Increased cerebrospinal fluid levels of GABA, testosterone and estradiol in women with polycystic ovary syndrome

Abstract

STUDY QUESTION

Do cerebrospinal fluid (CSF) concentrations of gamma-aminobutyric acid (GABA), testosterone (T) and estradiol (E2) differ in women with polycystic ovary syndrome (PCOS) as compared to eumenorrheic, ovulatory women (EW)?

SUMMARY ANSWER

Women with PCOS displayed higher CSF levels of GABA and E2, and possibly T, than EW.

WHAT IS KNOWN ALREADY

The chronic anovulation characteristic of PCOS has been attributed to increased central GnRH drive and resulting gonadotropin aberrations. Androgens are thought to regulate GABA, which in turn regulates the neural cascade that modulates GnRH drive.

STUDY DESIGN, SIZE, DURATION

This cross-sectional observational study included 15 EW and 12 non-obese women with PCOS who consented to a lumbar puncture in addition to 24 h of serum blood collection at 15-min intervals.

PARTICIPANTS/MATERIALS, SETTING, METHODS

In total, 27 women were studied at a the General Clinical Research Center (GCRC) at the University of Pittsburgh. Serum analytes included T, E2 and androstenedione. CSF analytes included GABA, glutamate, glucose, T and E2.

MAIN RESULTS AND THE ROLE OF CHANCE

Women with PCOS had higher CSF GABA as compared to EW (9.04 versus 7.04 μmol/L, P < 0.05). CSF glucose and glutamate concentrations were similar between the two groups. CSF T was 52% higher (P = 0.1) and CSF E2 was 30% higher (P < 0.01) in women with PCOS compared to EW. Circulating T was 122% higher (P < 0.01) and circulating E2 was 75% higher (P < 0.01) in women with PCOS than in EW.

LIMITATIONS REASONS FOR CAUTION

The study is limited by its small sample size and the technical limitations of measuring CSF analytes that are pulsatile and have short half-lives.

WIDER IMPLICATIONS OF THE FINDINGS

Women with PCOS displayed significantly higher circulating levels of T and E2, significantly higher CSF levels of E2, and higher levels of CSF testosterone, although the latter was not statistically significant. A better understanding of the central milieu informs our understanding of the mechanisms mediating increased the GnRH drive in PCOS and lends a new perspective for understanding the presentation, pathogenesis and potential health consequences of PCOS, including gender identity issues.

STUDY FUNDING/COMPETING INTEREST(S)

No conflicts of interest. The study was funded by NIH grants to SLB (RO1-MH50748, U54-HD08610) and NIH RR-00056 to the General Clinical Research Center of the University of Pittsburgh.

TRIAL REGISTRATION NUMBER

NCT01674426.

Introduction

Women with polycystic ovary syndrome (PCOS) display an androgenized phenotype that generally includes hirsutism or acne, truncal adiposity, hyperlipidemia, insulin resistance and increased gonadotropin-releasing hormone (GnRH) drive, which manifests in the circulation as increased LH pulse frequency relative to eumenorrheic, ovulatory women (EW) (Kazer et al., 1987; Waldstreicher et al., 1988; Berga et al., 1993). The maximal GnRH pulse frequency of women with PCOS is about one pulse per hour, which is similar to that seen in men and male monkeys (Matsumoto and Bremner, 1984; Plant, 1986), while eumenorrheic women in the follicular phase have a GnRH pulse frequency of one pulse every 90 min (for reviews, see Kalro et al., 2000; McCartney et al., 2002). Studies show that higher GnRH pulse frequencies concomitantly increase LH and reduce FSH secretion (Yen et al., 1970; Morales et al., 1996; Spratt et al., 1986; Taylor et al., 1997). In women with PCOS, FSH remains chronically below the threshold needed to initiate and sustain folliculogenesis (Van Der Meer et al., 1998), while chronically elevated LH drives increased theca cell secretion of androgens (McCartney and Marshall, 2016).

