The Effect of Ethanol on Human Brain Metabolites Longitudinally Characterized by Proton MR Spectroscopy

Abstract

The effect ethanol exerts on the human brain has not yet been addressed by longitudinal magnetic resonance (MR) spectroscopic experiments. Therefore, we longitudinally characterized cerebral metabolite changes in 15 healthy individuals by proton magnetic resonance spectroscopy (¹H-MRS) subsequent to the ingestion of a standard beverage (mean peak blood alcohol concentration (BAC): 51.43 ± 10.27 mg/dL). Each participant was examined before, over 93.71 ± 11.17 mins immediately after and 726.36 ± 94.96 mins (12.11 ± 1.58 h) past per os alcohol exposure. Fronto-mesial and cerebellar ethanol concentrations over time were similar as determined by the LCModel analysis of spectral data. Alcohol-induced changes of fronto-mesial creatine, choline, glucose, inositol and aspartate levels for 5.79 ± 2.94 mins upon ingestion as well as cerebellar choline and inositol levels for 8.64 ± 2.98 mins past exposure. Closely associated with ethanol concentrations, supratentorial creatine, choline, inositol and aspartate levels decreased after ethanol administration, whereas glucose levels increased. Similarly, infratentorial choline and inositol concentrations were negatively correlated with ethanol levels over time. There were no changes in N-acetyl-aspartate levels upon alcohol exposure. Furthermore, no influence of ethanol on brain water integrals was detected. Ethanol consumption may directly increase oxidative stress and the neuronal vulnerability to it. In addition, our results are compatible with ethanol-induced cell membrane modifications and alternative energy substrate usage upon alcohol exposure.

Introduction

Ethanol (Eth) consumption may result in a broad variety of cerebral alterations. In particular, Eth is known to affect brain metabolism. In vivo characterization of brain metabolites can readily be performed by proton magnetic resonance spectroscopy (¹H-MRS). Unlike aspartate (Asp) and glucose (Glc), at 1.5 T, cerebral metabolite levels of Eth (Mendelson et al, 1990), N-acetyl-aspartate (NAA), choline-containing compounds (Cho), inositol (Ins) and total creatine (Cr) can be reliably assessed.

In humans, various ¹H-MRS studies have explored chronic effects of Eth on cerebral metabolism by comparing alcoholic versus healthy subjects, light versus heavy drinkers or alcoholic subjects under certain conditions, for example, abstinence (Bartsch et al, 2007; Bendszus et al, 2001; Bloomer et al, 2004; Ende et al, 2005; Meyerhoff et al, 2004; Parks et al, 2002; Schweinsburg et al, 2003, 2001; Seitz et al, 1999). Presently, there are no data on acute Eth effects on cerebral metabolism of healthy subjects available.

Regarding the effects of chronic alcohol consumption on human brain metabolism, most investigations report reduced levels of NAA (Bartsch et al, 2007; Bendszus et al, 2001; Bloomer et al, 2004; Meyerhoff et al, 2004; Schweinsburg et al, 2003) and Cho (Bartsch et al, 2007; Bendszus et al, 2001; Bloomer et al, 2004; Ende et al, 2005; Parks et al, 2002) in alcoholics compared with those in healthy subjects and alcoholics before and after abstinence, respectively. Findings of Eth effects on cerebral Ins levels are highly ambiguous ranging from reduction to elevation (Meyerhoff et al, 2004; Parks et al, 2002; Schweinsburg et al, 2001; Seitz et al, 1999). Most studies have revealed no effects of Eth on cerebral Cr levels (Parks et al, 2002; Seitz et al, 1999).

In rat brain, MR spectroscopic findings of acute metabolite changes after intravenous, intragastrical and intraperitoneal Eth application revealed stable (Adalsteinsson et al, 2006) as well as elevated NAA levels (Hirakawa et al, 1994), reduced Cho and stable Cr concentrations (Adalsteinsson et al, 2006). For chronic Eth effects, NAA is referred to as stable (Braunova et al, 2000), Cho as stable (Braunova et al, 2000) or varying over time (Lee et al, 2003), Ins as decreased and Cr as decreased (Braunova et al, 2000) or stable (Lee et al, 2003).

The purpose of this study was to longitudinally delineate the acute impact of Eth on cerebral metabolites in human brain. Emphasis was placed on the assessment of water integrals to ensure that alterations of metabolite levels upon Eth consumption were not feigned by shifts in brain water content. For further reduction of false-positive findings, a specific cascade of statistical analyses was applied for every metabolite concentration time series.

On the basis of the findings of the Eth experiment, we performed a separate Glc infusion experiment to validate the capability of the applied MRS analysis method (LCModel) to extract even those metabolites from the spectral data set, which are challenging to detect at 1.5 T like Glc.

Materials and methods

The study was approved by the Medical Ethics Committee (Faculty of Clinical Medicine, University of Wuerzburg, Germany), and all participants gave written informed consent before enrollment. Volunteers were granted allowance for participating in the study.

