Risk Factors, Prognosis, Influence on the Offspring, and Genetic Architecture of Perinatal Depression Classified Based on the Depressive Symptom Trajectory

Ohseto, Hisashi; Takahashi, Ippei; Narita, Akira; Obara, Taku; Ishikuro, Mami; Kobayashi, Natsuko; Kikuchi, Saya; Li, Xue; Noda, Aoi; Murakami, Keiko; Tamiya, Gen; Sugawara, Junichi; Tomita, Hiroaki; Kuriyama, Shinichi

doi:https://doi.org/10.1155/2024/6622666

Depression and Anxiety

On this page

Abstract Introduction Materials and Methods Results Discussion Conclusions Data Availability Additional Points Conflicts of Interest Acknowledgments Supplementary Materials References Copyright Related Articles

Research Article | Open Access

Volume 2024 | Article ID 6622666 | https://doi.org/10.1155/2024/6622666

Risk Factors, Prognosis, Influence on the Offspring, and Genetic Architecture of Perinatal Depression Classified Based on the Depressive Symptom Trajectory

Hisashi Ohseto,¹Ippei Takahashi,¹Akira Narita,^1,2Taku Obara,^1,2,3Mami Ishikuro,^1,2Natsuko Kobayashi,³Saya Kikuchi,³Xue Li,¹Aoi Noda,^1,2,3Keiko Murakami,^1,2Gen Tamiya,^1,2,4and Junichi Sugawara^1,2,3,5 et al.

Academic Editor: Francesco Craig

Received22 Aug 2023

Revised26 Jan 2024

Accepted24 Feb 2024

Published15 Mar 2024

Abstract

This study is aimed at revealing the risk factors, prognosis, influence on offspring, and genetic architecture of perinatal depression (PD) classified based on the depressive symptom trajectory. Pregnant women with no history of major depressive disorder (MDD) were recruited and followed up with their offspring from 1 to 5 years postpartum. Using four self-report questionnaires in the perinatal period, PD was classified into four subtypes: pregnancy, early postpartum, late postpartum, and chronic PD. Risk factors, depressive symptom trajectory from 1 to 5 years postpartum, and child behavior problems were compared among the four PD subtypes. Genome-wide association studies (GWASs) were conducted for each subtype. The relationships between the PD subtypes and polygenic risk scores (PRS) for MDD, a psychiatric disorder, and premenstrual syndrome (PMS), a hormonal disorder, were examined. Among 12,338 participants, 1,145 (9.3%) developed pregnancy PD, 856 (6.9%) developed early postpartum PD, 382 (3.1%) developed late postpartum PD, and 1,048 (8.5%) developed chronic PD. Depressive symptoms decreased to 61.0%–73.3% in the 5 years postpartum. The relationship between risk factors and PD varied based on the PD subtype. Additionally, chronic PD increased the risk of child behavior problems by 2- to 3-fold. The GWASs uncovered five significant variants in different loci depending on PD subtypes, suggesting a subtype-specific genetic architecture. The PRS for MDD was related to pregnancy, early postpartum, and chronic PD, while that for PMS was related to late postpartum PD. It was concluded that PD is heterogeneous depending on the depressive symptom trajectory. Thus, specific prevention and treatment strategies are needed.

1. Introduction

Perinatal depression (PD) is characterized by depressive symptoms emerging during pregnancy or postpartum [1], affecting approximately 15%–20% of women [2, 3], resulting in poor family relationships [4], poor infant care [5], and suicidal ideation [6]. Regarding the influence of PD on the offspring, large studies have noted child behavioral problems [7, 8]; however, this observation is inconsistent [9]. The pathophysiological mechanism of PD is expected to be similar to that of major depressive disorders (MDD). However, environmental and physiological changes specific to the perinatal period make it more complex [10]. Clinically, PD is often treated similarly to MDD; however, given its distinct biological background and impact on mothers and children, treatment based on PD-specific etiology is warranted [11].

Focusing on the depressive symptom trajectory is one approach to elucidate the mechanisms of PD. Studies that focused on environmental risk factors for PD classified based on the depressive symptom trajectory revealed its environmental heterogeneity [12, 13]. Given the unique stressors and physiologic conditions at each perinatal stage, assuming that the etiology and biological pathways also differ at each stage is reasonable. PD subtypes may be evident as the depressive symptom trajectory because they are expected to relate closely to perinatal hormone concentration trends or unique environmental factors in each stage of the perinatal period.

Genetically, the heritability of PD was 54% and 44% in twins and siblings [14], which were greater than those of MDD [15]. The largest genome-wide association study (GWAS) of PD [16], with 6,333 cases, observed only one significant genome-wide loci mapped on GULP1, which had not been previously reported to be associated with MDD. Additionally, the second largest GWAS [17], with 3,804 cases, observed no significant loci. In 2023, GWAS meta-analysis [18] was performed in 20 cohorts worldwide with 18,770 cases, but no significant loci were observed. A rational interpretation is that PD has various subtypes; i.e., it is heterogeneous. Thus, it is conceivable that genetic effects have been diluted in previous GWASs and have not been significant despite PD having higher heritability than MDD. Classifying PD into more homogeneous subtypes may increase the detection power in genomic studies.

In a genomic study [17], PD subtypes classified based on the depressive symptom trajectory were variably associated with polygenic risk score (PRS) for other psychiatric disorders. Additionally, previous studies [19] have highlighted that premenstrual syndrome (PMS), primarily caused by sex hormone disturbances, and MDD share a common etiology, suggesting that perinatal sex hormone disturbances may be involved in PD development. Therefore, we hypothesized that certain PD subtypes are genetically associated with the psychopathic predisposition as measured by PRS for MDD, while other PD subtypes would be associated with susceptibility to sex hormone disturbance as measured by PRS for PMS.

