Inflammatory Biomarkers, Microbiome, Depression, and Executive Dysfunction in Alcohol Users

Alcohol-related disorders (ARD) are highly prevalent among Latin American-Caribbean countries. Mental disorders are common comorbidities in individuals with ARD. However, the etiology of the association between ARD and mental disorders remains unclear. We examined the association of inflammatory cytokines, microbiome, and other biomakers with measures of depression, social anxiety, and executive functions. We observed a significant increase in cytokine and chemokine expression levels in saliva and plasma in the alcohol group (AG) samples. Also, the salivary bacterial composition in the AG revealed an abundance of Prevotella. Depression symptomatology was markedly higher in the AG, but social anxiety levels were negligible. AG also exhibited executive dysfunctions, which negatively correlated with increased plasma levels of pro-inflammatory cytokines and increased salivary concentrations of Prevotella bacteria. Our study suggests that chronic alcohol use correlates with executive dysfunction, immune system dysregulation, and dysbiosis of the salivary microbiota. Additional studies are needed to understand the role of the microbiome and inflammation in alcohol use and mental comorbidities.


Introduction
Alcohol drinking is an important social activity in Latin American-Caribbean (LAC) culture; as a result, alcohol-related disorders (ARD) are highly prevalent among LAC countries [1][2][3]. A 2008 survey conducted in the Puerto Rican population aged 15-74 years reflected that 76.8% had used alcohol at least once during the individual's lifetime, and 48.8% reported using alcohol in previous 12 months [4]. According to the study, 12.1% of the participants reported suffering from alcohol abuse and 4.8% had alcohol dependence. A 2012 survey conducted with licensed drivers aged 16 years and older revealed that 22% of the participants admitted having driven a motor vehicle after consuming alcohol. Also, 33% of driving fatalities reported in Puerto Rico in 2016 were attributed to alcohol use; 66% of them also presented a speeding factor [5]. These findings suggest impairments in drivers' executive functions (i.e., cognitive modulation of goal-directed behaviors) [6]. In addition to the aspects above, multiple problems arise because of excessive alcohol use such as progression to illegal drugs, consumption of multiple substances, job loss, domestic or interpersonal violence, cancer, cirrhosis, and mental disorders.
The NIH Toolbox Cognition Battery [53] is a computer-based adaptive test that assesses cognition in the following areas: executive function, attention, episodic memory, language, processing speed, and working memory. The test generates several scores, including fully adapted national standard scores, which are used to classify participants' cognitive abilities based on the expected mean (100 points) and standard deviation (15 points). This study focused on the executive functions measured using the Flanker Inhibitory Control and Attention Test and the Dimensional Change Card Sort Test. These tests are related to the ability to inhibit automatic response tendencies that may interfere with achieving a goal and the ability to shift among multiple aspects of a strategy or task, respectively [53].

Assessment of Social Anxiety
The Liebowitz Social Anxiety Scale Self-Report (LSAS-SR) is a valid and reliable scale [51] for determining if an individual feels fear or anxiety about his/her diverse social situations and if he/she tries to avoid those situations. It has 24 items, each one of which is rated separately for fear/anxiety; (0 = none, 1 = mild, 2 = moderate, 3 = severe) and avoidance behavior (0 = never, 1 = occasionally, 2 = often, 3 = usually). Scores of <55, 55-65, 65-80, 80-95, >95 are considered as no to negligible, moderate, marked, severe, and very severe levels of social anxiety, respectively.

Depression Symptomatology
Depression symptoms were assessed using the nine-item Patient Health Questionnaire (PHQ-9). This self-report scale is a valid and reliable measure of depression severity [50]. Each response is scaled by how often the described symptom bothered the individual during the last two weeks (0 = not at all, 1 = several days, 2 = more than half the time, 3 = nearly every day). Total scores were used to assess depression severity as follow; ≤4 (minimal), 5-9 (mild), 10-14 (moderate), 15-19 (moderately severe), and ≥20 (severe, major depression).