Participants

Eth experiment: We studied 15 healthy subjects (7 females, 8 males; age 23.9 ± 4.6 years, range 20 to 38 years; body weight 71.9 ± 9.8 kg, range 60 to 95 kg). Female (66.7 ± 7.8 kg) and male (76.5 ± 9.4 kg) participants differed in body weight (two-sided Student's t-test, P = 0.049), but not in age. Exclusion criteria were current or past substance abuse other than nicotine, any other psychiatric disorder, neurologic and medical illness. Of 15 subjects, 13 consumed 3.25 ± 1.48 drinks at 1.69 ± 0.46 days/week for 7.75 ± 2.96 years and 2 of 15 abstained from alcohol for entire recollection. A drink was defined as containing 13 g of pure alcohol. Of 15 healthy subjects, 13 were light smokers (3.67 ± 2.52 cigarettes/day at 4 ± 3.00 days/week; 8.33 ± 1.53 pack-years) and 12 of 15 participants were non-smokers. Screening for exclusion criteria included a detailed collection of past medical history and the CAGE questionnaire (Mayfield et al, 1974), a reliable tool in diagnosing alcohol-related disorders. All subjects had normal CAGE questionnaire ratings and were free of medication. The subjects were asked not to drink alcohol 24 h before the study and caffeine-containing beverages 4 h before the study and not to eat fatty meals at the day of study.

Glc experiment: Five healthy subjects (two females and three males; age 26.2 ± 3.4 years, range 22 to 28 years; body weight 77.9 ± 17.94 kg, range 60 to 108 kg) participated in the Glc infusion experiment. Exclusion criteria were the same as those applied in the Eth experiment. Two of five participants were light smokers (4 ± 1.41 cigarettes/day at 4 ± 4.23 days/week; 7.5 ± 3.54 pack-years) and three of five healthy subjects were non-smokers. All subjects were free of medication. Eight hours before the study, participants had to start fasting and to avoid carbohydrate drinks.

Localized 1H-MRS

Localized ¹H-MRS consisted of single voxel spectroscopy measurements by a point-resolved spectroscopy SE sequence at echo time (TE) of 30msecs (repetition time (TR) = 2.500 msecs, vector size of 1,024, scan time of 2 mins 15 secs). All MR measurements were performed on a 1.5-T Magnetom Vision system (Siemens Medical, Erlangen, Germany) running on Numaris/4 (VA25A) with Quantum gradients using an eight-channel head coil.

For every voxel, two spectra with and without chemical-shift selective water suppression (Haase et al, 1985) were obtained (50 versus 10 acquisitions) to allow for eddy current correction and water scaling. Fixed scanner calibration was confirmed by external in vitro standards.

In the Eth experiment, voxels were placed supratentorially (voxel size = 25 × 25 × 25 mm³) in the left fronto-mesial lobe and infratentorially (voxel size = 22 × 22 × 25 mm³) in the left cerebellar hemisphere to capture putative spatial differences in susceptibility of the brain parenchyma to Eth. In the Glc experiment, voxels were placed only supratentorially. As described earlier (Bendszus et al, 2001), reproducible positioning was achieved by setting the infratentorial voxel with its medial edge close to the fourth ventricle, and with its superior edge below the tentorium, and by placing the supratentorial voxel with its inferior edge at the callosomarginal sulci and with its posterior edge at the central sulci (Figure 1A).

Figure 1

Voxel placement (A). Upper row: Fronto-mesial (i.e., supratentorial) MRS voxel placement. Lower row: Cerebellar (i.e., infratentorial) volume of interest. Exemplary spectra at single subject level (B). First and second rows represent Eth experiment. Data fit of baseline (black line) versus maximum (blue line) ethanol level (1st row). Data (gray line) and data fit (red line) of difference spectra. Third and fourth rows represent Glc experiment. Data fit of baseline (black line) versus maximum (blue line) glucose level (3rd row). Data (gray line) and data fit (red line) of difference spectra.

Protocol

Eth experiment: Before Eth ingestion, a baseline set of infratentorial and supratentorial spectral data and one blood sample to evaluate blood alcohol concentration (BAC) were obtained for each participant. Subsequently, subjects drank 300 mL of a standard beverage containing 0.65 g Eth (females) and 0.75 g Eth (males) per kg body weight within 4.5 ± 1.8 mins through plastic tubing. Throughout the following 93.71 ± 11.17 mins, seven sets of infratentorial and supratentorial spectral data and three blood samples were acquired. During this time, the subjects stayed within the scanner, enabling ¹H-MRS data acquisitions based on the initial shim. At for 726.36 ± 94.96 mins (12.11 ± 1.58 h) after Eth exposure, one additional infratentorial and supratentorial spectral data set, as well as another blood sample, were taken.

Glc experiment: The setup of the Glc infusion experiment equals those used for intravenous Glc tolerance testing. Before Glc infusion, a set of supratentorial ¹H-MRS data and the baseline blood Glc concentration were assessed. Subsequently, a bolus of 300 mg Glc per kg body weight was intravenously infused through plastic tubing within 3 mins. Simultaneously, ¹H-MRS was performed and continued for 34 ± 0.71 mins. Parallel to the assessment of spectral data, six additional blood Glc levels were retrieved.