Androgens have been shown to increase GnRH pulse frequency via multiple mechanisms (Sullivan and Moenter, 2005; Pielecka et al., 2006), including altered γ-aminobutyric acid (GABA)ergic regulation of the neural cascade that regulates GnRH release (for review, see Moore and Campbell, 2016). While GABA is typically thought to function as an inhibitory neurotransmitter, there is evidence that signaling through the GABA_A receptor exerts a predominantly excitatory effect on GnRH neurons (DeFazio et al., 2002). Additionally, women treated with valproic acid, an anti-epileptic medication that increases central GABA levels, develop PCOS-like symptoms (Löscher, 1999). This observation provides a clinical correlate that demonstrates the potential impact of GABA on reproductive function.

The effects of GABA on GnRH secretion may vary with developmental stage, hormonal milieu and receptor subtype expression (Terasawa, 1998; Perrot-Sinal et al., 2001; Leonhardt et al., 1995). If women with PCOS have higher levels of androgens or estrogens in the cerebrospinal fluid (CSF), this milieu could facilitate GABAergic stimulation of GnRH neurons, leading to a feedback loop that sustains the PCOS phenotype.

Given the above considerations and the chronically elevated levels of androgens of ovarian and adrenal origin in women with PCOS, we hypothesized that the CSF of PCOS women would contain higher levels of androgens and GABA relative to that of EW. High levels of testosterone (T) would be expected to promote the function of GABA as an excitatory neurotransmitter on GnRH neurons. To test our hypothesis, we measured CSF levels of T and GABA. Finally, CSF levels of glutamate, glucose and estradiol (E2) provided analytes for comparing the central milieu of both groups.

Materials and Methods

Participants

This study protocol was reviewed and approved by the institutional review boards of the University of Pittsburgh School of Medicine and Magee-Womens Hospital. All subjects gave written informed consent prior to participation. Two groups of women, those with irregular or absent menstrual cycles and those with eumenorrhea, were recruited by advertisement. To reduce heterogeneity, all recruited subjects met the following inclusion criteria: (1) gynecologic age, defined as years since menarche, of >5 and <25 years; (2) chronological age >17 and <35 years; (3) nonsmokers; (4) no concurrent medications or drugs, including psychotropic agents, habitual alcohol, or illicit drug use; (5) no medical, neurologic, or ophthalmologic disease, except acuity problems; (6) not exercising >10 h/week nor running >10 miles/week; (7) within 90–110% of ideal body weight according to the 1983 Metropolitan Life Table for women; (8) without weight loss or gain of >10 lbs. within a year preceding or since the onset of amenorrhea; (9) not meeting criteria for an eating disorder or other major psychiatric disorder; (10) at least 12 months postpartum; (11) not lactating for at least 6 months; (12) day-awake and night-asleep schedule; (13) negative screening pregnancy test and willing to use barrier contraception for the duration of the study; (14) normal thyroid stimulating hormone (TSH) and thyroxin levels; and (15) not using hormonal contraception for at least 8 weeks.

A diagnosis of PCOS was made using the NIH criteria (Zawadzki and Dunaif, 1992) that included a history of oligomenorrhea (<5 menses per year since menarche) and hyperandrogenism, either clinical (acne or hirsutism) or biochemical (androstenedione >200 ng/dL or total T > 2.0 nmol/L). All other causes of oligomenorrhea were excluded. Women with other secondary causes of hyperandrogenism and/or anovulation, including late-onset congenital adrenal hyperplasia, ovarian or adrenal androgen-secreting tumors, Cushing's syndrome or disease, acromegaly, premature menopause, hypothalamic amenorrhea and hyperprolactinemia were identified and excluded using screening laboratory studies and appropriate imaging, as indicated by the clinical presentation of the prospective subject. Serum progesterone levels were also documented prior to and following the GCRC visit among women with PCOS to determine ovarian secretory activity and to screen for subsequent ovulation. Subjects were excluded if progesterone suggested recent ovulation.

In EW, ovulation was confirmed if a midluteal progesterone concentration exceeded 30 nmol/L (9.43 ng/mL) in the current and preceding menstrual cycle. EW were studied during the early follicular phase (EFP), Days 3–5 from last menses. There were 19 EW and 19 women with PCOS who met the inclusion criteria, but not all consented to the study procedures. The final group of participants included 15 EW and 12 women with PCOS.