Here, we conducted a comprehensive study to elucidate PD’s pathogenesis and confirmed that PD subtypes with different depressive symptom trajectory have different genetic and environmental etiologies. Four subtypes of PD, pregnancy, early postpartum, late postpartum, and chronic PD, were diagnosed prospectively. Five-year depressive symptom trends and child behavior problems at age five were compared. Additionally, with a focus on the genetic architecture of PD, we conducted GWAS and PRS analysis for each subtype.

2. Materials and Methods

2.1. Participants

The Tohoku Medical Megabank Project Birth and Three-Generation (TMM BirThree) Cohort Study [20] was initiated in July 2013 after the Great East Japan Earthquake to collect information on perinatal exposure, subsequent disorder development, and genomics to achieve personalized medicine and health management. From over 50 obstetric clinics and hospitals in the Miyagi Prefecture, Japan, 23,406 pregnant women were recruited. We excluded 232 participants who withdrew consent, 369 with MDD history or currently taking selective serotonin reuptake inhibitors, 192 with abortion or stillbirth, and 9,918 who did not complete all four perinatal depressive symptom questionnaires. Regarding the 357 participants who participated in the TMM BirThree Cohort Study more than once, only the first participation was included. Finally, 12,338 pregnant women were included in this study. All study participants provided informed consent, and ethics approval was obtained from the Ethics Committee of the Tohoku Medical Megabank Organization (2022-4-110).

2.2. Measures

Participants self-reported educational attainment (junior high school/high school, vocational school/junior college, or university/graduate), household income (≤4 million yen, 4–6 million yen, or ≥6 million yen), family history of MDD, acceptance of pregnancy at registration (good: “very happy”/“unexpected but happy,” or poor), social isolation (the Lubben Social Network Scale (LSNS-6) [21]<12 or not), marital status (unmarried or married), and neuroticism trait via the short-form Eysenck Personality Questionnaire Revised [22]. Overweight (prepregnancy body mass kg/m² or not), neonate sex, parity (primipara or not), multiple births, delivery mode, and pregnancy complications (fetal growth restriction, threatened preterm delivery, and hypertensive disorders of pregnancy) were obtained from the medical records. Participants responded to the child behavior checklist (CBCL) [23] when their children were 5 years old, from which three composite scales (internalizing, externalizing, and total problems) were calculated. Participants answered the Kessler Psychological Distress Scale (K6) [24] twice during pregnancy, the Edinburgh Postnatal Depression Scale (EPDS) [25] at 1 and 6 months postpartum, and K6 annually thereafter. K6 measures psychological distress using six questionnaires ranging from 0 to 24 points. EPDS measures postnatal depressive symptoms consisting of 10 questionnaires of 0–30 points [24, 25]. In this study, individuals with K6 of ≥9 and an were considered depressed based on previous studies [26, 27]. Due to the nature of our cohort, some participants had suffered psychological and physical damage from the earthquake, possibly confounding the relationship between risk factors and PD or PD and child development. Therefore, house damages (total or partial, or none) and acquaintance deaths due to the earthquake were also collected through a self-reported questionnaire.

2.3. PD Subtypes

According to the depressive symptom trajectory measured by the four depressive symptom questionnaires, four PD subtypes were defined: pregnancy, early postpartum, late postpartum, and chronic PD (Figure 1(a)). Subtype-specific trajectories in the perinatal period were identified (Figure 1(b)). This definition was concordant with a previous study [12] that successfully demonstrated the heterogeneity of the environmental etiology of PD. Pregnancy PD was diagnosed when a pregnant woman was considered depressed at either or at both measurements during pregnancy but not at the postpartum measurements. Early postpartum PD was diagnosed when a pregnant woman without depression during pregnancy was considered depressed at the 1-month postpartum measurement, regardless of whether she still showed symptoms at the 6-month postpartum measurement or not. Late postpartum PD was diagnosed when a pregnant woman without depression during pregnancy and at the 1-month postpartum measurement had become depressed at the 6-month postpartum measurement. Chronic PD was diagnosed when a pregnant woman became depressed at both or either of the two measurements during pregnancy and at both or either the two postpartum measurements. Pregnant women who were not depressed at any of the four scheduled measurements were considered healthy. Using these definitions, the trajectories of all participants were categorized into four PD subtypes, or as healthy (Figure 1(a)).

(a)

(b)

Figure 1

Definitions of the four PD subtypes and the depressive symptom trajectory from the perinatal period to 5 years postpartum. (a) According to four measurements of depressive symptoms during the perinatal period, four PD subtypes, pregnancy, early postpartum, late postpartum, and chronic PD, were defined. The “○” symbol represents “depressive,” and the “-” symbol represents “not depressive.” (b) Participants responded to K6 twice during pregnancy, EPDS at 1 month and 6 months postpartum, and K6 annually thereafter. Each line represents PD subtypes and healthy participants. The -axis represents the time of the questionnaire, and the -axis represents the mean questionnaire score in each group.

2.4. Genotyping and Quality Control

This is described in detail in the Supplementary Note. Briefly, participants were genotyped on two platforms: Affymetrix Axiom Japonica Array v2 (JPA v2) [28] and Japonica Array NEO (JPA NEO) [29]. Participants genotyped on the first platform were cohort A, and those genotyped on the second were cohort B. In the genome-wide association study (GWAS), cohort A included 6,562 individuals with 11,737,866 variants, whereas cohort B included 5,098 individuals with 11,747,153 variants after imputation and standard quality control.

