010). In a spinal cord injury animal model, progesterone was shown to prevent upregulation of N-methyl-D-aspartate receptor subunits, an important component that induces pain produced by tactile stimuli (Coronel et al., 2011). Allopregnanolone, a progesterone metabolite, is also shown to increase neuroinhibitory effect by upregulation of GABA_A receptors (Twyman and Macdonald, 1992). It is unclear, however, what the underlying mechanism is for the interaction of progesterone with the function of estradiol at the level of the neural circuit.

There have been limited examinations on the synergistic effects of progesterone and estradiol on neuronal properties. Co-treatment of estradiol and progesterone was shown to reduce the BDNF levels in the hippocampus, while treatment with estradiol alone did not affect the levels of BDNF (Gibbs, 1999). Co-treatment of estradiol and progesterone was also shown to significantly enhance voltage-gated calcium conductance in hippocampal pyramidal neurons (Joëls and Karst, 1995). However, examinations in animal models on the probable estradiol-progesterone combined facilitatory effect on spinal neuronal networks are nonexistent. Correlative evidence seems to indicate that administration of progesterone on estrogen-primed animals is shown to facilitate lordosis (Whalen, 1974; Kow et al., 1979), a reflex response triggered by a touch stimulus (Kow and Pfaff, 1976), possibly through upregulation of estrogen-inducible progesterone receptors (Monks et al., 2001). Our in vivo data indicates that estradiol, with the presence of progesterone, may have a neurophysiologic effect at the spinal level, manifested by the finding that an increase in estradiol concentration in the luteal phase was associated with an increase of reflex latency. While the increase in estradiol concentration during the follicular phase was not associated with any reflex properties, a significant effect emerged during the luteal phase, as well as a trend toward significance for estradiol × progesterone effect. This finding may suggest the important role of progesterone in enhancing the effect of estradiol. This potential interaction warrants further examination.

It has been suggested the interneuron circuits involved in the flexion reflex are partly shared with the interneurons responsible for the central pattern generator for locomotion (Windhorst, 2021). Specifically, the afferents in the glabrous skin of the foot have an important role in sensing the foot’s contact with the ground and may contribute to standing balance and movement control (Kennedy and Inglis, 2002). In disease states such as in spinal cord injury and traumatic brain injury, administration of estradiol and progesterone has been proposed as a potential therapeutic option (Wright et al., 2007; Shvetcov et al., 2023) due to their important role as neuroprotective agent and the reported recovery of locomotor function in animal models (Thomas et al., 1999; Sribnick et al., 2003).

Our findings on the effect of estradiol on the flexion reflex latency would likely have functional significance during states where estradiol concentration is higher than those observed during the natural menstrual cycle. Recall that our regression coefficient between estradiol (ng/ml) and reflex latency (ms) was estimated to be 0.23, leading to a functionally insignificant within-cycle change in latency of 0.092 ms due to estradiol (a maximum change of 0.4 ng/mL). However, these effects may be significant in the states where higher estradiol levels are expected. During pregnancy, for example, the estradiol level has been reported to reach 30 ng/mL (Schock et al., 2016), leading to a probable change of a latency up to 7 ms during this state. The functional significance in the latency change of this magnitude is still under debate. However, in the context of a pathological state, it has been shown that individuals with chronic ankle instability expressed a 7 ms difference in TA latency between the affected and unaffected sides in response to perturbation (Sousa et al., 2019). Hence, our results may provide a basis for future examinations in physiological states in which the estradiol concentration is significantly higher than the average change observed during the menstrual cycle.

While the findings in this study contribute in several ways to our understanding of the effect of sex hormones on polysynaptic reflex networks, these results are limited to healthy pre-menopausal women and may need to be examined in females with neurological injuries (for example, spinal cord injury). In addition, the complex interactions between sex hormones, and other reproduction-associated hormones (luteinizing hormone, follicle-stimulating hormone, relaxin), and neurotransmitters, such as glutamate, serotonin, and dopamine (Barth et al., 2015) were not considered in this study. Both animal and human studies have shown that the interneurons involved in the nociceptive component of the flexion reflex (RIII response) are modulated by dopaminergic, cholinergic and GABAergic neural transmission (Mondrup and Pedersen, 1984; Sandrini et al., 1986b; Clemens and Hochman, 2004; Gerdelat-Mas et al., 2007). It is likely that during the luteal phase, the observed estradiol-mediated latency changes of flexion reflex may be attributed to the changes in neurotransmitters. In addition, and despite of the reported conflicting findings, a state of anxiety may modulate the nociceptive component of flexion reflex (RIII response) (French et al., 2005). In this study, multiple ways were implemented to minimize participants’ anxiety during testing. Participants were instructed to choose a comfortable position for their head and arms, fully relax and remain awake, and their muscle activation levels were monitored by the investigators. In addition, the testing was conducted in a quiet examination room with dim lighting. However, future examination incorporating measures of anxiety, such a questionnaire, and heart rate or skin conduction rate monitoring, may be required.

To our knowledge, this is the first study to investigate the neuromodulatory effect of sex hormones specifically on the polysynaptic spinal network activated by innocuous stimulation of sensory fibers on the sole of the foot and the posterior tibial nerve. Together with supraspinal commands, external inputs such as LTMR, and internal inputs from muscle spindles and the Golgi tendon organ, the sensation of pain constructs a sensory feedback circuit in the spinal cord and plays important roles in the control of motor neuron activation (Böhm and Wyart, 2016). Acute modulation of the sex hormones may alter the polysynaptic spinal circuit and may have implication for the restoration of motor function.