Significant outcomes
During the menstrual cycle, there are highly significant changes in the load of gut commensal Gram-negative bacteria in serum with peaks at the end of the cycle.
Increased load of gut commensal Gram-negative bacteria at the end of the menstrual cycle is associated with premenstrual symptoms including fatigue, physio-somatic and anxiety symptoms, breast swelling and food cravings.
These changes may be driven by progesterone affecting transcellular, paracellular and vascular pathways.
Limitations
It would have been even more interesting if we had measured the gut microbiome and stool assays including indicants of the transcellular, paracellular and vascular pathways.
Introduction
Premenstrual syndrome (PMS) is defined as a constellation of physical, emotional and/or behavioural symptoms appearing during the luteal phase of the menstrual cycle and improving after the onset of menses (Deuster et al., Reference Deuster, Adera and South-Paul1998; Dickerson et al., Reference Dickerson, Mazyck and Hunter2003). However, there is no consensus definition for PMS and different diagnostic criteria have been proposed (see Table 1). The American College of Obstetricians and Gynecologists (ACOG) proposed that women with PMS must have at least one affective and one physical symptom appearing 5 days prior to menses for at least three menstrual cycles (American College of Obstetricians and Gynecologists, 2014). Moreover, the symptoms must be relieved within 4 days after the onset of menses without recurrence until at least day 13 of the menstrual cycle (American College of Obstetricians and Gynecologists, 2014). Another gold standard method used to diagnose PMS includes measurement of the Daily Record of Severity of Problems (DRSP): women with a total DRSP score ≥70 on day −5 to −1 of menses and having at least a 30% difference between pre- and postmenstrual scores are diagnosed with PMS (Endicott et al., Reference Endicott, Nee and Harrison2006; Biggs & Demuth, Reference Biggs and Demuth2011; Qiao et al., Reference Qiao, Zhang, Liu, Luo, Wang, Zhang and Ji2012). In a recent study, two new case definitions were identified, namely 1) peri-menstrual syndrome (PeriMS), which refers to women with increasing DRSP ratings during the peri-menstrual period (day 1 + day 2 + day 24–28); and 2) menstrual cycle-associated symptoms (MCAS), which delineates women with increased DRSP ratings all over the menstrual cycle (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019). Furthermore, we verified that the diagnosis of PMS according to Biggs and Demuth (Reference Biggs and Demuth2011) as well as the diagnoses of PeriMS and MCAS, but not the ACOG-based PMS diagnosis, were externally validated by levels of the sex hormones oestradiol and progesterone (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019). In addition, a diagnosis of PMS according to Biggs and Demuth (Reference Biggs and Demuth2011) was only predicted by lower steady-state levels of progesterone in the luteal phase (Biggs & Demuth, Reference Biggs and Demuth2011), while the PeriMS and MCAS diagnoses were significantly related to both sex hormones (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019). Lower steady-state levels of progesterone averaged over the luteal phase coupled with decreasing progesterone levels during the luteal phase also predicted changes in severity of the DRSP as well as alterations in severity of its four subdomains, namely a) depressive symptoms; b) fatigue and physio-somatic symptoms; c) increased appetite and craving combined with breast tenderness and swelling; and d) anxiety (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019). Therefore, we concluded that the diagnosis of PeriMS comprises the most accurate diagnostic criteria to describe changes in different symptoms dimensions in the periMS period and that the latter are at least in part mediated by sex hormones. Furthermore, it appeared that PeriMS is associated with a relative luteal phase deficiency or corpus luteum deficiency (Roomruanwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019).
Table 1. Definition of four different diagnoses used in the current study to diagnose ‘premenstrual’ syndrome
DRSP, daily record of severity of problems.
Recently, evidence indicates that increased translocation of Gram-negative gut commensal bacteria may play a pathophysiological role in major depression (Maes et al., Reference Maes, Kubera and Leunis2008, Reference Maes, Kubera, Leunis and Berk2012, Reference Maes, Kubera, Leunis, Berk, Geffard and Bosmans2013a; Martin-Subero et al., Reference Martin-Subero, Anderson, Kanchanatawan, Berk and Maes2016; Slyepchenko et al., Reference Slyepchenko, Maes, Machado-Vieira, Anderson, Solmi, Sanz, Berk, Kohler and Carvalho2016, Reference Slyepchenko, Maes, Jacka, Köhler, Barichello, McIntyre, Berk, Grande, Foster, Vieta and Carvalho2017), fatigue and physio-somatic symptoms (Maes et al., Reference Maes, Mihaylova and Leunis2007, Reference Maes, Ringel, Kubera, Anderson, Morris, Galecki and Geffard2013b, Reference Maes, Leunis, Geffard and Berk2014; Maes & Leunis, Reference Maes and Leunis2008), anxiety/stress (Gareau et al., Reference Gareau, Silva and Perdue2008; Galley & Bailey, Reference Galley and Bailey2014; Keightley et al., Reference Keightley, Koloski and Talley2015; Roomruangwong et al., Reference Roomruangwong, Kanchanatawan, Sirivichayakul, Anderson, Carvalho, Duleu, Geffard and Maes2017a; Sgambato et al., Reference Sgambato, Miranda, Ranaldo, Federico and Romano2017) and postpartum depression (Roomruangwong et al., Reference Roomruangwong, Kanchanatawan, Sirivichayakul, Anderson, Carvalho, Duleu, Geffard and Maes2017b, Reference Roomruangwong, Anderson, Berk, Stoyanov, Carvalho and Maes2018). However, it remains unclear whether bacterial translocation of Gram-negative bacteria could play a role in PMS or PeriMS and its four symptom domains.