Experimental protocol

Subjects were admitted to the General Clinical Research Center at the University of Pittsburgh prior to 0800 h. Height and weight were measured, and body mass index (BMI) was calculated. Percent body fat was estimated using skin fold thickness at three sites: triceps, abdominal and suprailiac. An indwelling intravenous catheter was inserted into the nondominant forearm. Starting at 0900 h, blood samples were obtained via the catheter at 15-min intervals for 24 h (as outlined previously in Berga et al., 1997). At 1100 h on the following day, a single lumbar puncture (LP) was performed by an obstetrical anesthesiologist; 25 mL were withdrawn in 1 mL aliquots and flash-frozen in a mixture of dry ice and ethanol and stored at –80°C until the assay (as in Berga et al., 2000). The last five aliquots withdrawn were utilized for measurement of analyses reported herein. The CSF was replaced with artificial CSF to avoid subjects developing a spinal headache. There were no significant complications from the LPs; one subject required a blood patch for treatment of spinal headache. Subjects remained at bed rest for 2 h after the LP and then were discharged with instructions to rest, not exercise, and drink plenty of fluids for 24 h.

All samples from a given participant were analyzed together in the same assay and in duplicate to reduce variability. LH levels were measured every 15 min across the 24-h sampling period, and FSH was determined at baseline by immunofluorometric assay (Wallac, Turku, Finland).

E2, T and androstenedione were measured in sera using a solid phase radioimmunoassay (Coat-A-Count, DPC, Los Angeles, CA). These laboratory methods have been described in an earlier report (Daniels and Berga, 1997), and the intra- and inter-assay coefficients of variation for these methods were <10%. Serum SHBG levels were determined using an immunoradiometric assay (IRMA) (DSL, Webster, TX). Free androgen index was calculated using the following formula: total T ÷ SHBG × 100. The between assay and within assay CVs were <10%.

Estradiol (E2) and testosterone (T) were measured in duplicate in CSF using an enzyme immunoassay (EIA) designed to detect steroid hormone levels in transudates (Salimetrics, State College, PA). All samples were batched and analyzed together in the same assay to reduce variability, which was <5%. The reported sensitivity of each assay was 1 pg/mL for estradiol and 1.5 pg/mL for testosterone.

GABA was measured in duplicate in CSF samples using high pressure liquid chromatography (Waters, Milford, MA) with a Model 5200 A Coulochem II electrochemical detector (ESA, Inc., Chelmsford, MA). GABA standards were prepared in CSF, derivatized to enhance detectability and analyzed. Background levels were subtracted and the standard curve was plotted. Linearity was determined between 1 μmol/L and 50 μmol/L. Regression analysis confirmed a significant correlation coefficient of 0.999 between theoretical and measured concentrations. Subsequently, unknown concentrations were calculated from the standard curve by integrating the peak using Millennium software. The sensitivity of this method was in the nanomolar range.

Glutamate levels were determined in CSF samples using HPLC using methods as previously published (Wagner et al., 2005; Kerr et al., 2003; Palmer et al., 1994). Glutamate was measured using an OPA-derivatized fluorescence detection method, 334 nm excitation, 424 nm emission (Model 474, Waters, Milford, MA). Specimens were loaded and run through a C18 microsorb column with a buffered methanol gradient at 1 mL/min. Known standard concentrations of glutamate were evaluated with each batch of samples analyzed.

Glucose was also measured to serve as a control analyte in CSF. Samples were analyzed using a Glucose Lactate Analyzer (YSI, Inc., Yellow Springs, OH).

For outcomes that followed a normal distribution (all of Table I except LH:FSH ratio), the Student's t-test was utilized to compare differences in means between groups. Mann Whitney U was utilized to compare outcomes that were not normally distributed (LH:FSH ratio in Table I and all of Table II). SPSS version 17 for Windows was utilized for statistical comparisons. LH pulse number and amplitude were determined by a computer-assisted algorithm, Cluster, using a peak width of 2, a nadir width of 1, a t-statistic of 3 for upstroke and downstroke, and a quadratic equation based on assay CV to estimate variance (Daniels and Berga, 1997; Veldhuis and Johnson, 1986).

Table I.

Demographic and biochemical characteristics (mean ± SEM) of EW and women with PCOS.