For both experiments, detailed temporal information of infratentorial and supratentorial data acquisition is available (Supplementary Offset Table).

Spectral Analysis

For ¹H-MR spectral analysis, the user-independent frequency-domain fitting routine by LCModel (version 6.1–4/LCMgui version 2.1–4) (Provencher 1993) was used. LCModel analyses in vivo spectra as a linear combination of complete model spectra of the individual metabolites in vitro, which are defined in basis sets. For spectral analysis of Eth, no relevant data set had been available. In a number of in vitro MR experiments, Eth was integrated into a new basis set for long TR (i.e., 2.500 msecs) and short TE (i.e., 30 msecs), which enables reliable ¹H-MRS detection of Eth. In addition to Eth, the basis set included spectral data of alanine, Asp, Cr, γ-aminobutyric acid, Glc, glutamine, glutamate, Ins, lactate, NAA, N-acetyl-aspartyl-glutamate, glycerylphosphocholine (GPC), phosphocholine (PC), guanidinoacetate, lipids and macromolecules.

For fitting macromolecular resonances, the LCModel method uses a priori knowledge on macromolecules, as assessed by analyzing metabolite-nulled spectra at 1.5 T, that is, macromolecular models are simulated. Intrinsic to the applied method, macromolecular models will not be perfectly correct in handling highly variable macromolecular signals. Therefore, as for every fitting routine applied on empirical data, residual errors of LCModel data fits and thus imperfect metabolite concentrations cannot be ruled out. However, to reduce the impact of macromolecules on other metabolite resonances, in this study, all statistical analyses on cerebral metabolites are based on difference spectra (see Figures 1B and 2).

Figure 2

Difference spectra of the Eth (A) and Glc experiments (B). Mean data fit of left fronto-mesial difference spectra at various offsets (t_+x in minutes). Data acquisition started at 5.79 ± 2.94 mins upon alcohol exposure (Eth experiment) and was synchronized with the onset of intravenous glucose application (Glc experiment). For better visualization, only 8/11 spectra of the Glc experiment are shown.

Using complete spectra, rather than just single peaks, two metabolites with overlapping peaks at one chemical shift can still be resolved by the LCModel method if they have different structures at other chemical shifts. In LCModel, eddy current correction is performed in the time domain, dividing the water-suppressed signal by the phase factor of the water signal for each data point. To estimate metabolite concentrations, LCModel adjusts spectral measurements using the unsuppressed water reference signal (water scaling).

Spectral quality was assessed applying the following criteria as provided by LCModel output parameters: (i) the signal-to-noise ratio (S/N) (defined by the ratio of the maximum in the spectrum-minus-baseline over the analysis window to twice the root-mean-square residuals); (ii) the line width (roughly estimated through full width at half-maximum (FWHM, p.p.m. value); (iii) the distribution of residuals (characterizing the data fit); and (iv) standard deviations (s.d.) (representing the estimation errors in metabolite quantification). In LCModel, the s.d. estimates are Cramer-Rao lower bounds. Standard deviation > 50% indicate that the metabolite concentration may range from zero to twice the estimated concentration. Thus, for the Eth experiment, only spectra featuring (i) S/N > 4, (ii) FWHM ≤ 0.07p.p.m., (iii) randomly distributed residuals ~0 and (iv) s.d. < 25% were selected for further analysis. As parameter of interest in the Glc experiment, the quality criterion (iv) was not applied to the spectral data. Exemplary supratentorial ¹H-MRS single-subject spectral data of the Eth and Glc experiments are shown in Figure 1B.

Statistical Analysis

Statistical data postprocessing encompassed the following subsequent analysis steps: (i) For each metabolite, concentration changes over time were tested for significance, using repeated measurements analyses of variance (rmANOVA). Thereafter, (ii) metabolite (met)-specific concentration time series (TS_met) were extracted over time and across subjects. Metabolites significantly changing in concentration after per os Eth ingestion were (iii) supplied to a linear regression analysis to estimate associations between TS_met and TS_Eth. The results described the amount of variability explained by the model fit [R²), the slope expressing the steepness of the association (b) and the F-statistic. Moreover, (v) a balanced ANOVA with replications as factor (rANOVA) tested for spatial differences between fronto-mesial and cerebellar data.

¹H-MRS data were explored for gender differences in metabolite concentrations by unbalanced two-way ANOVAs, with replications and gender as factors. For spatial analysis of ¹H-MRS signal, fronto-mesial and cerebellar S/N data were corrected for different voxel volume and analyzed by two-tailed Student's t-tests.

The results of ANOVAs and regression analyses were conservatively adjusted for multiple comparisons by Bonferroni correction. For statistical inferences as performed in this study, P_corr values <0.05 were considered significant. Statistical procedures were performed in Matlab (version 7.0.1., The MathWorks, Natick, USA).

