Serotonin is the main tryptophan metabolite associated with psychiatric comorbidity in abstinent cocaine-addicted patients

The lack of effective treatments and a high rate of relapse in cocaine addiction constitute a major health problem. The present study was conducted to examine the expression of tryptophan-derived metabolites in the context of cocaine addiction and psychiatric comorbidity, which is common in addicted subjects. Abstinent patients with cocaine use disorder (CUD) and control subjects were recruited for a cross-sectional study. Participants were assessed with a semi-structured diagnostic interview (PRISM) based on DSM-IV-TR for substance and mental disorders. Plasma concentrations of tryptophan metabolites and their association with relevant CUD-related variables and psychiatric comorbidity were explored. We observed decreased plasma kynurenic acid concentrations in the cocaine group, however no associations between CUD-related variables and tryptophan-derived metabolites were found. In contrast, 5-HT concentrations were increased in CUD-patients and the diagnosis of different psychiatric disorders in the cocaine group was related to higher plasma 5-HT concentrations compared with non-comorbid patients. Therefore, while changes in plasma kynurenic acid concentrations appear to be directly associated with lifetime CUD, changes in 5-HT concentrations are associated with psychiatric comorbidity. These results emphasize the need to find potential biomarkers for a better stratification of cocaine-addicted patients in order to develop therapeutic approaches to prevent cocaine relapse.

Plasma concentrations of TRP-derived metabolites. The effect of lifetime CUD on the plasma concentrations of metabolites related to TRP metabolism (concentrations and ratios) in the sample was investigated using ANCOVA with 'diagnosis of CUD' as main factor and controlling for 'sex' , 'age' and 'BMI' . A base-10 logarithmic transformation of estimated marginal means was conducted for KA, 5-HT and ratios.
Characteristics related to pattern of cocaine use and lack of effects on TRP metabolism in abstinent CUD-patients. The cocaine group was characterized using variables related to pattern of use as shown in Table 1. Cocaine users were recruited in abstinence from cocaine for 222 days [mode: 30 days (range: 2,553)], with problematic cocaine use for 7 years [mode: 1 year (range: 32 years)] and a number of DSM-IV criteria for CUDs of 8-9 (severe CUD). In addition to cocaine, there was a high prevalence of patients diagnosed with other substance use disorders (68%).
We evaluated the relative effects of abstinence, problematic cocaine use and DSM-IV criteria for CUDs on plasma TRP-derived metabolites in the cocaine group using ANCOVA and controlling for 'sex' , 'age' and 'BMI' . However, we found no effects or associations between these descriptive variables and the expression of these metabolites (data not shown).  www.nature.com/scientificreports www.nature.com/scientificreports/ psychiatric comorbidity and effects of mental disorders on plasma concentrations of TRP-derived metabolites in abstinent CUD-patients. The cocaine group exhibited a high prevalence of psychiatric comorbidity (about 82%), considering comorbid mental disorders and comorbid substance use disorders, such that common mental disorders were diagnosed in 59% of CUD-patients. We divided patients into comorbid and non-comorbid patients according to lifetime diagnosis of mental disorders for further characterization (Table 2). We found that abstinent CUD-patients diagnosed with comorbid mental disorders had higher psychotropic medication use (p < 0.001) and higher prevalence of other substance use disorders (p < 0.05) than in patients with no comorbid mental disorders.
Because TRP and 5-HT have been linked to psychiatric disorders, we studied the effect of common 'comorbid mental disorders' on TRP and metabolites in the plasma of the cocaine group. These analyses were controlled for 'sex' , 'age' and 'BMI' but also for 'psychiatric medication' (during the last year) as confounding variables. A base-10 logarithmic transformation of estimated marginal means was conducted for QA and ratios.
With regard to the ratios, we found no main effect of 'comorbid mental disorders' (Table S3).
Comorbid mental disorders and plasma 5-HT concentrations in abstinent CUD-patients. As a continuation of the assessment of the effects of comorbid mental disorders on plasmatic 5-HT concentrations, we evaluated the contribution of the most prevalent mental disorders in the cocaine group to the observed effect:  www.nature.com/scientificreports www.nature.com/scientificreports/ mood disorders (N = 28), anxiety (N = 22), psychotic disorders (N = 16) and personality disorders (i.e., borderline and antisocial personality disorders) (N = 38).
We analyzed separately the effects of mental disorders on 5-HT concentrations using ANCOVA and including as primary factor the diagnosis of each comorbid mental disorder with 3 categories/subgroups [non-comorbid subgroup or lack of any mental disorder (N = 41), non-mental disorder subgroup or lack of the particular mental disorder and comorbid mental disorder subgroup or diagnosis of the particular mental disorder]. Again, 'sex' , 'age' , 'BMI' and 'psychiatric medication' were controlled as confounding variables. Estimated marginal means of 5-HT concentrations for each subgroup were represented in Fig. 3.
