The effect of inflammation, SARS-CoV-2 infection, age and mental health on serotonin, and kynurenine and catecholamine pathway metabolites

Background: A high prevalence of mental disorders following COVID-19 has been described. It is therefore essential to elucidate underlying biological mechanisms linking SARS-CoV-2 infection and mental health. The kynurenine and catecholamine metabolic pathways are modulated by inflammation and can affect systemic levels of serotonin and dopamine. Their activity may hence link physical disorders with mental health. We investigated factors that affect kynurenine and catecholamine pathway activity in SARS-CoV-2 infection and recovery. Methods: The cross-sectional SIMMUN (n = 165) and longitudinal INCOV cohort (n = 167, Su et al. 2022) were analyzed. Demographic and clinical characteristic, inflammatory markers, SARS-CoV-2 infection, symptoms of depression and anxiety (HADS), and mental stress (PSS-4) served as explanatory variables. Blood serotonin and markers of kynurenine (kynurenine/tryptophan ratio), and catecholamine pathway activity (dopamine 3-O-sul-fate, phenylalanine/tyrosine ratio) were modeled by multi-parameter linear regression. Results: In the SIMMUN cohort, the inflammatory marker neopterin ( β = 0.47 [95% CI: 0.34 – 0.61]), SARS-CoV-2-positivity (0.42 [0.16 – 0.68]), mental stress (0.18 [0.055 – 0.31]), and age (0.26 [0.12-0.39]) were positively associated with the kynurenine/tryptophan ratio. The phenylalanine/tyrosine ratio was lower in SARS-CoV-2-positive than uninfected participants ( (cid:0) 0.38 [ (cid:0) 0.68 to (cid:0) 0.08]). In the INCOV cohort, markers of inflammation were associated with lower serotonin (IL6: (cid:0) 0.22 [ (cid:0) 0.38 to (cid:0) 0.053]) and dopamine 3-O-sulfate levels (interferon-gamma: (cid:0) 0.15 [ (cid:0) 0.26 to (cid:0) 0.036]). Serotonin (0.76 [0.34 – 1.