This hypothesis is conceivable since sex hormones may modulate gut permeability (Edwards et al., Reference Edwards, Cunningham, Dunlop and Corwin2017). Furthermore, studies in pregnancy and postpartum, which are periods of dramatic changes in sex hormonal state, have reported altered gut functions and bacterial composition (Brantsaeter et al., Reference Brantsaeter, Myhre, Haugen, Myking, Sengpiel, Magnus, Jacobsson and Meltzer2011; Koren et al., Reference Koren, Goodrich, Cullender, Spor, Laitinen, Backhed, Gonzalez, Werner, Angenent, Knight, Backhed, Isolauri, Salminen and Ley2012). These hormonal changes may affect gut contractility thereby increasing gut transit time (Mayer et al., Reference Mayer, Savidge and Shulman2014), which may constitute an adaptive response to allow a better absorption of nutrients during pregnancy (Edwards et al., Reference Edwards, Cunningham, Dunlop and Corwin2017). Furthermore, pregnancy is accompanied by decreased gut permeability and a lowered bacterial translocation as indicated by significantly decreased IgA responses to Gram-negative bacteria, suggesting that pregnancy (with relatively high levels of oestrogen and progesterone) could attenuate bacterial translocation (Roomruangwong et al., Reference Roomruangwong, Kanchanatawan, Sirivichayakul, Anderson, Carvalho, Duleu, Geffard and Maes2017a,b). Another study found an increased susceptibility to Listeria monocytogenes infection during pregnancy leading to adverse obstetrics outcomes including preterm delivery or stillbirth, which were partly modulated by elevated oestrogen and progesterone levels (Garcia-Gomez et al., Reference Garcia-Gomez, Gonzalez-Pedrajo and Camacho-Arroyo2013). In patients with irritable bowel syndrome , sex hormones may affect peripheral and central regulatory processes of the brain–gut axis, leading to alterations in visceral sensitivity, intestinal barrier function and immune activation of intestinal mucosa (Mulak et al., Reference Mulak, Tache and Larauche2014). Cyclical changes of ovarian hormones during the menstrual cycle can arguably modulate gastrointestinal (GI) functions including small intestinal transit, gastric emptying and mucosal blood flow (Heitkemper et al., Reference Heitkemper, Cain, Jarrett, Burr, Hertig and Bond2003; Longstreth et al., Reference Longstreth, Thompson, Chey, Houghton, Mearin and Spiller2006). Lowered levels of ovarian hormone levels during menses are associated with exacerbations of GI symptoms including abdominal discomfort, bowel habit changes and bloating (Whitehead et al., Reference Whitehead, Cheskin, Heller, Robinson, Crowell, Benjamin and Schuster1990; Moore et al., Reference Moore, Barlow, Jewell and Kennedy1998; Mulak & Taché, Reference Mulak and Taché2010). However, there are no data whether changes in sex hormones during the menstrual cycle are associated with increased bacterial translocation.
Hence, the current study was carried out to examine whether increasing plasma IgA levels to lipopolysaccharides (LPS) of Gram-negative bacteria during the menstrual cycle could be associated with the pathophysiology of PMS or PeriMS and whether those associations could be related to alterations in sex hormones during the menstrual cycle.
Methods
Participants
Forty female participants aged 18–45 years were recruited by word of mouth at the King Chulalongkorn Memorial Hospital during the period of April–May 2018, including 20 women with subjective complaints of PMS and 20 women without such complaints. Participants comprised hospital’s staffs or friends/relatives of hospital’s staffs and women accompanying patients to the hospital. Inclusion criteria were: 1) women aged 18–45 years; 2) having a regular menstrual cycle with a cycle length of 27–30 days during the past year; 3) being able to read and write in Thai; 4) willing to have four blood samples drawn at day 7 (T1), day 14 (T2), day 21 (T3) and day 28 (T4) of the menstrual cycle; and 5) able to complete the DRPS ratings for all consecutive days of the menstrual cycle. Exclusion criteria for both groups were: 1) those with a lifetime history of psychiatric illness (including major depression, bipolar disorder, schizophrenia and obsessive compulsive disorder); 2) those with a history of medical illness, including type 1 diabetes and autoimmune/immune-inflammatory disorders (including rheumatoid arthritis, inflammatory bowel disease, psoriasis and multiple sclerosis); 3) pregnant women or women who are currently using hormonal contraceptive agents; and 4) women who are currently using any psychotropic medications. The study was approved by the Ethics Committee of the Faculty of Medicine, Chulalongkorn University, Bangkok, Thailand (IRB No.611/60, COA No. 1111/2017). Written informed consent was obtained from all participants prior to the study. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Clinical assessments
All participants were requested to complete a demographic and clinical data questionnaire, that is, menstrual information, age, education, height, weight, a history of substance use and life style, and they were evaluated by an experienced psychiatrist before enrolment in the study to rule out other medical and/or psychiatric conditions. All participants completed the DRSP during all consecutive days of their menstrual cycle starting on day 1 of menses to assess the severity of PMS symptoms. The DRSP consists of 21 items + 3 functional impairment items commonly used to assess PMS (Endicott et al., Reference Endicott, Nee and Harrison2006). All items are rated from 1 to 6 (1 = not at all, 2 = minimal, 3 = mild, 4 = moderate, 5 = severe, 6 = extreme). The DRSP is a self-report instrument that rates both the ‘presence’ and ‘severity’ of premenstrual symptoms and that can be used to reliably screen for a DSM-IV diagnosis of premenstrual dysphoric disorder (Biggs & Demuth, Reference Biggs and Demuth2011). The presence of PMS was considered when the total DRSP score was ≥70 on day −5 to −1 of menses and when there was a 30% difference between premenstrual (day −5 to −1) and postmenstrual (day 6–10) scores (Endicott et al., Reference Endicott, Nee and Harrison2006; Biggs & Demuth, Reference Biggs and Demuth2011; Qiao et al., Reference Qiao, Zhang, Liu, Luo, Wang, Zhang and Ji2012). In addition, participants were also categorised in those who had PeriMS with increased DRSP ratings during the peri-menstrual period (day 1+ day 2 + day 24–28) and MCAS (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019). We also computed scores of the four subdomains of the DRSP, namely a) depressive dimension; b) physio-somatic component; c) increased appetite and craving combined with breast tenderness and swelling; and d) anxiety dimension (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019).
