Academic Editor: Michael H. Dahan
This is an open access article under the CC BY 4.0 license.
Background: Evidence from observation studies has implied an
association between polycystic ovary syndrome (PCOS) and risk of depression.
Nevertheless, it remains elusive if the identified correlation is causal or owing
to biases in observation researches. Hence, we utilized a bidirectional
two-sample Mendelian randomization (MR) method to evaluate the potential causal
relationship between PCOS and depression. Methods: Genetic instruments
for PCOS and depression were acquired from two large genome-wide association
studies (GWASs). MR analyses were completed via the inverse-variance weighted
(IVW) method and weighted median approaches. The underlying pleiotropy was tested
by MR-Egger regression, and leave-one-out method was used to evaluate the
stability of MR results. Results: Using the IVW analyses (odds ratio
(OR) = 1.07, 95% confidence interval (CI) = 1.01–1.06, p
Polycystic ovary syndrome (PCOS) is a commonly seen endocrine disturbance which is the major cause of anovulatory infertility and affects up to 15% of women of reproduction age. It is characterized by androgenism, ovulatory dysfunction, obesity, menstrual dysfunction, metabolic and psychiatric abnormalities [1, 2]. Hyperandrogenemia is present in 15%–45% of patients with PCOS [3], such as acne, hairiness, and increased level of free testosterone in peripheral blood. Moreover, it is known that women with PCOS are more likely to suffer from psychological problems, especially anxiety and depression [4].
The pathogenesis of PCOS remains unclear. Clinical observations and animal experiment data suggest the assumption that PCOS is inherited and induced by developmental programming of normal genetic mutations. Those genes would be magnified by exposing to in-utero androgen and stimulated by various post-natal life-style and environment factors. Chemistry substances that disrupt endocrine harbor the possibility to affect the developmental programming of PCOS susceptible genes [5].
Depression is a more and more commonly seen disease which constitutes remarkable health-care challenge. In 2008, WHO considered severe depression the 3rd cause of disease burden across the globe and forecasted that the problem will rank first by 2030 [6]. Depression influences females more than males [7] and is the most commonly seen psychological issue for females with PCOS. Females with PCOS present an 8-fold greater incidence of depression in contrast to females with no PCOS [8]. The different pathophysiologic causal links inducing depression are insulin resistance (IR), disturbances in the hypothalamic pituitary adrenal (HPA) axis and hyperandrogenism [2].
Although previous studies provided a suggestive link between PCOS and the risk of depression, residual confounding and reverse causation are difficult to eliminate in observational studies. Recently, there is an alternative method to investigate the potential causal association unbiasedly, two-sample Mendelian randomization, which depends on gene mutations as instrumental variables (IVs) to assess the causal association between an exposure and a result [9]. After random location in the process of meiosis, single-nucleotide polymorphisms (SNPs) used as instrumental variables (IVs) remains stable and non-modified throughout a life-time of environment exposure, which makes them independent of confounding factors or reverse causation [10].
Based on gene mutations, Mendelian randomization (MR)-offered proofs of PCOS-related causality pertaining to several physical disorders have been found [11]. Moreover, previous Mendelian randomized study also explored the causal associations between PCOS and psychiatrical disorder and discovered that PCOS increased the risk of obsessive-compulsive disorder, but was not associated with the other four psychiatric disorders (anxiety disorders, bipolar disorders, severe depression disorders, or schizophrenia) [12]. On the basis of the release of a new database with a larger sample size, we conducted this updated bidirectional Mendelian randomized analyses to assess whether the underlying causality between PCOS and depression could have different outcomes.
The genome-wide association study (GWAS) dataset of PCOS was retrieved from the European ancestry. Using a meta-analysis technique, an adequate sample size from seven original datasets was obtained (10,074 cases and 103,164 controls) [13], which yielded better statistical power in identifying the associated SNPs. The diagnosis criteria of PCOS were on the foundation of the NIH or Rotterdam standards, or self-reporting in different original datasets. More specific information regarding the cohorts, genotyping, quality control, and imputation can be viewed in previous studies [13].
