Distinct Genetic Profiles in Postpartum Depression With Different Trajectory of Illness


 BackgroundPostpartum depression (PPD) is a common and highly heritabledisorder in the postnatal period of new mothers. The development of PPD is shown to affectneurodevelopment in children and recent evidence suggests thatthe trajectory of PPDisalso associated with children’s neurodevelopment and mental conditions. Thus, early identification and intervention for individuals at high risk of PPD are urgently needed.Additionally, it is not clear whether genetic factors affect thetrajectory of PPD. Therefore, using a polygenic risk score (PRS) approach, we investigated if PRS for depression (Depression-PRS) and bipolar disorder (Bipolar-PRS) are associated with the development and clinical course of PPD.Methods Usingrecent large genome-wide association studies(GWAS) of depression and bipolar disorder as discovery cohorts, we calculatedDepression-PRS and Bipolar-PRS in each individual. Then, we investigated the possible association between Depression-PRS and Bipolar-PRS with the development andtrajectory of PPD insubjects from the Hamamatsu Birth Cohort for mothers and children (n = 136). Depressive symptoms were assessed using the Edinburgh Postpartum Depression Scale. Gene-set enrichment analyses were used to identify pathways underlying these conditions. ResultsDepression-PRS was significantly higher in subjects with PPD than in those without PPD(t = -3.283, P = 0.002)and logistic analysis showed that Depression-PRS significantly increases therisk of developing PPD(OR [SE] = 2.274 [0.585], P = 0.002). Furthermore, Depression-PRS was positively associated with continuity of PPD (β [SE]=1.621 [0.672]; P = 0.032).Gene-set enrichment analyses revealed that pathways such as“response to hormone”(β[SE] -2.285[1.002], P < 0.001) and “epigenetic regulation”(β[SE] 2.831 [1.317], P < 0.001) were involved in the continuity of PPD. ConclusionThese preliminary findings indicate that the genetic component plays an important role not only in the development but also inthe continuity of PPD. A polygenic risk score approach could be useful to identify subjects at risk for PPD, especially for persistent PPD,who needcareful monitoring and intervention after delivery.


Abstract
BackgroundPostpartum depression (PPD) is a common and highly heritabledisorder in the postnatal period of new mothers. The development of PPD is shown to affectneurodevelopment in children and recent evidence suggests thatthe trajectory of PPDisalso associated with children's neurodevelopment and mental conditions. Thus, early identi cation and intervention for individuals at high risk of PPD are urgently needed.Additionally, it is not clear whether genetic factors affect thetrajectory of PPD. Therefore, using a polygenic risk score (PRS) approach, we investigated if PRS for depression (Depression-PRS) and bipolar disorder (Bipolar-PRS) are associated with the development and clinical course of PPD.
Methods Usingrecent large genome-wide association studies(GWAS) of depression and bipolar disorder as discovery cohorts, we calculatedDepression-PRS and Bipolar-PRS in each individual. Then, we investigated the possible association between Depression-PRS and Bipolar-PRS with the development andtrajectory of PPD insubjects from the Hamamatsu Birth Cohort for mothers and children (n = 136). Depressive symptoms were assessed using the Edinburgh Postpartum Depression Scale. Gene-set enrichment analyses were used to identify pathways underlying these conditions.
ResultsDepression-PRS was signi cantly higher in subjects with PPD than in those without PPD(t = ConclusionThese preliminary ndings indicate that the genetic component plays an important role not only in the development but also inthe continuity of PPD. A polygenic risk score approach could be useful to identify subjects at risk for PPD, especially for persistent PPD,who needcareful monitoring and intervention after delivery.