2.5. Statistical Analysis

Basic characteristics were compared among PD subtypes using the chi-square test for categorical variables and analysis of variance for continuous variables. The mean K6 scores at 1, 2, 3, 4, and 5 years postpartum were plotted for each PD subtype along with perinatal depressive symptom scores (K6 and EPDS) to reveal the long-term trajectory of depressive symptoms.

In risk factor analysis, risk factors for PD (age, educational attainment, household income, family history of MDD, acceptance of pregnancy, social isolation, marital status, neuroticism trait, house damages due to the earthquake, acquaintance deaths due to the earthquake, overweight, neonate sex, primipara, multiple births, delivery mode, and pregnancy complications) were compared using multinomial logistic regression analysis with healthy controls and each PD subtype as a case. In analyzing child behavior, CBCL scores were dichotomized based on whether the standardized -score [23] was ≥64 for the three composite scales. Multiple logistic regression was applied with adjustment for those in risk factor analysis after excluding multiple births. Multiple imputation [30] was applied to missing values in risk factors, and the results were subsequently integrated.

GWASs were performed for each PD subtype and for “any PD” defined as any PD subtypes using GCTA (1.94.0) [31] with a generalized linear mixed model adjusted for age and ten principal components in cohorts A and B. Twenty-eight outlier individuals detected by first and second principal components were excluded. The results from both cohorts were meta-analyzed using METAL [32]. Genome-wide statistical significance was set at the conventional value threshold of . Clumping was applied using PRSice-2 (2.3.5) [33] to determine independent variants. GWAS results were reported for all subtypes for significant or top-hit variants. For each reported variant, perinatal depressive symptom trajectories (K6 scores during pregnancy and EPDS scores at the 1- and 6-month postpartum measurements) were plotted according to genotypes to identify longitudinal effects. Fine mapping was performed using FOCUS for significant or top-hit variants [34]. FOCUS is a fine-mapping method that estimates posterior inclusion probability (PIP) for each gene to explain observed association signals from the transcriptome-wide association study utilizing GWAS summary data, expression weights, and the linkage disequilibrium among all SNPs in a region of association. Precomputed expression quantitative trait locus weights were used according to the original study [34]. The heritability of all PD subtypes and genetic correlations of all pairs of PD subtypes were estimated using LDSC [35].

PRS were calculated separately in cohorts A and B for MDD and PMS (Supplementary Note). The prevalence of each PD subtype was compared with the PRS quintiles using multiple logistic regression, with quintile 1 (lowest genetic load) as a reference, which was adjusted for age and five principal components. After that, the results for cohorts A and B were merged using inverse-variance weighting, and only merged results were reported. P for trends were compared among PRS with different value thresholds (0.5, 0.1, 0.05, 0.01, and 0.005) for inclusion of variants to confirm the relationship’s robustness. Additionally, to determine the effect of the genetic load of each phenotype on the trajectory of perinatal depressive symptoms, K6 scores during pregnancy and EPDS scores at the 1- and 6-month postpartum measurements were plotted for each PRS quintile after merging both cohorts, considering that the same quintiles had the same genetic load.

A series of sensitivity analyses were performed. First is to investigate the validity of the PD definition. Risk factor and PRS analyses were re-examined based on two additional PD definitions. According to previous studies, one defines “depressed” as K6 of ≥5, and the other defines it as K6 of ≥13 [24, 36, 37]. Second, those who became pregnant at the time they responded to the K6 at the 6 months postpartum were excluded, and risk factor and PRS analyses were performed again. Third, to account for the effect of current maternal depression on CBCL responses, the K6 score at the five-year postpartum measurement was added to the covariates in the child behavior analysis. Fourth, PRS for MDD was calculated from summary statistics of MDD GWAS meta-analysis, including the Japanese population [38], as opposed to the main analysis that included only the European population (Supplementary Note), and PRS analysis was performed again.

All analyses were conducted in R (4.1.0) unless otherwise noted. We considered two-tailed values of 0.05 as significant.

3. Results

The timing of the four questionnaire responses (median [interquartile range]) was at 17.7 [14.4–21.0] and 27.3 [24.4–30.3] weeks of gestation and 4.4 [4.0–4.9] and 27.7 [26.0–34.1] weeks postpartum. Out of 12,338 participants, 1,145 (9.3%) developed pregnancy PD, 856 (6.9%) developed early postpartum PD, 382 (3.1%) developed late postpartum PD, and 1,048 (8.5%) developed chronic PD. Most covariates, except for acquaintance deaths and house damages during the earthquake, multiple births, delivery mode, and threatened preterm delivery differed among healthy participants and those with PD (Table 1). Regarding prognosis, at five years postpartum, the follow-up rates between PD and healthy participants were similar. Depressive symptom scores decreased to 61.0%–73.3% of that in the first year. However, participants with PD still exhibited high depressive symptoms (Figure 1(b) and Supplementary Table 1). Chronic PD scored approximately 3–5 points higher than other PD subtypes (Supplementary Table 1).