Introduction
The immune system and the brain interact at multiple levels with influences on one system having consequences on the other (Dantzer et al., 2008). Physical disorders such as infections, cardiovascular disease, frailty, autoimmune illness or malignancy are often accompanied by symptoms of depression, anxiety and mental stress (Dickens and Creed, 2001;Hüfner et al., 2022;Hüfner et al., 2015;Mayerl et al., 2020;Mitchell et al., 2011;Rajan et al., 2020;Renault et al., 1987). Inflammation interferes with key metabolic processes such as the kynurenine and catecholamine pathway, which can affect serotonin and dopamine availability, respectively. Inflammation was thus proposed as an important link between mental and physical health (Bonaccorso et al., 2002;Brown et al., 2021;Dantzer et al., 2008;Gostner et al., 2020;Hüfner et al., 2015;Valdiglesias et al., 2018).
Serotonin is synthesized from the essential amino acid tryptophan (TRP) in reactions catalyzed by tryptophan hydroxylase (TPH) and aromatic L-amino acid decarboxylase. Most systemic TRP is catabolized along the kynurenine pathway. The conversion of TRP to kynurenine (KYN) by tryptophan 2,3-dioxygenase (TDO) or indoleamine 2,3-dioxygenases (IDO1, IDO2) is the first, rate-limiting step of the kynurenine pathway and may thereby control TRP levels subsequently available for serotonin synthesis (Badawy, 2017;Lukić et al., 2022). Importantly, IDO1 is stimulated by inflammatory signals (Badawy, 2017;Robinson et al., 2003) and TDO activity is increased by glucocorticoids which are also released as a reaction to mental stress (Knox, 1951;Maes et al., 1990). Hence, activity of IDO and TDO, which can be assessed by the ratio of KYN to TRP (KYN/TRP), is modulated by inflammatory and hypothalamic-pituitary-adrenal axis signals. Some kynurenine pathway metabolites such as KYN can additionally pass the blood-brain barrier, the catabolite quinolinic acid (QUIN) can interfere e.g. with glutamatergic signaling and exert neurotoxicity (Badawy, 2017;Brown et al., 2021;Schwarcz et al., 2012). High KYN and KYN/TRP, decreased TRP as well as low levels of associated neurotransmitter serotonin were reported for numerous physical disorders and are associated with symptoms of mental disorders (Capuron et al., 2011;Cervenka et al., 2017;Fellendorf et al., 2022;Hüfner et al., 2021Hüfner et al., , 2019Hüfner et al., 2015;Hunt et al., 2020;Marx et al., 2020).
The synthesis of catecholamine neurotransmitters such as dopamine via the catecholamine pathway involves the conversion of phenylalanine (PHE) to tyrosine (TYR) catalyzed by phenylalanine hydroxylase (PAH) followed by hydroxylation of TYR by tyrosine 3-hydroxylase (TH). PAH, TH and the serotonin-synthesizing TPH enzyme depend on tetrahydrobiopterin (BH4) as an essential cofactor (Meiser et al., 2013;Neurauter et al., 2008). During inflammation, BH4 is oxidized to its inactive derivative and consumed in a competitor reaction of nitric oxide formation (Neurauter et al., 2008;Rahimian et al., 2022). This results in a functional BH4 deficiency and ultimately a reduced biosynthesis of dopamine and serotonin (Capuron et al., 2011;Geisler et al., 2013;Vancassel et al., 2018). Accordingly, changes in PHE and TYR levels have been reported in different physical disorders such as cancer, infections and inflammatory conditions and associated with depression and anxiety (Capuron et al., 2011;Geisler et al., 2013;Hüfner et al., 2015;Vancassel et al., 2018).
SARS-CoV-2 virus is the causal pathogen of coronavirus disease 2019 . Apart from sustained physical disability, mental disorders in millions of individuals contribute the persistent burden of the COVID-19 pandemic (Al-Aly et al., 2021;Huang et al., 2021;Hüfner et al., 2022;Sahanic et al., 2023;Staudt et al., 2022). Increased kynurenine pathway activity has recently been identified in acute COVID-19 and recovery and is associated with infection severity (Bizjak et al., 2022;Dewulf et al., 2022;Lionetto et al., 2021;Santiago-Mujika et al., 2022). Theoretical considerations (Bower et al., 2022) and preliminary experimental data suggest that a highly active kynurenine pathway as well as reduced biosynthesis of serotonin and catecholamine neurotransmitters may contribute to symptoms of mental stress, anxiety and depression in COVID-19 patients (Kucukkarapinar et al., 2022;Matits et al., 2023).
The interaction of mental health, inflammation, kynurenine and catecholamine pathway activity and systemic levels of serotonin and dopamine in COVID-19 has not been investigated in clinical populations. To address this question, we analyzed the effects of demographic, clinical, psychometric (anxiety, depression, mental stress), inflammation-and SARS-CoV-2-related factors on the kynurenine pathway activity and associated serotonin levels as well as the catecholamine pathway in two cohorts of SARS-CoV-2 patients: the cross-sectional SIMMUN cohort and the published longitudinal INCOV cohort (Su et al., 2022).

Materials and Methods
Details on study cohorts, procedures and analysis are provided in Supplementary Methods.

Ethics statement
The SIMMUN study was conducted in accordance with the Declaration of Helsinki and European Data Policy. All participants gave written informed consent prior to enrollment. Participants' data were analyzed in anonymized form. This study was approved by the ethics committee of the Medical University Innsbruck (Austria, approval number: 1132/ 2020). No approval by the ethics committee was required for the published INCOV dataset (Su et al., 2022).

SIMMUN study
Individuals tested for SARS-CoV-2 via PCR at the University Hospital of Innsbruck and patients of the University Clinic for Psychiatry I and II (Innsbruck, Austria) undergoing routine SARS-CoV-2 PCR were invited to participate. The study was conducted between June 2020 and May 2021. The inclusion criteria were age of 18-70 years, proficiency in German, residence in the study region (Tyrol, Austria), and a SARS-CoV-2 PCR conducted at the study site. The exclusion criteria were active SARS-CoV-2 infection (< 14 days after diagnosis), pregnancy, active malignancy, organ transplantation, surgery in the past 3 months, inflammatory illness and oral corticosteroid treatment. Only individuals with complete explanatory and response variables presented in Supplementary Table S1 and Supplementary Fig. S1 were analyzed. Significant differences between the analyzed and excluded participants are listed in Supplementary Table S2.

