Plasma concentrations of granulocyte colony-stimulating factor (G-CSF) in patients with substance use disorders and comorbid major depressive disorder

Granulocyte colony–stimulating factor (G-CSF) has raised much interest because of its role in cocaine addiction in preclinical models. We explored the plasma concentrations of G-CSF in patients diagnosed with substance use disorder (SUD) and highly comorbid psychiatric disorders. In particular, we investigated the association between G-CSF concentrations and comorbid major depressive disorder (MDD) in patients with cocaine and alcohol use disorders (CUD and AUD, respectively). Additionally, patients with MDD but not SUD were included in the study. Three hundred and eleven participants were enrolled in this exploratory study: 136 control subjects, 125 patients with SUD (SUD group) from outpatient treatment programs for cocaine (N = 60, cocaine subgroup) and alcohol (N = 65, alcohol subgroup), and 50 patients with MDD but not SUD (MDD group) from primary-care settings. Participants were assessed based on DSM-IV-TR criteria, and a blood sample was collected to examine the plasma concentrations of G-CSF. G-CSF concentrations were negatively correlated with age in the entire sample (r = − 0.233, p < 0.001) but not in the patients with MDD. G-CSF concentrations were lower in patients with SUD than in controls (p < 0.05), specifically in the cocaine subgroup (p < 0.05). Patients with SUD and comorbid MDD had lower G-CSF concentrations than patients with SUD but not comorbid MDD or controls (p < 0.05). In contrast, patients with MDD but not SUD showed no differences compared with their controls. The negative association between G-CSF concentrations and age in the sample was not observed in patients with MDD. G-CSF concentrations were decreased in patients with SUD and comorbid MDD but not in patients with MDD. Therefore, G-CSF may be useful to improve the stratification of patients with dual diagnosis seeking treatment. Further investigation is needed to explore the impact of sex and type of drug on the expression of G-CSF.

www.nature.com/scientificreports/ Control subjects (N = 136) were recruited from two different sources, a multidisciplinary staff cohort of volunteers working at the Spanish National Public Health System (i.e., Hospital Regional Universitario de Málaga, Málaga, Spain) and a second cohort obtained from volunteers donating data and plasma to the National Biobank of DNA (Banco Nacional de ADN Carlos III, Salamanca, Spain). Control subjects were divided into two groups based on the age, sex, and body mass index (BMI) of patients with SUD (N = 92, control group for SUD) and MDD (N = 44, control group for MDD). Notably, the control group for SUD was sex-balanced and showed no statistically significant differences in age and BMI with the SUD group, and the control group for MDD showed no statistically significant differences in sex, age and BMI with the MDD group.
Patients with SUD (N = 125, SUD group) were recruited from outpatient treatment programs for alcohol (N = 65, alcohol subgroup with AUD) and cocaine (N = 60, cocaine subgroup with CUD), and patients with MDD (N = 50, MDD group) were recruited from primary-care settings at the Spanish National Public Health System (Hospital Regional Universitario de Málaga, Málaga, Spain; Centro Provincial de Drogodependencias, Málaga, Spain; and Hospital Universitario 12 de Octubre, Madrid, Spain).
Participation in the study was voluntary, and the participants had to meet eligibility based on the following inclusion criteria: 18 years of age or older (up to 65 years) for all patients, diagnosis of AUD or CUD (a minimum of 4 weeks of abstinence) for the SUD group and diagnosis of MDD (last two months of depressive symptoms) for the MDD group. Lifetime SUD and MDD were diagnosed based on the DSM-IV-TR criteria. The exclusion criteria included personal history of chronic inflammatory diseases or cancer, infectious diseases, severe mental disorders precluding evaluation, pregnancy or breastfeeding and less than 4 weeks of abstinence from any drug, except for nicotine and caffeine. Psychiatric comorbidity was accepted for the SUD and MDD groups (i.e., additional SUD and mental disease, respectively). Regarding healthy control subjects, the exclusion criteria were identical to those for the groups of patients but also included the use of psychotropic medication during the last year and the diagnosis of psychiatric disorders based on the DSM-IV-TR criteria.
