If and how obesity and elevated androgens in women with polycystic ovary syndrome (PCOS) affect their offspring’s psychiatric health is unclear. Using data from Swedish population health registers, we showed that daughters of mothers with PCOS have a 78% increased risk of being diagnosed with anxiety disorders. We next generated a PCOS-like mouse (F₀) model induced by androgen exposure during late gestation, with or without diet-induced maternal obesity, and showed that the first generation (F₁) female offspring develop  anxiety-like behavior, which is transgenerationally transmitted through the female germline into the third generation of female offspring (F₃) in the androgenized lineage. In contrast, following the male germline, F₃ male offspring (mF₃) displayed anxiety-like behavior in the androgenized and the obese lineages. Using a targeted approach to search for molecular targets within the amygdala, we identified five differentially expressed genes involved in anxiety-like behavior in F₃ females in the androgenized lineage and eight genes in the obese lineage. In mF₃ male offspring, three genes were dysregulated in the obese lineage but none in the androgenized lineage. Finally, we performed in vitro fertilization (IVF) using a PCOS mouse model of continuous androgen exposure. We showed that the IVF generated F₁ and F₂ offspring in the female germline did not develop anxiety-like behavior, while the F₂ male offspring (mF₂) in the male germline did. Our findings provide evidence that elevated maternal androgens in PCOS and maternal obesity may underlie the risk of a transgenerational transmission of anxiety disorders in children of women with PCOS.

Materials and Methods

Ethical Approvals

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. The cohort study was approved by the regional ethical review board in Stockholm, Sweden (diary number 2013/862-31/5; 2016/1214-32). The requirement for informed consent was waived because the study was register-based, and the included individuals were not identifiable at any time.

Animal experiments were done in accordance with the legal requirements of the European Community (SJVFS 2017:40) and the directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes and approved by the Stockholm Ethical Committee for Animal Research (Dnr. No. 10798-2017). Animal care and procedures were controlled by Comparative Medicine Biomedicum (KM-B), Karolinska Institutet, Stockholm, Sweden.

Register-based cohort study population

Using the unique personal identification number assigned to each individual in Sweden at birth or at immigration, several nationwide longitudinal registers containing health and sociodemographic data until December 31, 2013, were linked, including the Swedish Medical Birth Register (MBR), National Patient Register (NPR), Prescribed Drug Register (PDR), Total Population Register (TPR), Migration Register, and the Cause of Death Register (CDR).

Women who had delivered at least one child were identified through the MBR. The NIH diagnostic criteria for PCOS were established in 1990 followed by the Rotterdam criteria 2003⁶². Women with PCOS were identified by having at least one PCOS International Classification of Diseases (ICD) code (ICD-9: 256E; ICD-10: E28.2) recorded in the MBR or in the NPR after the age of 13 and between 1990 and 2013. Women with a concurrent diagnosed condition that could cause symptoms similar to PCOS were excluded to ensure specificity ^{4, 8}. This yielded a total of 12,955 mothers with PCOS. Time of PCOS diagnosis (before or after pregnancy) was not taken into account since elevated testosterone levels have been reported to be present throughout life in women with PCOS ⁵. Each mother with PCOS was then matched to up to 10 comparison mothers without a PCOS diagnosis randomly selected from the general population on her birth year and county of residence within the year of diagnosis.

Children born from 1995 to 2007 to mothers with PCOS and matched unaffected mothers were identified. All children from each mother were included. Children were excluded if they were born outside of Sweden, adopted, stillborn or died on the day of birth, or had congenital malformations. This yielded a total of 8,864 children born to women with PCOS and 89,431 children born to women without PCOS from the general population. The birth years of 1995 to 2007 were chosen so that the children had complete coverage in the outpatient register (which began in 2001) from the age-eligible for anxiety diagnoses (age 6). Children were followed until December 31, 2013, yielding an age range of 6 to 18.9 years at the end of follow-up.

Outcome classification. Diagnoses of anxiety recorded after the age of 6 were identified by ICD-10 codes F40.0, F40.1, F41.0, F41.1, F43.

