Changes in circulating metabolome precede alcohol-related diseases in middle-aged men: a prospective population-based study with a 30-year follow-up

UEF//eRepository

DSpace https://erepo.uef.fi

Rinnakkaistallenteet Terveystieteiden tiedekunta

2020

Changes in circulating metabolome precede alcohol-related diseases in middle-aged men: a prospective

population-based study with a 30-year follow-up

Kärkkäinen, Olli

Wiley

Tieteelliset aikakauslehtiartikkelit

© 2020 by the Research Society on Alcoholism All rights reserved

http://dx.doi.org/10.1111/acer.14485

https://erepo.uef.fi/handle/123456789/24082

Downloaded from University of Eastern Finland's eRepository

1

Changes in Circulating Metabolome Precede Alcohol-Related Diseases in Middle-Aged Men: A 1

Prospective Population-Based Study with a 30-Year Follow-Up 2

3

Olli Kärkkäinen ^1, *, Anton Klåvus ², Ari Voutilainen ², Jyrki Virtanen ², Marko Lehtonen ¹, Seppo Auriola ¹, 4

Jussi Kauhanen ², Jaana Rysä ¹ 5

6

1 School of Pharmacy, University of Eastern Finland, Kuopio, Finland 7

2 Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio, Finland 8

9

* correspondence: Olli Kärkkäinen, School of Pharmacy, University of Eastern Finland, Yliopistonranta 1, 10

70210 Kuopio, Finland; email: olli.karkkainen@uef.fi 11

12

This study was funded by the Finnish Foundation of Cardiovascular Research. The authors also want to 13

thank Biocenter Finland and Biocenter Kuopio for supporting the core LC-MS laboratory facility.

14

2 Abstract

15

Background: Heavy alcohol use has been associated with altered circulating metabolome. We investigated 16

whether changes in the circulating metabolome precede incident diagnoses of alcohol-related diseases.

17

Methods: This is a prospective population-based cohort study where the participants were 42-60-year-old 18

males at baseline (years 1984-1989). Subjects who received a diagnosis for an alcohol-related disease 19

during the follow-up were defined as cases (n = 92, mean follow-up of 13.6 years before diagnosis).

20

Diagnoses were obtained through linkage with national health registries. We used two control groups:

21

controls who self-reported similar levels of alcohol use as compared to cases at baseline (alcohol-controls, 22

n = 92), and controls who self-reported only light drinking at baseline (control-controls, n = 90). A non- 23

targeted metabolomics analysis of baseline serum samples was performed.

24

Results: There were significant differences between the study groups in the baseline serum levels of 64 25

metabolites: in amino acids (e.g. glutamine [FDR corrected q-value = 0.0012]), glycerophospholipids (e.g.

26

lysophosphatidylcholine 16:1 [q = 0.0008]), steroids (e.g. cortisone [q = 0.00001]), and fatty acids (e.g.

27

palmitoleic acid [q = 0.0031]). The main finding was that after controlling for baseline levels of self-reported 28

alcohol use and the biomarker of alcohol use, gamma-glutamyl transferase, and when compared to both 29

alcohol-control and control-control group, the alcohol-case group had lower serum levels of asparagine 30

(Cohen’s d = -0.48 [95% CI -0.78 to -0.19] and d = -0.49 [-0.78 to -0.19], respectively) and serotonin (d = - 31

0.45 [-0.74 to -0.15], and d = -0.46 [-0.75 to -0.16], respectively), with no difference between the two 32

control groups (asparagine d = 0.00 [-0.29 to 0.29], and serotonin d = -0.01 [-0.30 to 0.29]).

33

Conclusions: Changes in the circulating metabolome, especially lower serum levels of asparagine and 34

serotonin, are associated with later diagnoses of alcohol-related diseases, even after adjustment for the 35

baseline level of alcohol use.

36

Key words: alcohol use disorder, alcohol dependence, metabolomics, asparagine, serotonin 37

3 Introduction

38

Alcohol use is one of the most important global causes of disease, disability and death, accounting for 5% of 39

the worldwide disease burden and associated with around 5% of global deaths (Griswold et al., 2018). The 40

World Health Organization has listed the reduction of harmful alcohol use by 10% as one of the key global 41

health goals (World Health Organization, 2010). However, it is challenging in the primary care setting to 42

identify those who will later in life develop alcohol-related disease; in these scenarios, documenting alcohol 43

use is often based on self-reports.

44

The circulating metabolome, measured from serum or plasma samples, could offer a novel opportunity to 45

identify those individuals who are at risk of developing alcohol-related diseases in the future, since alcohol 46

use has been associated with alterations in the circulating metabolome in humans (for a review see 47

(Voutilainen and Kärkkäinen, 2019)). For example, heavy alcohol use has been associated with higher levels 48

of glutamate, tyrosine, phosphatidylcholine diacyls, and fatty acids like palmitoleate (FA 16:1), as well as 49

with lower levels of glutamine, serotonin, and phosphatidylcholine acyl-alkyls when compared to low 50

drinking subjects (Harada et al., 2016; Jaremek et al., 2013; Lacruz et al., 2016; Lehikoinen et al., 2018;

51

Pallister et al., 2016; van Roekel et al., 2018; Würtz et al., 2016; Zheng et al., 2014). Furthermore, some of 52

these changes in the circulating metabolome have been linked to alcohol-related health risks e.g. changes 53

in liver function (Bajaj et al., 2017; Lian et al., 2011), and structural changes in the white and grey matter 54

(Heikkinen et al., 2019; Shen et al., 2019). Moreover, some changes in the circulating metabolome have 55

been associated with treatment outcomes. For example, high glutamate levels at the baseline, seem to 56

predict favorable responses to acamprosate treatment (Hinton et al., 2017; Nam et al., 2015). However, 57

most of the studies so far have been cross-sectional or case-control studies, where persons with heavy 58

alcohol use history are compared to light drinking controls, where it is impossible to say which of the 59

changes in the metabolome precede the diagnosis of alcohol-related disease (Voutilainen and Kärkkäinen, 60

2019).

61

4

Here our aim was to find alterations in the circulating metabolome that would precede a future diagnosis of 62

alcohol-related diseases, such as alcohol dependence or alcoholic liver cirrhosis. We conducted a non- 63

targeted metabolomics analysis on serum samples from a Finnish population-based cohort with 29-34 years 64

of follow-up. In addition to the actual cases, who were diagnosed for alcohol-related diseases during the 65

follow-up, we analyzed samples from matched controls, who had reported similar levels of alcohol use at 66

baseline, as well as controls who had reported only light drinking at baseline.

67

68

Materials and Methods 69

Study subjects 70

This analysis is part of the Kuopio Ischaemic Heart Disease Risk Factor Study (KIHD), which is an ongoing 71

prospective population-based cohort study in middle-aged men from Eastern Finland. The KIHD cohort 72

consisted of 2682 male participants (83% of those eligible) who were aged 42, 48, 54, or 60 years at the 73

baseline examinations conducted between the years 1984–1989 (Salonen et al., 1991). The study was 74

approved by the Research Ethics Committee of the University of Kuopio. The study was performed in 75

compliance with Declaration of Helsinki and written informed consent was acquired from all participants.

76

The KIHD study utilizes the annually updated linkage data from the national Care Register for Health Care 77

(formerly the Hospital Discharge Registry), maintained by the Finnish Institute for Health and Welfare, 78

including diagnoses of inpatient health care admissions (License THL/93/5.05.00/2013) and from the 79

National Death Registry of the Statistics Finland (License TK-53-1770-16). For this study, we constructed an 80

outcome variable, an alcohol-related disease, based on codes provided in the 8th, 9th, and 10th revisions of 81

the Internal Classification of Diseases (ICD) that have been at use during the follow-up years. ICD codes 82

used for selection of cases were: ICD-10: F10.X, G62.1, G31.2, I42.6, K29.2, K70-K70.4, K70.9, and K86.0;

83

ICD-9: 291, 305.0, 303.0, 303.9, 357.5, 425.5, 535.3, and 571.0-571.3; and ICD-8: 30300-30320, 57100. The 84

Finnish social security number was used as a personal identifier to link the registry-based outcomes with 85

the subjects of the study. Anyone who had an alcohol-related disease only based on the National Death 86

5

Register, was not classified as a case, but excluded from the potential controls. As an end point, we 87

considered the first alcohol-related diagnosis in the Care Register for Health Care during the follow-up 88

(cases could have additional diagnoses later). The registry data were obtained from baseline to the end of 89

the year 2017 (29-34 years of follow-up depending on the year of baseline examination).