Risk factor analysis revealed that each PD subtype has a different relationship with risk factors and is largely consistent across sensitivity analyses (Supplementary Table 2). Older age was negatively associated with pregnancy and chronic PD, and lower household income was positively associated with early postpartum PD. However, no relationship was observed between education level and PD. Moreover, family history of MDD was positively associated with early postpartum and chronic PD, while poor acceptance was positively associated with PD, especially chronic PD. Social isolation and neuroticism traits were positively related to all PD subtypes. The relationship between unmarried status and PD changed across sensitivity analyses, with a relatively large confidence interval (CI). Acquaintance death was related to late postpartum PD. House damages from the earthquake, maternal overweight, multiple births, and cesarean section were unassociated with the PD subtypes. Male neonate was negatively associated with early postpartum PD. Primipara was a risk factor, except for late postpartum PD. Among the obstetric conditions, threatened prenatal delivery and hypertensive disorders of pregnancy were positively related to pregnancy and early postpartum PD and pregnancy PD, respectively, with moderate variation across sensitivity analyses.

After excluding multiple births and missing response cases, 6,618 mother-child pairs were eligible for child behavior analysis. All PD subtypes were positively related to all CBCL domain scores (Table 2). For instance, chronic PD was associated with a 2- to 3-fold risk of child behavior problems (odds ratio (OR) and 95% CI: 2.59 (2.01 to 3.33), 2.94 (2.27 to 3.80), and 3.26 (2.53 to 4.20) for internalizing, externalizing, and total problems, respectively), which remained significant after adjusting for the maternal depressive state when responding to CBCL. Early and late postpartum PD more strongly affected internalizing than external problem. However, this was not the case in pregnancy and chronic PD (Table 2).

In the GWAS meta-analysis (Table 3), five variants located on an intron of CBR3-AS1 (rs138801403), an intron of PTPRD (rs1853229), an intron of CDH12 (rs1075046), an intron of RBFOX1 (rs4786119), and an intergenic between LOC100129620 and PLPPR4 (rs56289435) were significant for the relationship with any PD subtype (), pregnancy PD (), late postpartum PD (), late postpartum PD (), and chronic PD (), respectively. The top-hit variant of early postpartum PD () was rs117741236, located on an intron of TGM5. Each variant exhibited a distinct longitudinal pattern specific to each genotype (Figure 2 and Supplementary Table 3). For instance, in the case of rs1853229, which was significant in pregnancy PD, participants who were heterozygous had higher depression scores during pregnancy than those who were homozygous for the major alleles. In fine mapping using FOCUS, the highest PIP for each variant was lnc-CBR3-1:11 in the aorta on chromosome 21 in any PD (), LOC101928797 in fibroblasts on chromosome 9 in pregnancy PD (), TUBGCP4 in basal ganglia () on chromosome 15 in early postpartum PD, CDH12 in dorsolateral prefrontal cortex on chromosome 5 in late postpartum PD, RBFOX1 in tibial nerve () on chromosome 16 in postpartum PD, and PLPPR4 in the testis () on chromosome 1 in chronic PD. Heritability estimates were 0.062 (), 0.037 (0.043), 0.010 (0.046), 0.078 (0.046), and 0.043 (0.046) for any PD subtype, pregnancy PD, early postpartum PD, late postpartum PD, and chronic PD, respectively. The genetic correlation was statistically significant only for any PD and chronic PD pairs (Supplementary Table 4; , , ).

PRS analysis revealed a significant positive relationship between genetic susceptibility to MDD and pregnancy, early postpartum, and chronic PD (Figure 3(a) and Supplementary Table 3), largely consistent with the different value thresholds (Figure 3(c)). Those with high PRS for MDD had high depressive symptoms throughout the perinatal period. In contrast, those with high PRS for PMS displayed no apparent symptom trend (Figure 3(b)). PRS for PMS with values 0.005 and 0.01 were related to late postpartum PD (Figure 3(c)). Each PD subtype was not associated with PRS for MDD from a meta-analysis including Japanese population (Supplementary Table 5). The results were largely consistent across sensitivity analyses (Supplementary Table 5).

(a)

(b)

(c)

Figure 3

PRS analysis. (a) The prevalence of each PD subtype was compared among PRS (left: MDD; right: PMS) quintiles using multiple logistic regression, considering quintile 1 (lowest genetic load) as a reference, adjusted for age and five principal components. The -axis is the PRS quintiles, and the -axis is the odds ratio. The asterisk indicates statistical significance (). (b) K6 scores obtained twice during pregnancy and EPDS scores obtained at 1 and 6 months postpartum were plotted by PRS (left: MDD; right: PMS) quintiles. The -axis represents the time of the questionnaires, and the -axis represents the mean difference between questionnaire scores in each quintile group and the healthy control. (c) The heat map indicates the P for trend calculated using PRS with different value thresholds (0.5, 0.1, 0.05, 0.01, and 0.005). The -axis indicates PRS phenotype and value thresholds, and the -axis represents PD subtypes.

4. Discussion

PD classified based on the depressive symptom trajectory had different relationships with risk factors such as socioeconomic status, maternal psychological status, and obstetric status. Each PD subtype, especially chronic PD, was a significant risk factor for child behavior problems at five years postpartum. The GWASs for each subtype identified four susceptibility loci that were not identified before subtyping. PRS analysis revealed the relationship between PRS for MDD and pregnancy, early postpartum, and chronic PD and between PRS for PMS and late postpartum PD.

The score trajectory displayed a gradual decrease in depressive scores for participants with PD. However, it remained at approximately 70% of the first year, even five years postpartum. Previous studies [39, 40] examining depressive symptoms for 1–3 years in dozens to hundreds of patients with PD also observed that depressive symptoms attenuated but remained at follow-up. This study confirms this result in a larger population over a prolonged follow-up, demonstrating the need for long-term care for patients with PD.