Collection of Biological Samples
Approximately 5 mL of saliva were collected from each participant. Initially, they were asked to rinse their mouth twice with room-temperature water. After five minutes, they were asked to spit into a conical 50 mL-volume, sterile conical centrifuge tube. Saliva samples were stored in aliquots at −85 • C until analysis [54].
A licensed medical technologist, experienced in phlebotomy, aseptically collected about 10 mL of venous blood into three separate EDTA-containing blood collection tubes (BD Biosciences, Franklin Lakes, NJ, USA). The blood samples were centrifuged for plasma separation and stored in aliquots at −85 • C until analysis.

Cytokines, Chemokines, Leptin, Insulin, and Cortisol Assays
We assessed the concentration of pro-inflammatory and anti-inflammatory cytokines and chemokines in saliva and plasma samples using the MILLIPLEX ® MAP Human Cytokine/Chemokine Magnetic Bead Panel Premixed 39-plex Immunology Multiplex Assay kit (EMD Millipore, Chicago, IL, USA) following manufacturer's instructions as previously described [55]. Briefly, a premixed cytokine bead suspension was incubated with undiluted saliva or plasma. A minimum of 50 microsphere events were acquired for each cytokine using the Luminex MAGPIX ® System with the xPONENT ® software (Luminex Corp., Austin, TX, USA). The median fluorescent intensity (MFI) was obtained for each cytokine using a five-parameter curve-fitting method of the program, and the software automatically calculated each cytokine concentration. Plasma and salivary concentrations of leptin, insulin, and cortisol were assayed using the Millipore Human Magnetic Bead Panel kit, following the manufacturer's instructions.

Microbiome Analysis by Next Generation Sequencing
We performed the metagenomic analysis of the bacterial gene coding for the16S ribosomal RNA according to the Illumina MiSeq 16S metagenomics sequencing library preparation procedure (part #15044223 rev. A; Illumina Inc., San Diego, CA, USA). Bacterial DNA was purified from 150 L of saliva using the QIAGEN blood DNA extraction kit following the manufacturer's recommendations (QIAGEN, Venlo, Limburg, The Netherlands). Once the DNA was purified, the conserved 16S V3 and V4 region was PCR amplified using the FastStart PCR master mix (Roche Diagnostics, Indianapolis, IN). The PCR conditions were as follows: 95 • C for 3 min, followed by 25 cycles (95 • C for 30 s, 55 • C for 30 s, 72 • C for 30 s), and a final elongation step at 72 • C for 5 min. The following primer pair suggested by Klindworth et al. [56] was used for the PCR: FWD: 5 TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG3 and REV: 5'GTCTCGGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVHHHTATCTAATCC3 .
Illumina AMPure XP beads were used to purify the amplicons. The concentration of the amplicons was adjusted to a concentration of 0.2 ng DNA/µL. We performed the Index PCR using the Illumina Nextera DNA preparation kit (catalog number FC-121-1031) under the same conditions as described above, except that the Nextera index 1 and index 2 primers were used in place of the forward and reverse primer pair described above. AMPure XP beads were used to purify the Nextera amplicons. The library data was then acquired using the Illumina MiSeq Reporter software. The acquired data were extracted, decompressed, and then analyzed using the QIIME software in a Linux platform (Ubuntu), as previously described by Caporaso et al. [57]. Sequences were de-multiplexed, quality filtered, and a Seqs.fna file was generated. This file was then used to create an operational taxonomic unit (OTU) table. The OTU table provided the taxonomic composition of the microbe community. Statistical analyses were then performed using the OTU table data.

Vitamin D Assay
Plasma and salivary concentration of vitamin D were assayed as its hydroxyl form using ENZO 25(OH) Vitamin D ® ELISA kit (ENZO, Farmingdale, NY, USA), following the manufacturer's instructions. Determination of optical density was performed in a BioTek Synergy HT (BioTek, Winooski, VT, USA) set to 450 nm.

Dietary Assessment
Nutrient data were collected using the Food Frequency Questionnaire (FFQ), developed by the Nutrition Assessment Shared Resource (NASR) of the Fred Hutchinson Cancer Research Center (FHCRC). Participants reported the frequency of consumption and portion size of 125 foods and beverages using a Likert-type scale referencing food intake patterns during the last month. The FFQs were processed for nutrient analysis through the University of Minnesota Nutrition Data Systems for Research (NDSR) software [58][59][60].