Ethics statements. Written informed consents were obtained from each participant after a complete description of the study. All the participants had the opportunity to discuss any questions or issues. The study and protocols for recruitment were approved by the Ethics Committee of the Hospital Regional Universitario de Malaga in accordance with the Ethical Principles for Medical Research Involving Human Subjects adopted in the Declaration of Helsinki by the World Medical Association (64th WMA General Assembly, Fortaleza, Brazil, October 2013) and Recommendation No. R (97) 5 of the Committee of Ministers to Member States on the Protection of Medical Data (1997), and Spanish data protection act [Regulation (EU) 2016/679 of the European Parliament and of the Council 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation). All data were given code numbers in order to maintain privacy and confidentiality.
Clinical assessments. All participants in the study were clinically evaluated by trained and experienced psychologists. The total sample was assessed using different psychiatric/clinical interviews based on the characteristics of each group, but the Spanish version of the Psychiatric Research Interview for Substance and Mental Disorders (PRISM) was typically used for all participants. The PRISM is a semistructured interview based on the DSM-IV-TR criteria with good psychometric properties in the evaluation of SUD and common comorbid psychiatric disorders in addicted patients 43,44 .
Therefore, the SUD group was specifically assessed with PRISM. The MDD group was assessed with the Beck Depression Inventory-II (BDI-II) to determine the severity of the MDD 45 but was also assessed with PRISM to dismiss the diagnosis of SUD. Moreover, the PRISM has demonstrated good to excellent validity and test-retest reliability for differentiating substance-induced MDD (i.e., MDD as a physiological consequence of the use of substances that may appear during active use, intoxication, or withdrawal) from primary MDD, with kappa ranging from 0.66 to 0.75 43,44,44,46 . Finally, the control group was assessed with the Spanish version of the Composite International Diagnostic Interview (CIDI) for the detection of psychiatric disorders 47 and PRISM (module 1: Overview for sociodemographic and physiological variables).
Collection of plasma samples. Blood extractions were conducted under the same conditions by experienced nurses in the morning after fasting for 8-12 h. Venous blood samples were extracted into 10-mL K 2 EDTA tubes (BD, Franklin Lakes, NJ, USA), and to obtain plasma, samples were centrifuged at 2200×g for 15 min (4 °C). Plasma was individually assayed by three rapid tests for detecting infectious diseases: HIV (Retroscreen HIV, QualPro Diagnostics-Tulip Group Ltd, Goa, India), hepatitis B (HBsAg Test, Toyo Diagnostics-Turklab Inc., Izmir, Turkey) and hepatitis C (Flaviscreen HCV, QualPro Diagnostics-Tulip Group Ltd). Infected samples were discarded following laboratory safety protocols. Each plasma sample was registered and stored at − 80 °C until determination of G-CSF.
Determinations of G-CSF. Plasma concentrations of G-CSF were determined using a selective enzymelinked immunosorbent assay (ELISA) according to the manufacturer's instructions: Human G-CSF (Granulocyte Colony Stimulating Factor 3) ELISA kit (#EH0149, Wuhan Fine Biotech Co., Ltd. Wuhan, China). The Human G-CSF ELISA kit indicated a sensitivity < 23.4 pg/mL and a range of 39.1-2500.0 pg/mL. The experimental detection range under the experimental conditions with eight ELISA determinations, using 96-well plates and samples appropriately diluted, was established between 7.2 and 2850.6 pg/mL. To perform the ELISA protocol, we used 100 μL of the samples in 1:1 and 1:3 dilutions into each cell, the plates were incubated 90 min at 37 °C. Subsequently, it was washed twice with the washing buffer, 100 μL of a solution of anti-G-CSF antibodies linked with a biotin molecule were added and it was subjected to a new incubation period of 60 min at . In all cases, the samples were run in duplicate and internal controls and calibration curve were included in each ELISA Kit. The concentration assigned to samples with an optical density (OD) lower than the limit of detection in the ELISA but higher than the background (zero values) (N = 26) was half of the minimum concentration that could be interpolated in the standard curve. Plasma concentrations of G-CSF were expressed as pg/mL. Tables 1 and 3  or mean ± standard deviation (mean ± SD). The significance of differences in categorical and normal continuous variables was determined using the chi-square test (or Fisher's exact test) and Student's t-test, respectively. Analyses of covariance (ANCOVA) were performed to evaluate the main effects and interaction of independent variables [e.g., "lifetime SUD diagnosis" (control and SUD; control, cocaine and alcohol), "lifetime MDD diagnosis" (control and MDD) and "sex" (men and women)] on plasma concentrations of G-CSF while controlling for "age" as a covariate. Because plasma concentrations of G-CSF showed a positively skewed distribution and did not pass the D' Agostino-Pearson normality test, raw data were log 10 -transformed to approximate a normal distribution and to ensure statistical assumptions of the ANCOVA. The estimated marginal means and 95% confidence intervals (95% CI) of the log 10 -transformed G-CSF concentrations are represented in the figures, and their back-transformation a into linear scale is included in the description of the results. Post hoc comparisons for multiple comparisons were performed using the Sidak 's correction test.