Risk factor analysis revealed a clear difference in PD subtypes, mostly consistent with other PD definitions in a sensitivity analysis. A previous study [12] that also focused on risk factors for PD classified based on the depressive symptom trajectory revealed largely consistent results with our result concerning them being risk or protective factors. However, the characteristics of PD subtypes differed. Older age was more associated with late postpartum PD than early postpartum PD in this study. Nonetheless, the relationship was reversed in the previous study [12]. After delivery, the mother’s social circumstance changes gradually, but her physiologic status rapidly normalizes. PD detected early may have more specific physiologic features; therefore, EPDS measurement timing is a critical factor in defining the pathophysiology, and its difference (at 6 weeks and 6 months postpartum in the previous study [12], whereas at 4 weeks and 6 months postpartum in this study) could explain the different results between the previous [12] and this study. It is also important to note the difference in population (Swedes vs. Japanese), measurement (EPDS only vs. K6 and EPDS), and statistical method (univariate vs. multivariate).

Child behavior analysis revealed the prominent risk of PD for child behavior problems, which was attenuated but remained significant after adjusting for maternal depressive symptoms when responding to CBCL. In a previous Japanese study of 1,199 children [41], children whose mothers had depressive symptoms both during pregnancy and postpartum had a fivefold increased risk of childhood emotional symptoms with CBCL, compared to a twofold increase with either pregnancy or postpartum alone. Moreover, a small previous study [42] observed that PD was related to child behavior problems, and this relationship was not necessarily dependent on the presence of current depressive symptoms in the mother, consistent with the findings of the current study.

The GWASs uncovered five significant variants. CBR3-AS1, which was related to any PD, was not significant in the previous GWASs for any psychiatric disorders to date. However, PTPRD was reported in the largest East Asian GWAS for MDD [43], and RBFOX1 was reported in the largest European GWAS for MDD [44]. CDH12 was reported in the GWAS for suicide [45], and PLPPR4 was reported in the GWAS for antidepressant response [46]. Despite the small number of cases compared to previous GWAS [17], the GWASs for each PD subtype could identify biologically relevant loci, suggesting that a GWAS with small but homogeneous subgroups can detect susceptibility loci [47]. Each variant was significant specifically for each subtype and had a characteristic depressive symptom trajectory based on genotype, reflecting each variant’s distinctive pathological mechanism. Fine mapping suggested the involvement of several genes; however, PIP was small and did not show strong evidence of causality. The small sample size of the current study makes distinguishing subtype-specific findings from random detection challenging. Therefore, we hope that larger studies with biological reproducibility will confirm our findings and uncover new loci in the future. Heritability was estimated at 0.01-0.078, which was relatively small compared with that in previous GWAS [18], estimated at 0.17 in the East Asian population. In addition to the small sample size, our assessment of PD was based solely on self-report and may be susceptible to environmental factors. Consequently, the estimates of genetic correlations were also difficult to interpret with high variance. These results indicate that attempts to split outcomes to create a more homogeneous population are subject to the challenge of decreasing sample size, especially in post-GWAS.

PRS analysis revealed the genetic relationship between MDD and pregnancy, early postpartum, and chronic PD, a notable finding in our study. The previous study [48] revealed that PRS for MDD was associated with prenatal and postpartum depressive symptoms and that perinatal depressive symptoms mediated 57.9%–88.0% of the effect of PRS on postpartum depression symptoms. Taken together with our result, genetic susceptibility to MDD affects pregnancy and early postpartum depressive symptoms, and late postpartum depressive symptoms are more dependent on prior depressive symptoms. PRS for PMS was not associated with pregnancy, early postpartum, or chronic PD, which is inconsistent with our hypothesis that perinatal sex hormone disturbance causes psychiatric symptoms, especially in pregnancies genetically susceptible to sex hormone disturbance. Additionally, the relationship between PRS for PMS and late postpartum PD may be explained by PMS symptoms after the postpartum menstrual cycle resumes, indicating the unique pathology of late postpartum PD.

Our study has at least three clinical and research implications. First, PD caused persistent maternal depressive symptoms and child behavioral problems even 5 years postpartum, suggesting the need for long-term care such as life support, environmental adjustment, and the affected mothers learning to cope with stress. Particularly, the high-risk mothers indicated in this study, such as young, low-income, unmarried, and socially isolated, require intensive care. Second, PD subtypes classified based on the depressive symptom trajectory have unique prognoses, risk factors, influence on the offspring, and genetic architecture that will provide clues to new prevention and treatment strategies. MDD medication, such as selective serotonin reuptake inhibitors, may be less effective in late postpartum PD than in other PD subtypes. Third, the depressive symptoms score trajectory differed across the perinatal period; this can be used in clinical practice to identify high-risk periods of PD and provide focused care.

The strength of this study is that it is the first to conduct GWAS and PRS analysis for PD classified based on the prospectively assessed depressive symptom trajectory. Only one previous study [17] focused on PD subtypes classified based on the depressive symptom trajectory and performed a GWAS and PRS analysis. However, diagnosis and identification of depressive symptom trajectory were retrospective, and the genetic relationship with PMS was not investigated. The prolonged follow-up and assessment of influence on the offspring in the current study allow for comprehensiveness. Our cohort recruited about half of the newborns in Miyagi Prefecture in 2016, which led to a relatively small sampling bias. The percentage of EPDS above the cutoff was similar to that of the general population [20].

