Cytokines and tryptophan metabolites can predict depressive symptoms in pregnancy

Abstract

Depression during and after pregnancy affects up to 20% of pregnant women, but the biological underpinnings remain incompletely understood. As pregnancy progresses, the immune system changes to facilitate fetal development, leading to distinct fluctuations in the production of pro-inflammatory factors and neuroactive tryptophan metabolites throughout the peripartum period. Therefore, it is possible that depression in pregnancy could constitute a specific type of inflammation-induced depression. Both inflammatory factors and kynurenine metabolites impact neuroinflammation and glutamatergic neurotransmission and can therefore affect mood and behavior. To determine whether cytokines and kynurenine metabolites can predict the development of depression in pregnancy, we analyzed blood samples and clinical symptoms in 114 women during each trimester and the postpartum. We analyzed plasma IL-1β, IL-2, -6, -8, -10, TNF, kynurenine, tryptophan, serotonin, kynurenic- quinolinic- and picolinic acids and used mixed-effects models to assess the association between biomarkers and depression severity. IL-1β and IL-6 levels associated positively with severity of depressive symptoms across pregnancy and the postpartum, and that the odds of experiencing significant depressive symptoms increased by >30% per median absolute deviation for both IL-1β and IL-6 (both P = 0.01). A combination of cytokines and kynurenine metabolites in the 2nd trimester had a >99% probability of accurately predicting 3rd trimester depression, with an ROC AUC > 0.8. Altogether, our work shows that cytokines and tryptophan metabolites can predict depression during pregnancy and could be useful as clinical markers of risk. Moreover, inflammation and kynurenine pathway enzymes should be considered possible therapeutic targets in peripartum depression.

Trial registration: ClinicalTrials.gov NCT02566980.

Conflict of interest statement

Dr. Achtyes has served on advisory boards for Alkermes, CAPNOS Zero (unpaid), Indivior, Janssen, Lundbeck/Otsuka, Neurocrine Biosciences, Roche, Sunovion and Teva and reports previous stock holdings in AstraZeneca, Johnson & Johnson, Moderna, and Pfizer. Dr. Achtyes has received research support from Alkermes, Astellas, Biogen, Boehringer-Ingelheim, InnateVR, Janssen, National Network of Depression Centers, Neurocrine Biosciences, Novartis, Otsuka, Pear Therapeutics, Takeda, and the Vanguard Research Group. All the other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

© 2022. The Author(s).

Figures

Fig. 1. Tryptophan metabolic pathways in the brain and periphery.

Abbreviations: TPH tryptophan hydroxylase, IDO indoleamine 2,3-dioxygenase. TDO tryptophan 2,3-dioxygenase. KATs kynurenine aminotransferases. KYNU kynureninase, HAAO 3-hydroxyanthranilate 3,4-dioxygenase, ACMSD aminocarboxymuconate-semialdehyde decarboxylase, QPRT quinolinic acid phosphoribosyltransferase, CSF cerebrospinal fluid, NMDAR N-methyl-D-aspartate receptor, NAD nicotinamide.

Fig. 2. ROC and PR curves of the prediction model comparisons.

ROC and PR curves of the prediction model comparisons (between full, age- only and null models) using biomarkers collected during the 2nd trimester to predict 3rd trimester EPDS ≥ 13. A using ordinal regression (cumulative probability that EPDS ≥ 13), B logistic regression (estimated probability EPDS ≥ 13). The full model outperforms the other two models in both the ROC and PR curves. The performance of the PR curve is important when the number of cases (EPDS ≥ 13) is not equal to non-cases (EPDS < 13). The leave-one-out cross-validation results are also consistent with the full model having the best predictive accuracy.

Fig. 3. Depression-warning cut off per age.

Bayesian models with flat priors were used to predict the probability of each 3rd trimester EPDS score based on individual marker levels and age. Plotted here are the minimal marker levels where the most probable EPDS score is ≥10 for women ages 18–44. As marker levels exceed these plotted thresholds, the probability that a given woman would score ≥ 10 on the EPDS scale in the 3rd trimester continues to increase asymptotically to 1. An EPDS score of 10 or greater is used clinically to screen women for PPD, thus these plots present a proof-of-concept for an advanced screening method where a woman’s age and measures of these 4 markers could jointly be used to flag someone as likely to score ≥ 10 on the EPDS assessment in the next trimester and therefore considered to be at risk of developing future PPD.