In summary, we observed higher 5-HT concentrations in abstinent CUD-patients diagnosed with lifetime mental disorders, particularly with mood, psychotic and personality disorders. In fact, the highest concentrations

Discussion
The identification of biomarkers in cocaine addiction is a clinical need that might help to stratify patients on the basis of their clinical phenotype and response to treatment. In this study we evaluated TRP metabolism in the plasma of abstinent cocaine users diagnosed with CUD to identify metabolites linked to the cocaine addict phenotype. We report a decrease in the plasma concentration of KA, a neuroprotective metabolite of the KYN pathway, which is independent of comorbid mental disorders and therefore attributable to CUD. We observed no significant differences between groups in TRP or KYN concentrations, and consequently there were no differences in KYN/TRP ratios indicating that TDO/IDO activity was unchanged in cocaine users. Finally, we also report an increase in plasma 5-HT concentration in abstinent CUD-subjects, which appears to be dependent on the presence of comorbid mood and personality disorders.
To date, studies examining differences in TRP metabolites between cocaine-dependent and control subjects are limited. A previous metabolomics study reported differences in various metabolites of the TRP, tyrosine and purine pathways in the plasma of cocaine-dependent subjects with at least two weeks of abstinence 15 . In contrast to our results, they reported no differences in 5-HT concentration although they did observe an increase in N-methyl-serotonin, a minor metabolite of 5-HT. The authors excluded patients diagnosed with mental disorders and since the changes we observe in 5-HT in the cocaine group depend on the presence of comorbid mental disorders, this could explain the difference in results. It is also very important to emphasize that we analyzed plasma 5-HT after a protracted period of abstinence.
In the periphery, 5-HT is synthesized by means of the rate-limiting enzyme tryptophan hydroxylase and secreted into blood by enterochromaffin cells of the intestine 16 . Since 5-HT is a charged molecule that cannot cross the blood-brain barrier, its concentration is regulated independently in each compartment. Given the role of SERT in regulating 5-HT concentration and the close relationship that exists between platelet SERT and plasma 5-HT it is reasonable to propose that platelet SERTs are involved in the elevation of plasma 5-HT in abstinent www.nature.com/scientificreports www.nature.com/scientificreports/ patients. Specifically, plasma membrane SERT, as well as 5-HT uptake, initially rise as platelets are exposed to increasing 5-HT concentration, but this initial response is followed by a second phase, in which higher concentrations of 5-HT decrease SERT levels 17 . In fact, it has been reported that a reduction in platelet SERT density in cocaine-dependent individuals could be the consequence of maintained high concentrations of 5-HT [18][19][20] .
The expression of SERT is controlled by a highly-conserved single gene (in humans SLC6A4 gene) and SERT expressed in platelets and 5-HT neurons appear to be identical 21 . Thus, the results obtained in the periphery may reflect changes that occur in the brain although any extrapolation must be made with due caution.
In this sense, genetic factors should be taken into account since the short variant of SERT polymorphism termed the 5-HT transporter-linked polymorphic region (h5-HTTLPR) reduces the transcriptional efficiency of the gene, resulting in decreased SERT expression 22,23 . This polymorphism has been linked to some psychiatric disorders such as major depressive disorder [24][25][26] although the findings are controversial. Whether a higher incidence of the short allele of the transporter gene was present in the CUD-patients with comorbid psychiatric disorder of our study needs to be explored.
In addition, regulation of SERT transcription in intestinal epithelium differs from that for neuronal SERT 27 , so we cannot rule out a possible role of this factor in the effect observed however, their involvement in regulating plasma 5-HT concentration is difficult to establish 28 .
Finally, because plasma 5-HT results from equilibrium among synthesis, platelet uptake and storage, and metabolism, we cannot rule out that 5-HT concentrations might be increased not only by changes in uptake, as suggested above, but also by changes in synthesis or metabolism. These possibilities should be explored in future studies. Interestingly, in abstinent CUD-patients we found a negative correlation between 5-HT concentration and age which appears to be independent of the presence or not of comorbid mental disorders, indicating that younger CUD-subjects showed greater increases in 5-HT concentration than older subjects. It is tempting to speculate that these greater increases in 5-HT in younger patients might be related to a shorter duration of cocaine abuse although other factors such as recent intensity of cocaine consumption might explain this observation. In any case, considering that both platelet 5-HT uptake and plasma 5-HT concentrations are clearly influenced by age, sex and season 29,30 , this hypothesis needs to be conclusively demonstrated.