	EW (n = 15)	PCOS (n = 12)	P value
Age (years)	25.6 ± 1.3	24.3 ± 1.0	0.4
BMI (kg/m2)	22.0 ± 0.7	22.7 ± 0.6	0.5
% Body fat	26.5 ± 1.4	33.4 ± 2.1	<0.01
LH pulse #/24 h	14.9 ± 0.7	18.5 ± 1.0	<0.01
LH pulse amplitude	1.8 ± 0.2	3.9 ± 0.3	<0.01
LH (IU/L)	3.7 ± 0.3	13.0 ± 0.8	<0.01
FSH (IU/L)	4.9 ± 0.4	5.3 ± 0.2	0.4
LH:FSH	0.8 ± 0.1	2.5 ± 0.2	<0.01

Statistically significant values are indicated in bold.

EW, eumenorrheic, ovulatory women; PCOS, polycystic ovary syndrome.

Table II.

Serum and CSF levels of sex steroids, GABA, and glutamate neurotransmitters and glucose control (mean ± SEM) in EW and women with PCOS.

	EW (n = 15)	PCOS (n = 12)	P value
Serum
Estradiol (E2) (pmol/L)	91.0 ± 7.0	159.6 ± 16.7	<0.01
Total T (pmol/L)	535.7 ± 61.7	1190.9 ± 161.6	<0.01
T:E2	5.89 ± 0.02	7.46 ± 0.03	0.6
Androstenedione (ng/dL)	127 ± 13.6	204 ± 10.5	<0.01
Sex hormone binding globulin (nmol/L)	199 ± 26.4	119 ± 18.0	<0.03
Free Androgen Index	0.34 ± 0.07	1.14 ± 0.18	<0.01
CSF
E2 (pmol/L)	20.6 ± 1.7	26.9 ± 1.4	<0.01
T (pmol/L)	120.7 ± 23.0	183.3 ± 32.2	0.1
T:E2	5.86 ± 0.07	6.81 ± 0.04	0.3
GABA (μmol/L)	7.04 ± 0.5	9.04 ± 0.8	<0.05
Glutamate (μmol/L)	0.24 ± 0.02	0.28 ± 0.02	0.2
Glucose (mg/dL)	58.6 ± 2.1	60.1 ± 2.0	0.6

Statistically significant values are indicated in bold.

CSF, cerebrospinal fluid; GABA, gamma-aminobutyric acid.

Results

Women with PCOS and EW were comparable in age and BMI, but percent body fat was higher in women with PCOS (P < 0.01, Table I). As expected, LH pulse frequency and amplitude and the LH:FSH ratio were higher in the PCOS women than in the EW group (P < 0.01, Table I). Women with PCOS also displayed elevated circulatory levels of total T (1190.9 ± 161.6 versus 535.7 ± 61.7 pmol/L, P < 0.01) and androstenedione (204 ± 10.5 versus 127 ± 13.6 ng/dL, P < 0.01), and lower sex hormone binding globulin levels (119 ± 18.0 versus 199 ± 26.4 nmol/L, P < 0.03) compared to EW (Table II). While the PCOS subjects displayed a higher free androgen index (P < 0.01, Table II), differences in testosterone to estradiol ratios (T:E2) in the serum and CSF were not statistically different between the two groups. CSF testosterone levels were 52% higher in PCOS as compared to EW (183.3 ± 32.2 versus 120.7 ± 23.0 pmol/L, P = 0.1), however wide range of T levels in both groups led to a large magnitude group difference that was not statistically significant, thus making it difficult to confidently reject the null hypothesis due to a lack of power. The greater variability in T levels in both groups likely reflects its short half-life in the circulation and the resulting pulsatile circulatory pattern. CSF estradiol levels were statistically lower in the EW group as compared to the PCOS group (20.6 ± 1.7 versus 26.9 ± 1.4 pmol/L, P < 0.01).

Data for the neurotransmitters and glucose control are shown in Table II. GABA levels in the CSF were 28% higher in PCOS (9.04 ± 0.8 μmol/L) compared to EW (7.04 ± 0.5 μmol/L) (P < 0.05). Glutamate and glucose levels were similar in the CSF of EW and women with PCOS.