Statistical analysis. Data in
Correlation analyses between plasma concentrations of G-CSF (log 10 -transformed data) and relevant SUDrelated variables (i.e., DMS-IV-TR criteria for SUD, duration of abstinence and duration of problematic substance use) were performed using the correlation coefficients of Pearson (r) and Spearman (rho) with continuous and categorical variables, respectively.
The statistical analyses were carried out with GraphPad Prism version 5.04 (GraphPad Software, San Diego, CA, USA) and IBM SPSS Statistics version 22 (IBM, Armonk, NY, USA). A p-value < 0.05 was considered statistically significant.

Results
Sociodemographic description of the sample. This study included 311 participants grouped into four experimental groups to explore plasma concentrations of G-CSF: the SUD group, the MDD group and their control groups. Table 1 shows the sociodemographic data for the total sample.
Although the control group for patients with SUD showed comparable age and BMI to the SUD group (mean age of 42.7 years and BMI of 25.2 kg/m 2 ), there were differences in the sex composition because the control group was originally sex-balanced in men and women and the SUD group was mainly composed of men (67%). The SUD group was divided into the cocaine subgroup and the alcohol subgroup. In this case, the comparison between both subgroups revealed significant differences in age and education level. Thus, the patients with CUD were younger than the patients with AUD, and the percentage of participants with elementary education was greater in the alcohol subgroup. www.nature.com/scientificreports/ A second control group showed sex, age and BMI comparable to those of the MDD group, and no differences were observed across sociodemographic variables. Unlike the SUD group, the MDD group was mostly composed of women (80%).
Plasma concentrations of G-CSF and age. An initial analysis was performed to explore the correlations between plasma concentrations of G-CSF and variables that could be included in the ANCOVA models as covariates or confounding variables (age and BMI). While log 10 -transformed G-CSF concentrations and BMI were not correlated, log 10 -transformed G-CSF concentrations and age were significantly correlated (r = − 0.233, p < 0.001) in the total sample. Interestingly, there were differences in this association based on the group [control group: r = − 0.307, p = 0.002; SUD group: r = − 0.165, p = 0.047; and MDD group: r = + 0.070, p = 0.707].
Plasma concentrations of G-CSF and lifetime SUD diagnosis. The association between lifetime SUD diagnosis and plasma concentrations of G-CSF was first investigated using two-way ANCOVA: raw data for G-CSF concentrations were log 10 -transformed for the analysis with "SUD diagnosis" [SUD (N = 125) and control (N = 92)] and "sex" as factors and "age" as an important covariate. ANCOVA revealed a significant main effect of "SUD diagnosis" on plasma concentrations of G-CSF [F (1, 212) = 5.915; p = 0.016] (Fig. 1A), and patients with SUD showed significantly lower G-CSF concentrations than control subjects [back-transformation of estimated marginal means: 1414.3 (95% CI 1095.3-1733.2) and 1999.6 (95% CI 1650.2-2349.0) pg/mL, respectively]. In contrast, we found no main effect of "sex" or interaction between the two factors on plasma concentrations of G-CSF.
To elucidate the contribution of each SUD subgroup, we performed a second two-way ANCOVA on log 10 -transformed G-CSF concentrations with "type of SUD diagnosis" [cocaine (N = 60), alcohol (N = 65) and control (N = 92)] as a factor. There was a significant main effect of "type of SUD diagnosis" on plasma concentrations of G-CSF [F (2,210)

Plasma concentrations of G-CSF and relevant SUD-related variables.