Assays
In all women, we sampled fasting blood at 8.00 a.m. at T1, T2, T3 and T4 for the assay of IgA directed to Gram-negative bacteria, oestradiol and progesterone. We described in detail elsewhere the assay to detect IgA antibodies directed to Gram-negative bacteria (Roomruangwong et al., Reference Roomruangwong, Kanchanatawan, Sirivichayakul, Anderson, Carvalho, Duleu, Geffard and Maes2017a). Briefly, LPS derived from Gram-negative bacteria were assayed, namely Hafnia alvei, Klebsiella pneumonia, Morganella morganii, Pseudomonas aeruginosa, Citrobacter koseri and Pseudomonas putida. Polystyrene 96-well plates (NUNC) were coated with 200 µl solution containing bacterial components at 4 µg/ml in 0.05 M carbonate buffer at pH 9.6. Well plates were incubated at 4°C for 16 h under agitation. Then, we added 200 µl blocking solution (PBS, Tween 20 0.05%, 5 g/l BSA) for 1 h and placed at 37°C. Following two washes with PBS, plates were filled up with 100 µl of sera diluted at 1 : 1000 in the blocking buffer A (PBS, 0.05% Tween 20, 2.5 g/l BSA) and incubated at 37°C for 105 min. After three washes with PBS-0.05% Tween 20, plates were incubated at 37°C for 1 h with peroxidase-labelled anti-human IgA secondary antibodies diluted, respectively, at 1 : 15 000 and 1 : 10 000 in the blocking buffer (PBS, 0.05% Tween 20, 2.5 g/l BSA). Afterwards, plates were washed three times with PBS-0.05% Tween 20 and incubated with the detection solution for 10 min in the dark. Chromogen detection solution (tetramethylbenzedine) was used for the peroxidase assay at 16.6 ml per liter in 0.11 M sodium acetate trihydrate buffer (pH 5.5) containing 0.01% H2O2. The reaction was stopped with 25 µl 2-N HCl. After addition of stop solution (H2SO4 or HCl), the obtained, proportional absorbance in the tested sample (compared to established concentration of respective antibodies), was measured at 450 nm with one alpha of correction at 660 nm.
The methods to assay both sex hormones were also described in detail previously (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019). In brief, we used an immunoassay for the quantitative determination of estradiol and progesterone using Cobas® 601. For estradiol, the two steps of assay included: 1) first incubation: incubating the sample (25 μl) with two estradiol-specific biotinylated antibodies, immune complexes are formed, the amount of which is dependent upon the analyte concentration in the sample; 2) second incubation: after addition of streptavidin-coated microparticles and an estradiol derivative labelled with a ruthenium complex, the still-vacant sites of the biotinylated antibodies become occupied, with formation of an antibody–hapten complex. The entire complex becomes bound to the solid phase via interaction of biotin and streptavidin, and the reaction mixture is aspirated into the measuring cell where the microparticles are magnetically captured onto the surface of the electrode. Unbound substances are then removed with ProCell/ProCell M. Application of a voltage to the electrode then induces chemiluminescent emission which is measured by a photomultiplier. Precision was determined using Elecsys reagents, samples and controls in a protocol (EP5-A2) of the Clinical and Laboratory Standards Institute (CLSI): two runs per day in duplicate each for 21 days (n = 84) with the intra-assay CV value of 1.2%. For progesterone, the two steps of assay included: 1) first incubation: incubating the sample (20 μL) with a progesterone-specific biotinylated antibody, immunocomplexes are formed, the amount of which is dependent upon the analyte concentration in the sample; 2) second incubation: after addition of streptavidin-coated microparticles and an progesterone derivative labelled with a ruthenium complex, the still-vacant sites of the biotinylated antibodies become occupied, with formation of an antibody–hapten complex. The entire complex becomes bound to the solid phase via interaction of biotin and streptavidin, and the reaction mixture is aspirated into the measuring cell where the microparticles are magnetically captured onto the surface of the electrode. Unbound substances are also removed with ProCell/ProCell M. Application of a voltage to the electrode then induces chemiluminescent emission which is measured by a photomultiplier. Results are determined via a calibration curve which is instrument specifically generated by two-point calibration and a master curve provided via the reagent barcode. Precision was also determined using Elecsys reagents, samples and controls in a protocol (EP5-A2) of the CLSI as in estradiol: two runs per day in duplicate each for 21 days (n = 84) with an intra-assay CV value of 2.3%.