90

Further criteria for selection were the existence of a serum sample from baseline and the availability of a 91

matched control (matching detailed described below). Subjects who already had an alcohol-related 92

diagnosis at baseline, or received one, or who died from an alcohol-related disease during the first year (a 93

365-days washout period) of follow-up were excluded from the study. This resulted in 92 subjects who 94

were selected for the metabolomics analysis (alcohol-case group, Figure 1 and Table 1). The mean follow- 95

up time for diagnosis of the alcohol-related disease was 13.6 years (4,957 days, SD = 9.4 years, 3,419 days), 96

with a minimum time of 377 days and maximum of 33.3 years (12,170 days).

97

We used two control groups: a) alcohol-control group and b) control-control group. Persons in the alcohol- 98

control group were of similar age and education level and had similar levels of self-reported alcohol intake 99

(measured in grams of ethanol per week) as persons in the alcohol-case group. We used self-reported 100

alcohol use as a measure of consumption since this is the most commonly recorded alcohol use related 101

variable recorded in regular visits to the healthcare services. Persons matched for age and education level, 102

but with no alcohol use or light alcohol use (< 12 grams per week) were selected as the low drinking control 103

group (control-control group). Ultimately, we had 92 subjects in the alcohol-case group, 92 subjects in the 104

alcohol-control group, and 90 subjects in the control-control group (Table 1). Serum samples collected at 105

baseline and stored at -80 °C were used in the non-targeted metabolomics analysis.

106

107

Figure 1 108

Table 1 109

110

6 Alcohol use and other measures at baseline

111

We used a structured self-report (quantity-frequency measure) to estimate the level of alcohol 112

consumption at baseline (Kauhanen et al., 1992). The self-report of alcohol use is still the most common 113

way to evaluate alcohol consumption in both epidemiologic studies and in regular primary health care 114

settings. The amount and frequency of consumption were enquired separately for different types of 115

alcoholic beverages. Finally, the mean weekly level of alcohol consumption for each individual was 116

calculated in grams of pure ethanol/week. Other alcohol use related measures from the baseline were 117

drink preference, frequency of hangovers, frequencies of being drunk, and parents’ drinking habits.

118

Other measures from the baseline included education level, body mass index, total leisure time, physical 119

activity, imprisonment, tobacco smoking, human population laboratory depression scale (HPL), and the use 120

of medication for anxiety, depression, neurosis or psychosis (Table 1). Dietary intakes were assessed with 4- 121

day food recording (Virtanen et al., 2014).

122

123

Non-targeted metabolomics 124

The workflow of the metabolomics analysis has been described in detail elsewhere (Klåvus et al., 2020).

125

Briefly, the analysis order of the serum samples was randomized. Samples were thawed on ice and then 126

100 μL of each sample was mixed by pipette with 400 μL of acetonitrile (LC-MS grade) in the well of a 127

Captiva ND filter plate (Agilent Technologies). The filter plate was centrifuged at 700 rcf at 4°C for 5 128

minutes. Then the protein-free filtrate was collected on a 96-well polypropylene plate. A pooled QC sample 129

was prepared for quality control and injected after every 12 analytical samples.

130

The samples were analyzed with two LC-MS methods. For more hydrophilic compounds, we used the liquid 131

chromatography quadrupole time-of-flight mass spectrometry system (UHPLC-qTOF-MS, Agilent 132

Technologies), which consisted of a 1290 LC system, a Jetstream electrospray ionization (ESI) source, and a 133

6540 UHD accurate-mass qTOF spectrometer. We applied hydrophilic interaction (HILIC) chromatography 134

7

(an Acquity UPLC BEH Amide column, 100 mm × 2.1 mm, 1.7 μm; Waters Corporation) and both positive 135

and negative ionization. For more lipophilic compounds, we utilized TUPLC chromatography and Thermo Q 136

Exactive™ Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Scientific). We used reversed-phase 137

chromatography (RP) column: Zorbax Eclipse XDB‐C18, particle size 1.8 µm, 2.1 × 100 mm (Agilent 138

Technologies) and both positive and negative ionization.

139

We used an open-source software MS-DIAL version 4.00 (Tsugawa et al., 2015) for peak picking, peak 140

alignment and metabolite identification. For peak picking, we used an accurate mass tolerance of 0.008 Da, 141

a minimum peak width of 8, and utilized a minimum peak height of 8000 (HILIC method) or 500000 (RP 142

method). For peak alignment, we limited the accurate mass tolerance of 0.01 Da and retention time 143

tolerance to 0.1 min. Peaks needed to be present in at least 10 % of all samples. This limit was set low to 144

allow detection of molecular features arising from external exposures, e.g. drugs and tobacco. After peak 145

picking and alignment, and before statistical analysis, preprocessing of the metabolomics data was done 146

according to the previously published protocol (Klåvus et al., 2020). In brief, the main preprocessing steps 147

included drift correction, flagging low-quality features and missing value imputation.

148

For molecular features with a false discovery rate (FDR) corrected p-value below 0.05 in the ANOVA, we 149

performed metabolite identification. Identifications were ranked according to the guidelines from Sumner 150

et al. (Sumner et al., 2007). Metabolites in level 1 were matched against mass, retention time, and MS/MS 151

spectra of fragmented ions from the in-house library of chemical standards built using the same instrument 152

and experimental conditions. Level 2 includes metabolites with matching exact mass and MSMS spectra 153

from public libraries (METLIN, Lipidmaps and Human Metabolome DataBase were used) or in the case of 154

lipids, the built-in MS-DIAL library version 4.00. In level 3, only the chemical group of the compound (but 155

not the exact compound) could be identified.

156

157

Statistical analysis 158

8

We used Kruskal-Wallis and Chi² tests to calculate p-values between the study groups for the background 159

variables. Post-hoc analyses of the background variables to compare alcohol-case group to the other study 160

groups were done using Bonferroni’s method.

161

In the metabolomics analysis, we used Welch’s one-way ANOVA to calculate p-values separately for each 162

molecular feature with FDR correction to account for multiple testing. With respect to the molecular 163

features with FDR corrected p-value below 0.05, we performed Welch’s t-test and calculated Cohen’s d- 164

effect sizes to evaluate differences between the study groups. We used partial least sum of squares 165

discriminant analysis (PLS-DA) to evaluate which molecular features best differentiated the study groups 166

from each other.

167

For those metabolites in which there were statistically significant differences between the alcohol-case and 168

alcohol-control groups, but no significant difference between alcohol-control and control-control groups, 169

we performed a logistic regression analysis with study group (alcohol-case vs alcohol-control) as the 170

dependent variable and the molecular feature as a predictor, with self-reported alcohol use and the 171

traditional alcohol biomarker i.e. gamma-glutamyl transferase (GGT) as covariates, to see if the metabolites 172

would offer additional information other than that available via these traditional measures.

173

We used Spearman’s method to perform the post-hoc correlation analyses between significantly altered 174

metabolite levels and the background characteristics measured at the baseline.

175

176

Results 177

Background characteristics of the subjects are shown in table 1. There were no significant differences 178

between the study groups in terms of age at baseline, educational level, body mass index, total leisure time 179

physical activity or social deviance as indicated by imprisonment. However, as expected, there were some 180

significant differences between the control-control group and the other study groups. Significant 181

differences between the alcohol-case and alcohol-control groups were observed in how often they 182

9

reported becoming drunk or having hangovers, with the alcohol-case group reporting higher frequencies.