INCOV study
Proteome, metabolome and clinical data are publicly available (Su et al., 2022). Samples with complete explanatory and response variables presented in Supplementary Fig. S1 and Supplementary Table S3 were analyzed.

SIMMUN study
The SIMMUN data were collected during a single on-site visit at median 139 days after the SARS-CoV-2 PCR (interquartile range: 119-157) and extracted from electronic patient records (Supplementary  Table S1). Demographic and clinical characteristics included age, sex, body mass index (BMI), mental disorders diagnosed by a physician, selfreported physical disorders, smoking and alcohol consumption, result and date of the SARS-CoV-2 PCR. Neutrophil/lymphocyte ratio (NLR) and serum concentration of neopterin (NEO) served as inflammatory markers. NLR was determined by a certified laboratory at the University Hospital of Innsbruck. NEO was measured by enzyme-linked immunosorbent assay (BRAHMS Diagnostics, Berlin, Germany). Serum TRP, KYN, PHE and TYR were determined by high-performance liquid chromatography (Neurauter et al., 2008;Widner et al., 1997). Immunoglobulin gamma against receptor binding domain S1/S2 protein (anti-RBD IgG) was quantified by ELISA (Deisenhammer et al., 2021). Laboratory measurements beyond the detection limits were substituted with the respective detection limit value. Mental stress was scored with the 4-item perceived stress scale (PSS-4) (Cohen et al., 1983). Clinically relevant symptoms of anxiety and depression were defined as ≥ 8 points in the hospital anxiety and depression scale (HADS) (Bjelland et al., 2002;Zigmond and Snaith, 1983).
Plasma samples were collected during acute and sub-acute infection, and recovery at median 10, 14, 64 days after diagnosis of SARS-CoV-2 infection via PCR, respectively (Supplementary Table S4).

Analysis endpoints
The first endpoint was to determine demographic, clinical, psychometric, inflammation-and SARS-CoV-2-related factors influencing serum markers of kynurenine and catecholamine pathway activity (TRP, KYN, KYN/TRP, PHE, TYR, PHE/TYR) by multi-parameter modeling in the SIMMUN cohort. The second endpoint was to investigate how systemic serotonin and dopamine availability (DA sulfate) are influenced by age, sex, BMI, cytokine markers of inflammation, infection status and timepoint of SARS-CoV-2 infection, kynurenine (TRP, KYN, QUIN) and catecholamine (PHE, TYR) pathway metabolite levels. This endpoint was addressed by multi-parameter robust linear modeling, time course modeling and correlation analysis in the INCOV cohort ( Supplementary  Fig. S1).

Statistical analysis
R version 4.2.3 was used for statistical analysis. Numeric variables were presented as medians with interquartile ranges and ranges. Categorical variables are presented as percentages and counts. Normality and variance homogeneity was assessed by Shapiro-Wilk and Levene test, respectively. In the SIMMUN cohort, non-normally distributed numeric variables were logarithm-or square root-transformed prior to modeling and parametric tests (Supplementary Table S1). Since most of the INCOV study variables were non-normally distributed, robust linear modeling and non-parametric testing were employed. Consistency of psychometric tools was investigated by global McDonald's ω (McDonald, 1999). Except for multi-parameter modeling, p values were corrected for multiple testing with the false discovery rate method (Benjamini and Hochberg, 1995) separately for each analysis task. Effects with p < 0.05 were considered significant.
Significance was determined by Mann-Whitney test with r effect size, two-tailed T test with Cohen's d effect size statistic or χ 2 test with Cramer's V effect size statistic. Correlation was assessed by Pearson's or Spearman's rank test and visualized as bubble or force-directed network plots (Csardi and Nepusz, 2006). Responses and numeric explanatory variables were normalized prior to modeling. In the SIMMUN cohort, effects of age, sex, BMI, physical and mental disorders, BMI, smoking and alcohol consumption, inflammation markers (NEO, NLR), SARS-CoV-2 infection, anti-RBD IgG titer, clinically relevant symptoms of depression and anxiety (HADS), mental stress (PSS-4) on kynurenine and catecholamine pathway activity markers were assessed by multi-parameter linear regression with backward elimination. In the INCOV cohort, effects of age, sex, BMI, cytokine markers of inflammation (IL6, IL10, TNF, IFNG), SARS-CoV-2 infection status and timepoint (uninfected, acute, sub-acute, recovery), kynurenine (TRP, KYN, QUIN) and catecholamine pathway metabolites (PHE, TYR) on plasma serotonin and DA sulfate were modeled by multiparameter robust linear regression (Huber, 2011). Explanatory performance, reproducibility and proper parameterization of the multi-parameter models was investigated by R 2 and root mean square error (RMSE) in 10-fold cross-validation. Differences in cytokines and metabolites between SARS-CoV-2 infection timepoints in the INCOV collective were investigated by robust linear modeling with uninfected subset or acute infection serving as baselines.