Despite these strengths and the sensitivity analysis series, this study has at least five limitations. First, our definition of PD has not been validated. K6 during pregnancy is not a common measure of depressive symptoms, and postpartum depressive symptoms have only been assessed twice (at 1 and 6 months postpartum). We confirmed the robustness of our result using a different K6 cutoff; however, our definition may miss the dynamics of depressive symptoms up to four weeks postpartum. Second, we used the GWAS results from a European rather than Japanese population to calculate the PRS for MDD in the main analysis. Thus, the PRS may not accurately reflect the polygenic loadings for MDD. The GWAS meta-analysis including the Japanese population has a smaller sample size (, including only 836 Japanese vs. 246,363 cases in the main analysis). Therefore, its validity is questionable. Third, less than 1% of participants were pregnant at the time they responded to the K6 at 6 months postpartum, which may have resulted in misclassification given that it is impossible to distinguish postpartum depressive symptoms from current pregnancy depressive symptoms. However, sensitivity analyses excluding these participants yielded comparable results, indicating robustness. Fourth, PD diagnosis and assessment of children depend on mothers’ self-reporting. Finally, some significant or top-hit variants had minor allele frequencies as low as ≤5%, making false positives more likely due to the available sample size.

5. Conclusions

PD subtypes classified based on the depressive symptom trajectory revealed different risk factors, prognosis, and influence on the offspring. The GWASs for each subtype identified four susceptibility loci that were not identified before subtyping. PRS for MDD was related to pregnancy, early postpartum, and chronic PD, while PRS for PMS was related to late postpartum PD, suggesting subtype-specific genetic architecture.

Data Availability

Summary GWAS statistics are publicly available at the Japanese Multi Omics Reference Panel website (https://jmorp.megabank.tohoku.ac.jp/). Individual genotyping results are available upon request after the approval of the Ethical Committee and the Materials and Information Distribution Review Committee of Tohoku Medical Megabank Organization.

Additional Points

Code Availability. The code supporting this study’s findings has been uploaded to GitHub. It is publicly accessible and can be found at https://github.com/HisashiOhseto/Genetic-architecture-of-perinatal-depression. Please contact the corresponding author for further inquiries or issues with accessing the code.

Conflicts of Interest

The authors declare that there is no conflict of interest regarding the publication of this paper.

Acknowledgments

This work was supported by the Japan Agency for Medical Research and Development (AMED), Japan (Grant Nos. JP19gk0110039, JP17km0105001, JP21tm0124005, and JP21tm0424601) and the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) KAKENHI (grant numbers 19H03894 and 22H03346).

Supplementary Materials

Supplementary Methods: Genotyping and quality control, polygenic risk score. Supplementary Tables: Supplementary Table 1: depressive symptom trajectory over a 5-year period. Supplementary Table 2: risk factor analysis using multinomial logistic regression. Supplementary Table 3: association of six GWAS significant or top-hit variants with PD subtypes. Supplementary Table 4: genetic correlations between PD subtypes. Supplementary Table 5: relationship between polygenic risk score for major depressive disorder and premenstrual syndrome and PD subtypes. (Supplementary Materials)

