Changes in the metabolic profile of human male postmortem frontal cortex and cerebrospinal fluid samples associated with heavy alcohol use

Heavy alcohol use is one of the top causes of disease and death in the world. The brain is a key organ affected by heavy alcohol use. Here, our aim was to measure changes caused by heavy alcohol use in the human brain metabolic profile. We analyzed human postmortem frontal cortex and cerebrospinal fluid (CSF) samples from males with a history of heavy alcohol use (n = 74) and controls (n = 74) of the Tampere Sudden Death Series cohort. We used a nontargeted liquid chromatography mass spectrometry‐based metabolomics method. We observed differences between the study groups in the metabolite levels of both frontal cortex and CSF samples, for example, in amino acids and derivatives, and acylcarnitines. There were more significant alterations in the metabolites of frontal cortex than in CSF. In the frontal cortex, significant alterations were seen in the levels of neurotransmitters (e.g., decreased levels of GABA and acetylcholine), acylcarnitines (e.g., increased levels of acylcarnitine 4:0), and in some metabolites associated with alcohol metabolizing enzymes (e.g., increased levels of 2‐piperidone). Some of these changes were also significant in the CSF samples (e.g., elevated 2‐piperidone levels). Overall, these results show the metabolites associated with neurotransmitters, energy metabolism and alcohol metabolism, were altered in human postmortem frontal cortex and CSF samples of persons with a history of heavy alcohol use.

Moreover, heavy alcohol use is associated with an altered circulating metabolic profile (for a review, see Voutilainen and Kärkkäinen 4 ). For example, increased levels of glutamate, tyrosine, and phosphatidylcholine diacyls, and decreased levels of glutamine, serotonin, and phosphatidylcholine acyl-alkyls have been associated with heavy alcohol use. [5][6][7][8] Furthermore, some of these changes in the circulating metabolic profile have been linked to structural changes in the white and grey matter. 8,9 Moreover, in rats, it has been shown that alcohol exposure alters the brain metabolic profile. 10 In living humans, in vivo 1 H magnetic resonance spectroscopy can be used to measure metabolites. Although this method has shown alterations in brain metabolite levels in individuals with alcohol use disorder, for example, decreased levels of choline containing compounds, the method is limited to only measuring a handful of metabolites and does not give a good picture of whole metabolic profile level changes (for a review, see Meyerhoff 11 ). Similarly, there are reports on changes in many metabolite classes, for example, ethanolamines, steroids, and neurotransmitters, in human postmortem brains measured with targeted methods. [12][13][14] However, to our knowledge, there is no published measurement of changes in the whole human postmortem brain metabolic profile associated with heavy alcohol use to get an overall view of altered metabolic processes. Therefore, our aim here was to measure changes in the human brain metabolic profile associated with heavy alcohol use by performing a nontargeted metabolomics analysis on human postmortem frontal cortex and cerebrospinal fluid samples.

| Subjects
We used samples from the Tampere Sudden Death Series (TSDS) cohort, which was collected from forensic medicine autopsies of people who suffered out of hospital death in the area of the Pirkanmaa Hospital District during 2010-2014. Frontal cortex (Broadman area 9) and CSF samples, a portion extracted via syringe, were collected from a total of 700 subjects, of which we selected 74 heavy alcohol users and 74 controls, stored at −80 C until use. Selection of the subjects was based on autopsy reports and medical records. Inclusion criteria for the heavy alcohol use group was diagnosis of alcohol-related diseases (ICD-10 codes F10.X, G31.2, G62.1, G72.1, I42.6, K70.0-K70.4, K70.9, and K86.0) or signs of heavy alcohol use in clinical or laboratory findings (e.g., increased levels of alcohol use biomarkers gamma-glutamyl transferase, mean corpuscular volume, and carbohydrate-deficient transferrin). Lack of these findings was inclusion criteria for the control group, most of whom had died due to cardiovascular system diseases. Causes of death and general background characteristics are shown in Table 1.
We only used samples from males, because there were only nine suitable female cases with a history of heavy drinking.  analysis. The mass spectrometry data processing was performed using MS-DIAL ver. 3.40. 16 The peak picking and peak alignment parameters were as follows: mass range from 40 to 1,000 (HILIC) or 1,600 (RP) Da, MS tolerance 0.005 Da, MS2 tolerance 0.01 Da, minimum peak width 10 scans, minimum peak height 10,000 (selected based on background noise level). Peaks needed to be detected in at least 70% of samples from one study group to be included in the final data matrix. Drift correction of the metabolomics data was done using results from the QC sample, according to a previously published protocol. 15 In brief, the molecular features were corrected for the drift pattern caused by the LC-MS procedures using regularized cubic spline regression, fit separately for each feature on the QC samples.

| Metabolomics analysis
The smoothing parameter was chosen from an interval between 0.5 and 1.5 using leave-one-out cross validation to prevent overfitting.
After the drift correction, feature quality was assessed, and lowquality features were flagged. Features were kept if their RSD* was below 20% and their D-ratio below 40%. In addition, features with classic RSD, RSD* and basic D-ratio all below 10% were kept. This additional condition prevents the flagging of features with very low values in all but a few samples. Missing values were imputed using random forest imputation. QC samples were removed prior to imputation to prevent them from affecting imputation.
Metabolite identification was focused on statistically significantly altered molecular features. Metabolite identification was based on exact mass, MSMS fragmentation, and retention time. Identifications were ranked according to common guidelines. 17

| Statistical analysis
We used Welch's t test (continuous variables) and χ 2 test (binominal variables) to evaluate differences between the study groups in background characteristics. We used Welch's t test and Cohen's d effect sizes to evaluate differences between the study groups in the metabolomics data. To account for multiple testing, we adjusted the α level by the number of principal components needed to explain 95% of the variation in the data. For correlations, we used Spearman's correlations.