Characteristic of the cohorts
Two cohorts of uninfected and SARS-CoV-2-infected individuals were analyzed (SIMMUN and INCOV). Out of 215 individuals enrolled in the SIMMUN study, 165 participants for whom the complete study variables were available, were analyzed ( Supplementary Fig. S1, Supplementary Table S1). The excluded individuals were characterized by more frequent mental disorders and more frequent clinically relevant symptoms of depression and anxiety (HADS), higher scores of mental stress (PSS-4) and less frequent SARS-CoV-2 infections than the analyzed participants (Supplementary Table S2). In the analyzed SIMMUN cohort, males represented 38% and the median age was 50 years. The SIMMUN study data were recorded at median 139 days after the SARS-CoV-2 PCR. None of the participants had received anti-SARS-CoV-2 vaccination prior to enrollment. SARS-CoV-2-positive individuals accounted for 39% of the analyzed participants. The gender and age structure of the SARS-CoV-2-negative and -positive subsets was comparable. In the entire SIMMUN cohort, 47% of participants were overweight or obese, and 51% reported a physical disorder; these figures were comparable between SARS-CoV-2-negative and -positive individuals. Mental disorders diagnosed by a physician affected 41% of participants and were significantly more common in SARS-CoV-2-negative (50%) than in SARS-CoV-2-positive individuals (28%, p = 0.021, effect size: V = 0.21) ( Table 1). The psychometric tools for assessment of perceived mental stress (PSS-4), and symptoms of depression and anxiety (HADS) displayed good-toexcellent internal consistency (ω = 0.74-0.96) (McDonald, 1999) ( Supplementary Fig. S2). Clinically relevant symptoms of anxiety (HADS ≥ 8) were more frequent in SARS-CoV-2-negative (43%) than in SARS-CoV-2-positive participants (20%, p = 0.013, effect size: V = 0.23). Clinically relevant symptoms of depression (HADS ≥ 8) were more common in SARS-CoV-2-negative (31%) than in SARS-CoV-2-positive participants (16%), yet this effect was not statistically significant. Scores of mental stress were comparable in both SARS-CoV-2 subsets (negative: median 6, positive: median 5 points). In 73% of SARS-CoV-2-infected SIMMUN study participants the infection was mild and treated on an outpatient basis ( Table 1).