References

L. M. Howard, E. Molyneaux, C. L. Dennis, T. Rochat, A. Stein, and J. Milgrom, “Non-psychotic mental disorders in the perinatal period,” The Lancet, vol. 384, no. 9956, pp. 1775–1788, 2014.
View at: Publisher Site | Google Scholar
S. Shorey, C. Y. I. Chee, E. D. Ng, Y. H. Chan, W. W. S. Tam, and Y. S. Chong, “Prevalence and incidence of postpartum depression among healthy mothers: a systematic review and meta-analysis,” Journal of Psychiatric Research, vol. 104, pp. 235–248, 2018.
View at: Publisher Site | Google Scholar
K. Tokumitsu, N. Sugawara, K. Maruo, T. Suzuki, K. Shimoda, and N. Yasui-Furukori, “Prevalence of perinatal depression among Japanese women: a meta-analysis,” Annals of General Psychiatry, vol. 19, no. 1, p. 41, 2020.
View at: Publisher Site | Google Scholar
G. Lilja, M. Edhborg, and E. Nissen, “Depressive mood in women at childbirth predicts their mood and relationship with infant and partner during the first year postpartum,” Scandinavian Journal of Caring Sciences, vol. 26, no. 2, pp. 245–253, 2012.
View at: Publisher Site | Google Scholar
J. F. Paulson, S. Dauber, and J. A. Leiferman, “Individual and combined effects of postpartum depression in mothers and fathers on parenting behavior,” Pediatrics, vol. 118, no. 2, pp. 659–668, 2006.
View at: Publisher Site | Google Scholar
J. J. Kim, L. M. La Porte, M. P. Saleh et al., “Suicide risk among perinatal women who report thoughts of self-harm on depression screens,” Obstetrics and Gynecology, vol. 125, no. 4, pp. 885–893, 2015.
View at: Publisher Site | Google Scholar
B. Avan, L. M. Richter, P. G. Ramchandani, S. A. Norris, and A. Stein, “Maternal postnatal depression and children's growth and behaviour during the early years of life: exploring the interaction between physical and mental health,” Archives of Disease in Childhood, vol. 95, no. 9, pp. 690–695, 2010.
View at: Publisher Site | Google Scholar
M. J. Walker, C. Davis, B. Al-Sahab, and H. Tamim, “Reported maternal postpartum depression and risk of childhood psychopathology,” Maternaland Child Health Journal, vol. 17, no. 5, pp. 907–917, 2013.
View at: Publisher Site | Google Scholar
S. S. Aoyagi and K. J. Tsuchiya, “Does maternal postpartum depression affect children's developmental outcomes?” Obstetrics and Gynaecology Research, vol. 45, no. 9, pp. 1809–1820, 2019.
View at: Publisher Site | Google Scholar
Y. Yu, H. F. Liang, J. Chen et al., “Postpartum depression: current status and possible identification using biomarkers,” Frontiers in Psychiatry, vol. 12, article 620371, 2021.
View at: Publisher Site | Google Scholar
P. Cuijpers and E. Karyotaki, “The effects of psychological treatment of perinatal depression: an overview,” Archives of Women’s Mental Health, vol. 24, no. 5, pp. 801–806, 2021.
View at: Publisher Site | Google Scholar
A. Wikman, C. Axfors, S. I. Iliadis, J. Cox, E. Fransson, and A. Skalkidou, “Characteristics of women with different perinatal depression trajectories,” Journal of Neuroscience Research, vol. 98, no. 7, pp. 1268–1282, 2020.
View at: Publisher Site | Google Scholar
H. Santos Jr., X. Tan, and R. Salomon, “Heterogeneity in perinatal depression: how far have we come? A systematic review,” Archives of Women’s Mental Health, vol. 20, no. 1, pp. 11–23, 2017.
View at: Publisher Site | Google Scholar
A. Viktorin, S. Meltzer-Brody, R. Kuja-Halkola et al., “Heritability of perinatal depression and genetic overlap with nonperinatal depression,” American Journal of Psychiatry, vol. 173, no. 2, pp. 158–165, 2016.
View at: Publisher Site | Google Scholar
P. F. Sullivan, M. C. Neale, and K. S. Kendler, “Genetic epidemiology of major depression: review and meta-analysis,” American Journal of Psychiatry, vol. 157, no. 10, pp. 1552–1562, 2000.
View at: Publisher Site | Google Scholar
T. D. Nguyen, A. Harder, Y. Xiong et al., “Genetic heterogeneity and subtypes of major depression,” Molecular Psychiatry, vol. 27, no. 3, pp. 1667–1675, 2022.
View at: Publisher Site | Google Scholar
J. Kiewa, S. Meltzer-Brody, J. Milgrom et al., “Perinatal depression is associated with a higher polygenic risk for major depressive disorder than non-perinatal depression,” Depression and Anxiety, vol. 39, no. 3, pp. 182–191, 2022.
View at: Publisher Site | Google Scholar
J. Guintivano, E. M. Byrne, J. Kiewa et al., “Meta-analyses of genome-wide association studies for postpartum depression,” American Journal of Psychiatry, vol. 180, no. 12, pp. 884–895, 2023.
View at: Publisher Site | Google Scholar
T. B. Pearlstein, “Hormones and depression: what are the facts about premenstrual syndrome, menopause, and hormone replacement therapy?” American Journal of Obstetrics and Gynecology, vol. 173, no. 2, pp. 646–653, 1995.
View at: Publisher Site | Google Scholar
S. Kuriyama, H. Metoki, M. Kikuya et al., “Cohort Profile: Tohoku Medical Megabank Project Birth and Three-Generation Cohort Study (TMM Bir Three Cohort Study): rationale, progress and perspective,” Internal Journal of Epidemiology, vol. 49, no. 1, pp. 18-19, 2020.
View at: Google Scholar
J. Lubben, E. Blozik, G. Gillmann et al., “Performance of an abbreviated version of the Lubben Social Network Scale among three European community-dwelling older adult populations,” The Gerontologist, vol. 46, no. 4, pp. 503–513, 2006.
View at: Publisher Site | Google Scholar
T. Hosokawa and M. Ohyama, “Reliability and validity of a Japanese version of the short-form Eysenck personality questionnaire—revised,” Psychological Reports, vol. 72, no. 3, pp. 823–832, 1993.
View at: Publisher Site | Google Scholar
Y. Funabiki and T. Murai, “Standardization of a Japanese version of the child behavior checklist for ages 1½-5 and the caregiver teacher report form,” Japanese Journal of Child and Adolescent Psychiatry, vol. 58, no. 5, pp. 713–729, 2017.
View at: Google Scholar
R. C. Kessler, G. Andrews, L. J. Colpe et al., “Short screening scales to monitor population prevalences and trends in non-specific psychological distress,” Psychological Medicine, vol. 32, no. 6, pp. 959–976, 2002.
