We hypothesized that women with PMDD would display altered steroid metabolic profiles/signatures compared with asymptomatic control women during controlled and standardized exposures to ovarian steroids. To test this, we used a steroid-based metabolomics platform to compare asymptomatic controls and women with PMDD in whom symptom remission and recurrence (that is, E2/P4 sensitivity) were confirmed by ovarian steroid manipulation.5 The main goal of this study was to examine the potential differences in the metabolic processing of standardized doses of ovarian steroids in women with PMDD and controls. Metabolomics tools enable identification and quantification of tens to thousands of compounds that represent changes in biochemical pathways31, 32 in response to treatment. Metabolomics strategies have also mapped global biochemical changes in depression, characterized the effects of selective serotonergic re-uptake inhibitors on metabolic pathways and defined several pathways implicated in individual variation in response to these medications.33, 34, 35
We studied 15 women with PMDD aged 23-48 years. All were medication-free, with regular menstrual cycles (range, 21-35 days), not medically ill and not pregnant. Women with PMDD were self-referred in response to newspaper advertisements or were referred by their physician. Before study entry PMDD was confirmed in these women based on criteria outlined in the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV). Women completed 3 months of prospective daily ratings using a four-item 100-mm visual-analog scale that confirmed the timing and severity of their self-reported menstrually related mood symptoms (irritability, sadness, anxiety and mood swings) as described previously. The mean score of at least one negative mood symptoms was at least 30% higher (relative to the range of the scale used by each woman) in the week before menstruation compared with the week after the cessation of menstruation in at least two of the three cycles. Functional impairment was characterized as a daily rating form (DRF) score of 2 (minimal) or higher on one of 4 questions related to functional impairment (that is, stayed at home or avoided social activities, had conflicts or problems with people, symptoms interfered with relationships at work or home or symptoms interfered with work productivity) in at least 3 days out of 7 days pre-menses. Finally, DRF ratings and the results of both a semi-structured interview and a self-report questionnaire (that is, the Menstrual Screening Questionnaire and the Menstrual Assessment Form, respectively) confirmed that all women met the required number of symptoms specified in the DSM. Women with significant negative mood symptoms (on the DRF) during the follicular phase of the menstrual cycle were excluded. Thus, the diagnostic criteria for PMDD were augmented by the severity criterion of a 30% or greater increase in the mean negative mood during the week before menses compared with the week after menses -- a more stringent criterion than that of DSM-IV or V.
We also recruited a group of 15 control women, none of whom had premenstrual mood symptoms using the same daily rating scales during a 2-month baseline period.
The women with PMDD had no current Axis I psychiatric diagnosis within the past 2 years per Structured Clinical Interview for DSM-IV (SCID), while controls had neither current nor past Axis I diagnoses (also confirmed by SCID).
This study was approved by the Central Neuroscience Institutional Review Board within the NIMH IRP. All women provided written informed consent, and all received payment for participation according to the NIH intramural guidelines.
Participants received six monthly injections of the GnRH agonist leuprolide acetate, Lupron (3.75 mg intramuscularly). Plasma follicle stimulating hormone (FSH), luteinizing hormone (LH), estradiol and progesterone levels were measured at study visits every 2 weeks to confirm adequate gonadal suppression. Following 3 months of Lupron alone, participants were randomly assigned in a double-blind, crossover manner to receive either E2 (100 mcg daily by skin patch, Noven Pharmaceuticals Jersey City, NJ, USA, Watson Pharmaceuticals, Parsippany, NJ, USA) or P4 (200 mg vaginal suppository twice daily, Upsher-Smith Laboratories; NIH Pharmacy, Bethesda, MD, USA) replacement lasting for 5 weeks (with a 2-week washout between hormone administration periods) while continuing to receive Lupron injections (Figure 1). All women used a patch and a suppository each day during the hormonal addback to maintain the patency of the double-blind. This standardized the exposures to physiologic levels of ovarian steroids to avoid the confound of varying levels of estradiol and progesterone across the normal menstrual cycle. All women used the appropriate placebo patch or suppository daily during the 12-week addback.