183

The alcohol-case group had higher levels of GGT when compared to the control groups. The light drinking 184

control-control group reported much lower levels of smoking as compared to the other two study groups at 185

the baseline. The alcohol-case group had on average higher scores in the Human Population Laboratory 186

Depression Scale than the two control groups. There were no significant differences between the study 187

groups in self-reported use of prescription drugs affecting the central-nervous system. In general, the 188

alcohol-case group had the least healthy dietary pattern especially compared to the control-control group, 189

with a higher intake of red meat and lower intakes of whole grains, vegetable margarines and oils, and 190

fruits, berries and vegetables. There were no significant differences in the intake of different food items in 191

the post-hoc analysis between alcohol-case and alcohol-control groups (Table 1).

192

In the metabolomics analysis, we measured a total of 4,347 molecular features (Supplementary table 1). Of 193

these, 862 had raw p-values below 0.05 (217 expected by chance) and 500 molecular features had FDR 194

corrected q-values below 0.05. Of these, 87 molecular features were significantly different between 195

alcohol-cases and alcohol-controls in the post-hoc analysis.

196

The identified and significantly altered metabolites are shown in Table 2. Overall, after correction for 197

multiple testing, statistically significant differences between the study groups were seen in levels of amino 198

acids, glycerophospholipids, steroid hormones, pyrimidines, and fatty acids, in microbiota derived 199

metabolites (e.g. 3-indolepropionic acid), as well as in metabolites from external exposures such as 200

caffeine, scoparone (fruit intake) and cotinine (a nicotine metabolite). Moreover, results from the 201

multivariate PLS-DA models are also shown in table 2 and in supplementary figure 1 they are in line with 202

the univariate analyses.

203

204

Table 2 205

206

10

Of the identified and significantly altered metabolites, the alcohol-case group had lower levels of 207

asparagine, serotonin, sphingomyelin (SM) 36:3 and higher levels of palmitoleic acid (FA 16:1), 208

docosatetraenoic acid (FA 22:4), and tyrosine when compared to both controls matched for self-reported 209

alcohol use and light drinking controls. In contrast, there was no significant difference between the alcohol- 210

control and control-control groups in these metabolites (Table 2). These six metabolites were selected for a 211

logistic regression analysis where self-reported alcohol use and a traditional alcohol biomarker, GGT, were 212

included as covariates to control for baseline alcohol use. In this model, only asparagine (p = 0.0433) and 213

serotonin (p = 0.0467) remained significant. Other selected metabolites (SM 36:3 (p = 0.1725), FA 16:1 (p = 214

0.2920), FA 22:4 (0.4602), and tyrosine (0.6133)), as well as used covariates’ self-reported alcohol use (p = 215

0.9120) and GGT (p = 0.1994), did not reach statistical significance in the logistic regression model.

216

Since serotonin has been associated with depression, we undertook a post-hoc Spearman correlation 217

analysis between measured serotonin levels and HPL score and found a small negative correlation between 218

these two variables (r = -0.13, p = 0.0275).

219

Finally, we calculated the Spearman correlation to investigate how well the now measured levels of 220

metabolites reflected the values measured when the samples had been freshly collected. From the 221

significantly altered metabolites, only levels of palmitoleic acid (FA 16:1) and docosatetraenoic acid (FA 222

22:6) had also been measured at the time of sample collection. For both of these metabolites, there were 223

large and significant correlations between the levels measured now and those measured in the past (FA 224

16:1 r = 0.70, p < 0.0001, and FA 22:6 r = 0.73, p < 0.0001).

225

226

Discussion 227

The main finding of the present study was that the alcohol-case group, who would later develop diseases 228

related to alcohol use, had lower levels of asparagine and serotonin at the baseline, when compared to 229

both controls matched for self-reported alcohol use (alcohol-control group) and light drinking controls 230

(control-control group, Figure 2).

231

11 232

Figure 2 233

234

Asparagine is biosynthesized when oxaloacetate (from the citric acid cycle) is first metabolized to aspartate 235

by aspartate aminotransferase (EC:2.6.1.1) and then to asparagine by asparagine synthase (EC:6.3.5.4). In 236

the first step, glutamate is also converted into α-ketoglutarate, and in the second step, glutamine is also 237

turned into glutamate. Previous studies have also described decreased circulating asparagine levels in 238

pregnant women who drink during pregnancy (Lehikoinen et al., 2018). It has been reported that 239

asparagine is able to prevent the increase in the triglyceride concentrations evoked by acute ethanol 240

administration, at least in rats (Lansford et al., 1962). Low asparagine levels have been linked with reduced 241

hepatic protein synthesis and liver damage (Kamal et al., 2019). Asparagine synthetase is also needed for 242

normal development and function of the brain (Ruzzo et al., 2013). Asparagine does not seem to be 243

catabolized in mammalian cells to make other amino acids, and when glutamine levels are low (which is the 244

case in both the alcohol-case and alcohol-control groups when compared to the control-control group, 245

Table 2), asparagine is needed for protein synthesis (Pavlova et al., 2018). Therefore, low levels of 246

asparagine could be associated with an increased likelihood of developing alcohol-induced damage to 247

organs, like the liver or the brain, an intriguing hypothesis in need of testing. Overall, more research is 248

needed into the possible role of asparagine in alcohol-related diseases.

249

Serotonin modulates many neural systems linked with alcohol use disorder from reinforcement learning to 250

social cognition and decreased serotonergic tone has been associated with heavy alcohol use (Kärkkäinen 251

et al., 2015; Marcinkiewcz et al., 2016). Moreover, drugs affecting the serotonergic system, e.g. the 5-HT3 252

receptor antagonist ondansetron, have been used in the treatment of alcohol use disorder (Johnson et al., 253

2011; Kenna et al., 2014; Marcinkiewcz et al., 2016). Lower circulating serotonin levels have been described 254

in relation to heavy alcohol use, measured both from serum samples as well as from platelets (Bailly et al., 255

1993; Lehikoinen et al., 2018; Pivac et al., 2004). In the present study, there were no significant alterations 256

12

in the levels of the serotonin precursors, tryptophan and phenylalanine, nor in the other main metabolite 257

of tryptophan, kynurenine (Supplementary table 1). There has been some discussion on whether low 258

serotonin levels precede or are caused by heavy alcohol use. Here we provide prospective evidence that 259

low circulating serotonin levels precede an alcohol-related disease diagnosis in middle-aged Finnish males, 260

and that the low serotonin levels are not evident in controls drinking at a similar level. This points to an 261

underlying dysfunction in the serotonin system preceding the health problems that associate with heavy 262

alcohol use.

263

In the present study, high self-reported alcohol use was associated with alterations in different amino acids, 264

glycerophospholipids, steroid hormones, pyrimidines, and fatty acids (Table 2). These results are in line 265

with previous studies showing that alcohol use alters the circulating metabolome in these compound 266

classes (Heikkinen et al., 2019; Jaremek et al., 2013; Lacruz et al., 2016; Lehikoinen et al., 2018; Pallister et 267

al., 2016; van Roekel et al., 2018; Voutilainen and Kärkkäinen, 2019; Würtz et al., 2016; Zheng et al., 2014);

268

it is also known that the levels of compounds in these classes are altered in organs like brain and liver in 269

association with alcohol use disorder (Kärkkäinen et al., 2016; Kashem et al., 2016; Schofield et al., 2017).

270

Moreover, alterations in metabolites associated with gut microbiota, like 3-indolepropionic acid and 271

hippuric acid, were evident between the study groups. These results indicate possible differences between 272

the study groups in the composition of gut microbiota, which is in line with reports of intestinal dysbiosis in 273

association with heavy alcohol use (Leclercq et al., 2014; Temko et al., 2017). Some of the changes are likely 274

from differences in diets. For example, scoparone is found in fruits like sweet orange (Afendi et al., 2012), 275

and the low scoparone levels in the alcohol-case group are possibly explained by their lower intake of fruits, 276

berries and vegetables (Tables 1 and 2).