Because there was a significant association between lifetime SUD diagnosis and plasma concentrations of G-CSF, we explored relevant variables related to cocaine and alcohol addiction: the severity of SUD (based on DSM-IV-TR criteria for AUD and CUD), duration of the last abstinence and problematic substance use. To this end, the SUD group was divided into alcohol and cocaine subgroups, and correlation analyses were performed between plasma concentrations of G-CSF (log 10 -transformed) and these variables ( Table 2). The cocaine subgroup had an average of 7.6 DSM-IV-TR criteria for CUD and 55.2 days of abstinence: We found no significant correlations between CUD-related variables and G-CSF concentrations. Regarding the alcohol subgroup, the patients had an average Psychiatric comorbidity and medication in patients with SUD. Lifetime SUD is often associated with psychiatric comorbidity (dual diagnosis). We assessed all patients with SUD, and the high prevalence of psychiatric comorbidity was confirmed as a primary characteristic of the sample: 64 patients with SUD (51.6%), including 19 patients with CUD (31.7%) and 45 patients with AUD (69.2%), were diagnosed with other mental disorders. The most prevalent mental disorders in the SUD group were as follows: 31.7%, mood disorders; 27%, anxiety disorders; 15%, cluster B personality disorders; and 4.8%, psychotic disorders. No significant differences were found in the sex proportions among these comorbid mental disorders. Table 3 shows main mental disorders diagnosed based on DSM-IV-TR criteria in the cocaine and alcohol subgroups.
As expected, the elevated prevalence of psychiatric comorbidities was associated with high use of psychoactive medications. Notably, 66.7% of patients with SUD were using medication during the last 12 months: antidepressants (40.7%), anxiolytics (40.7%), anticonvulsants (22.8%), antipsychotics (8.9%) and disulfiram (39%). To investigate the potential effects of psychiatric medication on G-CSF concentrations, we performed two-way ANCOVA with "psychiatric medication" (nonmedicated and medicated patients with SUD) and "sex" as factors and "age" as a covariate. However, the analysis revealed no main effects of "psychiatric medication" on log 10 -transformed G-CSF concentrations (Table 4). In addition, no significant main effect of "sex" on G-CSF concentrations or interaction between factors was observed.

Plasma concentrations of G-CSF and comorbid MDD in patients with SUD.
Because we observed no significant effect of "psychiatric medication", we analyzed G-CSF concentrations in the SUD group based on the most prevalent comorbid mental disorders using two-way ANCOVA with "sex" as a factor while controlling for "age" but not for "psychiatric medication". The analyses revealed a significant association between Table 2. Correlation analyses between plasma concentrations of G-CSF and relevant SUD-related variables in the SUD group. a Log 10 -transformed data. *Denotes significant difference (p < 0.05).  www.nature.com/scientificreports/ log 10 -transformed G-CSF concentrations and "comorbid MDD diagnosis" but not with other comorbid mental disorders (data not shown). Therefore, there was a significant main effect of "comorbid MDD diagnosis" on plasma concentrations of G-CSF [F (2, 210) = 5.270; p = 0.006] (Fig. 2A). The post hoc test revealed that patients with SUD and comorbid MDD had significantly lower G-CSF concentrations than control subjects (p < 0.01) [back-transformation of estimated marginal means: 934.8 (95% CI 388.7-1480.8) pg/mL and 2007.5 (95% CI 1661.5-2353.5) pg/mL, respectively]. Regarding the "sex" factor, there was no main effect or interaction on plasma concentrations of G-CSF.  Because comorbid MDD can be classified according to DSM-IV-TR criteria into primary and substance-induced MDD, we explored whether the low plasma concentrations of G-CSF in patients with SUD and comorbid MDD were associated with a specific type of MDD. However, the statistical analysis revealed no main effect of "type of comorbid MDD" on log 10 -transformed G-CSF concentrations (Table 5) and, therefore, no differences among patients with SUD and primary MDD and/ or substance-induced MDD.