Data Analyses
Statistical analyses were performed using IBM SPSS (v 24.0, Armonk, NY, USA). Descriptive statistics included means, medians, standard deviation, and interquartile ranks; these were assessed for all continuous variables. Bivariate analysis was used to analyze differences in sex, age, and education. Chi-square was used to compare differences between groups; p-values < 0.05 were considered statistically significant. Mann-Whitney U tests were performed to compare the differences between two independent groups. Multiple logistic regression adjusted odds ratio was used to assess the association between alcohol use and inflammation, depression, anxiety, and executive functions. Spearman Rho was used to assess correlations among depression, social anxiety, executive functions, microbiota, and inflammatory markers taking into consideration the source of the sample (saliva or plasma).
The microbiota structure and statistical comparison of the metagenomic samples were evaluated using the MicrobiomeAnalyst online platform [61]. Briefly, data were filtered to improve downstream statistical analyses by using a minimum count at four with prevalence and low variance (based on inter-quantile range) at 20% and at 10%, respectively. Normalization was performed by rarefying to the minimum library size (2846) and scaling to the cumulative sum scaling (CSS) method. The abundance profiling (phylum/genus) were based on the sum of their counts across the groups. The core microbiome analysis, which is the evaluation of a group of taxa that were in a high fraction of the population, was performed at a sample prevalence of 20% and a relative abundance of 0.2%. The beta-diversity analysis was performed by using the Bray-Curtis index distance method with a permutational MANOVA (PERMANOVA) statistical method.
The alpha-diversity profiling and significance testing were evaluated by using Shannon metrics and Mann-Whitney/Kruskal-Wallis, respectively. The univariate statistical comparison at the genus level was evaluated by using the Mann-Whitney U Test statistical method with an adjusted p-value cutoff of 0.05. The linear discriminant analysis effect size (LEfSe) and random forest methods were used to perform biomarker analysis [62,63]. The LEfSe, which employs a non-parametric Kruskal-Wallis rank-sum test, was obtained with an adjusted p-value cutoff of 0.05 and a log of the linear discriminant analysis (LDA) score of 1. Random forest analysis was at the genus level with 500 bootstraps, randomness setting on, and seven numbers of predictors to try. Table 1 summarizes the general characteristics of the study participants in the alcohol (AG) and control (CG) groups. The self-descriptive information was obtained through individual interviews. The average age of the participants was 41.6 years old for the alcohol group and 37.9 years old for the control group. All participants were of Hispanic ethnicity. The majority of study participants were unemployed (36.7% of AG, 50.0% of CG) or retired (13.3% of AG and 20.0% of CG). The AG (43.3%) had more participants with lower education than the CG (10.0%). There was a higher number of participants in the AG who reported smoking as compared to the CG (63.3% vs. 10.0% in CG). Those who reported smoking also had higher risks of drinking: OR = 15.5 (95% CI: 3.01-80.04; p < 0.01). Both groups reported having medical conditions (60.0% for AG; 55.0% for CG) but not statistically significant differences were found. The most prevalent medical conditions were hypertension (23.3% for AG; 45.0% for CG) and hypothyroidism (6.6% for AG; 20.0% for CG). The participants in both groups were receiving medication to treat these conditions. Another medical condition reported was asthma (13.3% for AG; 5.0% for CG) although symptoms were absent at the time of assessment. We also explored the history of psychiatric treatment. The AG showed a higher prevalence of depression (30% vs. 10%) and anxiety (37% vs. 10%) disorders than the CG. Participants from both groups reported using antidepressants or anxiolytics. Most of the study participants from the AG reported having a sleep disorder (80% vs. 35% from the CG); alcohol consumption increased the odds of also having a sleep disorder OR = 7.4 (95% CI: 2.1-26.7, p < 0.05). Only 7% of the AG reported a diagnosis of alcohol dependence disorder. However, most of AG participants (73%) reported having a family history of alcohol dependence, which increases the odds of drinking alcohol OR = 3.3 (95% CI: 1.01-11.11, p < 0.05). The mean of breath alcohol concentration for the AG was 0.01 (SD = 0.03), whereas the mean for the CG was 0. AUDIT median score for the AG was 18 (IQR = 15.5-28.0). Most of the participants in AG showed a pattern of harmful and hazardous drinking (33%) or possible alcohol dependence status (43%), according to AUDIT scores. Table 2 summarizes the scores and severity levels of depression and anxiety, as measured by the study instruments. The PHQ-9 median scores reflected statistically significant differences among groups; the AG showed more depression symptoms than the CG (8.5 vs. 4; p ≤ 0.01). A considerable portion of the participants from both groups reported having minimal or mild depression (57% and 85% for AG and CG, respectively); 43% of the AG versus 15% of CG indicated symptoms corresponding to moderate to severe depression. However, LSAS-SR scores showed that the majority of study participants have a negligible level of social anxiety.