277

Limitations of the analysis include the long storage time between sample collection and analysis, which 278

could affect the measured metabolome. Previously, it has been reported that long-term storage changes 279

the levels of lipids, fatty acids and amino acids, and for example levels of asparagine seem to increase in 280

plasma samples undergoing a long storage (Wagner-Golbs et al., 2019). However, for the compounds that 281

13

were measured both in the present study and at the time of sample collection, there were large and 282

significant correlations between the measurements. Furthermore, all samples were stored under the same 283

conditions. Therefore, the differences between study groups in the present analysis were most likely 284

present already at the baseline. Moreover, only a limited number of alcohol use biomarkers were at use in 285

the time of baseline, e.g. phosphatidylethanol analyses were not available. It was not possible to analyze 286

this compound now since we did not have the whole blood samples that are needed for its analysis.

287

Furthermore, because of the cohort demographics, we analyzed only samples from middle-aged males.

288

Therefore, the present results might not be generalizable to women nor to individuals with early onset 289

alcoholism, and a separate analysis would need to be done for these populations. Moreover, the statistical 290

power of the present study is limited due to the comparatively small sample sizes. For this reason, we were 291

not able to conduct subgroup analyses (e.g. diagnoses of alcohol dependence vs. diagnoses of alcohol- 292

related organ damage) and this limitation could have restricted our ability to detect small effect size 293

differences between the study groups. The study contains diagnoses from a limited time window meaning 294

that some subjects in control groups could be diagnosed with an alcohol-related disease after the follow-up 295

period. However, the follow-up was long, 29-34 years, and the average time to diagnosis was 13.6 years, 296

with a standard deviation of 9.4, giving us a mean + 2 standard deviations of 32.4 years. Thus, we believe 297

that the overall follow-up was long enough in this sense to ensure the reliability of our results.

298

In conclusion, we observed that the circulating metabolome is different in the group of individuals who 299

later in their life developed alcohol-related diseases, as compared to controls who reported a similar level 300

of alcohol use but would not be diagnosed for alcohol-related diseases during the next 29 to 34 years. The 301

most striking finding was that decreased circulating levels of asparagine and serotonin precede alcohol- 302

related diseases even when controlling for the baseline level of drinking.

303

14 Acknowledgement

304

We thank Miia Reponen for technical assistance with the mass spectrometry analyses, Kimmo Ronkainen 305

for initial construction of alcohol variables and registry linkages, and Ewen MacDonald for proof-reading 306

the English language.

307

This study was funded by the Finnish Foundation of Cardiovascular Research. The authors also want to 308

thank Biocenter Finland and Biocenter Kuopio for supporting the core LC-MS laboratory facility.

309

310

Conflict of interest 311

OK and AK are co-owners of Afekta Technologies Ltd. a company providing metabolomics analysis services.

312

313

Author contributions 314

OK, JV, JK and JR designed the study. AV did data management and linking of the registry data. ML and SA 315

supervised the mass spectrometry analysis. AK did preprocessing of the metabolomics data. OK performed 316

the statistical analysis, compound identification and results interpretation of the metabolomics data. OK 317

wrote the first draft and all authors have commented and approved the final version.

318

15 References

319

Afendi FM, Okada T, Yamazaki M, Hirai-Morita A, Nakamura Y, Nakamura K, Ikeda S, Takahashi

320

H, Altaf-Ul-Amin M, Darusman LK, Saito K, Kanaya S (2012) KNApSAcK family

321

databases: integrated metabolite-plant species databases for multifaceted plant research.

322

Plant Cell Physiol 53:e1.

323

Bailly D, Vignau J, Racadot N, Beuscart R, Servant D, Parquet PJ (1993) Platelet serotonin levels

324

in alcoholic patients: Changes related to physiological and pathological factors. Psychiatry

325

Research 47:57–68.

326

Bajaj JS, Kakiyama G, Zhao D, Takei H, Fagan A, Hylemon P, Zhou H, Pandak WM, Nittono H,

327

Fiehn O, Salzman N, Holtz M, Simpson P, Gavis EA, Heuman DM, Liu R, Kang DJ,

328

Sikaroodi M, Gillevet PM (2017) Continued Alcohol Misuse in Human Cirrhosis is

329

Associated with an Impaired Gut-Liver Axis. Alcohol Clin Exp Res 41:1857–1865.

330

Griswold MG, Fullman N, Hawley C, Arian N, Zimsen SRM, Tymeson HD, Venkateswaran V,

331

Tapp AD, Forouzanfar MH, Salama JS, Abate KH, Abate D, Abay SM, Abbafati C,

332

Abdulkader RS, Abebe Z, Aboyans V, Abrar MM, Acharya P, Adetokunboh OO, Adhikari

333

TB, Adsuar JC, Afarideh M, Agardh EE, Agarwal G, Aghayan SA, Agrawal S, Ahmed MB,

334

Akibu M, Akinyemiju T, Akseer N, Asfoor DHA, Al-Aly Z, Alahdab F, Alam K, Albujeer

335

A, Alene KA, Ali R, Ali SD, Alijanzadeh M, Aljunid SM, Alkerwi A, Allebeck P, Alvis-

336

Guzman N, Amare AT, Aminde LN, Ammar W, Amoako YA, Amul GGH, Andrei CL,

337

Angus C, Ansha MG, Antonio CAT, Aremu O, Ärnlöv J, Artaman A, Aryal KK, Assadi R,

338

Ausloos M, Avila-Burgos L, Avokpaho EF, Awasthi A, Ayele HT, Ayer R, Ayuk TB,

339

Azzopardi PS, Badali H, Badawi A, Banach M, Barker-Collo SL, Barrero LH, Basaleem H,

340

Baye E, Bazargan-Hejazi S, Bedi N, Béjot Y, Belachew AB, Belay SA, Bennett DA,

341

Bensenor IM, Bernabe E, Bernstein RS, Beyene AS, Beyranvand T, Bhaumik S, Bhutta ZA,

342

Biadgo B, Bijani A, Bililign N, Birlik SM, Birungi C, Bizuneh H, Bjerregaard P, Bjørge T,

343

16

Borges G, Bosetti C, Boufous S, Bragazzi NL, Brenner H, Butt ZA, Cahuana-Hurtado L,

344

Calabria B, Campos-Nonato IR, Campuzano JC, Carreras G, Carrero JJ, Carvalho F,

345

Castañeda-Orjuela CA, Castillo Rivas J, Catalá-López F, Chang J-C, Charlson FJ,

346

Chattopadhyay A, Chaturvedi P, Chowdhury R, Christopher DJ, Chung S-C, Ciobanu LG,

347

Claro RM, Conti S, Cousin E, Criqui MH, Dachew BA, Dargan PI, Daryani A, Das Neves J,

348

Davletov K, De Castro F, De Courten B, De Neve J-W, Degenhardt L, Demoz GT, Des

349

Jarlais DC, Dey S, Dhaliwal RS, Dharmaratne SD, Dhimal M, Doku DT, Doyle KE, Dubey

350

M, Dubljanin E, Duncan BB, Ebrahimi H, Edessa D, El Sayed Zaki M, Ermakov SP,

351

Erskine HE, Esteghamati A, Faramarzi M, Farioli A, Faro A, Farvid MS, Farzadfar F,

352

Feigin VL, Felisbino-Mendes MS, Fernandes E, Ferrari AJ, Ferri CP, Fijabi DO, Filip I,

353

Finger JD, Fischer F, Flaxman AD, Franklin RC, Futran ND, Gallus S, Ganji M, Gankpe

354

FG, Gebregergs GB, Gebrehiwot TT, Geleijnse JM, Ghadimi R, Ghandour LA, Ghimire M,

355

Gill PS, Ginawi IA, Giref AZZ, Gona PN, Gopalani SV, Gotay CC, Goulart AC, Greaves F,

356

Grosso G, Guo Y, Gupta R, Gupta R, Gupta V, Gutiérrez RA, Gvs M, Hafezi-Nejad N,

357

Hagos TB, Hailu GB, Hamadeh RR, Hamidi S, Hankey GJ, Harb HL, Harikrishnan S, Haro

358

JM, Hassen HY, Havmoeller R, Hay SI, Heibati B, Henok A, Heredia-Pi I, Hernández-

359

Llanes NF, Herteliu C, Hibstu DTT, Hoogar P, Horita N, Hosgood HD, Hosseini M, Hostiuc