Statistics
We used analysis of contingency tables (χ 2 test) and analysis of variance (ANOVA) to assess associations between categorical variables and differences in continuous variables between diagnostic groups, respectively. Generalised estimating equation (GEE) analysis, repeated measures, was used to check effects of time, diagnosis and time × diagnosis interaction on the IgA levels, while adjusting for age, cycle length, age of menarche and duration of menses. Using GEE analyses, repeated measurements, we also examined the relationships among the IgA levels to Gram-negative bacteria and either the DRSP values over time (T1, T2, T3 and T4) or changes in sex hormones during the menstrual cycle. Furthermore, we used a distributed lag model to predict the DRPS values over time (dependent variable) by lagged (1 week) values of the IgA responses to Gram-negative bacteria and we computed the ΔIgA responses as current IgA values − lagged IgA values obtained 1 week earlier, which denotes the changes in IgA values the last week before blood sampling. We also use steady-state hormonal levels, namely the sum of the z scores of the progesterone hormone levels at T2, T3 and T4 (zT2 + T3 + T4). Tests were two-tailed and a p-value of 0.05 was considered for statistical significance. All statistical analyses were performed using IBM SPSS windows version 25.
Results
Demographic and clinical data
Table 2 shows the demographic and clinical data in participants with and without PMS. There were no significant differences in age, years of education, age of menarche, cycle length, duration of menses, total DRSP scores and BMI between groups.
Table 2. Demographic and clinical data of women with and without PMS
PMS, premenstrual syndrome; BMI, body mass index; DRSP, daily record of severity of problems.
All results are shown as mean (SD).
PMS: diagnosis according to the criteria of the American College of Obstetricians and Gynecologists.
Table 3 shows the DRSP score and subscores at the four different time points, T1, T2, T3 and T4. Thus, there were highly significant variations in those scores all over the menstrual cycle with higher total DRSP and physio-somatic scores at T4 compared to the other time points, and higher at T1 compared to T2 and T3, while T3 showed higher scores than T2. In addition, depression scores were higher at T4 than at T2 and T3, at T1 than at T2, while there were no differences between T2 and T3. Breast-craving and anxiety symptoms were higher at T4 than at T1, T2 and T3, while lowest scores were detected at T2.
Table 3. Measurements of DRSP and subdomains, and plasma levels of oestradiol and progesterone during the menstrual cycle
DRSP, daily record of severity of problems.
Menstrual cycle-associated changes in IgA levels to Gram-negative bacteria
In Table 4, we examine the effects of time on IgA and ΔIgA (i.e. actual value − value 1 week earlier) responses to the Gram-negative bacteria. The data were analysed using GEE analysis considering effects of time, time × PMS diagnosis (according to the four definitions) and PMS diagnosis, while adjusting for age, cycle length, age of menarche and duration of menses. There were highly significant effects of time on the six IgA and ΔIgA levels to Gram-negative bacteria. Table 4 shows differences in ΔIgA responses to the six Gram-negative bacteria at the four different time points of the menstrual cycle. Peak ΔIgA levels for all Gram-negative bacteria were detected at T4. The lowest ΔIgA responses were detected at T1 (for C. koseri) or T2 (for all other bacteria). The ΔIgA responses were significantly higher at T4 than at T1, T2 or T3 for P. putida, H. Alvei, P. aeruginosa and M. morganii and significantly higher at T4 than T1 and T2 for C. koseri and Klebsiella pneumoniae. There were no significant differences between any of the ΔIgA values between T1 and T2. The ΔIgA values at T3 occupied an intermediate position with values which were often significantly different from T2 and T4. Fig. 1 shows the mean ΔIgA values (in z scores) across the four time points. As an index of the overall LPS load, we computed a z unit-weighted composite score, namely the sum of all z ΔIgA values. Table 4 shows that there were highly significant differences in this overall index with significantly higher values at T4 than the other three time points while the values were higher at T3 than T2 and no differences between T1 and T2 could be established. GEE analyses showed that the effects of time on IgA directed to Gram-negative bacteria were highly significant and that peak levels were obtained at T4 with lows at T2 or T3 (not significantly different) while IgA levels to LPS at T1 occupied an intermediate position. There were no significant effects of diagnosis (using the four diagnostic criteria) or the interaction term time × diagnosis on the IgA or ΔIgA to LPS of Gram-negative bacteria. GEE analyses showed that there were significant and positive effects of age on the ΔIgA levels to H. alvei (W = 17.87, df = 1, p < 0.001), K. pneumoniae (W = 4.51, df = 1, p = 0.034) and P. aeruginosa (W = 5.46, df = 1, p = 0.019). There were also significant and positive effects of cycle length on ΔIgA to H. alvei (W = 5.71, df = 1, p = 0.017) and P. putida (W = 7.60, df = 1, p = 0.006).
Table 4. Results of GEE analysis, that is, effects of time on the ΔIgA responses directed against LPS of six Gram-negative bacteria as dependent variables
GEE, generalised estimating equation.
Results are shown as mean (±SE) and as z scores.
Δ: computed as actual value − values 1 week earlier.
Fig. 1.
Prediction of DRSP symptoms by IgA response to Gram-negative bacteria
Table 5 shows the associations between total DRSP and subdomain scores (as dependent variables) and changes in IgA responses to Gram-negative bacteria during the menstrual cycle (explanatory variables). We used GEE analysis, repeated measures, to analyse these associations and entered the actual measurements of IgA responses as well as the Δ responses in the analyses. We detected that the changes in DRSP were significantly associated with the Δ (but not actual) IgA levels of H. alvei, M. morganii or P. putida. We also found significant associations between changes in the severity of fatigue and physio-somatic and breast-craving symptoms with the ΔIgA levels to LPS of the same three bacteria, while changes in ΔIgA responses to C. koseri also predicted breast-craving symptoms. Changes in anxiety symptoms were predicted by ΔIgA responses to H. alvei.