Covariates. Highest attained maternal education level (primary and secondary education, upper secondary education, post-secondary/post-graduate education) was used as a proxy for the child’s socioeconomic status. Maternal and paternal lifetime history of psychiatric disorders was determined by any recorded psychiatric diagnosis in the NPR. Region of birth (Nordic/non-Nordic) was extracted from the TPR. The child’s sex and year of birth were extracted from the MBR. Data on maternal body mass index (BMI) was available in the MBR, however, due to a high proportion of missing values (13%) combined with data not missing at random (missingness was associated with the outcome) it was not included in the analysis so as not to introduce bias.

Animals

The number of mice used for behavioral testing and for gene expression analyses is given in the figure legend and/or text. Details of breeding of females and male sibling have previously been described¹⁴ and is detailed in Table S3. In brief, twenty-one days old C57Bl/6J mice were obtained from Janvier Labs (Le Genest-Saint-Isle, France). All mice were maintained under a 12-h light/dark cycle and in a temperature-controlled room with ad libitum access to water and a diet. After one week of acclimatization, female mice were randomly divided into two groups and fed either 1) a CD (Research Diets, D12328) comprising 11% fat, 73% carbohydrates [0% sucrose], and 16% proteins or 2) HFHS diet (Research Diets, D12331) comprising 58% fat, 26% carbohydrates [17% sucrose], and 16% proteins) for six weeks. A female in the proestrus or estrus phase, determined by vaginal cytology, was mated overnight with a male. Females were checked daily for post-copulatory plugs, and a plug on the morning after mating was considered embryonic day (E) 0.5. Details of the number of animals used for 1) phenotypic testing and 2) for breeding to generate F₁, F_2, and F₃ in each group have previously been described ¹⁴.

Prenatal androgen exposure

The CD and HFHS groups were in random order subdivided and subjected to daily injections subcutaneously (s.c.) in the inter-scapular area from E16.5 to E18.5 with 50 µl of a solution containing: 1) a mixture of 5 µl benzyl benzoate (B6630; Sigma-Aldrich) and 45 µl sesame oil (S3547; Sigma-Aldrich, St. Louis, Missouri, USA)  i.e. vehicle, or 2) 250 µg DHT (5α androstane-17β-ol-3-one, A8380; Sigma-Aldrich, St. Louis, Missouri, USA) dissolved in a mixture of 5 µl benzyl benzoate and 45 µl sesame oil i.e. prenatal androgen exposure by DHT. In brief, prior to mating with male mice, F₀ mothers were fed CD or HFHS-diet for 6 weeks¹⁴. During the embryonic day (E)16.5-E18.5 pregnant mice were injected subcutaneously with dihydrotestosterone (DHT), a non-aromatizable androgen, or sesame oil alone (vehicle) resulting in four experimental groups; MatCD+Vehicle (control), MatCD+DHT (androgenized), MatHFHS+Vehicle (obese), and MatHFHS+DHT (obese and androgenized). These are well-established models of prenatal androgen exposure and diet-induced obesity^{13, 14}. F₁ female offspring were mated with unrelated healthy males fed CD to generate F₂ and thereafter to generate F₃ female and male (siblings) offspring (female germline) (Fig. 2A). Furthermore, F₁ males were mated with unrelated health females fed CD to generate mF₂ and mF₃ males (male germline) (Fig. 2A and Table S3). Anxiety-like behavior was assessed in the elevated plus maze and in the open field at adult age in F₁, F₂ and F₃ female offspring and in their F₂ and F₃ male siblings (fF₂ and fF₃), as well as in F₁, mF₂ and mF₃ male offspring. Compromised embryonic development of second-generation (F₂) offspring in the obese and androgenized female lineage¹⁴ prevented us from investigating transgenerational transmission of anxiety-like behavior in F₃ offspring in the combined lineage.

Mouse-breeding scheme and feeding paradigm to generate F₁ to F₃ generations

All offspring, F₁, F_2, and F₃ were weaned at postnatal day 21 onto control diet: CD+Veh, CD+DHT, HFHS+Veh, and HFHS+DHT. To generate F₂, a subset of F₁ female and male offspring were mated with unrelated males and females fed the control diet, respectively, and a subset of F₂ female and male offspring were mated with unrelated males and females, respectively, to generate F₃. Remaining F₁, F₂ and F₃ female and male offspring were subjected to phenotypic behavioral testing as described below.