Plasma concentrations of G-CSF and comorbid MDD in the cocaine and alcohol subgroups. Then, we investigated the association between G-CSF concentrations and comorbid MDD diagnosis in the SUD group while separately considering cocaine and alcohol subgroups. For each subgroup, two-way ANCOVA was used with "comorbid MDD diagnosis" and "sex" as factors while controlling for "age". The analysis in the cocaine subgroup showed a significant main effect of "comorbid MDD diagnosis" on plasma concentrations of G-CSF [F (2, 145) = 3.564; p = 0.031] (Fig. 2B). However, the post hoc test revealed no significant differences between patients with CUD. In addition, there was no main effect of "sex" or interaction between factors on plasma concentrations of G-CSF.

Plasma concentrations of G-CSF and MDD diagnosis.
Given that there was a significant association between G-CSF concentrations and comorbid MDD diagnosis in the SUD group, we examined plasma concentrations of G-CSF in patients from primary-care settings who were diagnosed with primary MDD but no lifetime SUD. Similar to the previous analysis, two-way ANCOVA was performed with "MDD diagnosis" and "sex" as factors while controlling for "age". However, there were no main effects of "MDD diagnosis", "sex" or interaction on log 10 -transformed G-CSF concentrations (Fig. 3A).
Finally, because many patients with MDD were using psychiatric medications during the study as part of their treatment [primarily antidepressants (36%) and anxiolytics (40%)] and recent research in our group has reported Table 5. Plasma concentrations of G-CSF and type of comorbid MDD in the SUD group. a Data were analyzed using ANCOVA (type of MDD, sex, age).  www.nature.com/scientificreports/ high plasma concentrations of antiinflammatory molecules (e.g., N-acylethanolamines) in patients with MDD and antidepressant treatment 48 , we investigated the effects of antidepressants and anxiolytics on G-CSF concentrations. However, the analyses revealed no significant main effects of "antidepressants" on log 10 -transformed G-CSF concentrations (Fig. 3B). There was also no main effect of "anxiolytics".

Discussion
In the present exploratory study, we examined G-CSF concentrations in the plasma of abstinent patients with lifetime SUD from outpatient treatment programs, patients with MDD from primary-care settings and control healthy subjects. In addition, all participants were characterized through different psychiatric/clinical assessments based on the DSM-IV-TR criteria. The main findings are summarized as follows: (1) Plasma concentrations of G-CSF were significantly and inversely correlated with age in the control group and the SUD group, but the correlation was nonsignificant in the MDD group; (2) G-CSF concentrations were significantly lower in the patients with SUD than in the control subjects, specifically in the cocaine subgroup; (3) The patients with SUD and comorbid MDD had significantly lower G-CSF concentrations than the patients with SUD but not comorbid MDD or control subjects; (4) G-CSF concentrations were significantly lower in the cocaine subgroup with comorbid MDD but not in the alcohol subgroup with comorbid MDD; and (5) The patients with MDD showed no differences in plasma concentrations of G-CSF compared with their control subjects. The results suggest a potential role of this immunomodulatory factor in MDD disorders associated with SUDs rather than in primary MDD disorders. Consistent with our findings, it is known that G-CSF is close and negatively correlated with age as a typical growth factor. Previous studies have reported G-CSF as a risk factor for cognitive impairment in preclinical models and as one of the 18 plasma signaling proteins that are collectively predictive of conversion from mild cognitive impairment to Alzheimer's disease 49,50 . Unlike the control group, our data showed fluctuations in the inverse association between G-CSF concentrations and age across the different groups of patients. Thus, the inverse association was attenuated in the patients with SUD and nonsignificant in the patients with MDD. This suggests that a history of MDD could be associated with disturbances in the immune system through immunomodulatory growth factors.