With regard to the changes observed in the KYN pathway, no association with comorbid disorders was found for any of the metabolites analysed. Previous studies had pointed to alterations of kynurenine pathway linked to depression, as well as to schizophrenic or bipolar patients. However, the results from human studies are not conclusive and sometimes contradictory since the findings are dependent on the stage of the illness, age, brain anatomic area involved and sometimes even contradictory between peripheral and central nervous system 31,32 . Furthermore, it must be considered that cocaine addiction in the patients of our study may be contributing to the lack of changes associated with comorbidity and, as mentioned below, one of the limitations of this study is that data from patients with mental disorders without cocaine addiction are missing.
It is interesting that the decrease in serum concentration of KA in CUD-patients occurs in the absence of changes in KYN concentration pointing to the possible involvement of a downstream imbalance in the KYN metabolic pathway. KA is formed from KYN by the action of kynurenine amino-transferases (KATs) and thus the ratio KA/KYN can be considered an index of the activity of this group of enzymes. Our results show a decrease in this KA/KYN ratio in abstinent CUD-patients compared with control subjects. These findings suggest that either KAT enzyme levels or activity is modified in abstinent cocaine-addicted subjects and could be a consequence of endogenous KAT inhibitors 33 or modifications of its properties by genetic mutations.
The reduced formation of KA may contribute to an imbalance in the neuroprotective and neurodegenerative pathways 10 . Previous studies have demonstrated that the concentration of 3HK (formed from KYN through the action of kynurenine 3-monooxygenase) is unaltered in CUD-patients 15 and because we found no change in the concentration of QA, the decrease in the neuroprotective pathway might permit greater activity from this neurotoxic metabolite. However, it is important to highlight that we have no evidence of how the plasma concentrations reflect CNS concentrations although a significant correlation between peripheral and CSF levels of QA and KA have been suggested 34 .
This study indicates that cocaine addiction and its medical consequences might be associated with dysregulation in the KYN pathway. A specific enhancement of KAT activity or KA concentrations in order to return to homeostatic levels could be beneficial in cocaine abstinence and, in particular, in lowering drug craving and seeking. In this sense, recent evidence indicates that increasing KA may be an effective strategy for inhibiting the rewarding effects of THC and alcohol-and cocaine-seeking behavior and relapse in animal models 11,13 . Limitations. These findings support the importance of monitoring TRP metabolites in the context of CUD and psychiatric comorbidity, but we are aware of the limitations of this cross-sectional study: (i) The sources for recruitment were different in the control group and the cocaine group, therefore several socio-demographic variables can influence the association between biochemical data and diagnosis of CUD (e.g., 'education' and 'occupation'); (ii) The recruitment of CUD-patients was conducted from outpatient programs and there are clinical variables that remain unknown and can affect the validity of results; (iii) Larger samples of abstinent CUD-patients and additional study groups should be included (e.g., subjects with mental disorders and no substance use); (iv) Longitudinal studies are needed to monitor changes in these metabolites during abstinence at different times in the same patients; and (v) Pharmacological and behavioral studies in animal models can help to elucidate the association between TRP metabolites and cocaine exposure.
In summary, our findings indicate the existence of changes in TRP metabolism in subjects with lifetime CUD. More specifically, KYN pathway dysregulation is a consequence of cocaine consumption while plasma 5-HT concentrations are related to comorbid psychiatric disorders in cocaine addiction. These results emphasize the need to find biomarkers that allow a better stratification of cocaine-addicted patients in order to develop therapeutic approaches to prevent cocaine relapse following periods of abstinence.

Materials and Methods
Ethics approval and consent to participate. Written informed consent was obtained from each participant after they had received a complete description of the present study and had been given the chance to discuss any questions or issues. The study and protocols for recruitment, privacy and confidentiality were approved by the Ethics Committee for Research of our institution (Comité de Ética de la Investigación Participants and recruitment. The present cross-sectional study was conducted in accordance with the STROBE (strengthening the reporting of observational studies in epidemiology) statement 35 . The procedure for recruiting participants study was a non-random convenience sampling (June 2014 to June 2016) according to the participation criteria.
One hundred white Caucasian participants were selected in the cocaine group from outpatient treatment programs for cocaine addiction at centers for drug dependence treatment in the province of Malaga (Centro Provincial de Drogodependencias and Centros Ambulatorios de Tratamiento) (Malaga, Spain) after confirming the eligibility criteria. Sixty healthy individuals from a multidisciplinary staff working at the Hospital Regional Universitario de Málaga (Malaga, Spain) were included in the control group.
Eligibility criteria. The participation was voluntary but all participants had to meet eligibility criteria. To be eligible, subjects had to be between 18 and 65 years of age. Exclusion criteria were as follows: (1) Personal history of chronic diseases and/or cancer; (2) Infectious diseases; (3) Incapacitating cognitive alterations; and (4) Pregnancy for women.