Background
Postpartum depression (PPD) is a common psychiatric disease observed among new mothers in the postnatal period, with a reported prevalence higher than 20% [1]. Recent studies, including our own, suggest that PPD affects children's neurodevelopment and their subsequent mental health [2][3][4][5][6]. Additionally, the trajectory of PPD has been reportedly associated with these children's behavioral outcomes [6][7][8]. Thus, early identi cation and intervention for individuals with high risk for PPD are urgently needed .
In the American Psychiatric Association's Diagnostic and Statistical Manual of Disorders, fth edition [9], PPD is de ned as a major depressive episode with onset during pregnancy or in the 4 weeks following delivery. However, many studies have de ned PPD as a depressive episode that occurs from 4 weeks to 12 months after childbirth [10]. While it is well known that PPD follows different trajectories among individuals, little research has been conducted to examine the factors in uencing these variabilities among different individuals [1].
Meanwhile, evidence suggests that genetic factors play an important role in the development of PPD [11,12]. Indeed, the heritability of PPD is estimated to be as high as 50% [13], which is much higher than that of major depressive disorder. More recently, other factors such as a past history of depression, social conditions, and hormonal levels have been discussed as possible factors that may in uence the individual differences related to the trajectory of PPD [14]. Little is known, however, about whether genetic factors are involved in the trajectory of PPD.
In this pilot study, we attempted to investigate whether genetic factors were associated with the trajectory of PPD using a polygenic risk score (PRS) approach capitalizing on the data of a part of our birth cohort study, the Hamamatsu Birth Cohort for Mothers and Children (HBC study). Notably, since recent systematic reviews have reported that some subjects with a high Edinburgh Postpartum Depression Scale (EPDS) score were later diagnosed as having a bipolar disorder [15,16], we used PRS for both the unipolar and bipolar disorders in our analysis.

Participants
Participants in this study (n = 136) were mothers who gave birth to their children in Japan between December 2007 and June 2011 with DNA genotyping data from an HBC study (total participants = 1138). Recruitment procedures are comprehensively described in our previous studies [17]. The study procedures were approved by the Hamamatsu University School of Medicine and University Hospital Ethics Committee (research ID:17-037 and 19-145) and written informed consent was obtained from each mother for the participation of her infant in our study. This study followed the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) reporting guidelines.

Measurement
Depressive symptoms were assessed by using the Japanese version of Edinburgh Postpartum Depression Scale (EPDS) at week 1 and 4, 3 months, and 10 months after delivery. The cut-off score for EPDS was 9, which is widely used and validated in the PPD study for Japanese subjects [18]. Mothers whose EPDS score was above the cut-off point within 3 months or 10 months postpartum were de ned as subjects with "PPD at any time point".
Mothers whose EPDS score was above cut-off point only within 3 months postpartum were de ned as subjects with "transient PPD." Mothers whose EPDS score was above the cut-off point at both within 3 months and at 10 months postpartum were de ned as subjects with "persistent PPD". Since it is di cult to identify the exact date of onset among mothers whose EPDS score was above the cut-off point only at 10 months postpartum, these mothers were assigned to subjects with "persistent PPD" in the analyses.

PRS analysis
We used PRSice-2 to generate PRS, according to the developers' protocol [21]. The summary GWAS data used to determine the PRS for depression (Depression-PRS) and bipolar disorder (Bipolar-PRS) were obtained from the Psychiatric Genomics Consortium (https://humandbs.biosciencedbc.jp/en/). To account for population strati cation, we included 4 principle components (PCs). The PCs were calculated based on the pruned data with PLINK 1.9 [22]. The criteria for SNP clumping was pairwise LD r 2 < 0.1 within 1 Mb window. PRS scores were calculated with P value thresholds at 0.05, which is widely accepted [23]. The number of SNPs used for calculating Depression-PRS and Bipolar-PRS were 126,868 and 290,127, respectively. Standardized PRS scores (mean = 0 and standard deviation = 1) were used for the analyses.
Gene-set analyses were conducted using PRSet function implemented in PRSice-2. The collection of gene-sets was obtained from the MSigDB database and GO gene sets (c5:Biological Process) were used for the analyses (http://software.broadinstitute.org/gsea/msigdb/index.jsp).

Statistical analysis
Demographic differences between subjects without PPD, with "transient PPD," and "persistent PPD" were tested, using the analysis of variance for continuous variables, and the chi-square test for categorical variables. Paired t-test was used for comparing Depression-PRS and Bipolar-PRS between subjects without PPD and subjects with "PPD at any time point." P-value was corrected by Bonferroni corrections for multiple testing (2 independent hypotheses: Depression-PRS and Bipolar-PRS) and the signi cance of P-value was set at 0.05.
The associations of PPD trajectory with Depression-PRS and Bipolar-PRS were analyzed using multinomial logistic regression analysis. Age at childbirth, household incomes, emotional support, maternal educational attainment (in years), and maternal pre-pregnancy body mass index (BMI) were included as covariates. Again, P-value was corrected by Bonferroni corrections for multiple testing (2 independent hypotheses: Depression-PRS and Bipolar-PRS) and the signi cance of P-value was set at 0.05.
For gene-set enrichment analysis, P value was corrected using 10000 permutation tests and the signi cance was set at 0.05. All statistical analysis was two-tailed and was conducted by Stata version 15.