| RESULTS
Based on background information, the heavy alcohol user group had lower brain weight and lower body mass index (BMI) when compared with controls (Table 1) Figure 1).
Identified metabolites that had a p value < 0.05 when comparing study groups are shown in Tables S1 (frontal cortex) and S2 (CSF).
Overall, we were able to identify more altered metabolites from the frontal cortex samples than from the CSF samples. Of note is that the Correlations between frontal cortex and CSF levels of metabolites, which had p values below 0.05 in both measured sample types, are shown in Table S3. There were significant correlations between frontal cortex and CSF samples in many of the altered metabolites, including 2-piperidone and 3-hydroxyisovaleric acid.

| DISCUSSION
In the present study, we found differences in the metabolic profile of  23 2-Piperidone is a substrate for cytochrome P450-2E1 (CYP2E1), which metabolizes 2-piperidone to 6-hydroxy-2-piperidone. 24 2-Piperidone has been suggested to be a biomarker of CYP2E1 activity. 24 Therefore, high 2-piperidone levels observed in the present study could be associated with alcohol-induced inhibition of 2-piperidone metabolism by blocking its access to the CYP2E1 enzyme. CYP2E1 has an important role in alcohol-induced changes in the function of the brain because it produces acetaldehyde, which has been associated with the behavioral effects of alcohol. 25 Further research is needed to understand the possible biological role of 2-piperidone in the brain and investigate if 2-piperidone levels are also increased in the plasma or urine in heavy alcohol users.
Moreover, the cholinergic system of the brain is considered to be important in the development of alcohol dependence. 26 We observed significantly decreased levels of acetylcholine and a trend in decreased choline levels in the frontal cortex samples, which is in line with a previous in vivo 1 H magnetic resonance spectroscopy study where decreased levels of choline containing molecules were measured in persons with alcohol use disorder. 20 In contrast, these postmortem brain results show a trend towards increased levels of many lysophosphatidylcholines (lysoPCs) and phosphocholine (Table S1).
This indicates that the choline balance has shifted towards lysoPCs, compared with free choline or acetylcholine. LysoPCs can increase NFkB activation 27 and the recruitment of microglia, 28 and increased levels have been reported in animal models of brain damage 29 and in human postmortem samples from persons with Alzheimer's disease. 30 F I G U R E 2 Significantly altered metabolites. Heavy alcohol use was associated with significant (corrected α level = 0.0005 to account for multiple testing) differences in metabolite levels in both postmortem frontal cortex and cerebrospinal fluid (CSF) samples when compared with controls. High 2-piperidone and 3-hydroxyisovaleric acid levels were observed in the heavy alcohol use group in both frontal cortex and CSF samples. Mean ion abundance and 95% confidence intervals are shown. Legend: d, Cohen's d effect size; GABA, gamma-aminobutyric acid; p, p value from Welch's t test Furthermore, in the present study, we observed significantly decreased levels of aspartic acid and trends toward decreased levels of both N-acetylaspartate (NAA) and N-acetylaspartylglutamic acid (NAAG) in postmortem frontal cortex brain samples (Table S1). This is in line with previous studies showing decreased NAA peaks (consisting of both NAA and NAAG in magnetic resonance spectroscopy) in frontal cortex and white matter in association with heavy alcohol use. 20,[31][32][33][34][35][36][37] Decreased aspartic acid levels could be associated with alcohol caused ketosis, 38 and NAA and NAAG levels are considered to be a sign of decreased neuronal viability and/or integrity. [39][40][41] However, NAA is also found in myelin, and low NAA levels have been associated with myelin damage. 42  The main limitation of the present study is that PMI could influence the results because many metabolic processes continue after death. However, after correcting for multiple testing, all metabolites with significant differences between the study groups did not significantly correlate with PMI. This indicates that PMI is likely not causing the differences between the study groups of significantly altered metabolites. Other possible confounding factors include that differences in medical diagnoses (there are no healthy controls in postmortem studied) and used medications have been different between the cases and controls. For example, the metabolomics analysis showed that the controls used more metoprolol and the cases more nordiazepam, which could affect the seen results, for example, the decreased GABA levels (Table S1). Moreover, future studies should also analyze samples from females and different brain regions, because alcohol use could influence the metabolic profile differently in different genders and brain regions.
In conclusion, we show here that postmortem frontal cortex and CSF samples from persons with a history of heavy alcohol use have an altered metabolic profile when compared with samples from control subjects. Most alterations could be associated with neurotransmitter and energy metabolism, but metabolites associated with gut microbiota were also altered. Further studies are needed in other brain regions and with preclinical models to understand the spatial and temporal aspects of how heavy alcohol use alters the brain metabolic profile with connection to alcohol-related diseases.

ACKNOWLEDGMENTS
We thank Miia Reponen for the excellent technical assistance with the mass spectrometry analyses. This study is funded by the Finnish

CONFLICT OF INTEREST
OK and KH are founders of Afekta Technologies Ltd., a company offering metabolomics analysis services. Other authors report no potential conflicts of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.