Prepubertal androgen exposure, in vitro fertilization and offspring generation

To generate the prepubertal PCOS-like mouse model, four week-old female mice were implanted subcutaneously with a 10-mm length pellet  of crystalline 5alpha-DHT (A8380-1G; Sigma-Aldrich, St. Louis, MO) and as control, a non-DHT containing pellet was inserted²⁸. After six weeks of pellet implantation, F0 donors were subjected to phenotypic behavioral testing. Then these female mice (10-week of age) were superovulated and oocytes were collected and added to capacitated spermatozoa to perform in vitro fertilization (IVF) in the Karolinska Center for Transgene Technologies (KCTT), Karolinska Institutet. Two-cell embryos were transferred into healthy surrogate mothers to generate F₁ female and male offspring. A subset of F₁ female and male offspring, respectively, underwent IVF to generate F₂ female and male offspring thus following both female and male germline. F₀ donors and remaining F₁ and F₂ female and male offspring were subjected to phenotypic behavioral testing as described below.

Assessment of behavior

All behavior experiments were performed at age of 16-17 weeks and during the light phase of the light-dark cycle, and mice were acclimatized in the testing room for 20-min before the test. In the EPM and OF tests, the mice were tracked automatically by an infrared digital camera using the EthoVision XT software (Noldus, Wageningen, the Netherlands). Innate fear/anxiety was assessed by using elevated plus maze as previously described¹³. Briefly, mice were placed into the middle of a four-armed maze facing an open arm and allowed to explore for 10-min. The number of arm entries and distance covered was tracked and analyzed by EthoVisionXT Software. We assessed anxiety-like behavior by calculating the percentage of time spent in the open and closed arms. For open field, the mouse was placed at the center of the arena and its movements were recorded for 10 min. We evaluated anxiety-like behavior and general locomotor activity by calculating the percentage of time spent in the center and the total distance traveled, respectively. The OF test was performed 4 days after the EPM.

Statistical Analysis

In the register-based cohort, children were followed from age 6, when they were eligible to receive a diagnosis of anxiety, to the date of diagnosis, death, emigration, age 18, or end of follow-up (December 31, 2013), whichever came first. The mean age and standard deviation at the end of follow-up were 11.2 ± 3.6 years. Associations between maternal PCOS and offspring anxiety were estimated as hazard ratios (HR) with 95% confidence intervals (CI) using stratified Cox regression models with attained age as the underlying time scale. In addition to the maternal matching criteria (maternal birth year and county of residence within the year of PCOS diagnosis), potential confounding variables were adjusted for including offspring sex and year of birth, maternal age at child’s birth, maternal education, maternal region of birth, and maternal and paternal lifetime history of psychiatric disorders (adjusted Model). Robust standard errors were used to account for dependence between observations since several children from the same family were included in the study population. The analysis was conducted first for all offspring combined, then stratified by offspring sex.

In the mice experiments, data were assessed for normality and variance (Kolmogorov Smirnov). Group allocation during experiments was not blinded to the investigators. However, data analyses were repeated by two or more investigators. The sample size in the mice experiments was based on differences in AGD between the control and the androgenized lineage in our previous studies^{13, 63}. Nine animals per group were required to detect a mean difference in AGD of 40.6% with a standard deviation (SD) of 0.1, a significance level of 0.05, and a power of 0.8. All data are presented as mean ± s.e.m, SD or as median and range. Differences between groups were determined by two-way ANOVA followed by Tukey’s post hoc test or Bonferroni post hoc test, and by one-way ANOVA followed by Tukey’s post hoc test or Bonferroni post hoc test. For the IVF experiment, we performed Student’s t-test in normalized data. Differences were considered statistically significant at P < 0.05. Statistical analyses for the register-based cohort study were performed using Stata statistical software version 15.1 (Stata Corps, Texas, USA), and in the animal studies by GraphPad Prism 8 (GraphPad Software Inc., CA, USA) and SPSS software v.26.0 (IBM, Armonk, NY, USA).