Inflammation, SARS-CoV-2 infection, age, mental stress and depression influence systemic activity of the kynurenine pathway
In the SIMMUN dataset, we initially searched for predictors of activity of the kynurenine and catecholamine pathways. TRP, KYN and KYN/TRP served as markers of the kynurenine pathway, and PHE, TYR and PHE/TYR were investigated as markers of the catecholamine pathway activity. The candidate explanatory variables were age, sex, BMI, the presence of physical and mental disorders, smoking, alcohol consumption, SARS-CoV-2 infection, anti-SARS-CoV-2 antibody titer, NEO and NLR as inflammatory markers, perceived mental stress (PSS-4), and clinically relevant symptoms of depression and anxiety (HADS) (Supplementary Table S1).
Meaningful multi-parameter models optimized by backward elimination could be established for TRP, KYN, KYN/TRP, TYR and PHE/TYR. The PHE model had a negligible explanatory performance, proved non-significant as compared with the intercept-only null model and was hence not further analyzed (Supplementary Table S6). The remaining models were characterized by good reproducibility and proper parameterization as indicated by comparable fit errors (RMSE) in the training dataset and cross-validation. The KYN and KYN/TRP ratio models had the best, moderate-to-substantial explanatory performance measured by cross-validated R 2 of 0.21 and 0.3, respectively. The crossvalidated R 2 metric values for the TRP, TYR and PHE/TYR models ranged from 0.1 to 0.15 indicative of weak-to-moderate explanatory value ( Supplementary Fig. S3).
By analysis of single significant predictors which had been identified by modeling, we could corroborate significant, positive, moderate-tostrong correlations of age and NEO with KYN and KYN/TRP in the SIMMUN cohort. This analysis also revealed a significant positive correlation of age with TYR and a significant negative correlation of age with PHE/TYR ( Supplementary Fig. S4 -S5, Supplementary Tables S7 -S8).

Inflammatory cytokines, SARS-CoV-2 infection timepoint, age and availability of biosynthesis precursors regulate systemic serotonin and dopamine levels
Next, we investigated factors affecting systemic levels of serotonin and a major circulating catabolite of dopamine, DA sulfate. This analysis was performed in the INCOV cohort with age, sex, BMI, markers of inflammation (IL6, IL10, TNF, IFNG), status and timepoint of SARS-CoV-2 infection, concentrations of kynurenine (TRP, KYN, QUIN) and catecholamine pathway metabolites (PHE, TYR) as candidate explanatory variables (Supplementary Table S3). Metabolites and cytokines were sampled in SARS-CoV-2-positive participants during acute and subacute infection, and recovery, at median 10, 14 and 64 days after infection diagnosis via PCR, respectively (Supplementary Table S4).
The multi-parameter robust regression model for plasma serotonin was characterized by moderate explanatory performance (cross-validated R 2 = 0.15). The explanatory performance of the DA sulfate model was weak (cross-validated R 2 = 0.053). Comparable fit errors (RMSE) in the training dataset and cross-validation indicated good reproducibility and proper parameterization of the models ( Fig. 2A).

Result summary
We investigated factors affecting the kynurenine as well as catecholamine pathway activity and serotonin availability in SARS-CoV-2 infection in two separate cohorts. Our analysis underlines the pivotal role of SARS-CoV-2-dependent and independent inflammation in serotonin, kynurenine, and catecholamine metabolism. Combining the results from both cohorts we report higher levels of the kynurenine pathway markers, KYN, KYN/TRP and QUIN, and lower concentration of the serotonin precursor TRP, in the presence of inflammation. This observation and the significant association of serotonin with TRP levels in the INCOV cohort suggest that possibly depletion of TRP via the kynurenine pathway could limit systemic serotonin availability in an inflammatory mileu (Badawy, 2017). Age and mental stress were identified as infection-independent predictors of high kynurenine pathway activity. Additionally, our data suggest that the catecholamine pathway and the resulting systemic dopamine availability are also affected through inflammation and SARS-CoV-2 positivity.
We note that explanatory performance of the analyses of serotonin availability and kynurenine pathway activity measured by R 2 and correlation coefficients were generally higher than explanatory values of the catecholamine pathway analyses. This indicates that the predictors identified by our analysis, inflammation, mental stress, age, and SARS-CoV-2-positivity affect the kynurenine pathway activity to a larger degree than catecholamine metabolism, in line with a recent metabolomic study in COVID-19 individuals (Thomas et al., 2020). Effects of inflammation on markers of the kynurenine and catecholamine pathway activity were stronger in the INCOV than in the SIMMUN cohort, which may reflect more severe infection in INCOV study participants. In the SIMMUN cohort, the inflammatory marker NLR was unexpectedly linked to lower KYN, which may be explained by a sustained low-grade neutropenia reported for up to 1 year after COVID-19 (Lin et al., 2022).