Table 5. Results of GEE analysis with the DRSP score and subdomains as dependent variables
GEE, generalised estimating equation; DRSP, daily record of severity of problems.
Lag progesterone: 1 week lagged values.
Δ: computed as actual value − values of 1 week earlier.
In addition, we have carried out a second series of GEE analysis whereby we entered ΔIgA responses to LPS together with the lagged progesterone values and the zT2 + T3 + T4 progesterone scores as explanatory variables (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019). Table 5 shows that after considering the effects of both progesterone values, the effects of ΔIgA values on the DRSP and anxiety scores were no longer significant. Nevertheless, the effects of ΔIgA responses to P. putida on physio-somatic and breast-craving symptoms remained significant. The oestradiol values were not significant in these GEE analyses and the effects of the IgA levels to different bacteria remained significant after introducing oestradiol data.
Associations between IgA responses to Gram-negative bacteria and sex hormones
In Table 6, we examine the effects of progesterone (explanatory variables) on the ΔIgA levels to Gram-negative bacteria (dependent variables). We used three different progesterone levels, namely the lagged progesterone values, the Δ changes and the steady-state progesterone values averaged over the second part of the cycle (zT2 + T3 + T4). The Δ changes in H. alvei, P. putida, C. koseri and P. aeruginosa were significantly associated with the lagged progesterone values (positively), the Δ changes in progesterone (positively) and zT2 + T3 + T4 (negatively). The ΔIgA responses to M. morganii and K. pneumoniae were significantly associated with the lagged progesterone data (again positively) and zT2 + T3 + T4 (again negatively). In addition, another z composite score denoting the ratio between steady-state progesterone/steady-state oestradiol values (computed as z(zT1 + zT2 + zT3 + zT4) progesterone − z(zT1 + zT2 + zT3 + zT4) oestradiol values) was significantly associated (inversely) with the ΔIgA data and could be used instead of the zT1 + T2 +T3 progesterone scores shown in Table 6 (same significance levels).
Table 6. Results of GEE analysis with the IgA directed to LPS of six Gram-negative bacteria as dependent variables
GEE, generalised estimating equation.
Lag progesterone: 1 week lagged values.
Δ: computed as actual value − values 1 week earlier.
Also, the IgA response to LPS of Gram-negative bacteria was significantly associated with the lagged progesterone data but the effects of progesterone were markedly less as compared with the ΔIgA data. Thus, the lagged progesterone levels were significantly associated with the IgA levels to LPS of C. koseri (W = 5.23, df = 1, p = 0.022), P. putida (W = 10.16, df = 1, p = 0.001), K. pneumonia (W = 4.36, df = 1, p = 0.037) and M. morganii (W = 4.64, df = 1, p = 0.031), but not H. alvei or P. aeruginosa.
Discussion
The first major finding of this study is that there are highly significant changes in the six IgA levels to Gram-negative bacteria during the menstrual cycle. Overall, peak changes in IgA levels to LPS of all bacteria were observed at T4 (day 28) with lows at T1 (day 7) or T2 (day 14). These results indicate that women exhibit common rhythms in IgA responses to LPS during the menstrual cycle and by inference that changes in LPS load in the plasma and, consequently, in bacterial translocation may ensue during the menstrual cycle. Phrased differently, our findings indicate increased LPS load at the end of the menstrual cycle with a corresponding reduction in LPS load of potentially harmful pathogens after menstruation. In this regard, Profet hypothesised that menstruation may help to clean the vaginal tract of pathogens (Profet, Reference Profet1993), although in 1993 there was no evidence for elevated pathogen load before menstruation.
To the best of our knowledge, there are no previous studies suggesting significant menstrual cycle-associated rhythms in LPS load. Previously, no dysfunctions in gut permeability were observed during the menstrual cycle in normal women using the lactulose/mannitol test, a less sensitive test to assess leaky gut (Torella et al., Reference Torella, Colacurci, De Franciscis, Cuomo, Gallo, Familiali, Carteni and de Margistris2007). Nevertheless, one study demonstrated a relationship between gut microbiota and an irregular menstrual cycle as indicated by a relative Prevotella-enriched microbiome, but lower Bacteroidales S24-7, Clostridiales, Ruminococcus and Lachnospiraceae (Sasaki et al., Reference Sasaki, Kawamura, Kawamura, Odamaki, Katsumata, Xiao, Suzuki and Tanaka2019). Prevotella is associated with increased gut permeability since it may degrade mucin (Brown et al., Reference Brown, Davis-Richardson, Giongo, Gano, Crabb, Mukherjee, Casella, Drew, Ilonen, Knip, Hyoty, Veijola, Simell, Simell, Neu, Wasserfall, Schatz, Atkinson and Triplett2011), whereas Clostridiales, Ruminococcus and Lachnospiraceae are butyrate-producing bacteria, which play a role in maintaining gut homeostasis (Hamer et al., Reference Hamer, Jonkers, Venema, Vanhoutvin, Troost and Brummer2008; Pryde et al., Reference Pryde, Duncan, Hold, Stewart and Flint2002) through providing energy sources to intestinal epithelial cells and producing anti-inflammatory effects (Inan et al., Reference Inan, Rasoulpour, Yin, Hubbard, Rosenberg and Giardina2000). Moreover, decreased mucin production may lead to a micro-inflammatory environment which may be associated with ovulatory disorders (Sasaki et al., Reference Sasaki, Kawamura, Kawamura, Odamaki, Katsumata, Xiao, Suzuki and Tanaka2019) as indicated by recent findings that inflammation may exert a detrimental effect on ovarian follicle growth and ovulation (Boots & Jungheim, Reference Boots and Jungheim2015).
