Associations among peripheral and central kynurenine pathway metabolites and inflammation in depression

Abstract

Kynurenine pathway (KP) metabolites are believed to be a link between inflammation and depression through effects on brain glutamate receptors. However, neither the relationship between plasma and cerebrospinal fluid (CSF) KP metabolites nor their association with inflammatory mediators is well-established in depression. Moreover, the clinical profile associated with combined activation of plasma inflammatory and kynurenine pathways is unknown. Accordingly, plasma and CSF-KP metabolites and inflammatory markers along with depressive symptoms and antidepressant treatment response were measured in 72 unmedicated depressed patients. Following bivariate analyses, component factors representing immune and kynurenine variables in the plasma and CSF were extracted and were used to examine directionality of associations in a path model. In addition, patients were clustered using individual markers that most accounted for the association between plasma immune and KP systems. Path analysis revealed a directional association extending from plasma inflammatory markers to plasma kynurenines, to CSF kynurenines. Among immune markers, plasma tumor necrosis factor (TNF) was robustly associated with plasma kynurenine (KYN) and KYN/tryptophan (TRP), which was in turn significantly associated with CSF KYN, kynurenic acid, and quinolinic acid. Clustering of patients based on plasma TNF and KYN/TRP yielded subgroups of high (N = 17) and low (N = 55) TNF-KYN/TRP groups. High TNF-KYN/TRP subjects exhibited greater depression severity, anhedonia, and treatment nonresponse. In conclusion, plasma-KP metabolites may mediate an inflammation-associated depressive symptom profile via CNS KP metabolites that can serve as a target for intervention at the level of inflammation, peripheral KYN metabolism, KYN transport to the brain, or effects of KP metabolites on glutamate receptors.

Figures

Fig. 1. Correlation heatmap of immune and kynurenine pathway markers.

The x- and y-axes represent quantile-normalized kynurenine and immune markers in the plasma and cerebrospinal fluid (CSF) compartments, and accompanying r-scale represents strength and directionality of Pearson’s product–moment correlation coefficients. Bonferroni-corrected p-values of the above correlation coefficients are provided in Supplementary Table S4. Abbreviations: AA anthranilic acid, CRP c-reactive protein, IL6 interleukin-6, IL6sr IL6-soluble receptor, IL10 interleukin-10, IL1b interleukin-1β, IL1ra IL1b-receptor antagonist, KYN kynurenine, KYN/TRP kynurenine/tryptophan, KYNA kynurenic acid, MCP1 monocyte chemoattractant protein-1, QA quinolinic acid, QA/KYNA quinolinic acid/kynurenic acid, TNF tumor necrosis factor, TNFR2 soluble tumor necrosis factor type-2, TRP tryptophan, 3HAA 3-hydroxy anthranilic acid, 3HKYN 3-hydroxy kynurenine.

Fig. 2. Bivariate associations between Plasma KYN/TRP and CSF kynurenine pathway metabolites.

The panel presents linear fits between individual immune and kynurenine pathway markers in the plasma and CSF. KYN/TRP is represented as a proxy for indoleamine-2-dioxygenase (IDO) activity, as it is responsible for generating kynurenine from tryptophan. a Demonstrates significant linear association between plasma kynurenine/tryptophan ratio (KYN/TRP, *1000 for graphic scaling) and CSF KYN/TRP [r (99% CI) = 0.77 (0.50–0.90)]. Similar relationships are demonstrated in (b) between plasma KYN/TRP and CSF kynurenine [KYN, r (99% CI) = 0.72 (0.42–0.88)]; c between plasma KYN/TRP and CSF anthranilic acid [AA, r (99% CI) = 0.60 (0.22–0.82)], and d between plasma KYN/TRP and CSF quinolinic acid [QA, r (99% CI) = 0.52 (0.08–0.79)]. All p-values were < 0.0001.

Fig. 3. Path diagram of the directional associations between plasma and CSF immune markers (IMM) and kynurenine pathway (KP) metabolites.

Path analysis was conducted to test the directionality of relationships among the extracted component factors controlling for age, race, and BMI (as detailed in the text). The broken arrows represent indirect effects and the filled arrows represent direct effects with increasing line widths signifying stronger effects. Although included in the analyses, covariates are excluded from this figure to improve clarity. Analytic details are provided in Supplementary Table S6A–D. Abbreviations: Plasma-IMM plasma immune component factor, Plasma-KP plasma kynurenine pathway component factor, CSF-IMM cerebrospinal immune component factor, CSF-KP cerebrospinal kynurenine pathway component factor, Std Cf standardized path coefficients (calculated by normalizing to variance). *Significant effects at p < 0.001.

Fig. 4. Clustering patients based on plasma TNF and KYN/TRP to associate with clinical characteristics.

a Demonstrates nonlinear, quadratic fit between plasma tumor necrosis factor (TNF) and kynurenine/tryptophan (KYN/TRP, *1000 for graphic scaling). The curved (quadratic/polynomial) fit (broken line) was significantly stronger than the linear fit (linear/nonlinear slope difference [F(1,69) = 11.3, p = 0.001]. The blue line shows the linear fit between Plasma KYN/TRP and Plasma TNF <5.67 pg/ml, whereas the red line shows the same fit at Plasma TNF >5.67 pg/ml. b Demonstrates the use of K-means clustering to subgroup patients into Low (n = 55) and High (n = 17) TNF-KYN/TRP groups. A connected scatterplot associating plasma TNF in the x-axis and KYN/TRP (*1000) in y-axes demonstrates significant separation between clusters. The High and Low TNF-KYN/TRP groups are represented using the filled red and blue circles, respectively. The cross-haired, hollow circles in the center of the groups represent the centroid values for the two clusters. c–f Present area under the curve estimates (AUC) obtained from receiver operating (ROC) curves to compare the stability of the association between TNF-KYN/TRP clusters and clinical variables across the “Discovery” (shaded purple/blue) and “Validation” (shaded orange) sets. ROC curves associating TNF-KYN/TRP group status (High vs. Low) with depression severity (IDS-SR scores) is presented in (c); with anhedonia (anhedonia subscale scores) in (d); and treatment nonresponse (MGH-ATRQ scores) in (e). f Presents comparison of AUC measures across Discovery and Validation sets. The insignificant differences in AUC between the randomly split Discovery and Validation subset, indicates satisfactory cross-validation.