360

M, Hu G, Huang H, Husseini A, Idrisov B, Ileanu BV, Ilesanmi OS, Irvani SSN, Islam

361

SMS, Jackson MD, Jakovljevic M, Jalu MT, Jayatilleke AU, Jha RP, Jonas JB, Jozwiak JJ,

362

Kabir Z, Kadel R, Kahsay A, Kapil U, Kasaeian A, Kassa TDD, Katikireddi SV, Kawakami

363

N, Kebede S, Kefale AT, Keiyoro PN, Kengne AP, Khader Y, Khafaie MA, Khalil IA,

364

Khan MN, Khang Y-H, Khater MM, Khubchandani J, Kim C-I, Kim D, Kim YJ, Kimokoti

365

RW, Kisa A, Kivimäki M, Kochhar S, Kosen S, Koul PA, Koyanagi A, Krishan K, Kuate

366

Defo B, Kucuk Bicer B, Kulkarni VS, Kumar P, Lafranconi A, Lakshmana Balaji A, Lalloo

367

R, Lallukka T, Lam H, Lami FH, Lan Q, Lang JJ, Lansky S, Larsson AO, Latifi A, Leasher

368

17

JL, Lee PH, Leigh J, Leinsalu M, Leung J, Levi M, Li Y, Lim L-L, Linn S, Liu S, Lobato-

369

Cordero A, Lopez AD, Lorkowski S, Lotufo PA, Macarayan ERK, Machado IE, Madotto F,

370

Magdy Abd El Razek H, Magdy Abd El Razek M, Majdan M, Majdzadeh R, Majeed A,

371

Malekzadeh R, Malta DC, Mapoma CC, Martinez-Raga J, Maulik PK, Mazidi M, Mckee M,

372

Mehta V, Meier T, Mekonen T, Meles KG, Melese A, Memiah PTN, Mendoza W, Mengistu

373

DT, Mensah GA, Meretoja TJ, Mezgebe HB, Miazgowski T, Miller TR, Mini G, Mirica A,

374

Mirrakhimov EM, Moazen B, Mohammad KA, Mohammadifard N, Mohammed S, Monasta

375

L, Moraga P, Morawska L, Mousavi SM, Mukhopadhyay S, Musa KI, Naheed A, Naik G,

376

Najafi F, Nangia V, Nansseu JR, Nayak MSDP, Nejjari C, Neupane S, Neupane SP,

377

Ngunjiri JW, Nguyen CT, Nguyen LH, Nguyen TH, Ningrum DNA, Nirayo YL, Noubiap

378

JJ, Ofori-Asenso R, Ogbo FA, Oh I-H, Oladimeji O, Olagunju AT, Olivares PR, Olusanya

379

BO, Olusanya JO, Oommen AM, Oren E, Orpana HM, Ortega-Altamirano DDV, Ortiz JR,

380

Ota E, Owolabi MO, Oyekale AS, P A M, Pana A, Park E-K, Parry CDH, Parsian H, Patle

381

A, Patton GC, Paudel D, Petzold M, Phillips MR, Pillay JD, Postma MJ, Pourmalek F,

382

Prabhakaran D, Qorbani M, Radfar A, Rafay A, Rafiei A, Rahim F, Rahimi-Movaghar A,

383

Rahman M, Rahman MA, Rai RK, Rajsic S, Raju SB, Ram U, Rana SM, Ranabhat CL,

384

Rawaf DL, Rawaf S, Reiner RC, Reis C, Renzaho AMN, Rezai MS, Roever L, Ronfani L,

385

Room R, Roshandel G, Rostami A, Roth GA, Roy A, Sabde YD, Saddik B, Safiri S,

386

Sahebkar A, Salama JS, Saleem Z, Salomon JA, Salvi SS, Sanabria J, Sanchez-Niño MD,

387

Santomauro DF, Santos IS, Santric Milicevic MMM, Sarker AR, Sarmiento-Suárez R,

388

Sarrafzadegan N, Sartorius B, Satpathy M, Sawhney M, Saxena S, Saylan M, Schaub MP,

389

Schmidt MI, Schneider IJC, Schöttker B, Schutte AE, Schwendicke F, Sepanlou SG, Shaikh

390

MA, Sharif M, She J, Sheikh A, Shen J, Shiferaw MS, Shigematsu M, Shiri R, Shishani K,

391

Shiue I, Shukla SR, Sigfusdottir ID, Silva DAS, Silva NTD, Silveira DGA, Sinha DN, Sitas

392

F, Soares Filho AM, Soofi M, Sorensen RJD, Soriano JB, Sreeramareddy CT, Steckling N,

393

18

Stein DJ, Sufiyan MB, Sur PJ, Sykes BL, Tabarés-Seisdedos R, Tabuchi T, Tavakkoli M,

394

Tehrani-Banihashemi A, Tekle MG, Thapa S, Thomas N, Topor-Madry R, Topouzis F, Tran

395

BX, Troeger CE, Truelsen TC, Tsilimparis N, Tyrovolas S, Ukwaja KN, Ullah I, Uthman

396

OA, Valdez PR, Van Boven JFM, Vasankari TJ, Venketasubramanian N, Violante FS,

397

Vladimirov SK, Vlassov V, Vollset SE, Vos T, Wagnew FWS, Waheed Y, Wang Y-P,

398

Weiderpass E, Weldegebreal F, Weldegwergs KG, Werdecker A, Westerman R, Whiteford

399

HA, Widecka J, Wijeratne T, Wyper GMA, Xu G, Yamada T, Yano Y, Ye P, Yimer EM,

400

Yip P, Yirsaw BD, Yisma E, Yonemoto N, Yoon S-J, Yotebieng M, Younis MZ, Zachariah

401

G, Zaidi Z, Zamani M, Zhang X, Zodpey S, Mokdad AH, Naghavi M, Murray CJL,

402

Gakidou E (2018) Alcohol use and burden for 195 countries and territories, 1990–2016: a

403

systematic analysis for the Global Burden of Disease Study 2016. The Lancet 392:1015–

404

1035.

405

Harada S, Takebayashi T, Kurihara A, Akiyama M, Suzuki A, Hatakeyama Y, Sugiyama D,

406

Kuwabara K, Takeuchi A, Okamura T, Nishiwaki Y, Tanaka T, Hirayama A, Sugimoto M,

407

Soga T, Tomita M (2016) Metabolomic profiling reveals novel biomarkers of alcohol intake

408

and alcohol-induced liver injury in community-dwelling men. Environ Health Prev Med

409

21:18–26.

410

Heikkinen N, Kärkkäinen O, Laukkanen E, Kekkonen V, Kaarre O, Kivimäki P, Könönen M,

411

Velagapudi V, Nandania J, Lehto SM, Niskanen E, Vanninen R, Tolmunen T (2019)

412

Changes in the serum metabolite profile correlate with decreased brain gray matter volume

413

in moderate-to-heavy drinking young adults. Alcohol 75:89–97.

414

Hinton DJ, Vázquez MS, Geske JR, Hitschfeld MJ, Ho AMC, Karpyak VM, Biernacka JM, Choi

415

D-S (2017) Metabolomics biomarkers to predict acamprosate treatment response in alcohol-

416

dependent subjects. Sci Rep 7:2496.

417

19

Jaremek M, Yu Z, Mangino M, Mittelstrass K, Prehn C, Singmann P, Xu T, Dahmen N,

418

Weinberger KM, Suhre K, Peters A, Döring A, Hauner H, Adamski J, Illig T, Spector TD,

419

Wang-Sattler R (2013) Alcohol-induced metabolomic differences in humans. Transl

420

Psychiatry 3:e276–e276.

421

Johnson BA, Ait-Daoud N, Seneviratne C, Roache JD, Javors MA, Wang X-Q, Liu L, Penberthy

422

JK, DiClemente CC, Li MD (2011) Pharmacogenetic Approach at the Serotonin Transporter

423

Gene as a Method of Reducing the Severity of Alcohol Drinking. AJP 168:265–275.