View at: Publisher Site | Google Scholar
J. L. Cox, J. M. Holden, and R. Sagovsky, “Detection of postnatal Depression,” British Journal of Psychiatry, vol. 150, no. 6, pp. 782–786, 1987.
View at: Publisher Site | Google Scholar
T. Okano, M. Murata, F. Masuji et al., “Validation and reliability of a Japanese version of the EPDS,” Archives of Psychiatric Diagnostics and Clinical Evaluation, vol. 7, pp. 525–533, 1996.
View at: Google Scholar
T. A. Furukawa, N. Kawakami, M. Saitoh et al., “The performance of the Japanese version of the K6 and K10 in the World Mental Health Survey Japan,” International Journalof Methods in Psychiatric Research, vol. 17, no. 3, pp. 152–158, 2008.
View at: Publisher Site | Google Scholar
Y. Kawai, T. Mimori, K. Kojima et al., “Japonica array: improved genotype imputation by designing a population- specific SNP array with 1070 Japanese individuals,” Journal of Human Genetics, vol. 60, no. 10, pp. 581–587, 2015.
View at: Publisher Site | Google Scholar
M. Sakurai-Yageta, K. Kumada, C. Gocho et al., “Japonica Array NEO with increased genome-wide coverage and abundant disease risk SNPs,” Journal of Biochemistry, vol. 170, no. 3, pp. 399–410, 2021.
View at: Publisher Site | Google Scholar
S. van Buuren and K. Groothuis-Oudshoorn, “mice: multivariate imputation by chained equations inR,” Journal of Statistical Software, vol. 45, no. 3, pp. 1–67, 2011.
View at: Publisher Site | Google Scholar
J. Yang, S. H. Lee, M. E. Goddard, and P. M. Visscher, “GCTA: a tool for genome-wide complex trait analysis,” American Journal of Human Genetics, vol. 88, no. 1, pp. 76–82, 2011.
View at: Publisher Site | Google Scholar
C. J. Willer, Y. Li, and G. R. Abecasis, “METAL: fast and efficient meta-analysis of genomewide association scans,” Bioinformatics, vol. 26, no. 17, pp. 2190-2191, 2010.
View at: Publisher Site | Google Scholar
S. W. Choi and P. F. O'Reilly, “PRSice-2: polygenic risk score software for biobank-scale data,” Gigascience, vol. 8, no. 7, article giz082, 2019.
View at: Publisher Site | Google Scholar
N. Mancuso, M. K. Freund, R. Johnson et al., “Probabilistic fine-mapping of transcriptome-wide association studies,” Nature Genetics, vol. 51, no. 4, pp. 675–682, 2019.
View at: Publisher Site | Google Scholar
B. K. Bulik-Sullivan, P. R. Loh, H. K. Finucane, S. Ripke, and J. Yang, “LD Score regression distinguishes confounding from polygenicity in genome- wide association studies,” Nature Genetics, vol. 47, no. 3, pp. 291–295, 2015.
View at: Publisher Site | Google Scholar
K. Sakurai, A. Nishi, K. Kondo, K. Yanagida, and N. Kawakami, “Screening performance of K6/K10 and other screening instruments for mood and anxiety disorders in Japan,” Psychiatry and Clinical Neurosciences, vol. 65, no. 5, pp. 434–441, 2011.
View at: Publisher Site | Google Scholar
R. C. Kessler, P. R. Barker, L. J. Colpe et al., “Screening for serious mental illness in the general population,” Archives of General Psychiatry, vol. 60, no. 2, pp. 184–189, 2003.
View at: Publisher Site | Google Scholar
S. Sakaue, M. Kanai, Y. Tanigawa et al., “A cross-population atlas of genetic associations for 220 human phenotypes,” Nature Genetics, vol. 53, no. 10, pp. 1415–1424, 2021.
View at: Publisher Site | Google Scholar
N. Vliegen, S. Casalin, P. Luyten et al., “Hospitalization-based treatment for postpartum depressed mothers and their babies: rationale, principles, and preliminary follow-up data,” Psychiatry, vol. 76, no. 2, pp. 150–168, 2013.
View at: Publisher Site | Google Scholar
A. L. Sutter-Dallay, O. Cosnefroy, E. Glatigny-Dallay, H. Verdoux, and N. Rascle, “Evolution of perinatal depressive symptoms from pregnancy to two years postpartum in a low-risk sample: the MATQUID cohort,” Journal of Affective Disorders, vol. 139, no. 1, pp. 23–29, 2012.
View at: Publisher Site | Google Scholar
M. Yamada, K. Tanaka, M. Arakawa, and Y. Miyake, “Perinatal maternal depressive symptoms and risk of behavioral problems at five years,” Pediatric Research, vol. 92, no. 1, pp. 315–321, 2022.
View at: Publisher Site | Google Scholar
D. M. Bagner, J. W. Pettit, P. M. Lewinsohn, and J. R. Seeley, “Effect of maternal depression on child behavior: a sensitive period?” Journal of the American Academy of Child and Adolescent Psychiatry, vol. 49, no. 7, pp. 699–707, 2010.
View at: Publisher Site | Google Scholar
O. Giannakopoulou, K. Lin, X. Meng et al., “The genetic architecture of depression in individuals of East Asian ancestry,” Journal of the American Medical Association Psychiatry, vol. 78, no. 11, pp. 1258–1269, 2021.
View at: Publisher Site | Google Scholar
D. M. Howard, M. J. Adams, T.-K. Clarke et al., “Genome-wide meta-analysis of depression identifies 102 independent variants and highlights the importance of the prefrontal brain regions,” Nature Neuroscience, vol. 22, no. 3, pp. 343–352, 2019.
View at: Publisher Site | Google Scholar
Q. S. Li, A. A. Shabalin, E. DiBlasi et al., “Genome-wide association study meta-analysis of suicide death and suicidal behavior,” Molecular Psychiatry, vol. 28, no. 2, pp. 891–900, 2023.
View at: Publisher Site | Google Scholar
Q. S. Li, C. Tian, G. R. Seabrook, W. C. Drevets, and V. A. Narayan, “Analysis of 23andMe antidepressant efficacy survey data: implication of circadian rhythm and neuroplasticity in bupropion response,” Translational Psychiatry, vol. 6, no. 9, article e889, 2016.
View at: Publisher Site | Google Scholar
A. Narita, M. Nagai, S. Mizuno et al., “Clustering by phenotype and genome-wide association study in autism,” Translational Psychiatry, vol. 10, no. 1, p. 290, 2020.
View at: Publisher Site | Google Scholar
V. Rantalainen, E. B. Binder, M. Lahti-Pulkkinen et al., “Polygenic prediction of the risk of perinatal depressive symptoms,” Depression and Anxiety, vol. 37, no. 9, pp. 862–875, 2020.
View at: Publisher Site | Google Scholar

Copyright

Copyright © 2024 Hisashi Ohseto et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies

Views

191

Downloads

99

Citations