Results

Signal-to-Noise Ratio

Corrected for different voxel sizes, supratentorial S/N data (21.44 ± 3.58) were higher than infratentorial S/N ratios (12.52 ± 2.98) (P < 0.001).

Gender Difference

There were no differences in infratentorial and supratentorial metabolite concentrations between female and male subjects, that is, no gender difference of Eth effects in the brain parenchyma.

Ethanol

At 1.19 p.p.m., the main resonance of the Eth triplet was detectable at 5.79 ± 2.94 and 8.64 ± 2.98 mins after alcohol exposure for supratentorial and infratentorial data, respectively (Figure 3). Ethanol concentrations culminated in 91.14 ± 11.69 mins (fronto-mesial parenchyma) and 74.43 ± 5.58 mins (cerebellar parenchyma) after alcohol exposure. Blood alcohol concentration measurements (mean peak BAC: 51.43 ± 10.27 mg/dL at 93.71 ± 11.27 mins after Eth exposure) correlated well with supratentorial TS_Eth (P = 0.004, b = 9.872, F = 47.740, R² = 0.386) and infratentorial TS_Eth (P = 0.023, b = 12.251, F = 17.138, R² = 0.893). Concentration changes over time were significant supratentorially (rmANOVA; P_corr < 0.001) and infratentorially (rmANOVA; P_corr < 0.001) (Table 1). Fronto-mesial and cerebellar Eth levels did not differ (rANOVA).

Figure 3

Eth and cerebral metabolites. Left column: Upon Eth exposure, fronto-mesial (solid line) Glc, Ins, Cr, Asp, and Cho and cerebellar (dashed line) Ins and Cho levels differed over time; s.d. depicted as gray envelopes. Right column: Regression analysis revealed a significant association between supratentorial (▲) TS_Eth and TS_GlC (positive) and between supratentorial TS_Eth and TS_Ins, TS_Cr, TSASP and TS_Cho (negative), respectively, and infratentorial (Δ) TS_Eth and TS_Ins and TS_Cho, respectively; s.d. depicted as error bars. (*1s.d. maximum value = 0.70, *2s.d. minimum value = −1.79, and *3s.d. maximum value = 0.33).

Table 1

Repeated measurements ANOVA of supratentorial and infratentorial metabolite concentrations

	Fronto-mesial level					Cerebellar level
	rmANOVA		Regression analysis			rmANOVA		Regression analysis
	Pcorr	Pcorr	b	F	R2	Pcorr	Pcorr	b	F	R2
Eth	0.000	NA	NA	NA	NA	0.000	NA	NA	NA	NA
Ins	0.000	0.012	−0.060	21.288	0.753	0.022	0.042	−0.125	21.288	0.565
Cho	0.000	0.009	−0.038	24.282	0.776	0.005	0.009	−0.089	23.933	0.774
Asp	0.000	0.021	−0.067	17.355	0.713	NS	NA	NA	NA	NA
Cr	0.000	0.003	−0.067	33.788	0.828	NS	NA	NA	NA	NA
Glc	0.000	0.003	−0.094	35.139	0.834	NS	NA	NA	NA	NA
H2O	NS	NA	NA	NA	NA	NS	NA	NA	NA	NA
NAA	NS	NA	NA	NA	NA	NS	NA	NA	NA	NA

Asp, aspartate; Cho, choline-containing compounds; Cr, creatine; Glc, glucose; Eth, ethanol; Ins, inositol; NA, not applied; NAA, N-acetyl-aspartate; NS, not significant; rmANOVA, repeated measurements analyses of variance.

For metabolites differing over time (P_corr < 0.05), linear regression analysis was performed describing the slope expressing the sleepness of the association (b), amount of variability in data explained by the model fit (R²) and F-statistics (F). Results were Bonferroni-corrected for multiple testing (P_corr).

Glucose

Eth experiment: Owing to the challenging ¹H-MRS detection and analysis of Glc at 1.5 T, we momentarily label the metabolite resonating at 3.43 p.p.m. after Eth exposure putative Glc (pGlc). In contrast to cerebellar levels, supratentorial pGlc concentrations revealed significant changes over time (rmANOVA, P_corr < 0.001) after Eth administration (Table 1). Regression analysis showed a positive correlation of supratentorial TS_pGlc and TS_Eth (P_corr= 0.003). Analyzed by rANOVA, supratentorial and infratentorial pGlc levels did not differ. Increasing supratentorial pGlc concentrations were first observed at 5.79 ± 2.94 mins after controlled Eth consumption (Figure 3).

Glc experiment: At 4.40 ± 0.55 mins after application of an intravenous glucose bolus, Glc was first detectable in the fronto-mesial parenchyma with a specific resonance at 3.43 p.p.m. In parallel, metabolite s.d. decreased from 486.80 ± 474.58% to 23.60 ± 4.16% and further decreased to 22.40 ± 5.68% at 34 ± 0.71 mins after Glc exposure. Blood Glc concentrations were in good accordance to TS_{Glc, iv}. However, at 12.60 ± 0.55 mins after intravenous Glc infusion, the blood Glc level started to decline, whereas brain Glc remained at a constant level (Figure 4).