The gene correlation estimates of depression were acquired from a meta-analysis [14]. It enrolled a total of 807,553 participants (246,363 cases and 561,190 controls) from three cohorts: 23andMe [15] (75,607 cases & 231,747 controls), UK Biobank [16] (127,552 cases & 233,763 controls), and PGC_139k [17] (43,204 cases & 95,680 controls), after excluding overlapping samples. All the samples were of European descent. In the 23andMe cohort, individuals with depression were defined via self-reporting using on-line investigation. Detailed description of these questions can be accessed in the publication by Hyde CL et al. [15]. In the UK Biobank cohort, depression was defined according to the self-report using a touchscreen survey. Participants with bipolar disorder, schizophrenia or personality disorder were excluded. Additionally, the enrolled samples were strictly confined to the “White British” to reduce the population architecture bias. The PGC_139k cohort used samples from the earlier released data of 23andMe and UK Biobank cohorts. The overlapping samples were removed from the combined cohort.
The genetic instruments of PCOS and depression were identified with a
genome-wide statistical threshold of p
Chr | Position | SNP | Effect Allele | Other Allele | EAF | Beta | SE | Gene | p value | F-statistic |
2 | 43561780 | rs7563201 | A | G | 0.4507 | –0.1081 | 0.0172 | THADA | 3.68E-10 | 39.49976 |
2 | 2.13E+08 | rs2178575 | A | G | 0.1512 | 0.1663 | 0.0219 | ERBB4 | 3.34E-14 | 57.66287 |
3 | 1.32E+08 | rs13164856 | T | C | 0.7291 | 0.1235 | 0.0193 | IRF1/RAD50 | 1.45E-10 | 40.94674 |
8 | 11623889 | rs804279 | A | T | 0.2616 | 0.1276 | 0.0184 | GATA4/NEIL2 | 3.76E-12 | 48.09121 |
9 | 5440589 | rs10739076 | A | C | 0.3078 | 0.1097 | 0.0197 | PLGRKT | 2.51E-08 | 31.0085 |
9 | 97723266 | rs7864171 | A | G | 0.4284 | –0.0933 | 0.0168 | FANCC | 2.95E-08 | 30.84216 |
9 | 1.27E+08 | rs9696009 | A | G | 0.0679 | 0.202 | 0.0311 | DENND1A | 7.96E-11 | 42.18732 |
11 | 30226356 | rs11031005 | T | C | 0.8537 | –0.1593 | 0.0223 | ARL14EP/FSHB | 8.66E-13 | 51.02956 |
11 | 1.02E+08 | rs11225154 | A | G | 0.0941 | 0.1787 | 0.0272 | YAP1 | 5.44E-11 | 43.16297 |
11 | 1.14E+08 | rs1784692 | T | C | 0.8237 | 0.1438 | 0.0226 | ZBTB16 | 1.88E-10 | 40.48563 |
12 | 56477694 | rs2271194 | A | T | 0.416 | 0.0971 | 0.0166 | ERBB3/RAB5B | 4.57E-09 | 34.21545 |
12 | 75941042 | rs1795379 | T | C | 0.2398 | –0.1174 | 0.0195 | KRR1 | 1.81E-09 | 36.24657 |
16 | 52375777 | rs8043701 | A | T | 0.815 | –0.1273 | 0.0208 | TOX3 | 9.61E-10 | 37.45675 |
SNP, single nucleotide polymorphisms; EAF, effect allele frequency; PCOS, polycystic ovary syndrome; Chr, chromosome; SE, standard error. |
Chr | Position | SNP | Effect allele | Other allele | Beta | Se | EAF | p value |
1 | 52274078 | rs7551758 | G | T | 0.0283 | 0.0043 | 0.5329 | 5.11E-11 |
1 | 72765116 | rs2568958 | A | G | 0.0382 | 0.0044 | 0.6042 | 2.90E-18 |
1 | 175913828 | rs10913112 | T | C | –0.0262 | 0.0045 | 0.378 | 4.53E-09 |
1 | 197704717 | rs17641524 | T | C | –0.03 | 0.0053 | 0.2101 | 1.50E-08 |