However, our results confirm that changes in the plasma concentrations of G-CSF might be attributed to the history of addiction in patients with AUD and CUD, although the nature of this relationship demands further research. While in preclinical studies this immunomodulatory growth factor helps to consolidate cocaine reward/ cocaine seeking behavior, in abstinent humans we observed that G-CSF concentrations are lowered in patients with CUD and with a nonsignificant trend toward a negative correlation with the severity of CUD. Because there is a lack of data on the acute effects of cocaine on circulating concentrations of G-CSF, we can only speculate on the nature of this finding. In accordance with the allostatic model set in place for most biological modulatory transmitters related to addiction 51 , whether G-CSF treatment in preclinical models of cocaine reward is associated with boosting or sustaining cocaine-seeking behavior 40 , it is reasonable to consider that G-CSF might be downregulated during a period of abstinence after chronic cocaine use in severe cocaine-addicted patients. However, these results regarding cocaine need to be conclusively determined. Regarding alcohol, G-CSF has been linked to alcohol-related liver disease, and a study is currently validating the efficacy of liver regeneration through G-CSF therapy 52 . Unfortunately, we have no data regarding accessory digestive organ diseases in the present sample to test this association. However, previous studies in patients with AUD from the same source as our group reported a high prevalence of alcohol-related liver and pancreas diseases 4,53 . In contrast, our study revealed a positive correlation between plasma concentrations of G-CSF and the duration of the last abstinence from alcohol. Similar to G-CSF, other inflammation-related signals in the plasma [such as the lipid messenger N-oleoilethanolamine (OEA)] were found to be related to the duration of alcohol abstinence. Thus, OEA decreases as the duration of alcohol abstinence increases in patients with AUD, and it has been proposed as a marker of abstinence 54 .
Our results also showed that plasma concentrations of G-CSF were lower in the patients with SUD and comorbid MDD but not in the patients with primary MDD without SUD. Long-lasting stress effects have been studied in the development of mood disorders 55 , with several forms of dysregulation observed in relation to proinflammatory status 56 . Recent studies from our group have described differences in other peripheral inflammatory mediators relative to the presence of comorbid psychiatric disorders in patients with AUD. In particular, decreased concentrations of the chemokine eotaxin-1 (CCL11) in the plasma of abstinent patients with AUD have been associated with comorbid mood and anxiety disorders 14 . Another study reported lower concentrations of IL-1β, CXCL12 and CCL11 in cocaine-induced MDD 57 . Although we observed lower concentrations of G-CSF in patients with CUD and comorbid MDD, these changes in the concentrations were not linked to the type of MDD (primary MDD, substance-induced MDD and both). How the interaction of SUD and MDD affects G-CSF must be elucidated with further research.
Furthermore, this immunomodulatory growth factor has also been studied in relation to memory functions in a preclinical model, showing that deficiencies in G-CSF concentrations in the hippocampus decreased spatial learning performance and memory formation 58 . Therefore, we consider that it would be interesting to measure the role of G-CSF in a neuropsychological cohort of patients with SUD and substance-induced memory deficits.

Conclusions and limitations
These findings support the importance of monitoring G-CSF in the context of SUD and psychiatric comorbidity, but we are aware of the limitations of this exploratory study: (1) The high prevalence of psychiatric comorbidity in patients with SUD increased the difficulty in the characterization of a specific comorbid disorder such as MDD; (2) There are additional uncontrolled biological and sociodemographic variables that were not included in the present study but could be a source of variability in G-CSF concentrations; (3)  www.nature.com/scientificreports/ based on statistical limitations such as compliance of parametric assumptions and/or appropriate transformations of raw data (e.g., using log 10 -transformation), calculation of sample size, number of independent variables/ covariates in the linear models, etc. Finally, it is important to replicate these findings with larger samples in cohorts from different geographical and cultural backgrounds, as well as to characterize G-CSF concentrations in longitudinal studies during critical stages in the course of MDD and SUD, such as substance abstinence and active substance use.
In conclusion, these findings support an association between lifetime SUD and plasma concentrations of G-CSF, with lowered concentrations in abstinent patients with CUD. Interestingly, we found lower plasma concentrations of G-CSF in association with comorbid MDD diagnosis in these cocaine-addicted patients. Additionally, plasma concentrations of G-CSF correlated with alcohol-related variables such as the duration of abstinence, an interesting variable related to a good prognosis for those with SUD. Moreover, a dysregulation in G-CSF concentrations was observed with comorbid MDD in patients with lifetime SUD but not with primary MDD without SUD. This association with comorbid MDD was independent of the type of MDD (comorbid substance-induced MDD or comorbid primary MDD). Further research is necessary to elucidate the role of G-CSF as a potential biological marker of immunomodulatory and inflammatory states in addiction, mood disorders and dual diagnosis.