Cognitive and Executive Function Abilities
Participants' cognitive and executive function abilities were measured using the NIH Toolbox Cognition Battery test. As shown in Table 3, participants from the AG exhibited significantly lower median standard scores in the NIH Toolbox Cognition Composite Score (75.5 vs. 100.0; p ≤ 0.001). Statistically significant differences among groups were also detected in measures of executive functions. The AG showed lower median standard scores in Flanker inhibitory (82.5 vs. 103.0, p ≤ 0.001) as well as in dimensional change card sort (95.0 vs. 107.0; p ≤ 0.001). These scores were re-coded in two intervals as follows: <85 and ≥85. Our results show that drinking alcohol increases the odds of having lower cognition ability (OR = 21.2; 95% CI: 3.9-113.9; p ≤ 0.001). Also, drinking alcohol increases the odds of lower abilities in executive functions: Flanker inhibitory score OR = 8.2 (95% CI: 1.9-35.1; p ≤ 0.01), and dimensional change card sort OR = 1.9 (95% CI: 1.4-2.5; p ≤ 0.05). Table 3. Scores on selected measures of NIH toolbox cognition battery.

Plasma and Salivary Levels of Cytokines
We determined the concentrations of the following cytokines in plasma and saliva: TNF-α, IL-12p70, MDC, IL-10, IFN-γ, TNF-β, IL-1β, IL-5, IL-2, IL-6, IL-17A, IL-4, IL-1RA, IL-13, IL-7, and GM-CSF among other analytes. The median concentration levels of plasma and saliva levels of cytokines with statistically significant differences are presented in Table 4. The AG showed significantly lower concentrations of salivary IL-12p70, and IL5, and significantly higher concentrations of TNF-β, IL-2, IL-4, and IL-13. Also, plasma levels of MDC were significantly higher in the AG.

Plasma and Salivary Levels of Insulin, Leptin, Cortisol, and Vitamin D
Analysis of plasma and saliva levels of insulin, leptin, cortisol, and vitamin D revealed no statistical difference of leptin, insulin and cortisol between the AG and CG groups (Table 5). Interestingly, the CG exhibited lower levels of vitamin D in plasma (45.8 pg/mL and 18.4 pg/mL for AG and CG, respectively). Vitamin D is significantly reduced in saliva for the AG (p = 0.001). In contrast, the blood level of vitamin D in the AG is significantly higher than the CG (p = 0.01). A significant positive correlation was detected between plasma and salivary leptin levels in the total sample as well as in AG and CG (r s = 0.57, p ≤ 0.001; r s = 0.59, p = 0.001; and r s = 0.67, p = 0.002, respectively; Figure 1A). We also found a significant positive correlation between plasma and salivary insulin levels (r s = 0.49, p ≤ 0.001; r s = 0.45, p = 0.015; and r s = 0.63, p = 0.004, respectively; Figure 1B).  A significant positive correlation was detected between plasma and salivary leptin levels in the total sample as well as in AG and CG (rs = 0.57, p ≤ 0.001; rs = 0.59, p = 0.001; and rs = 0.67, p = 0.002, respectively; Figure 1A). We also found a significant positive correlation between plasma and salivary insulin levels (rs = 0.49, p ≤ 0.001; rs = 0.45, p = 0.015; and rs = 0.63, p = 0.004, respectively; Figure  1B).

Diet
Analysis of nutrients that participants consumed in the last month showed statistically significant differences in reports of alcohol intake (16.75 g for AG; 0.04 g for CG). Reports of other nutrients did not show statistically significant differences among groups (Supplementary Table S1). The Shannon alpha-diversity did not reach significant differences (p value= 0.94) (Figure 4, panels A and B) between the two populations evaluated. Similar finding was observed in beta-diversity analysis (f value = 0.9019, R-squared = 0.014) (Figure 4, panel C and D). Nevertheless, a significant difference between the groups was only observed in Prevotella genus by using Mann-Whitney U Test statistical method (p value = 0.015) ( Figure 5). The Random forest prediction ( Figure 6) and LEfSe evaluation (Figure 7) suggest that Prevotella genus might be a possible biomarker of alcohol use in saliva.