Secondly, the immune characteristics of the female reproductive tract may share some similarities with those of the gut (Shacklett & Greenblatt, Reference Shacklett and Greenblatt2011). There are significant differences in microbiota in the female reproductive tract between the phases of the menstrual cycle (Chen et al., Reference Chen, Song, Wei and Zhong2017). For example, increased presentation of Lactobacillus species, Sphingobium sp., Propionibacterium acnes and Pseudomonas sp. during the proliferative (day 1–14) and secretory (day 15–28) phases, whereas P. acnes appeared to be more abundant during the secretory phase. Overall, the proliferative phase appeared to be associated with increased bacterial proliferation when compared to the secretory phase as indicated by higher pyrimidine and purine metabolism, aminoacyl-tRNA and peptidoglycan biosynthesis, whereas during secretory phase, porphyrin, arginine and proline metabolism were increased, as well as the degradation of benzoate, nitrotoluene and biosynthesis of siderophore. Studies in primates also found that vaginal microbial ecologies are highly affected by the menstrual cycle, especially during the estrous phase (Keane et al., Reference Keane, Ison and Taylor-Robinson1997; Narushima et al., Reference Narushima, Itoh, Sankai, Takasaka, Otani and Yoshikawa1997; Gajer et al., Reference Gajer, Brotman, Bai, Sakamoto, Schutte, Zhong, Koenig, Fu, Ma, Zhou, Abdo, Forney and Ravel2012). In humans, high midcycle oestrogen levels are associated with increased Lactobacillus proliferation (Boskey et al., Reference Boskey, Telsch, Whaley, Moench and Cone1999, Reference Boskey, Cone, Whaley and Moench2001), whereas increased mucosal secretions are associated with growth of Candida (Schwebke & Weiss, Reference Schwebke and Weiss2001). High levels of oestrogen and progesterone during midcycle are associated with higher stability of microbial communities (Gajer et al., Reference Gajer, Brotman, Bai, Sakamoto, Schutte, Zhong, Koenig, Fu, Ma, Zhou, Abdo, Forney and Ravel2012), whereas there is a lower prevalence, intensity and diversity of microbiota during menstruation (Stumpf et al., Reference Stumpf, Wilson, Rivera, Yildirim, Yeoman, Polk, White and Leigh2013).
The second major finding of our study is that there were significant associations between the Δ changes in the IgA responses to LPS and the DRSP scores and its subdomains. Thus, the Δ changes in H. alvei, M. morganii and P. putida were significantly associated with changes in the total DRSP scores, physio-somatic symptoms and breast-craving symptoms, while H. alvei was also associated with anxiety. As such, the Δ changes in IgA responses to LPS of Gram-negative bacteria are associated with all symptom domains of the DRSP, except depression. Our current findings extent those of previous studies indicating that IgA levels to Gram-negative bacteria are significantly correlated with physio-somatic symptoms in depression and CFS/ME (Maes et al., Reference Maes, Kubera and Leunis2008; Maes & Leunis, Reference Maes and Leunis2008). Gut microbiota also influences the host’s appetite and food intake by modulating nutrient sensing and appetite and satiety-regulating systems (Turnbaugh et al., Reference Turnbaugh, Ley, Mahowald, Magrini, Mardis and Gordon2006; Huang & Douglas, Reference Huang and Douglas2015; Leitao-Goncalves et al., Reference Leitao-Goncalves, Carvalho-Santos, Francisco, Fioreze, Anjos, Baltazar, Elias, Itskov, Piper and Ribeiro2017; van de Wouw & Schellekens, Reference van de Wouw and Schellekens2017). In animal studies, essential amino acids and the concerted action of the commensal bacteria Acetobacter pomorum and Lactobacilli significantly modulate food choice, especially towards amino acid-rich food (Leitao-Goncalves et al., Reference Leitao-Goncalves, Carvalho-Santos, Francisco, Fioreze, Anjos, Baltazar, Elias, Itskov, Piper and Ribeiro2017). Studies in patients with anorexia nervosa found significantly lower alpha (within-sample) diversity in taxa abundance between admission and after discharge from hospital when compared to healthy controls, while severity of depression, anxiety and eating problems were associated with the composition and diversity of the intestinal microbiota (Kleiman et al., Reference Kleiman, Watson, Bulik-Sullivan, Huh, Tarantino, Bulik and Carroll2015). There are also profound microbial perturbations in patients with anorexia nervosa with higher levels of mucin degraders and members of Clostridium clusters I, XI and XVIII and lowered levels of the butyrate-producing Roseburia sp., while in anorexia nervosa patients with restrictive and binge/purging subtypes distinct perturbations in microbial community compositions were observed (Mack et al., Reference Mack, Cuntz, Gramer, Niedermaier, Pohl, Schwiertz, Zimmermann, Zipfel, Enck and Penders2016).