Discussion

To the best of our knowledge, this is the first study to interrogate central levels of neurotransmitters and sex steroids in women with PCOS and to compare these to levels in EW. We confirmed our hypothesis that women with PCOS have increased CSF concentrations of GABA compared to EW. As expected, women with PCOS had higher levels of circulatory T. Although CSF levels of T were 52% higher in PCOS, the wide variation in CSF T levels rendered the result not statistically significant (P = 0.1) and we could neither confirm that part of our hypothesis nor reject the null hypothesis due to a lack of power. The shorter circulatory half-life of T (as compared to that of E2) may explain, at least in part, the wide variation in T levels and therefore the large SEMs observed in both groups. Thus, despite a larger magnitude increase in CSF levels of T (52%) as compared to E2 (30%) in women with PCOS as compared to EW, the statistical significance of the group difference was less for T than for E2. However, the possibility remains that the increase in CSF T is of significance despite the lack of statistical significance. Additional studies will be needed to clarify whether CSF androgens are elevated in women with PCOS. Circulatory and CSF levels of E2 were clearly higher in women with PCOS. The T:E2 ratios did not differ between the two groups in either the CSF or the serum, which likely indicates comparable aromatase activity in PCOS women and EW.

The most parsimonious explanation for the increased CSF and circulatory E2 is increased androgen precursor. While the actual magnitude of the increase in PCOS was greater for CSF T (52%) than for CSF E2 (30%), both T and E2 likely impact CNS, including hypothalamic, functions. Increased central and circulatory estrogen levels suppress FSH to levels below those needed to initiate folliculogenesis (Van Der Meer et al., 1998, Archer et al., 1988). Additionally, both suppressed FSH and elevated LH levels contribute to the altered LH:FSH ratio characteristic of PCOS (Yen et al., 1970). Further, elevated LH stimulation drives theca cell secretion of the androgens androstenedione and T that can have direct and indirect effects on the CNS-hypothalamic-pituitary-gonadal axis. To provide a physiological comparison, we studied eumenorrheic women in the EFP. We recognize that LH pulse frequency is slower in the EFP than in later stages of the follicular phase and that estradiol levels are lowest in the EFP than in later stages of the menstrual cycle. However, the goal of the study was to compare steroid and GABA levels in the CSF of eumenorrheic women with those of anovulatory women with PCOS rather than to examine the correlation between GABA and CSF steroid levels and LH pulsatility. The purpose of determining LH pulsatility was to support the diagnosis of PCOS.

Chronically elevated circulatory levels of T and E2 and CSF levels of E2, and the possibly elevated CSF levels of T, likely play a role in the pathogenesis of PCOS. Administration of dihydrotestosterone (DHT), a non-aromatizable androgen, to mice increased GnRH firing activity. The frequency and amplitude of GnRH peaks were approximately doubled compared to controls (Pielecka et al., 2006). Despite there being little evidence for a direct action of sex steroids on GnRH neurons, androgens appear to alter GnRH activity through a complex hypothalamic cascade that involves GABAergic input to neuropeptide Y (NPY) and kisspeptin/neurokinin B/dynorphin (KNDy) neurons (Moore and Campbell, 2016). Further, DHT interferes with negative feedback from progesterone (Pielecka et al., 2006), which may explain why the GnRH drive in women with PCOS is resistant to suppression by progestin and progesterone feedback (Daniels and Berga, 1997; Pastor et al., 1998; Eagleson et al., 2000). DHT treatment also increases GABA transmission to GnRH neurons (Sullivan and Moenter, 2004; Sullivan and Moenter, 2005), which fits with our observation that women with PCOS had higher levels of CSF GABA. A wide body of evidence suggests that GABA has a primarily excitatory effect on GnRH neurons (for review, see Herbison and Moenter, 2011), so stimulation of this pathway by excess testosterone could increase GnRH pulse frequency and consequently elevate LH and suppress FSH release (Spratt et al., 1986).

Like T, E2 may regulate GnRH release via modulation of the GABA-kisspeptin system (see reviews by De Bond and Smith, 2014). Acute intravenous administration of kisspeptin has been shown to increase LH, FSH and T in males and females (Ratnasabapathy and Dhillo, 2013). Kisspeptin neurons integrate sex steroid and metabolic signals to modulate GnRH drive. These neurons control positive and negative sex steroid feedback of GnRH release, including the midcycle LH surge (Smith, 2013; Chan, 2013). E2 potentiates the kisspeptin-mediated signal to GnRH via both GABAergic and glutamatergic transmission (Pielecka-Fortuna et al., 2008; Pielecka-Fortuna and Moenter, 2010).