Diet
Analysis of nutrients that participants consumed in the last month showed statistically significant differences in reports of alcohol intake (16.75 g for AG; 0.04 g for CG). Reports of other nutrients did not show statistically significant differences among groups (Supplementary Table S1).     or log-transformed count (panel B), represents the abundance distribution of Prevotella genus. The Mann-Whitney/Kruskal-Wallis method with adjusted p-value cut-off of 0.05 was used to perform statistical analyses. Figure 6. Random forests ranked by contributions to classification accuracy (mean decreased accuracy). The random forest analysis, which is an algorithm suitable method for high dimensional data analysis, was performed by using the ensemble of classification trees by random feature selection from a bootstrap sample at each branch. The 15 more important genera were classified based in their mean decreased in accuracy (high = red, low = green).

Cytokines and Microbiome Correlations with Executive Functions
We found that salivary concentrations of Prevotella showed a significant negative correlation with Flanker (rs = −0.408, p = 0.  The random forest analysis, which is an algorithm suitable method for high dimensional data analysis, was performed by using the ensemble of classification trees by random feature selection from a bootstrap sample at each branch. The 15 more important genera were classified based in their mean decreased in accuracy (high = red, low = green).

Cytokines and Microbiome Correlations with Executive Functions
We found that salivary concentrations of Prevotella showed a significant negative correlation with Flanker (r s = −0.408, p = 0.031) and Dimensional (r s = −0.389, p = 0.037) in the AG but not in the CG  The differences between the groups were evaluated by using the linear discrimination analysis effect size (LEfSe). The LEfSe method, which is a biomarker discovery, uses Kruskal-Wallis sum test and linear discrimination analysis to identify significant difference abundance and estimated the effect size, respectively. The bar graph shows the LDA scores of (at most) top 25 significant bacteria. The colors (alcohol users = red, control = blue) represent which group was more abundant compared to the other group at genus level. The p value cut-off was 0.1 with an adjusted false discovery rate.  The differences between the groups were evaluated by using the linear discrimination analysis effect size (LEfSe). The LEfSe method, which is a biomarker discovery, uses Kruskal-Wallis sum test and linear discrimination analysis to identify significant difference abundance and estimated the effect size, respectively. The bar graph shows the LDA scores of (at most) top 25 significant bacteria. The colors (alcohol users = red, control = blue) represent which group was more abundant compared to the other group at genus level. The p value cut-off was 0.1 with an adjusted false discovery rate.