Moreover, the associations found in our study between changes in IgA to LPS of Gram-negative bacteria and breast symptoms may be explained by possible effects of the gut microbiome on breast symptoms via the modulating effects of oestrogen. Plottel and Blaser (Reference Plottel and Blaser2011) proposed the ‘estrobolome’ as the aggregate of enteric bacterial genes whose products are capable of metabolising oestrogens (Plottel & Blaser, Reference Plottel and Blaser2011). Under normal conditions, oestrogens and their metabolites are conjugated in the liver through glucuronidation or sulfonation to allow for biliary excretion (Zhu & Conney, Reference Zhu and Conney1998). Conjugated oestrogens are excreted in bile, urine and feces (Raftogianis et al., Reference Raftogianis, Creveling, Weinshilboum and Weisz2000). Nevertheless, approximately 65% of estradiol is recovered in bile, 10–15% is found in feces while a significant proportion of oestrogens is reabsorbed into the circulation (Sandberg & Slaunwhite, Reference Sandberg and Slaunwhite1957; Adlercreutz & Martin, Reference Adlercreutz and Martin1980; Adlercreutz & Jarvenpaa, Reference Adlercreutz and Jarvenpaa1982). This reabsorption of hepatically conjugated oestrogens is mediated by deconjugation processes by gut bacteria with β-glucuronidase activity such as the Clostridium leptum and Clostridium coccoides cluster, and the Escherichia/Shigella bacterial group (Gloux et al., Reference Gloux, Berteau, El Oumami, Beguet, Leclerc and Dore2011; Kwa et al., Reference Kwa, Plottel, Blaser and Adams2016; Fernandez & Reina-Perez, Reference Fernandez and Reina-Perez2018). Thus, a deconjugating enzyme-enriched estrobolome could promote reabsorption of free oestrogens thereby increasing oestrogen levels, which may contribute to breast tissue changes (Kwa et al., Reference Kwa, Plottel, Blaser and Adams2016; Fernandez & Reina-Perez, Reference Fernandez and Reina-Perez2018).
In the current study, we also found a significant association between the anxiety subdomain of the DRSP and increased LPS load in the plasma. These findings extent our previous results that increased IgA responses to P. aeruginosa at the end of term pregnancy are associated with anxiety 4–6 weeks after delivery (Roomruangwong et al., Reference Roomruangwong, Kanchanatawan, Sirivichayakul, Anderson, Carvalho, Duleu, Geffard and Maes2017a). It is plausible that the above associations between increasing LPS load and symptom domains including anxiety and physio-somatic symptoms may be explained by low-grade immune-inflammatory responses induced by LPS activation of the toll-like receptor-4 complex, a receptor of the innate immune system which upon activation causes release of reactive oxygen species, cytokines and nitric oxide (Lucas & Maes, Reference Lucas and Maes2013). This theory is corroborated by findings that increased root canal endotoxin in subjects with chronic apical periodontitis is associated with increased nitro-oxidative stress and depressive symptoms (Gomes et al., Reference Gomes, Martinho, Barbosa, Antunes, Povoa, Baltus, Morelli, Vargas, Nunes, Anderson and Maes2018). Moreover, repeated and intermittent administration of LPS may induce depressive-like behaviours in the rodent in association with increased microglial activation and increased levels of nuclear factor-kB, superoxide and cytokine production, lowered tryoptophan and increased neurotoxic tryptophan catabolites (Kubera et al., Reference Kubera, Curzytek, Duda, Leskiewicz, Basta-Kaim, Budziszewska, Roman, Zajicova, Holan, Szczesny, Lason and Maes2013; Rodrigues et al., Reference Rodrigues, de Souza, Lima, da Silva, Costa, Dos Santos, Miyajima, de Sousa, Vasconcelos, Barichello, Quevedo, Maes, de Lucena and Macedo2018). Administration of LPS to humans not only induces the levels of pro- and anti-inflammatory cytokines, but also lowers mood, and induces anxiety and social disconnection (Eisenberger et al., Reference Eisenberger, Inagaki, Mashal and Irwin2010; Grigoleit et al., Reference Grigoleit, Kullmann, Wolf, Hammes, Wegner, Jablonowski, Engler, Gizewski, Oberbeck and Schedlowski2011). All in all, our findings may indicate that variations in LPS of Gram-negative bacteria during the menstrual cycle with peaks at the end of day 28 of the menstrual cycle could play a pathophysiological role in premenstrual and PeriMS symptoms.