Because reproduction requires energy, a tight link between reproduction and metabolism is expected. For instance, leptin has long been known to regulate GnRH drive. While leptin does not appear to regulate GnRH neurons directly, it may do so indirectly (Martin et al., 2014; Yan et al., 2014). Although our participants were all within 90–110% ideal body weight and BMIs were similar in both groups, body composition differed and PCOS women had a higher percentage of body fat than EW. Women with PCOS have higher leptin levels (Chapman et al., 1997). Higher levels of E2 and leptin could increase in kisspeptin signaling, which in turn would stimulate GnRH release.

Our study does not allow us to determine the plasticity of the GnRH pulse generator after chronic exposure to androgens (Sullivan and Moenter, 2004) and did not examine other mechanisms that might contribute to increased GnRH drive in PCOS. High insulin levels might synergistically interact with androgens to alter GnRH connectivity and increase GnRH pulsatility (Berga, 2009). For instance, ovarian steroids alter potassium channel functioning in hypothalamic GnRH neurons and may modulate GnRH output via this mechanism (Huang et al., 2008). Not surprisingly, K(ATP) channels also mediate responsiveness to glucose and metabolic inhibition (Zhang et al., 2007). While our women with PCOS and EW differed in body composition, they had comparable BMIs, so our findings may not be generalizable to overweight and obese women with PCOS.

Increased central and circulatory androgens, possibly mediated by aromatization to estrogens, and estrogens per se likely impact more than metabolism or ovarian function (Franks and Berga, 2012). The extent to which other features of PCOS derive from chronically elevated central or circulatory ambient sex steroid exposures remains an open question. For instance, women with PCOS display enhanced athletic prowess that has been attributed to increased androgen levels (Rickenlund et al., 2003). Both androgens and estrogens alter metabolism and modify fat deposition and energy utilization and stress reactivity. Androgens or high sex steroid levels may alter gender identity (Kowalcyzk et al., 2012; Manlove et al., 2008), predispose to mental health disorders (Cesta et al., 2016) and alter cognitive (Rees et al., 2016) or emotional processing (Marsh et al., 2013).

In summary, lean women with PCOS display altered CSF levels of GABA, E2 and possibly T. Taken together, the chronic exposure of the CNS of women with PCOS to an altered hormonal milieu likely contributes to both the faster GnRH pulse frequency and accompanying chronic anovulation and may alter neural and somatic functions other than reproductive drive per se. Further, the finding of increased GABA levels in the CSF of women with PCOS may explain, at least in part, the altered emotional and cognitive processing and the increased prevalence of gender identity issues and mental health disorders in this population (Chiapponi et al., 2016; Deligiannidis et al., 2016; Michopoulos et al., 2013). Further investigation into mechanisms underlying increased GnRH drive could provide promising targets for treatment of reproductive dysfunction and other health sequela associated with PCOS.

Authors’ roles

Jennifer Kawwass, MD: analysis of data, drafting, critical review and final approval of manuscript. Kristen Sanders, MS: re-analysis of data, critical review and final approval of manuscript. Lisa Rohan, PhD: technical development of HPLC measurement of GABA and glutamate, critical review and final approval of manuscript. Tammy Loucks, MPH, DrPH and Sarah Berga, MD: study design and data acquisition, analysis of data, critical review of manuscript, final approval of manuscript.

Funding

This work was supported in part by grants to SLB from NIH R01-MH50748, from NICHD/NIH as part of the Specialized Cooperative Centers Program in Reproduction Research through cooperative agreement U54 HD08610 (Project I-PI SLB, Program Director Tony Plant), and by grant NIH RR-00056 to the General Clinical Research Center of the University of Pittsburgh.

Conflict of interest

None declared. SLB serves on the Editorial Board of the Journal of Clinical Endocrinology and Metabolism, the Editorial Board of Menopause, on the Subspecialty Advisory Board for the American Journal of Obstetrics and Gynecology, and as Associate Editor for Human Reproduction Update.