Effects of inflammation and infection on the kynurenine pathways and systemic levels of serotonin
IDO1, together with TDO and IDO2, catalyzes the first reaction of the kynurenine pathway, the TRP -KYN conversion, which catabolizes > 90% of TRP and might thereby limit serotonin availability (Badawy, 2017;Lukić et al., 2022). Expression and activity of IDO1 is induced by inflammatory stimuli such as IL6, TNF and IFNG (Badawy, 2017;Robinson et al., 2003). Consequently, increased kynurenine pathway activity was reported for multiple inflammatory conditions (Badawy, 2017). In COVID-19, levels of kynurenine pathway metabolites in blood and urine were found to correlate positively with inflammatory markers

days after SARS-CoV-2 infection diagnosis via PCR), plasma levels of metabolites of the kynurenine (tryptophan [TRP], kynurenine [KYN], quinolinic acid [QUIN]) and catecholamine pathways (phenylalanine [PHE], tyrosine [TYR]), and plasma concentrations of cytokine markers of inflammation (interleukin-6 [IL6], interleukin-10 [IL10], tumor necrosis factor-alpha [TNF] and interferon-gamma [IFNG]) on plasma concentrations of serotonin and dopamine 3-O-sulfate (DA sulfate) were modeled by multi-parameter robust linear regression. (A) Model fit error (root mean square error [RMSE]) and fraction of explained variance (R 2 ) of the robust linear models assessed in the training dataset and infection timepoint-stratified 10-fold cross-validation (CV). (B) Estimates of model coefficients (β) with 95% confidence intervals presented in Forest plots. Numbers of complete observations are indicated in the plot captions.
Model intercepts (baseline) are indicated in the X axis titles. (Dewulf et al., 2022;Santiago-Mujika et al., 2022), disease severity (Ceballos et al., 2022;Dewulf et al., 2022;Lionetto et al., 2021) and were implicated in persistent COVID-19 symptoms (Bizjak et al., 2022;Cysique et al., 2022;Matits et al., 2023). These findings were summarized in a recent metaanalysis demonstrating an increased KYN/TRP ratio in COVID-19, particularly in severe manifestations (Almulla et al., 2022). Using high-throughput metabolomics, tryptophan metabolism via serotonin and kynurenine pathways was identified as the most prominently affected metabolic system in acute COVID-19. In this study, TRP and serotonin were demonstrated to decrease and KYN was shown (right). Numbers of complete observations per timepoint are indicated in the X axes. to increase in an IL6-dependent way (Thomas et al., 2020). Of note, IL6 was also identified in our present analysis of the INCOV cohort as a predictor of lower plasma serotonin. We demonstrated significant, positive associations of serotonin and TRP levels in acute and sub-acute infection in the INCOV cohort. This suggests that kynurenine pathway may especially efficiently compete for TRP with systemic serotonin synthesis in the highly inflammatory milieu of early COVID-19. In turn, increasing TRP and serotonin during infection recovery in the INCOV cohort may reflect decreasing IDO1 activity and re-routing of TRP to the serotonin biosynthesis (Badawy, 2017;Lukić et al., 2022). The multi-parameter modeling in the SIMMUN cohort revealed an additional, inflammation-independent effect of SARS-CoV-2-positivity at a median of 139 days after infection on KYN and KYN/TRP. While its mechanism remains unclear, an analogical sustained increase in KYN in the absence of the inflammatory marker C-reactive protein was described in long-term COVID-19 recovery (Bizjak et al., 2022). Age was identified in the present analysis as another predictor of high kynurenine pathway activity and also low serotonin availability, which may be a result of a chronic low-grade inflammation observed in elderly (Capuron et al., 2011;Sorgdrager et al., 2019). Furthermore, significantly lower levels of serotonin were observed in male participants of the INCOV cohort. This may reflect a generally stronger inflammatory response to SARS-CoV-2 in males (Ceballos et al., 2022), which was also evident in our data (INCOV, IL6: β = 0.69 [95% CI: 0.073-1.3], male versus female, robust linear modeling).