The third major finding of our study is that many, but not all, associations between ΔIgA responses to LPS and symptom domains disappeared after introducing progesterone and changes in progesterone levels in the GEE analyses, although the effects of P. putida on physio-somatic symptoms and breast-craving remained significant. This may be explained as the increments in IgA responses to LPS are largely predicted by increasing progesterone levels coupled with lowered steady-state progesterone levels or a relative increase in oestradiol steady-state levels versus those of progesterone. Progesterone receptors are present in colon epithelial cells where they interact with progesterone and modulate the colonic transit time (Guarino et al., Reference Guarino, Cheng, Cicala, Ripetti, Biancani and Behar2011). The colonic transit time is longer during the luteal phase (high progesterone) when compared to the follicular phase (low progesterone) (Wald et al., Reference Wald, Van Thiel, Hoechstetter, Gavaler, Egler, Verm, Scott and Lester1981; Jung et al., Reference Jung, Kim and Moon2003). Progesterone also impairs smooth muscle contraction (Xiao et al., Reference Xiao, Biancani and Behar2009; Li et al., Reference Li, Ling, Biancani and Behar2012) and downregulates the barrier function of tight junctions which may contribute to cytoskeletal remodeling (Someya et al., Reference Someya, Kojima, Ogawa, Ninomiya, Nomura, Takasawa, Murata, Tanaka, Saito and Sawada2013) in uterine endometrium. Progesterone promotes endometrial remodeling via modifications of actin fibres architecture, which leads to cell membrane reshaping and movement (Pfaendtner et al., Reference Pfaendtner, Lyman, Pollard and Voth2010; Shortrede et al., Reference Shortrede, Montt-Guevara, Pennacchio, Finiguerra, Giannini, Genazzani and Simoncini2018; Svitkina, Reference Svitkina2018). Moreover, progesterone controls actin polymerisation, branching and focal adhesion complex formation via membrane-organising extension spike protein and focal adhesion kinase (Sanchez et al., Reference Sanchez, Flamini, Genazzani and Simoncini2013; Shortrede et al., Reference Shortrede, Montt-Guevara, Pennacchio, Finiguerra, Giannini, Genazzani and Simoncini2018). Adhesion assembly in uterine epithelial cells is regulated by progesterone while oestrogens concentrate talin and paxillin (Kaneko et al., Reference Kaneko, Lecce and Murphy2009). Progesterone also induces dickkopf homologue 1 (DKK1) and forkhead box O1 (FOXO1), resulting in inhibition of Wnt/β-catenin signalling in the human endometrium (Wang et al., Reference Wang, Hanifi-Moghaddam, Hanekamp, Kloosterboer, Franken, Veldscholte, van Doorn, Ewing, Kim, Grootegoed, Burger, Fodde and Blok2009). Moreover, oestrogens play a role in the tight junctions in the gut by decreasing zonula occludens 1 mRNA and protein expression thereby increasing gut permeability (Zhou et al., Reference Zhou, Zhang, Ding, Luo, Yuan, Bansal, Gilkeson, Lang and Jiang2017). Moreover, oestrogens increase mucin protection in intestinal epithelial cells thereby decreasing gut permeability (Diebel et al., Reference Diebel, Diebel, Manke and Liberati2015). As such, increasing progesterone levels in the luteal phase may possibly affect the tight and adherens junctions of the paracellular pathway, the transcellular (talin) and the vascular barrier (catenin) pathways, which all protect against bacterial translocation (Maes et al., Reference Maes, Sirivichayakul, Kanchanatawan and Vojdani2019). Moreover, lowered steady-state levels of progesterone may be associated with upregulated progesterone receptors (Saracoglu et al., Reference Saracoglu, Aksel, Yeoman and Wiebe1985), which may increase sensitivity of, for example, colon muscle cells to progesterone (Cheng et al., Reference Cheng, Pricolo, Biancani and Behar2008). As a consequence, relatively small increments in progesterone coupled with upregulated progesterone receptors and relatively higher oestradiol steady-state levels could contribute to increased gut permeability and, in turn, bacterial translocation thereby stimulating IgA production 5–7 days later (Cerutti, Reference Cerutti2008). As such, changes in progesterone during the menstrual cycle coupled with a relative corpus luteum insufficiency (Roomruangwong et al., Reference Roomruangwong, Carvalho, Comhaire and Maes2019) may drive menstrual cycle-associated increments in IgA responses to LPS and thus PMS/PeriMS symptoms. Nevertheless, no studies have examined the effects of sex hormones on the gut tight and adherens junctions and the gut vascular barrier.
The current findings should be interpreted within its limitations. First, it would have been even more interesting if we had measured the gut microbiome and stool assays including direct indicants of gut dysbiosis (Simeonova et al., Reference Simeonova, Ivanovska, Murdjeva, Carvalho and Maes2018). Second, we enrolled a relatively small sample to detect associations between the biomarkers and PMS or PeriMS classifications. Nevertheless, the strengths of the study are that we examined associations over time between biomarker measurements and clinical data during the menstrual cycle. Interestingly, while the repeated measurements in IgA responses were significantly associated with those in symptoms, no associations could be detected between LPS data and any of the diagnoses of PMS or PeriMS. This indicates that research in PMS or PeriMS should always examine the associations over time between biomarkers and affective, fatigue and physio-somatic symptoms because a diagnosis of PMS/PeriMS is a limited aspect of peri-menstrual symptoms that cannot capture those associations over time.
In conclusion, during the menstrual cycle there are significant changes in IgA responses to LPS of Gram-negative bacteria with peaks in the late luteal phase and lows from week 1 to ovulation. Increments in progesterone during the menstrual cycle superimposed on lowered steady-state progesterone levels during the cycle may drive those menstrual cycle-associated alterations in IgA responses to LPS thereby contributing to severity of peri-menstrual, physio-somatic, anxiety, food cravings and breast swelling symptoms.
Acknowledgement
The laboratory assays were supported by Center for Medical Diagnostic Laboratories (CMDL), Faculty of Medicine, Chulalongkorn University.
Author contributions
CR and MM made the design of the study. CR recruited and screened the participants. MM performed statistical analyses. MG performed analyses. AC contributed in a meaningful way to the intellectual content of this paper. All authors agreed upon the final version of the paper.
Financial support
This research has been supported by 1) the Ratchadaphiseksomphot Fund, Faculty of Medicine, Chulalongkorn University, grant number RA61/016; 2) Chulalongkorn University; Government Budget; and 3) the Ratchadaphiseksomphot Fund, Chulalongkorn University.
Conflict of interest
The authors have no conflict of interest with any commercial or other association in connection with the submitted article.