1 | 49675276 | rs354155 | C | G | –0.0449 | 0.0075 | 0.0923 | 1.75E-09 |
1 | 67132262 | rs7538938 | C | T | 0.0251 | 0.0043 | 0.5599 | 7.29E-09 |
1 | 18122009 | rs4141983 | C | T | –0.0264 | 0.0046 | 0.326 | 9.69E-09 |
2 | 208049581 | rs2111592 | A | G | 0.0263 | 0.0046 | 0.3141 | 1.35E-08 |
2 | 212618440 | rs72948506 | A | G | 0.0265 | 0.0047 | 0.2975 | 1.71E-08 |
3 | 158171455 | rs35469634 | G | A | –0.0241 | 0.0044 | 0.5774 | 3.28E-08 |
3 | 61255413 | rs843812 | A | G | 0.0248 | 0.0044 | 0.4117 | 1.41E-08 |
3 | 49214303 | rs9831648 | T | G | –0.0292 | 0.0052 | 0.7739 | 1.59E-08 |
3 | 117515519 | rs66511648 | C | T | 0.0297 | 0.0048 | 0.284 | 6.03E-10 |
3 | 115977242 | rs76954012 | A | T | 0.0412 | 0.0074 | 0.0931 | 2.41E-08 |
5 | 103972357 | rs30266 | A | G | 0.0366 | 0.0046 | 0.3271 | 1.43E-15 |
5 | 87630769 | rs247910 | G | A | 0.0237 | 0.0043 | 0.457 | 4.71E-08 |
5 | 164487555 | rs7725715 | A | G | 0.029 | 0.0043 | 0.5343 | 1.61E-11 |
6 | 27182377 | rs150186873 | C | A | 0.0704 | 0.012 | 0.0327 | 4.51E-09 |
6 | 28366151 | rs2232423 | G | A | –0.062 | 0.007 | 0.1056 | 1.14E-18 |
6 | 165117329 | rs9364755 | G | A | 0.0283 | 0.0051 | 0.2262 | 3.49E-08 |
6 | 67000001 | rs2214123 | G | A | –0.0261 | 0.0045 | 0.6466 | 8.56E-09 |
6 | 142996618 | rs2876520 | G | C | 0.026 | 0.0043 | 0.4688 | 2.24E-09 |
7 | 82448100 | rs2522831 | C | T | 0.024 | 0.0043 | 0.4739 | 2.11E-08 |
7 | 109100414 | rs4730387 | A | T | 0.0238 | 0.0043 | 0.4659 | 4.12E-08 |
7 | 117625599 | rs150346963 | T | C | 0.0283 | 0.0044 | 0.4118 | 1.16E-10 |
7 | 12250402 | rs3807865 | A | G | 0.031 | 0.0044 | 0.4105 | 1.09E-12 |
7 | 2086814 | rs10235664 | C | T | –0.027 | 0.0049 | 0.2529 | 4.68E-08 |
7 | 38724868 | rs59082935 | T | C | 0.0363 | 0.0066 | 0.1342 | 3.07E-08 |
9 | 37182655 | rs62535714 | A | G | 0.0339 | 0.0058 | 0.1639 | 4.69E-09 |
9 | 11203149 | rs1931388 | G | A | –0.0295 | 0.0044 | 0.4042 | 1.68E-11 |
9 | 25232978 | rs59283172 | A | G | –0.039 | 0.007 | 0.1081 | 2.41E-08 |
9 | 119731359 | rs2418449 | C | T | –0.0281 | 0.0048 | 0.281 | 4.25E-09 |
10 | 106610839 | rs1021363 | G | A | –0.03 | 0.0045 | 0.6434 | 2.29E-11 |
11 | 61471678 | rs198457 | T | C | –0.0315 | 0.0056 | 0.1886 | 1.90E-08 |
11 | 88756779 | rs4497414 | C | T | 0.0291 | 0.0044 | 0.44 | 2.93E-11 |
11 | 113365141 | rs4936276 | C | G | 0.0278 | 0.0044 | 0.622 | 3.57E-10 |
12 | 52352301 | rs61914045 | A | G | 0.0309 | 0.0054 | 0.2034 | 7.96E-09 |
13 | 31790053 | rs9529218 | T | C | –0.034 | 0.0054 | 0.2031 | 2.23E-10 |
13 | 53860655 | rs9536381 | T | C | 0.0255 | 0.0046 | 0.3259 | 2.62E-08 |
13 | 80921519 | rs508502 | T | C | –0.0264 | 0.0048 | 0.2992 | 3.56E-08 |
14 | 42097937 | rs1950829 | G | A | –0.0297 | 0.0043 | 0.5173 | 4.74E-12 |
14 | 103997525 | rs754287 | A | T | –0.0289 | 0.0045 | 0.3664 | 1.31E-10 |
14 | 75125540 | rs7152906 | C | T | 0.0258 | 0.0043 | 0.5196 | 1.87E-09 |
15 | 88945878 | rs28541419 | G | C | –0.0292 | 0.0052 | 0.2308 | 1.76E-08 |
16 | 13800430 | rs12919291 | C | G | 0.0327 | 0.0055 | 0.1884 | 3.09E-09 |
18 | 35155910 | rs4799949 | T | C | –0.0292 | 0.0046 | 0.6684 | 1.40E-10 |