Effects of inflammation and infection on the catecholamine pathway and systemic dopamine availability
BH4 availability is another link between neurotransmitter metabolism and inflammation. Activity of GTP cyclohydrolase I, the enzyme catalyzing the rate-limiting step of BH4 synthesis, is strongly stimulated by IFNG and other inflammatory stimuli. However, during inflammation, BH4 availability for enzymes of the catecholamine and serotonin synthesis is reduced by oxidation and nitric oxide synthesis (Geisler et al., 2013;Meiser et al., 2013;Neurauter et al., 2008). As previously reported (Luporini et al., 2021), we observed a higher plasma PHE during acute and sub-acute infection in the INCOV cohort. We also found an association of metabolites of catecholamine pathway with inflammatory markers. Taken together, these observations are consistent with the inflammatory BH4 deficiency leading to a reduced PHE -TYR conversion and impaired catecholamine biosynthesis during acute and sub-acute SARS-CoV-2 infection. By contrast, the increased DA sulfate during infection recovery in the INCOV cohort and the significantly lower PHE/TYR at median 139 days after SARS-CoV-2 infection in the SIMMUN cohort may indicate a restored BH4 homeostasis and efficient dopamine synthesis (Geisler et al., 2013;Goldstein et al., 1999;Meiser et al., 2013;Neurauter et al., 2008). Contrary to the reported suppression of catecholamine synthesis with age (Peters, 2006), we observed higher blood levels of TYR and lower PHE/TYR in older participants of the SIMMUN study. However, since we could not replicate this effect in the INCOV cohort, its relevance remains questionable. Analogically, although low DA sulfate levels were associated with obesity in the INCOV cohort, we could not observe any significant effects of obesity on markers of catecholamine pathway activity in the SIMMUN cohort.

Interaction of mental health with activity of the kynurenine and catecholamine pathways, and systemic serotonin and dopamine availability
Peripheral and central nervous system inflammation has been identified as a neuroimmune mechanism contributing to mental stress (Hüfner et al., 2020;Kim and Won, 2017) and mental disorders such as depression or anxiety (Brown et al., 2021;Schwarcz et al., 2012;Vancassel et al., 2018). Mechanistically, sickness behavior which accompanies acute inflammation and shares many similarities to depression, was postulated to involve the kynurenine and catecholamine pathway (Dantzer et al., 2008;Maes et al., 2012;Vancassel et al., 2018). Additionally, kynurenine pathway metabolites can contribute to sickness behavior and depression via interference with neuronal signaling, e.g. with glutamatergic receptors (Brown et al., 2021;Schwarcz et al., 2012). Markers of kynurenine pathway activity such as TRP, KYN or KYN/TRP were associated with symptoms of depression in recent metaanalyses (Fellendorf et al., 2022;Hunt et al., 2020;Marx et al., 2020). Inflammatory markers were also associated with elevated PHE/TYR, which may indicate a reduced dopamine availability, and proposed to contribute to depression in cancer (Hüfner et al., 2015) and trauma (Hüfner et al., 2019). The potential role of inflammation and TRP depletion in persistent somatic symptoms and mental disorders in COVID-19 convalescents was proposed in recent hypothesis papers (Bower et al., 2022;Eroglu et al., 2021). There is first clinical evidence for concomitantly elevated markers of inflammation, kynurenine pathway activity and lowered TRP availability in COVID-19 patients suffering from persistent symptoms (Bizjak et al., 2022), cognitive impairment (Cysique et al., 2022), or symptoms of anxiety, depression and mental stress (Kucukkarapinar et al., 2022;Matits et al., 2023). In the SIMMUN cohort, clinically relevant symptoms of depression (HADS) along with the inflammatory marker NEO were associated with decreased TRP. Additionally, mental stress levels (PSS-4) together with NEO were identified as predictors of higher KYN/TRP. In the INCOV cohort, inflammation was also found to affect levels of circulating serotonin, TRP, and kynurenine pathway metabolites KYN and QUIN. Hence, the additive effects of inflammation, infection, mental stress and age may disrupt the homeostasis of the kynurenine and catecholamine pathways and predispose to depressive or anxiety disorders in COVID-19 patients (Bower et al., 2022;Dantzer et al., 2008).
Most evidence on the relevance of the peripheral neurotransmitter levels in mental disorders is provided by observational studies, whereas in vivo experimental reports are scarce (Brown et al., 2021). In the periphery, the liver, mesenteric organs and blood vessels are the main sites of TRP, serotonin and dopamine metabolism (Badawy, 2017;Goldstein et al., 1999;Lukić et al., 2022;Meiser et al., 2013). Although kynurenine pathway metabolites were postulated to pass the blood-brain barrier (Brown et al., 2021;Schwarcz et al., 2012;Vancassel et al., 2018), mental disorders are not consistently paralleled by uniform and reproducible changes in dopamine, serotonin, and kynurenine pathway metabolites in the central nervous system (Brown et al., 2021;Clark et al., 2016;Miller et al., 2008). Clinical evidence for the serotonin theory of depression was also questioned in a recent systematic review (Moncrieff et al., 2022). Whether there is a causal association of inflammatory and infection-related factors with mental disorders frequently observed in COVID-19 patients (Al-Aly et al., 2021;Huang et al., 2021;Hüfner et al., 2022;Sahanic et al., 2023), needs further validation.