Discussion
One of the main aims of this study was to examine plasma and saliva samples for inflammation and microbiome analyses in individuals with alcohol use living in Puerto Rico. The observed significant positive correlations between plasma and salivary IL-10, and between plasma and salivary IL-13 in alcohol users, suggest the possibility of using saliva as a non-invasive method for assessing these cytokines in this population. A recent study found significant correlations between the levels of IL-13 and IL-10, suggesting an interdependency between these two cytokines [64]. Periodontitis is a gum disease characterized by tissue breakdown caused by pathogenic bacteria of dental plaque, which activate the host immune-inflammatory response against these periodontal pathogens [65]. The AG showed higher salivary IL-10 and IL-13 than the CG. IL-10 has been identified to be increased in the process of a hangover after alcohol consumption [66]. IL-10 has also been associated with the development of periodontitis [67]. In periodontal diseases, IL-10 and IL-13 are involved in modulating the immune response and in the suppression of pro-inflammatory cytokines by monocytes/macrophages, respectively [68,69]. To our knowledge, there is little information on any associations between IL-13 in the oral cavity and alcohol use. Regardless, there have been studies that have found that alcohol consumption increases IL-13 signaling in airways [70,71].
On the other hand, these two anti-inflammatory cytokines, IL-10 and IL-13, have been identified as elevated in oral squamous cell carcinoma [72,73]. These cytokines have also been suggested as biomarkers for squamous cell carcinoma disease progression [72]. The results described above are not unexpected since alcohol consumption has been associated with oral squamous cell cancer [74]. Lastly, anti-inflammatory cytokines like IL-10 seem to be implicated in binge drinking. During binge drinking, IL-10 levels decreased in the basolateral amygdala. Interestingly, infusion of IL-10 decreased binge drinking behavior [75]. Further study of IL-13 could lead to defining what the role of this cytokine in the setting of alcohol use is.
Other significant positive correlations were observed between plasma and salivary levels of leptin and insulin in all samples and between groups. A recent review reported a significant correlation between serum and salivary insulin [76]. Similar correlations have been found between plasma and salivary leptin [77]; however, the need for additional studies to validate the use of saliva for these purposes is highlighted.
We also explored the associations among blood and saliva biomarkers (pro-inflammatory cytokines, vitamin D, insulin, leptin, and cortisol) with measures of executive functions depression, and social anxiety. Increased plasma levels of pro-inflammatory MDC, IL-6, IL-17A, INF-γ, and GM-CSF were negatively correlated to measures of executive functions in alcohol users, suggesting a possible role of these cytokines in their reduced ability to inhibit (restrain interferences for achieving a goal) and shifting (change attention between tasks). Besides, the negative correlation observed in salivary levels of IL-1β and measures of executive functions implies a protective factor: lower levels of these salivary biomarkers correlated with increased ability to inhibit and shifting.
Our results revealed that the AG exhibited executive dysfunctions, mainly a reduced ability to self-control and to shift responses based on rules or contingencies [78]. This finding supports previous studies reporting that alcohol use is linked to impairment in cognitive functioning [79,80]. However, it is crucial to take into consideration that deficits in executive functions may precede alcohol use [81] while conflictive results have been reported [82]. Consistent with the literature, co-morbid depression and anxiety [4,7-11] and polysubstance use [83] were more frequently reported in the AG. Both depression and polysubstance abuse are also linked to impairment in executive functions [79,84,85]. However, no social anxiety was present, which can be associated with the Latin American-Caribbean culture that promotes alcohol drinking as a social activity [86].
Chronic alcohol use has been reported to impair the immune response functions through effects on multiple organs such as the liver, lung, stomach, intestine, and brain [87,88]. According to our results, alcohol users showed increased plasma macrophage-derived chemokine (MDC). MDC, also referred to as chemokine CCL22, has a dual function: inflammatory and homeostatic; it stimulates Th2 response, Th2 cell migration, and T reg migration [89]. The Th2 cells produce IL-4, IL-5, IL-6, IL-9, IL-10, and IL-13 cytokines, and evoke strong antibody responses and eosinophil accumulation, but inhibit several functions of phagocytic cells [90]. Also, MDC has been associated with alcoholism, phobia, and interpersonal sensitivity [91].
The CG showed significantly increases levels of salivary IL-5 and IL-12p70 compared to the AG. It should be noted that although not statistically different, many of the CG participants reported having medical conditions that might impact their immunology profile. IL-5 plays a critical role in adaptive immune responses since it regulates eosinophils, cells that play a central role in the pathogenesis of allergic diseases, and induces cell maturation, survival, and activation [92,93]. IL-12p70 has a critical pro-inflammatory role in innate immunity and plays a vital role in the initiation of the protective immune response [92]. Interestingly, a recent study found increased levels of both cytokines after exposure to psychological stress in adults [94].
The AG presented significantly increased levels of salivary cytokines TNF-β, IL-2, IL-4, and IL-13, suggesting immune dysregulation in the oral cavity. Previous studies showed associations of increased risk for oral cancer with TNF-β polymorphisms genes and increased expression of salivary IL-4 and IL-13 [72,95]. Other studies comparing salivary cytokines of clinical samples and controls found increased expression of IL-2 and IL-4 in smokers and in patients with oral lichen planus (a chronic inflammatory disease that affects the oral mucosa) [96,97]. Besides, increased salivary IL-4 also has been associated with periodontitis [97][98][99]. However, data about such salivary cytokines (TNF-β, IL-2, IL-4, and IL-13) in alcohol users were not found. A previous study with alcohol users assessed differences in salivary concentrations of TNF-α, IL-1α, and INF-γ compared to controls, showing significantly increased expression of all cytokines in alcohol users [100]. To the best of our knowledge, data from the current study provides new evidence identifying potential salivary biomarkers associated with alcohol use. The inflammatory profile revealed in the AG is different from other studies reporting that increases in serum/plasma TNF-α, IL-1, IL-6, IL-12, IL1-β, and IL-8 are associated with chronic alcohol consumption or alcohol-related liver diseases [16,101,102]. This difference might be due to genetics, age, cultural, and behavioral aspects (e.g., diet and lifestyle) [103].
Vitamin D deficiency in plasma was found in the CG, which has been described to be inversely associated with physiological dysregulation associated with chronic exposure to stress and adverse health outcomes [104]. Deficiency of vitamin D has also been found to independently predict cognitive impairment [105]. Although previous studies have related alcohol consumption with lower levels of vitamin D in plasma, we found higher levels of this vitamin in AG [106][107][108]. Nevertheless, vitamin D deficiency has been linked with other factors like inadequate sun exposure, limited oral intake, severe liver disease, renal insufficiency, and impaired intestinal absorption, among other risk factors [109]. A possible explanation is that some of the factors mentioned above may influence the level of vitamin D in our control group.
Although blood has been considered to be bacteria-free or 'sterile', a recent review describes that many bacteria can survive in a dormant form in blood in several non-communicable diseases such as diabetes mellitus and cardiovascular diseases [110]. A diversified microbiome has also been described in the blood of healthy donors [111]. The origin of these bacteria is suggested to be from the gut microbiota as well as the oral microbiota [110]. Leclercq et al. identified the alcohol-induced dysbiosis of gut microbiota and subsequent inflammatory responses as the likely causes of depressive disorders and alcohol craving [16]. Similarly, variation in the human oral microbiome caused by alcohol use has been related to systemic inflammatory and autoimmune disorders [112,113]. We found that the genus Prevotella was significantly more abundant in AG saliva. A similar finding was observed by Fan et al., who identified increased Prevotella in the oral microbiome of drinkers [113]. The observed differences in Prevotella among our groups may suggest a possible dysbiosis in the salivary microbiota of the alcohol users.
Interestingly, reduced executive functions were correlated to increased salivary concentrations of Prevotella in AG as well as in all samples. Prevotella has been documented as associated with periodontitis, cavities, oral cancer, and root canal procedures [114][115][116][117][118]. Other studies also have reported associations between oral diseases and reduced cognitive function [119][120][121]. To our knowledge, this is the first time that Prevotella is reported as inversely associated with executive function in the scenario of the oral microbiome and alcoholism. The few studies revealing a positive correlation between cognitive function and Prevotella assessed this bacteria in the gut and its antibody in serum [122,123]. Our study is possibly the first to report any correlation between oral Prevotella and cognitive function and in particular, to report a negative correlation.