Figure 4

Glucose (Glc experiment). Fronto-mesial Glc levels (○) differed over time after glucose infusion (P_corr < 0.001). Blood Glc levels (*) were in good accordance with TS_GlC. Metabolite s.d. (☐) decreased from 486.80% ± 474.58% to 23.60% ± 4.16% after 4.40 ± 0.55 mins.

Importantly, cerebral pGlc and Glc resonances peaked at 3.43 p.p.m. and could be resolved in both experiments even at a single subject level (Figure 1B). Figure 2 shows the mean data fits of difference spectra, as assessed in the Eth (Figure 2A) and Glc (Figure 2B) experiments.

Thus, the Glc experiment provides strong evidence for Glc to having originated the resonance peaks at 3.43 p.p.m., as detected in the Eth experiment. Most probably, Glc and pGlc are identical.

Aspartate

After the Eth application, rmANOVA yielded significant changes in supratentorial Asp concentration (P_corr < 0.001) over time, but not infratentorially (Table 1). Regression analysis showed a positive correlation between supratentorial TS_Asp and TS_Eth (P_corr = 0.021). Compared by rANOVA, supratentorial and infratentorial Asp data did not differ. The decrease in supratentorial Asp concentration was apparent at 5.79 ± 2.94 mins after alcohol exposure (Figure 3).

Inositol

For Ins, metabolite concentration of the fronto-mesial as well as the cerebellar parenchyma changed over time after Eth ingestion (rmANOVA, supratentorial P_corr < 0.000; infratentorial P_corr = 0.022) (Table 1). Regression analysis showed a negative correlation of TS_Ins and TS_Eth in the fronto-mesial and cerebellar brain tissue (supratentorial P_corr= 0.012; infratentorial P_corr = 0.042). Testing supratentorial versus infratentorial ¹H-MRS data, rANOVA revealed no difference in Ins concentrations. Fronto-mesial and cerebellar Ins concentrations started to subside at 5.79 ± 2.94 and 8.64 ± 2.98 mins after Eth exposure, respectively (Figure 3).

Choline-Containing Compounds

After Eth exposure, Cho concentrations changed in the supratentorial (rmANOVA, P_corr < 0.000) and in the infratentorial level (rmANOVA, P_corr = 0.005) over time. TS_cho of both the fronto-mesial and cerebellar parenchyma were negatively correlated with TS_Eth (Table 1) after alcohol ingestion (supratentorial P_corr = 0.009; infratentorial P_corr = 0.009). By rANOVA, no differences between supratentorial and infratentorial Cho levels were detected. The reduction of supratentorial and infratentorial Cho levels was first observed at 5.79 ± 2.94 and 8.64 ± 2.98 mins, respectively, after standardized alcohol consumption (Figure 3).

Creatine

Upon Eth application, supratentorial Cr levels changed over time (rmANOVA, P_corr < 0.000), whereas statistical analyses revealed no differences in infratentorial Cr concentrations (Table 1). Regression analysis showed a negative correlation between supratentorial TS_Cr and TS_Eth (P_corr = 0.003). For supratentorial versus infratentorial Cr levels, rANOVA revealed no difference. Fronto-mesial Cr concentrations started to decrease at 5.79 ± 2.94 mins after alcohol ingestion (Figure 3).

N-Acetyl-Aspartate

For NAA, rmANOVA revealed no changes in concentrations of the fronto-mesial and cerebellar parenchyma over time after standardized Eth exposure (Table 1).

Water

There were no changes of mean supratentorial or infratentorial parenchymal water content over time, as determined by the integral of unsuppressed water within the corresponding voxel of interest after Eth ingestion (Table 1, Figure 5).

Figure 5

Water integrals. Upon Eth ingestion, changes in the supratentorial (solid line) and infratentorial (dashed line) areas under unsupressed water peaks (Δ[AUWP]) were not significant; s.d. are depicted as error bars.

Discussion

Not addressed by longitudinal ¹H-MRS experiments yet, the principal goal of this study was to characterize human brain metabolites in vivo subsequent to Eth ingestion. According to our findings, Eth explicitly affects human brain physiology within a short time (Eth experiment), and ambitious to assess at 1.5 T, the findings on Glc were validated (Glc experiment).

Characterizing metabolite signal changes in the setting of our study, it is important to note that T1 relaxation effects cannot be ruled out. Subsequently, spectral data fits of the CH₂ Eth group might be distorted due to an imperfect estimation of the CH₃ moiety. As a consequence, the fit of adjacent metabolite resonances might be affected and a systematic error may be introduced. However, all metabolite concentration time series are based on difference spectra, that is, on changes relative to metabolite baseline values. Therefore, inference on the qualitative nature of metabolite changes ought to be valid, but interpretation in terms of absolute concentrations may not and are avoided.