18 | 53099012 | rs12967143 | C | G | –0.0345 | 0.0047 | 0.7012 | 2.53E-13 |
18 | 77580712 | rs7241572 | A | G | 0.0323 | 0.0054 | 0.2047 | 2.43E-09 |
18 | 50861409 | rs1367635 | C | T | 0.0253 | 0.0043 | 0.5148 | 4.35E-09 |
20 | 44692598 | rs13037326 | T | C | 0.031 | 0.0049 | 0.2597 | 2.40E-10 |
SNP, single nucleotide polymorphisms; EAF, effect allele frequency; PCOS, polycystic ovary syndrome; Chr, chromosome; SE, standard error; IV, instrumental variants. |
Three methods including the inverse variance weighted (IVW), weighted median value, and MR Egger regression were utilized to assess the bilateral causal association between PCOS and depression (Fig. 1). The IVW approach hypothesizes that all the IVs are effective. It combines the effects of IVs and then yields an overall weighted effect. The weighted median estimator [20] can produce stable causality estimates even when 50% IVs are not valid. Two methods were undertaken to identify potential pleiotropy. First, we applied MR Pleiotropy Residual Sum and Outlier (MR-PRESSO) method to identify potential pleiotropic outliers, and MR-PRESSO conducts a global test of heterogeneity to identify potential horizontal pleiotropy [21]. Then, the intercept test from MR-Egger [22] was also applied to assess the directional pleiotropy. The intercept term significantly away from zero in statistics indicates the presence of pleiotropy and violation of basic MR assumptions.
Principles of mendelian randomization study for PCOS and depression. SNP, single nucleotide polymorphisms; PCOS, polycystic ovary syndrome.
Leave-one-out (LOO) analysis was used to identify the potential influential SNPs
in the causality estimates between PCOS and depression. p
As shown in Fig. 2, genetically predicted PCOS was associated with a 1.04-fold increased risk of depression by the IVW method (95% confidence interval (CI) = 1.01–1.06, p = 0.003). This increased risk was also replicated by the weighted median approach (OR = 1.04, 95% CI = 1.00–1.08, p = 0.03), suggesting a potential risky role of PCOS in the suffering of depression. No pleiotropic signs (MR-Egger intercept = –0.004, p = 0.733; MR-PRESSO global test p = 0.179) were observed (Table 3).
The forest plot for bidirectional Mendelian randomization. PCOS, polycystic ovary syndrome; OR, odds ratios; IVW, inverse-variance-weighting; CI, confidence interval.
Exposure | Outcome | MR-Egger intercept | MR-PRESSO global test | |
The estimates of egger intercept | p | p | ||
PCOS | Depression | –0.004 | 0.733 | 0.179 |
Depression | PCOS | –0.003 | 0.961 | 0.319 |
PCOS, polycystic ovary syndrome. |
As displayed in Fig. 3A, with the increase of the SNP effect on PCOS, the SNP effect on depression increased as well. The LOO analyses disclosed that no influential SNP existed in the PCOS-depression causal association.
The scatter plot for bidirectional Mendelian randomization. The subplots (A) represents the causal effects of PCOS on depression; The subplots (B) represents the causal effects of depression on PCOS.