Participant characteristics
A summary of participant characteristics is provided in Table 1. A total of 136 participants were analyzed and 25 subjects developed PPD after delivery (18.4%). Twenty subjects (14.7%) developed "transient PPD" and 5 subjects (3.7%) developed "persistent PPD". There was no difference in the participant characteristics among groups.  Depression-PRS, Bipolar-PRS, and development of PPD Depression-PRS was signi cantly higher in subjects assigned to "PPD at any time point" group compared to subjects without PPD (t = -3.283, df = 134, P = 0.0026) (Fig. 1a, Table 2). On the contrary, there was no difference in Bipolar-PRS between these two groups (Fig. 1b, Table 3). However, Bipolar-PRS was not associated with PPD trajectory (Fig. 2b, Table 3).

Gene-set enrichment analysis of PPD trajectory
To obtain biological insights related to PPD trajectory, we conducted a gene-set enrichment analysis between the "transient PPD" and "persistent PPD" groups.

Discussion
In the present study, rst, we con rmed that the genetic risks of major depression, as measured using the PRS, are associated with the development of PPD in our cohort study. Next, we found that the genetic risk of major depression affects the trajectory of PPD.
Using PRS analysis, we found that Depression-PRS was signi cantly higher in subjects with "PPD at any time point" than in subjects without PPD. Furthermore, Depression-PRS was higher in subjects with "persistent PPD" than in subjects with "transient PPD." To our knowledge, this is the rst study demonstrating that genetic factors are involved in the development and trajectory of PPD using the PRS approach.
The prevalence of PPD in the current study was 18%, which is similar to the results of previous studies [1,24]. Moreover, approximately 20% of mothers who developed PPD showed persistent symptoms 10 months postpartum, supporting the notion that there are strong individual differences in PPD trajectories [25]. The present results suggest that there might be biological differences between transient PPD and persistent PPD. Lapato et al. [26] reported that women who experienced their rst symptoms of major depression in the peripartum have different genetic backgrounds, as compared to those who have recurrent major depression and happen to develop symptoms during the peripartum. In our gene-set analyses, we found that genes associated with hormone response and epigenetic changes were involved in the persistence of PPD. In conjunction with our current results, several studies have shown that reproductive hormones, such as lactogenic hormones and thyroid function, were related to the development of PPD [14,27]. More recently, it has been proposed that hormones such as glucocorticoid cause treatment-resistant depression through epigenetic changes in the brain [28]. It is possible that different courses of PPD are due to different severities of the illness between transient and persistent PPD groups. However, there was no difference in the mean EPDS scores between subjects with "transient PPD" and "persistent PPD" (t = -1.510, df = 18, P = 0.148). Taken together, we can safely conclude that the trajectory of PPD is mainly affected by genetic factors.
In the current study, contradictory to the prior study by Munk et al. [26], which reported that the diagnosis of women with initial diagnosis of major depression was more likely to change to bipolar disorder in 15 years of follow-up periods, in our study, Bipolar-PRS was not associated with the development of PPD [16]. While these ndings were inconsistent with ours, the differences may be explained by the variations in our subject characteristics. That is, while all subjects in the study by Munk et al. were clinical patients, and were referred to psychiatric hospitals, our participants were drawn from a general population.
There are a few limitations that should be considered in our study. First, the sample size was relatively small; thus, we had only 25 subjects with PPD in the analysis. Second, we did not have information about whether our participants with depressive symptoms had taken their medication and received psychological treatment for PPD. Third, our cases with PPD were ascertained using a screening tool, the EPDS, and not via a clinical interview conducted by a research clinician. Future studies using a larger sample size with PPD assessed via a structured clinical interview for the DSM and replications in different cohorts are needed.

Conclusions
In this pilot study, using a part of data from a longitudinal cohort study, we demonstrated that genetic factors for depression are associated with the development and trajectory of PPD. Our study provided initial evidence that a polygenic risk score approach could be used for early identi cation of individuals with a high risk for PPD and for a prediction of the trajectory of PPD. We hope that these initial ndings will eventually lead to effective interventions for PPD.