424

Kamal N, Koh C, Samala N, Fontana RJ, Stolz A, Durazo F, Hayashi PH, Phillips E, Wang T,

425

Hoofnagle JH, for the Drug-Induced Liver Injury Network (2019) Asparaginase-induced

426

hepatotoxicity: rapid development of cholestasis and hepatic steatosis. Hepatol Int 13:641–

427

648.

428

Kärkkäinen O, Häkkinen MR, Auriola S, Kautiainen H, Tiihonen J, Storvik M (2016) Increased

429

steroid hormone dehydroepiandrosterone and pregnenolone levels in post-mortem brain

430

samples of alcoholics. Alcohol 52:63–70.

431

Kärkkäinen O, Laukkanen V, Haukijärvi T, Kautiainen H, Tiihonen J, Storvik M (2015) Lower

432

[3H]Citalopram Binding in Brain Areas Related to Social Cognition in Alcoholics. Alcohol

433

and Alcoholism 50:46–50.

434

Kashem MA, Ahmed S, Sultana N, Ahmed EU, Pickford R, Rae C, Šerý O, McGregor IS, Balcar

435

VJ (2016) Metabolomics of Neurotransmitters and Related Metabolites in Post-Mortem

436

Tissue from the Dorsal and Ventral Striatum of Alcoholic Human Brain. Neurochem Res

437

41:385–397.

438

Kauhanen J, Julkunen J, Salonen JT (1992) Coping with Inner Feelings and Stress: Heavy Alcohol

439

Use in the Context of Alexithymia. Behavioral Medicine 18:121–126.

440

Kenna GA, Zywiak WH, Swift RM, McGeary JE, Clifford JS, Shoaff JR, Vuittonet C, Fricchione

441

S, Brickley M, Beaucage K, Haass-Koffler CL, Leggio L (2014) Ondansetron Reduces

442

20

Naturalistic Drinking in Nontreatment-Seeking Alcohol-Dependent Individuals with the LL

443

5′-HTTLPR Genotype: A Laboratory Study. Alcohol Clin Exp Res 38:1567–1574.

444

Klåvus A, Kokla M, Noerman S, Koistinen VM, Tuomainen M, Zarei I, Meuronen T, Häkkinen

445

MR, Rummukainen S, Farizah Babu A, Sallinen T, Kärkkäinen O, Paananen J, Broadhurst

446

D, Brunius C, Hanhineva K (2020) “notame”: Workflow for Non-Targeted LC–MS

447

Metabolic Profiling. Metabolites 10:135.

448

Lacruz ME, Kluttig A, Tiller D, Medenwald D, Giegling I, Rujescu D, Prehn C, Adamski J, Frantz

449

S, Greiser KH, Emeny RT, Kastenmüller G, Haerting J (2016) Cardiovascular Risk Factors

450

Associated With Blood Metabolite Concentrations and Their Alterations During a 4-Year

451

Period in a Population-Based Cohort. Circ Cardiovasc Genet 9:487–494.

452

Lansford EM, Hill ID, Shive W (1962) Effects of Asparagine and other Related Nutritional

453

Supplements upon Alcohol-Induced Rat Liver Triglyceride Elevation. The Journal of

454

Nutrition 78:219–222.

455

Leclercq S, Matamoros S, Cani PD, Neyrinck AM, Jamar F, Stärkel P, Windey K, Tremaroli V,

456

Bäckhed F, Verbeke K, de Timary P, Delzenne NM (2014) Intestinal permeability, gut-

457

bacterial dysbiosis, and behavioral markers of alcohol-dependence severity. Proc Natl Acad

458

Sci USA 111:E4485–E4493.

459

Lehikoinen AI, Kärkkäinen OK, Lehtonen MAS, Auriola SOK, Hanhineva KJ, Heinonen ST (2018)

460

Alcohol and substance use are associated with altered metabolome in the first trimester

461

serum samples of pregnant mothers. European Journal of Obstetrics & Gynecology and

462

Reproductive Biology 223:79–84.

463

Lian J, Liu W, Hao S, Guo Y, Huang H, Chen D, Xie Q, Pan X, Xu W, Yuan W, Li L, Huang J

464

(2011) A serum metabonomic study on the difference between alcohol- and HBV-induced

465

liver cirrhosis by ultraperformance liquid chromatography coupled to mass spectrometry

466

plus quadrupole time-of-flight mass spectrometry. Chin Med J 124:1367–1373.

467

21

Marcinkiewcz CA, Lowery-Gionta EG, Kash TL (2016) Serotonin’s Complex Role in Alcoholism:

468

Implications for Treatment and Future Research. Alcohol Clin Exp Res 40:1192–1201.

469

Nam HW, Karpyak VM, Hinton DJ, Geske JR, Ho AMC, Prieto ML, Biernacka JM, Frye MA,

470

Weinshilboum RM, Choi D-S (2015) Elevated baseline serum glutamate as a

471

pharmacometabolomic biomarker for acamprosate treatment outcome in alcohol-dependent

472

subjects. Transl Psychiatry 5:e621–e621.

473

Pallister T, Jennings A, Mohney RP, Yarand D, Mangino M, Cassidy A, MacGregor A, Spector

474

TD, Menni C (2016) Characterizing Blood Metabolomics Profiles Associated with Self-

475

Reported Food Intakes in Female Twins. PLoS ONE 11:e0158568.

476

Pavlova NN, Hui S, Ghergurovich JM, Fan J, Intlekofer AM, White RM, Rabinowitz JD,

477

Thompson CB, Zhang J (2018) As Extracellular Glutamine Levels Decline, Asparagine

478

Becomes an Essential Amino Acid. Cell Metabolism 27:428-438.e5.

479

Pivac N, Mück-Šeler D, Mustapić M, Nenadić-Šviglin K, Kozarić-Kovačić D (2004) Platelet

480

serotonin concentration in alcoholic subjects. Life Sciences 76:521–531.

481

Ruzzo EK, Capo-Chichi J-M, Ben-Zeev B, Chitayat D, Mao H, Pappas AL, Hitomi Y, Lu Y-F, Yao

482

X, Hamdan FF, Pelak K, Reznik-Wolf H, Bar-Joseph I, Oz-Levi D, Lev D, Lerman-Sagie T,

483

Leshinsky-Silver E, Anikster Y, Ben-Asher E, Olender T, Colleaux L, Décarie J-C, Blaser

484

S, Banwell B, Joshi RB, He X-P, Patry L, Silver RJ, Dobrzeniecka S, Islam MS, Hasnat A,

485

Samuels ME, Aryal DK, Rodriguiz RM, Jiang Y, Wetsel WC, McNamara JO, Rouleau GA,

486

Silver DL, Lancet D, Pras E, Mitchell GA, Michaud JL, Goldstein DB (2013) Deficiency of

487

Asparagine Synthetase Causes Congenital Microcephaly and a Progressive Form of

488

Encephalopathy. Neuron 80:429–441.

489

Salonen JT, Salonen R, Seppänen K, Rauramaa R, Tuomilehto J (1991) HDL, HDL2, and HDL3

490

subfractions, and the risk of acute myocardial infarction. A prospective population study in

491

eastern Finnish men. Circulation 84:129–139.

492

22

Schofield Z, Reed MA, Newsome PN, Adams DH, Günther UL, Lalor PF (2017) Changes in

493

human hepatic metabolism in steatosis and cirrhosis. WJG 23:2685.

494

Shen Q, Heikkinen N, Kärkkäinen O, Gröhn H, Könönen M, Liu Y, Kaarre O, Zhang Z, Tan C,

495

Tolmunen T, Vanninen R (2019) Effects of long-term adolescent alcohol consumption on

496

white matter integrity and their correlations with metabolic alterations. Psychiatry Research:

497

Neuroimaging 294:111003.

498

Sumner LW, Amberg A, Barrett D, Beale MH, Beger R, Daykin CA, Fan TW-M, Fiehn O,

499

Goodacre R, Griffin JL, Hankemeier T, Hardy N, Harnly J, Higashi R, Kopka J, Lane AN,

500

Lindon JC, Marriott P, Nicholls AW, Reily MD, Thaden JJ, Viant MR (2007) Proposed

501

minimum reporting standards for chemical analysis: Chemical Analysis Working Group

502

(CAWG) Metabolomics Standards Initiative (MSI). Metabolomics 3:211–221.