The estimates of genetically forecasted depression on PCOS were displayed in Fig. 2. In the IVW analyses, the OR was 1.07 (95% CI = 0.76–1.50, p = 0.706), which did not suggest a risky role of depression in the occurrence of PCOS. The estimate from the weighted-median method (OR = 0.90, 95% CI = 0.55–1.47) was also insignificant (p = 0.674). There were no signs of pleiotropy (MR-Egger intercept = –0.003, p = 0.961; MR-PRESSO global test p = 0.319) in Table 3.
The scatter plot visualizing the SNPs-depression association against SNPs-PCOS association is displayed in Fig. 3B. The LOO analyses indicated that no influential SNP existed in the depression-PCOS association.
Herein, our team completed a two-sample MR and identified a potential causal association of PCOS with depression via the biggest GWAS data set to date. This study suggested that PCOS might increase the risk of depression. In contrast, we did not find evidence that depression may increase the risk of PCOS.
Previously, observational researches have unveiled the underlying correlation
between depression and PCOS. Açmaz et al. [23] discovered that PCOS
group (n = 86) with infertility had higher depression scores. Altinkaya
et al. [24] found that Beck Depression Inventory scoring was remarkably
greater in patients with PCOS (n = 83) in contrast to normal controls [24]. An
Iranian case control study found that there existed a remarkable diversity
between PCOS (n = 742) and controls (n = 798) in depression (18.9% vs. 7.9%;
p
Despite the fact that the biofunction of PCOS in the depression progression remains elusive, some researches have offered reasonable elucidation in this regard, one of which is hyperandrogenism [28, 29]. Second, a positive association between insulin resistance and depression was found [30, 31]. The randomized control trial (RCT) of Greenwood et al. [32] (738 PCOS females) potently reveals that insulin resistance is a causation factor for PCOS-related depression. In addition, PCOS is considered a proinflammation status featured by elevated contents of proinflammation biomarkers. Hence, there exists a probability of an inflammation association between depression and PCOS [33]. It is probable that the inflammation biomarkers in PCOS can cross the blood-brain barrier, inducing the progression of depression [33]. While the above theory-wise elucidation is reasonable, more researches are needed to reveal the potential causal link between PCOS and depression.
The major contribution of the present research is the utilization of the MR
method, which has been extensively utilized to explore the causality of PCOS with
the risks of other diseases. In addition, our team obtained the SNPs of
depression and PCOS via the biggest GWAS datasets to date. The SNPs herein were
remarkably related to PCOS at genome-wide significance, hence decreasing
potential breach of the first hypothesis of MR. In addition, the F-statistics for
the IVs all satisfied the liminal value of F-statistics
Nonetheless, there are certain limitations of our research. Firstly, as our analyses were limited by European individuals, the results might not be extended to other races. Nevertheless, this remarkably decreased the underlying influences of population stratification bias as well. Secondly, while the MR method may offer a non-biased outcome because of the diminished confounders, the gene–milieu and gene–gene interplay might influence the progression of depression or PCOS inevitably. Thirdly, the MR analyses of PCOS and depression were on the foundation of summary statistics with comparatively smaller sample size, and the underlying side effects of PCOS on the risks of depression ought to be further investigated in bigger sample size. In addition, as the related data of all PCOS phenotypes were unavailable, our team merely investigated the correlation between PCOS and depression, and our team did not stratify the outcomes herein as per the diverse PCOS phenotypes. More researches highlighting the correlation of PCOS with depression are still needed.
To sum up, the present research offered evidence to suggest potential causality between PCOS and an elevated risk of depression amongst European individuals. Nevertheless, the accurate roles and the potential biology processes of PCOS in the progression of depression require deeper explorations.
FQ and YF designed the study. XZ and YT conducted data collection. XZ and YG conducted data analysis. XZ and MD carried out table arrangement and picture drawing. YM conducted the supervision of data analysis. YF conducted supervision. XZ and YT wrote the original draft. YF and FQ reviewed the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.
Not applicable.
Not applicable.
This research was funded by National Natural Science Foundation of China (No. 82074476 and 81874480).
The authors declare no conflict of interest. FQ is serving as one of the Guest editors of this journal. We declare that FQ had no involvement in the peer review of this article and has no access to information regarding its peer review. Full responsibility for the editorial process for this article was delegated to MHD.
Publisher’s Note: IMR Press stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.