Shortly after ingestion, Eth impact on brain physiology resulted in significant metabolite concentration changes of supratentorial Cr, Ins, Cho, Glc and Asp. Moreover, TS_met of these metabolites correlated with supratentorial TS_Eth. Infratentorially, Eth significantly affected cerebellar Ins and Cho levels. Here, their concentration time series correlated with Eth levels over time as well. All Eth-induced changes of brain metabolites were fully reversible.

On average, Eth was detected first at 5.79 and 8.64 mins upon per os ingestion within the fronto-mesial and cerebellar volume of interest, respectively, leaving a 2.85-min offset, mainly caused by the subsequent collection of supratentorial and infratentorial data. In addition, supratentorial and infratentorial Eth concentrations culminated at different times. Besides the subsequent acquisition of fronto-mesial and cerebellar MRS data for Eth kinetics, there is much dispersion in time, because of interindividually different ingestion and absorption rates. These individual disparities defy data correction in time through postprocessing. Thus, varying onsets of fronto-mesial and cerebellar Eth peaks rather reflect intrinsic physiology of each individual than true effects of spatially distinct Eth kinetics.

In agreement with earlier studies (Fein and Meyerhoff 2000; Mendelson et al, 1990), supratentorial and infratentorial TS_Eth significantly paralleled BAC levels, underlining the high sensitivity and specificity of our ¹H-MRS measurements.

Considerable differences in supratentorial and infratentorial S/N values (corrected for different voxel sizes) cast doubt on the comparability of fronto-mesial and cerebellar absolute metabolite values. However, statistic analyses revealed no difference in fronto-mesial and cerebellar metabolite levels. This finding indicates similar supratentorial and infratentorial metabolite kinetics after per os Eth exposure.

Ethanol and Alternative Energy Substrate Utilization

In principle, cerebral Eth can be enzymatically metabolized to acetaldehyde by catalase (Zimatkin and Buben 2007) or cytochrome P4502E1 (Upadhya et al, 2000). However, contribution of alcohol dehydrogenase to Eth metabolism in the brain is still unclear (Deitrich et al, 2006). Acetaldehyde is metabolized to acetate (Act) by aldehyde dehydrogenase (Erwin and Deitrich, 1966) in the brain. Outside the brain, Act is converted into acetyl-CoA by liver enzymes. Producing Act, Eth decomposition presents an alternative cerebral energy substrate to Glc. For energy utilization, brain cells are able to quickly adapt to Act (Waniewski and Martin, 1998). A modified monocarboxylate (MCT; ketone bodies, lactate, pyruvate) compartmentation model of neurons and astrocytes may explain the physiologic underpinnings of an alternative energy substrate utilization upon Eth exposure. The original MCT compartment model of neurons and astrocytes (Cerdan et al, 2006) describes two kinetically different MCT pools for cellular energy supply, fed by Glc and lactate (or other MCTs). Essentially, degradation of both, lactate and Glc, requires cytosolic NAD⁺. This competition is subject to substrate concentration. For example, in the presence of high intracellular lactate concentrations, Glc metabolism may be reduced or stopped. Our modified MCT compartmentation concept suggests that a different and/or additional substrate Glc has to compete with for enzymatic degradation: Ethanol. Eventually, decomposition of Eth would form a new MCT pool composed of Act serving for cellular energy supply. Thus, glycolysis in neurons and astrocytes might be hampered or disabled upon alcohol exposure by cytosolic competition for NAD + owing to the need for Eth decomposition. ¹H-MRS findings of our Eth experiment, that is cerebral Glc accumulation upon Eth exposure, support this notion, although, for detailed characterization and validation, our idea of a modified MCT compartmentation surely needs further study.

Although challenging, ¹H-MRS detection of Glc at 1.5 T is possible, in particular if the physiologic metabolite equilibrium is altered, for example, by an exogenic substance such as Eth. However, working on difference spectra, this study characterizes relative metabolite changes rather than absolute values upon alcohol consumption.

At low field strength, the complex multiplet Glc spectrum collapses into two multiplets centered at 3.43 and 3.8 p.p.m. (Govindaraju et al, 2000). In contrast to the Glc multiplet peaking at 3.43 p.p.m., the moiety resonating at 3.8 p.p.m. is not amenable to visual inspection of spectral data because of strong overlapping with other metabolite signals in vivo. Findings of our Glc infusion experiment provide strong evidence for Glc to cause differing spectral shapes at 3.43 p.p.m., as occurred in the Eth experiment, suggesting that LCModel's nearly model-free constrained regularization method is sensitive to detect even concentration changes of those metabolites upon Eth exposure, which are per se difficult to assess at 1.5 T.