503

Temko JE, Bouhlal S, Farokhnia M, Lee MR, Cryan JF, Leggio L (2017) The Microbiota, the Gut

504

and the Brain in Eating and Alcohol Use Disorders: A ‘Ménage à Trois’? Alcohol and

505

Alcoholism 52:403–413.

506

Tsugawa H, Cajka T, Kind T, Ma Y, Higgins B, Ikeda K, Kanazawa M, VanderGheynst J, Fiehn O,

507

Arita M (2015) MS-DIAL: data-independent MS/MS deconvolution for comprehensive

508

metabolome analysis. Nat Methods 12:523–526.

509

van Roekel E, Trijsburg L, Assi N, Carayol M, Achaintre D, Murphy N, Rinaldi S, Schmidt J,

510

Stepien M, Kaaks R, Kühn T, Boeing H, Iqbal K, Palli D, Krogh V, Tumino R, Ricceri F,

511

Panico S, Peeters P, Bueno-de-Mesquita B, Ardanaz E, Lujan-Barroso L, Quirós J, Huerta J,

512

Molina-Portillo E, Dorronsoro M, Tsilidis K, Riboli E, Rostgaard-Hansen A, Tjønneland A,

513

Overvad K, Weiderpass E, Boutron-Ruault M-C, Severi G, Trichopoulou A, Karakatsani A,

514

Kotanidou A, Håkansson A, Malm J, Weijenberg M, Gunter M, Jenab M, Johansson M,

515

Travis R, Scalbert A, Ferrari P (2018) Circulating Metabolites Associated with Alcohol

516

23

Intake in the European Prospective Investigation into Cancer and Nutrition Cohort.

517

Nutrients 10:654.

518

Virtanen JK, Mursu J, Tuomainen T-P, Voutilainen S (2014) Dietary Fatty Acids and Risk of

519

Coronary Heart Disease in Men: The Kuopio Ischemic Heart Disease Risk Factor Study.

520

Arterioscler Thromb Vasc Biol 34:2679–2687.

521

Voutilainen T, Kärkkäinen O (2019) Changes in the Human Metabolome Associated With Alcohol

522

Use: A Review. Alcohol and Alcoholism 54:225–234.

523

Wagner-Golbs A, Neuber S, Kamlage B, Christiansen N, Bethan B, Rennefahrt U, Schatz P, Lind L

524

(2019) Effects of Long-Term Storage at −80 °C on the Human Plasma Metabolome.

525

Metabolites 9:99.

526

World Health Organization (Ed.) (2010) Global strategy to reduce the harmful use of alcohol.

527

Geneva, World Health Organization.

528

Würtz P, Cook S, Wang Q, Tiainen M, Tynkkynen T, Kangas AJ, Soininen P, Laitinen J, Viikari J,

529

Kähönen M, Lehtimäki T, Perola M, Blankenberg S, Zeller T, Männistö S, Salomaa V,

530

Järvelin M-R, Raitakari OT, Ala-Korpela M, Leon DA (2016) Metabolic profiling of

531

alcohol consumption in 9778 young adults. Int J Epidemiol 45:1493–1506.

532

Zheng Y, Yu B, Alexander D, Steffen LM, Nettleton JA, Boerwinkle E (2014) Metabolomic

533

patterns and alcohol consumption in African Americans in the Atherosclerosis Risk in

534

Communities Study. Am J Clin Nutr 99:1470–1478.

535 536 537

24 Figure Legends:

538

539

Figure 1: Simple schema of the study flow. The mean follow-up before the diagnosis of an alcohol-related 540

disease was 13.6 years (SD = 9.4 years). We excluded subjects who already had an alcohol-related diagnosis 541

at baseline or received one during the first year of follow-up (wash-out period).

542

543

25 544

Figure 2: Decreased asparagine and serotonin levels precede diagnosis of alcohol-related disease.

545

Alcohol-case group (those who were diagnosed with alcohol-related diseases during the follow-up) had 546

significantly decreased levels of asparagine and serotonin when compared to either alcohol-control group 547

(who reported similar levels of alcohol use as the alcohol-case group at baseline) or control-control group 548

(who reported only light or no drinking). Mean ion abundance and 95% confidence intervals are shown.

549

False discovery rate was used to correct for multiple testing and corrected q-values are shown for the 550

Welch’s ANOVA comparisons (asparagine q = 0.0097, and serotonin q = 0.0154). In the post-hoc analysis 551

using Welch’s t-test, there were significant differences identified between the alcohol-case group and both 552

alcohol-control (asparagine p = 0.0012, d = -0.48; serotonin p = 0.0030, d = -0.45) and control-control group 553

(asparagine p = 0.0012, d = -0.49; serotonin p = 0.0026, d = -0.46). In a logistic regression model, where self- 554

reported alcohol use and measurements of gamma-glutamyl transferase were used as covariates, both 555

asparagine (p = 0.0433) and serotonin (p = 0.0467) remained statistically significant between the alcohol- 556

case and alcohol-control groups.

557

26 Table 1: Background information

Alcohol-case Alcohol-control Control-control p

Number of subjects 92 92 90

Age at the baseline (years; mean±SD) 53.4 ±4.5 53.3 ±4.4 53.2 ±4.4 0.8410^a

Education level (n, %) 0.9994^b

Part of elementary school 9 10 % 9 10 % 9 10 %

Elementary school 48 52 % 50 54 % 50 56 %

Junior high school and/or vocational school 30 33 % 28 30 % 26 29 %

Senior high school or above 5 5 % 5 5 % 5 6 %

Body mass index 27.5 ±4.2 27.2 ±3.6 27.1 ±3.7 0.8172^a

Total leisure time physical activity (hours/year) 375 ±353 349 ±309 397 ±329 0.5862^a

Imprisonment (n, %) 0.1630^b

Never 85 92 % 92 100 % 88 98 %

Yes, during past 12 months 1 1 % 0 0 % 0 0 %

Yes, 1-10 years ago 3 3 % 0 0 % 1 1 %

Yes, more than 10 years ago 3 3 % 0 0 % 1 1 %

Self-reported alcohol use (grams/week; mean ±SD) 188 ±174.6 158.5 ±137.9 0.0*** ±0.1 <0.0001^a

Drink preference (n, %) <0.0001^b

Abstainer 0 0 % 0 0 % 84 93 %

Beer 20 22 % 23 25 % 1 1 %

Wine 2 2 % 1 1 % 1 1 %

Strong wine 5 5 % 7 8 % 0 0 %

Liquor 65 71 % 61 66 % 4 4 %

How many times hungover /year (mean ±SD) 14.5 ±21.9 8.7* ±15.6 0.0*** ±0.0 <0.0001^a How many times drunk /year (mean ±SD) 23.5 ±28.3 14.0** ±19.0 0.0*** ±0.0 <0.0001^a Binge drinking (When you drink, how many drinks do you drink at a time?, drink = 12 g of alcohol, n, %) <0.0001^b

0-2 drinks 13 14 % 16 17 % 90 100 %

3-6 drinks 36 39 % 44 48 % 0 0 %

>7 drinks 43 47 % 32 35 % 0 0 %

Parents’ drinking habits (0 = teetotaler, 5 = drank a lot of alcohol)

Father 1.45 ±1.37 1.63 ±1.25 1 ±1.37 0.0979^a

Mother 0.18 ±0.41 0.27 ±0.72 0.06 ±0.27 0.0119^a

27

gamma- Glutamyl transferase (GGT, mean±SD) 69.9 ±79.8 40.2** ±60.0 21.1*** ±12.3 <0.0001^a First alcohol-related diagnosis during follow-up (n, %)^c