Women completed a modified DRF, daily to measure symptom severity during the 6-month hormone-manipulation study. A significant recurrence of PMDD symptoms was defined by a weekly average DRF score of greater than three (moderate severity) in irritability, anxiety or sadness. Asymptomatic controls had no affective symptoms during the 6 months (that is, no weekly average score ⩾2). At each clinic visit, all women completed the Rating for Premenstrual Tension self-ratings, a self-report rating that measures mood, behavior and physical symptoms on a 36-point scale, with scores ⩾ 10 consistent with PMDD symptoms.
Serum was obtained and stored at -80 °C for metabolomics analysis after at least 8 weeks of GnRH agonist-induced ovarian suppression (hypogonadal state and PMDD symptoms in remission), after 2-3 weeks of estradiol and after 2-3 weeks of progesterone addback (when symptoms recur in PMDD but not in control women).
We tested 49 metabolites with a focus on estradiol, progesterone and pregnenolone biosynthesis (see Table 1 and Supplementary Table S1, and the Kyoto Encyclopedia of Genes and Genomes: http://www.genome.jp/kegg/pathway.html). We selected direct precursors or metabolites of estradiol (estrone, estriol and so on), progesterone (pregnenolone, pregnenolone sulfate (3b-hydroxy-5-pregnen-20-one-3-SO), allopregnanolone (3a-hydroxy-5a-pregnan-20-one)) and testosterone (androstenedione, dehydroepiandrosterone (DHEA)) and metabolites with neuroactive potential. Of particular interest were steroids with direct actions at the GABA receptor, including pregnenolone and DHEA (direct or indirect positive modulators) or their sulfated metabolites, pregnenolone sulfate and dehydroepiandrosterone-SO (DHEAS; direct or indirect negative modulators). The sulfation of several neuroactive steroids converts agonist actions at GABA receptors to antagonist effects, whereas removal of the sulfate group has the opposite effect.
Finally, we measured levels of cortexolone, cortisol, corticosterone and cortexone to evaluate possible alternate pathways of steroid metabolism (analogous to the diversion of cholesterol precursors from corticosteroids toward sex steroid synthesis in congenital adrenal hyperplasia).
Serum samples were extracted and subjected to ultraperformance liquid chromatography tandem mass spectrometry analysis for measurement of neurosteroids in serum samples as described previously (see Supplementary Methods for detail).
To compare demographic and clinical characteristics of women with PMDD and control women (that is, age, body mass index (BMI), days of exposure to Lupron, E2 and P4, storage times, parity and race), we used Student's t-tests and χ-tests (or Fisher's exact test when appropriate), with a significance threshold of P<0.05, uncorrected for multiple comparisons. We employed analysis of variance with repeated measures (ANOVA-R) to examine potential differences in serum levels of estradiol and progesterone with diagnosis (that is, PMDD and control women) as a between-group factor and hormone condition (that is, Lupron alone, estradiol addback and progesterone addback) as the within-subject factor. Finally, we also used ANOVA-Rs to examine differences in symptom severity across hormone conditions or between diagnoses.
We did not use metabolite data with greater than 40% of values below the individual assay's limits of detectability (LOD) in any of the two diagnostic groups or in any of the three hormone conditions. These non-detectable metabolites were distributed across both diagnostic groups and all hormone conditions. Thus, metabolites meeting this criterion in any hormone condition were not included in analyses examining the change in the metabolite level from Lupron to E2 or Lupron to P4, or the differences across diagnoses in basal metabolite levels within each hormone condition.
To analyze differences in metabolite concentrations between the women with PMDD and control women, we used the open-source statistical software, R (R Foundation for Statistical Computing, Vienna, Austria; ISBN 3-900051-07-0 http://www.R-project.org/).
Women were first stratified by diagnostic group (PMDD and controls), and within each diagnostic group we tested for significant metabolite differences between each of the three hormone conditions using a non-parametric, paired Wilcoxon rank-sum test. Subsequently, changes in metabolite levels between PMDD and control women (that is, diagnosis-related differences in the 'delta' metabolite level) from Lupron to E2 and from Lupron to P4 were tested using an unpaired Wilcoxon rank-sum test. All results were corrected for multiple comparisons using a false-discovery rate approach (q<0.2).