Study Limitations
The small study sample size and the use of a case-control design limit the ability to draw firm conclusions. Therefore, a larger sample size, and a longitudinal design with more clearly defined and tightly controlled groups are needed. Possible confounders such as drug use, tobacco use, depression, anxiety, body mass index, and vitamin D deficiency may increase the inflammatory responses in the samples. Notably, the AG reported a higher frequency of tobacco use compared to the CG and might be confounding the study results. Another significant limitation of the study was the absence of a previous oral health status evaluation between AG and CG groups before sample collection. Periodontitis and dental caries, as well as other oral diseases that modulate the expression of pro-inflammatory cytokines, may drive cytokine expression between the groups [124]. The inclusion of a previous evaluation of the oral cavity in the current study could have increased the reliability of the results. Also, other factors in the methodology, such as the inclusion criteria for participants, sampling, storage conditions, and storage time before samples evaluation, may have some adverse effects in the assessment of cytokines concentrations [125].

Conclusions
The results of the present study support previous knowledge that alcohol use correlates with executive and immune system dysfunctions and dysbiosis of the salivary microbiota. Alcohol users showed upregulation of plasma MDC, confirming recent findings [91]. This study also identified potential alcohol-associated salivary inflammatory biomarkers (TNF-β, IL-2, IL-4, and IL-13), that may be important in its pathophysiology. Given the detrimental effects of alcohol in the individual's mental and physiological health, these results warrant further investigation to understand the pathophysiology of alcohol use and mental comorbidities, and the use of salivary biomarkers in the identification of alcohol-related diseases.