Cocaine group: Cocaine users were required to have been diagnosed with lifetime CUD (i.e., cocaine abuse or dependence) exclusively by intranasal use and also to be abstinent from abused drugs (except for nicotine and caffeine) for at least 2 weeks before testing. Lifetime CUD was determined through psychiatric interviews, while the abstinence from abused drugs (cocaine, amphetamine, opiates, barbiturates, phencyclidine and cannabis) was checked weekly by urine analysis using a V-Twin Drug Testing System (Siemens AG, Erlangen, Germany) in the outpatient settings. Subsequently, rapid plasma analyses were conducted to verify abstinence from cocaine and alcohol.
Control group: The control group had no history of mental disorders or substance use disorders (lifetime abuse or dependence) and was sex-balanced and matched for age and body mass index with the cocaine group. No psychotropic medication had been used during the last year.
Exclusions. Thus, 110 cocaine users and 65 controls were initially recruited for this observational study based on the eligibility criteria. However, 15 participants (10 cocaine users and 5 controls) were excluded after examining for eligibility in the interviews or rapid plasma analyses.
Clinical evaluations. Prior to the clinical assessments, the 'Trail Making Test' (TMT) Part B was administered to all participants as a memory and attention-screening test to detect cognitive alterations 36 .
All abstinent CUD-subjects were evaluated according to the 'Diagnostic and Statistical Manual of Mental Disorders-4th Edition-Text Revision' (DSM-IV-TR) criteria, using the Spanish version of the 'Psychiatric Research Interview for Substance and Mental Diseases' (PRISM) 37,38 . Lifetime prevalence was used to present the frequency of substance use disorders, non-substance use disorders or common mental disorders. DSM-IV-TR criteria for substance dependence and abuse were used to diagnose substance use disorders and determine the severity of CUD 14,39 .
Control subjects were initially evaluated by PRISM (for substance screening and abuse and dependence) and the Spanish version of 'Dual Diagnosis Screening Instrument' (DDSI) to detect psychiatric disorders 40 . All the interviews were performed by experienced psychologists who had received training for TMT, PRISM and DDSI. Laboratory methods. Collection and rapid analyses of plasma samples. Blood samples were collected (09:00-11:00 h AM) after fasting for 8-12 h (prior to the psychiatric interviews). Venous blood was extracted into 10 mL K2 EDTA tubes (BD, Franklin Lakes, NJ, USA) and was immediately processed to obtain plasma. Blood samples were centrifuged at 2,200 × g for 15 min (4 °C) and individually assayed to detect infectious diseases by 3 rapid tests for HIV, hepatitis B and hepatitis C purchased from ALL.DIAG (Strasbourg Cedex, France). Samples displaying infection were discarded following safety protocols. Plasma analyses for cocaine metabolite (Benzoylecgonine Specific Direct ELISA Kit purchased from Immunalysis, Pomona, CA, USA) were also performed to confirm cocaine abstinence. Additionally, the blood alcohol concentration was measured according to the alcohol oxidase reaction using an Analox AM1 analyzer (Analox Instruments, Stourbridge, UK).
QA was determined using a commercially available ELISA immunoassay (Cloud-Clone Corp., Houston, USA) according to instructions of the manufacturer.
The ratios of KYN or 5-HT to TRP and KA or QA to KYN were used as a measure of TRP degradation and of KA and QA formation, respectively. Statistical analysis. All data for tables are expressed as number and percentage of subjects [N (%)] or mean and standard deviation [mean (SD)]. The significance of differences in categorical and normal continuous variables was determined using Fisher's exact test (Chi-square test) and Student's t-test, respectively.
Statistical analysis of concentrations and ratios was performed using analysis of covariance (ANCOVA) as a general linear model. The ANCOVA models included CUD-related variables as categorical and continuous independent variables to explore their association with TRP, KYN, KA, QA and 5-HT concentrations. Because the size of the sample restricts the number of independent variables in the ANCOVA, only biological variables ('sex' , 'age' , 'BMI') and 'psychiatric medication' were used as covariates. Socio-demographic characteristics were not used as covariates in the study because they are multilevel categorical variables whose interpretation is complex, they do not meet parametric assumptions and, therefore, decrease the statistical efficiency. To ensure statistical assumptions of a general linear model, log(10)-transformation was used for positive skewed distributions. Estimated marginal means [95% confidence intervals (95% CI)] were expressed and represented in the figures, using back-transformation if there was logarithmic transformation. The post hoc comparisons were performed using Sidak's correction test.
Additionally, correlation analyses were performed using the Pearson's coefficient (r).