Limitations
Our study has limitations. First, incompatibility of study designs and variable sets of the SIMMUN and INCOV studies precluded direct comparison and validation of our findings in the classical training -test cohort analysis setting. In particular, we were not able to validate effects of early infection on the metabolic pathways of interest in the SIMMUN cohort. Conversely, we could not validate the effects of stress, depression and anxiety in the INCOV cohort due to unavailability of psychometric data. Second, relevance of the systemic kynurenine and catecholamine pathway activity for the central nervous system and mental health is controversial. We give a balanced discussion of this controversy to account for this fact. Third, the SIMMUN cohort had a limited sample size and a highly variable SARS-CoV-2 PCR test -sampling interval, which likely diminished effect sizes of the SARS-CoV-2 infection. Fourth, a selection bias occurred in the SIMMUN cohort due to enrichment in patients with mental disorders. In the INCOV cohort there was a selection bias towards hospitalized COVID-19 patients. Hence, none of the cohorts is representative for the general population. Fifth, both cohorts were recruited during initial phases of the pandemic and do not include vaccinated patients. Similarly, the analyzed cohorts were exposed to wild-type-like SARS-CoV-2 variants and do not allow to assess effects of infection with highly transmissible but less virulent omicron pathogens. For these reasons, validation studies with further cohorts are urgently needed.

Conclusions
SARS-CoV-2-dependent and -independent inflammation is associated with changes in the activity of the kynurenine and catecholamine pathway which may lead to effects on systemic availability of serotonin and dopamine. Those effects can be further amplified by advanced age, mental stress and depression. Further research is needed to explore the mechanistic interplay of SARS-CoV-2 infection, inflammation and mental health. It remains to be investigated, if and how this mechanism may contribute to mental disorders following SARS-CoV-2 infection.

Funding
The study was supported by the Science Fund of the Land Tirol (grant number GZ71134 to Katharina Hüfner).

Declaration of Competing Interest
Katharina Hüfner has received research grants from Austria Wirtschaftsservice (AWS) and the Science Fund of the Land Tirol as well as lecturer's honoraria from Forum Medizinische Fortbildung (FOMF), the Hospital of Schwaz and the Austrian Society for Biological Psychiatry related to the topic of COVID-19.
Piotr Tymoszuk owns the Data Science as a Service Tirol enterprise and works a free-lance data scientist and biostatistician. He has received an honorarium for statistical data analysis, bioinformatic and scientific writing services.
Bernhard Holzner has intellectual property rights to the CHES software tool used for questionnaire data collection.
Jens Lehmann reports consultancy for Evaluation Software Development GmbH, the software company developing CHES.
All other authors report no conflicts of interest related to the current manuscript.

Data Availability
Anonymized patient data will be made available upon request to the corresponding author. The INCOV cohort data are publicly available at https://data.mendeley.com/datasets/96v329bg7g/1. The analysis pipeline code is available at https://github.com/PiotrTymoszuk/ simmun.