Consistent with the idea of a modified MCT compartmentation concept and supported by earlier FDG PET findings (Volkow et al, 2006), our study provides evidence for a cerebral energy switch, that is, an alternative energy substrate usage by the brain entailed through alcohol obviously commencing already at ~6 mins upon ingestion. The reversible Michaelis-Menten transport model for Glc (Equation (1)) describes the

$$[\mathrm{G}\mathrm{l}\mathrm{c}{]}_{\mathrm{b}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}}=\frac{([\mathrm{G}\mathrm{l}\mathrm{c}{]}_{\mathrm{b}\mathrm{l}\mathrm{o}\mathrm{o}\mathrm{d}}\times \frac{Vmax}{{\mathrm{G}\mathrm{M}\mathrm{R}}_{\mathrm{G}\mathrm{l}\mathrm{c}}})-K_{\mathrm{m}}-[\mathrm{G}\mathrm{l}\mathrm{c}{]}_{\mathrm{b}\mathrm{l}\mathrm{o}\mathrm{o}\mathrm{d}}}{1+\frac{V_{max}}{{\mathrm{C}\mathrm{M}\mathrm{R}}_{\mathrm{G}\mathrm{l}\mathrm{c}}}}$$

(1)

physiologic underpinnings of our findings, assuming stable transport kinetics for cerebral Glc and constant global energy demands of the brain in the presence of Eth. Corresponding to Equation (1), an increase of, for example, 50% in brain Glc ([Glc]_brain) is associated with a decrease of ~25% in brain Glc consumption (Michaelis-Menten constant K_m for GLUT-1 = 1.2 mmol/L, maximum Glc transport rate V_mra = 0.57 μmol/g per min) (de Graaf et al, 2001). The theoretical background is illustrated in Figure 6 for different blood Glc concentrations [Glc]_blood, describing the relation between CMR_Glc and [Glc]_brain at different nutritional states of individuals.

Figure 6

Relation between cerebral metabolic utilization rate of Glc (CMR_Glc) and brain Glc concentration [Glc]_brain. Absolute (A) and relative (B) values of Glc kinetics based on the reversible Michaelis-Menten model (see Equation (1), V_max = 0.57 μmol/g per min, K_M = 1.2 mmol/L; de Graaf et al, 2001) for plasma glucose levels of 5 mmol/L (solid line), 7.5 mmol/L (dotted line) and 10 mmol/L (dashed line).

Thus, in accordance with the reversible Michaelis-Menten transport model, accumulation of cerebral Glc subsequent to Eth ingestion seems to correspond to a decrease in cerebral Glc consumption, for example, explained by the modified MCT compartmentation concept.

Please note that as Eth ingestion is known to lower blood Glc levels (Huang and Sjoholm, 2008), increased cerebral Glc levels are most likely not directly linked to alcohol consumption itself through subsequent elevation of blood Glc concentration.

Ethanol and Oxidative Stress

In the brain, decomposition of Eth to Act results in NADH⁺ accumulation. Presented to mitochondrial metabolism, NADH⁺ can be metabolized to NAD⁺ for cytosolic replenishment and thus stable redox potentials. The essential NAD⁺ transport system into the mitochondrion is the malate-aspartate shuttle (MAS). As a critical part of the MAS (McKenna et al, 2006), transaminase reactions between Asp and glutamate, which result in reduced Asp levels, characterize an increased flux through the MAS (Mangia et al, 2007), that is, activation. Spectroscopic assessment of Asp at 1.5 T, as performed in our study, is sophisticated and quantitative statements in absolute terms are not possible. Therefore, the description of Asp signal changes is only qualitative, in that TS_Asp and TS_Eth are negatively associated. Consistent with the metabolite interactions described by the MAS model, negatively correlated TS_Eth and TS_Asp would indirectly reflect Eth-induced accumulation of NADH⁺ and the subsequent need for mitochondrial reoxidation of NADH⁺ to NAD⁺ to grant stable redox potentials.

Thus, in line with other reports (Chen et al, 2008), the in vivo findings of our study suggest cerebral oxidative stress (due to NADH⁺ accumulation as precipitated by Eth decomposition) shortly upon alcohol exposure. In this context, further study on Asp at higher field strengths is needed to confirm our findings.

Ethanol and Cerebral Vulnerability

As part of the high-energy metabolism in the cytosol of neurons and glia, Cr is reversibly phosphorylated to phosphocreatine (PCr) in the equilibrium reaction, Cr + ATP ↔ PCr + ADP + H, which is enzymatically catalyzed by Cr kinase (Wyss and Kaddurah-Daouk, 2000). This reaction is considered essential for temporal buffering of high-energy phosphates and pH, which is necessary for proper cell functioning. Using ¹H-MRS, Cr and PCr resonance peaks cannot be sufficiently resolved. Creatine as referred to here means total Cr comprising both Cr and PCr. Therefore, results of this study grant no detailed insight into the Cr/PCr equilibrium reaction. However, in contrast to recent animal studies on acute Eth effects (Adalsteinsson et al, 2006; Braunova et al, 2000; Hirakawa et al, 1994; Lee et al, 2003), results of our study provide important evidence for Eth to affect Cr. As reviewed in detail (Wyss and Kaddurah-Daouk, 2000), there is ample evidence for neuroprotective effects of Cr. For example, there is evidence for Cr to protect against free radical generation and to regulate a mitochondrial protein complex, which is implicated in cell death (O'Gorman et al, 1997). Vice versa, the absence of Cr implies greater possibility of neuronal damage in case of hypoxia, oxidative stress conditions (Carter et al, 1995) or neurodegenerative diseases (Smith et al, 1995). In this context, a negative correlation of TS_Eth and TS_Cr, that is reduced Cr levels after per os alcohol consumption suggests elevated cerebral vulnerability. Producing oxidative stress (see the section Ethanol and Oxidative Stress), continuing Eth application could therefore result in vicious circles of neurotoxicity, which may contribute to the functional and morphologic correlates of chronic alcohol consumption.