Alcohol dependence 29 32% - - - -

Alcohol abuse 21 23% - - - -

Alcoholic cirrhosis of liver/hepatitis/hepatic failure 10 11% - - - -

Alcohol-induced acute/chronic pancreatitis 8 9% - - - -

Other alcohol-related diagnoses 24 26% - - - -

Smoker (n, %) <0.0001^b

Current 39 42 % 45 49 % 13 14 %

Previous 43 47 % 32 35 % 29 32 %

Never 10 11 % 15 16 % 48 53 %

Cigarettes / year of smoking (mean ±SD) 248 ±389 284 ±392 72** ±205 <0.0001^a

HPL Depression Scale (mean ±SD) 3.1 ±2.6 2.1** ±2.0 2.0** ±2.2 <0.0043^a

Using medication for (n, %)

Anxiety 3 3 % 1 1 % 3 3 % 0.5489^b

Depression 4 4 % 0 0 % 2 2 % 0.1313^b

Neurosis 3 3 % 1 1 % 3 3 % 0.5489^b

Psychosis 9 9 % 3 3 % 3 3 % 0.0834^b

Food intakes (g/day; mean ±SD)

Red meat 156 ±77 149 ±74 128* ±65 0.0290^a

Dairy 666 ±325 697 ±379 785 ±361 0.0299^a

Butter 36 ±30 30 ±23 39 ±32 0.2945^a

Fish 69 ±55 67 ±67 46 ±66 0.0008^a

Whole grains 142 ±85 139 ±61 174* ±80 0.0035^a

Vegetable margarines and oils 15 ±16 17 ±16 21* ±18 0.0087^a

Fruits, berries, vegetables 161 ±113 212 ±139 289*** ±206 <0.0001^a

Potatoes 152 ±81 137 ±76 184 ±111 0.0035^a

HPL, Human Population Laboratory Depression Scale; ^a, Kruskal-Wallis test; ^b, Chi-squared; ^c, Diagnosis codes from ICD-9 and ICD-8 grouped with closest match in the ICD-10 and diagnoses related to similar organs grouped (only the first alcohol-related diagnosis reported, other diagnoses possible during follow-up); *, p < 0.05; **, p < 0.01; ***, p < 0.001 in the post-hoc comparison with the alcohol-case group.

28

Table 2: Identified and significantly altered metabolites (ascending order based on p-value in the alcohol-case vs alcohol-control comparison).

Alcohol-case vs Alcohol-control Alcohol-case vs Control-control Alcohol-control vs Control-control

95% CI for d 95% CI for d 95% CI for d

Metabolite FDR q d^a lower upper p VIP d^b lower upper p VIP d^c lower upper p VIP

Asparagine 0.0097 -0.48 -0.78 -0.19 0.0012 2.33 -0.49 -0.78 -0.19 0.0012 1.31 0.00 -0.29 0.29 0.9936 1.16 Serotonin 0.0154 -0.45 -0.74 -0.15 0.0030 2.26 -0.46 -0.75 -0.16 0.0026 1.32 -0.01 -0.30 0.29 0.9703 0.37 PC 33:2 (15:0_18:2) 1.32E-10 -0.44 -0.73 -0.14 0.0036 2.59 -1.26 -1.58 -0.94 8.30E-15 2.98 -0.89 -1.19 -0.58 1.38E-08 2.54 FA 16:1 0.0031 0.43 0.13 0.72 0.0050 2.23 0.62 0.33 0.92 4.25E-05 1.57 0.29 0.00 0.58 0.0560 0.94 Pyroglutamic acid 2.66E-06 -0.38 -0.67 -0.09 0.0104 1.77 -0.95 -1.26 -0.65 1.80E-09 1.41 -0.62 -0.92 -0.33 5.12E-05 1.20 Caffeine 0.0395 0.40 0.11 0.69 0.0111 2.03 0.05 -0.24 0.34 0.7203 2.24 -0.39 -0.68 -0.10 0.0121 1.83 Scoparone 0.0007 -0.38 -0.67 -0.09 0.0118 1.91 -0.70 -1.00 -0.40 6.64E-06 0.42 -0.33 -0.62 -0.03 0.0298 1.54 LysoPC 16:1 0.0008 0.37 0.07 0.66 0.0166 1.90 0.69 0.40 0.99 1.71E-05 1.95 0.39 0.10 0.69 0.0093 0.99 Tyrosine 0.0256 0.35 0.06 0.64 0.0201 1.90 0.52 0.23 0.82 0.0007 1.60 0.15 -0.14 0.44 0.3292 1.23 Glu-Val 0.0003 -0.34 -0.63 -0.05 0.0213 1.44 -0.74 -1.04 -0.44 1.58E-06 1.20 -0.45 -0.74 -0.15 0.0033 0.67 FA 22:4 0.0470 0.33 0.04 0.62 0.0262 1.81 0.46 0.16 0.75 0.0024 1.88 0.18 -0.11 0.48 0.2217 1.80 gamma-Glutamylleucine 1.36E-04 -0.33 -0.62 -0.04 0.0263 2.01 -0.77 -1.07 -0.47 6.35E-07 1.28 -0.50 -0.79 -0.20 0.0010 0.90 PC 15:0_HODE 9.99E-08 -0.32 -0.61 -0.03 0.0313 1.71 -1.05 -1.36 -0.74 4.00E-11 1.87 -0.70 -1.00 -0.40 4.70E-06 1.85 SM 36:3 0.0062 -0.32 -0.61 -0.03 0.0317 1.96 -0.59 -0.88 -0.29 1.11E-04 2.53 -0.23 -0.52 0.06 0.1224 1.96 Cortisone 0.00001 0.31 0.01 0.60 0.0397 1.68 0.88 0.58 1.19 1.33E-08 1.55 0.54 0.25 0.84 0.0003 0.77 PC 33:2 1.07E-08 -0.31 -0.60 -0.01 0.0403 1.34 -1.09 -1.40 -0.78 6.46E-12 2.03 -0.88 -1.18 -0.57 2.24E-08 1.68 Cotinine 0.0060 -0.30 -0.60 -0.01 0.0409 1.99 0.29 -0.01 0.58 0.0570 2.75 0.59 0.30 0.89 1.16E-04 2.56 Hippuric acid 0.0008 -0.28 -0.57 0.01 0.0565 1.75 -0.70 -1.00 -0.40 5.54E-06 0.88 -0.42 -0.71 -0.13 0.0053 2.12 PC 32:1 (16:0_16:1) 0.0018 0.29 0.00 0.58 0.0585 1.27 0.64 0.35 0.94 1.01E-04 1.71 0.39 0.09 0.68 0.0122 1.43 LysoPC 15:0 6.71E-07 -0.28 -0.57 0.01 0.0604 1.76 -0.98 -1.29 -0.67 5.73E-10 1.48 -0.71 -1.01 -0.41 3.62E-06 1.12 L-Glutamine 0.0012 -0.28 -0.57 0.01 0.0640 1.55 -0.67 -0.97 -0.38 1.01E-05 2.56 -0.33 -0.63 -0.04 0.0256 2.28 SM 32:2 0.0004 -0.27 -0.56 0.02 0.0686 1.41 -0.73 -1.03 -0.43 2.27E-06 1.64 -0.49 -0.78 -0.19 0.0012 1.09 AC 12:1 0.0152 0.27 -0.02 0.56 0.0691 1.53 0.54 0.25 0.84 0.0003 2.10 0.27 -0.02 0.56 0.0672 1.95 PC 35:2 7.80E-06 -0.27 -0.56 0.02 0.0735 1.52 -0.88 -1.18 -0.57 1.60E-08 1.41 -0.70 -1.00 -0.40 6.09E-06 1.02 PC 32:0 0.0446 0.26 -0.03 0.55 0.0861 1.88 0.48 0.18 0.77 0.0016 2.32 0.24 -0.05 0.53 0.1054 2.07 AC 16:1 0.0151 0.24 -0.05 0.53 0.1038 0.93 0.54 0.24 0.83 0.0004 1.27 0.30 0.01 0.59 0.0431 1.03 AC 4:0 0.0495 -0.24 -0.53 0.05 0.1124 1.51 -0.48 -0.77 -0.18 0.0020 1.55 -0.28 -0.57 0.01 0.0663 1.10 AC 18:1 0.0245 0.22 -0.07 0.51 0.1372 1.14 0.51 0.21 0.80 0.0008 1.37 0.30 0.00 0.59 0.0479 1.63