Ethanol and Water Shift

As an organic osmolyte, neuroglial Ins serves as a regulatory component for the maintenance of constant cell volume (Wolfson et al, 2000). As phosphorylated derivative, Ins is additionally involved in neuronal receptor messenger systems (Downes and Macphee, 1990). However, Ins moieties contributing to the second messenger system are most probably far too small for spectroscopic in vivo detection.

Prevention of Eth-induced cell swelling by transmembrane Ins flux mediated through volume-sensitive organic osmolyte channels (Fisher et al, 2002) might result in excessive substrate consumption and consequently lead to decreased Ins levels (Braunova et al, 2000). In agreement with this rationale, a negative correlation of TS_Eth and TS_Ins suggests successful osmotic regulation, because there was no difference in brain water content before and upon Eth ingestion. Vice versa, maintenance of equilibrated cerebral water indicates that metabolite level changes after per os Eth administration were not due to shifts in brain water.

Ethanol and Membrane Modifications

Using ¹H-MRS, Cho resonance peaking at 3.2 p.p.m. is mainly composed of PC and GPC with minor contributions from free choline (Boulanger et al, 2000). Phosphocholine and GPC resonance peaks cannot be resolved by ¹H-MRS. Therefore, PC, GPC and free choline are referred to as Cho, which is substrate for the synthesis of cell membranes and neurotransmitters (Cousins, 1995) and highly concentrated in glial cells (Urenjak et al, 1993). Thus, Eth-induced negative correlation of TS_Eth and TS_Cho can theoretically reflect changes in membrane turnover. Supporting our findings, decreasing Cho levels have been shown in the rat brain after Eth application (Adalsteinsson et al, 2006). Vice versa, longitudinal studies on detoxified alcohol-dependent subjects revealed increasing Cho levels (Bartsch et al, 2007; Bendszus et al, 2001; Ende et al, 2005) after initially reduced concentrations (Bendszus et al, 2001; Ende et al, 2005). Our findings suggest a parallel between acute Eth effects on Cho and Cho changes in alcohol-dependent subjects upon detoxification: In both cases, Cho concentrations initially declined and gradually recovered. Thus, Eth seems to trigger reversible membrane modifications even in healthy subjects shortly upon alcohol exposure, whereas cell membrane alterations in alcohol-dependent subjects prevail and persist until detoxification.

Ethanol and Neuronal Integrity

NAA is present in neurons, neuroglial precursors and immature oligodendrocytes, and absent in mature glial cells (Urenjak et al, 1993). There is a positive correlation between cerebral NAA concentration and neuronal density and viability (Miller, 1991). Thus, NAA is regarded as a neuronal marker, and decreased NAA concentrations are considered indicative of neuronal loss or dysfunction. In line with longitudinal animal studies (Adalsteinsson et al, 2006; Braunova et al, 2000), our study provides no evidence for Eth to affect NAA shortly after per os alcohol consumption.

Creatine and Metabolite Ratios

Considering it a stable metabolite, many studies apply Cr as reference for relative metabolite quantification. Noteworthy, this method is adopted by many ¹H-MRS studies on Eth effects. Importantly, using metabolite ratios for quantification without definite a priori knowledge about its kinetics on certain experimental conditions may entail detrimental consequences. For this study, results would not be properly interpretable owing to striking correlations between Eth and Cr concentrations after Eth ingestion. Using unsuppressed water as internal reference rather than metabolite ratios, metabolite quantification seems to yield unbiased concentrations.

In summary, this study longitudinally characterized the acute effects of a single Eth exposure on human brain metabolism by ¹H-MRS. Findings indicate adaptation to alternative energy substrate usage, elevation of oxidative stress level and cerebral vulnerability and alterations of cell membranes shortly after alcohol ingestion. There was no evidence for changes in neuronal integrity and brain water content after Eth consumption.

Future longitudinal MR studies are needed to further unravel the biochemical underpinnings of how Eth affects the human brain metabolism. For example, characterization of the Cr/PCr equilibrium reaction by phosphorous MRS might yield crucial information about the impact of Eth on cerebral high-energy metabolism and the neurotoxic effects of Eth on the brain. Besides contributing to the MAS transaminase reaction, description of glutamate by ¹H-MRS using field strengths higher than 1.5 T might grant a more detailed insight into the concept of Eth as oxidative stressor.

Our findings suggest that longitudinal MRS experiments offer an interface between metabolic mechanisms of the human brain and neuroimaging and thus might broaden our understanding of cellular processes in vivo.