Translational Magnetic Resonance Spectroscopy Reveals Excessive Central Glutamate Levels During Alcohol Withdrawal in Humans and Rats

s

cfhiltpmmgTfwst

R

ARCHIVAL REPORT

Translational Magnetic Resonance SpectroscopyReveals Excessive Central Glutamate Levels DuringAlcohol Withdrawal in Humans and RatsDerik Hermann, Wolfgang Weber-Fahr, Alexander Sartorius, Mareen Hoerst, Ulrich Frischknecht,Nuran Tunc-Skarka, Stephanie Perreau-Lenz, Anita C. Hansson, Bertram Krumm, Falk Kiefer,Rainer Spanagel, Karl Mann, Gabriele Ende, and Wolfgang H. Sommer

Background: In alcoholism, excessive glutamatergic neurotransmission has long been implicated in the acute withdrawal syndrome andas a key signal for dependence-related neuroplasticity. Our understanding of this pathophysiological mechanism originates largely fromanimal studies, but human data are needed for translation into successful medication development.

Methods: We measured brain glutamate levels during detoxification in alcohol-dependent patients (n � 47) and in healthy control subjects(n � 57) as well as in a rat model of alcoholism by state-of-the-art 1H-magnetic magnetic resonance spectroscopy at 3 and 9.4 T, respectively.

Results: We found significantly increased glutamate levels during acute alcohol withdrawal in corresponding prefrontocortical regions oftreatment-seeking alcoholic patients and alcohol-dependent rats versus respective control subjects. The augmented spectroscopic gluta-mate signal is likely related to increased glutamatergic neurotransmission because, enabled by the high field strength of the animal scanner,we detected a profoundly elevated glutamate/glutamine ratio in alcohol-dependent rats during acute withdrawal. All dependence-inducedmetabolic alterations normalize within a few weeks of abstinence in both humans and rats.

Conclusions: Our data provide first-time direct support from humans for the glutamate hypothesis of alcoholism, demonstrate thecomparability of human and animal magnetic resonance spectroscopy responses, and identify the glutamate/glutamine ratio as potentialbiomarker for monitoring disease progression.

itl

chaacararbsptetpAs

gihdc

M

P

Key Words: Alcoholism, chronic intermittent alcohol vapor expo-ure, glutamate-glutamine cycle, human, neuroimaging, rat model

T he most prominent theory of alcoholism, the glutamate the-ory, describes that enhanced glutamate-mediated neuronalexcitability during withdrawal and protracted abstinence

ontributes to craving and relapse process (1). This theory, builtundamentally on preclinical findings, is lacking direct evidence inumans. For example, numerous brain microdialysis experiments

n animals demonstrated augmented extracellular glutamate (Glu)evels that correspond with the intensity of the withdrawal symp-omatology (2– 4), but direct measurements of brain Glu levels inatients during alcohol withdrawal are lacking, although some hu-an findings indicate indirectly the role of elevated central gluta-ate (Glu) levels in acute withdrawal by the effectiveness of anti-

lutamatergic compounds in alleviating withdrawal symptoms (5).he importance of repeated cycles of intoxication and withdrawalor the emergence of long-term plasticity in protracted abstinence

ith immediate relevance for relapse to excessive alcohol use isuggested by the animal literature (6 – 8). Although the link be-ween early withdrawal phenomena and subsequent relapse liabil-

From the Departments of Addiction Medicine (DH, UF, FK, KM), Neuroimag-ing (MH, NT-S, GE), Psychopharmacology (SP-L, ACH, RS, WHS), Psychia-try and Psychotherapy (AS), and Biostatistics (BK) and the ResearchGroup for Translational Neuroimaging (WW-F, AS), Central Institute forMental Health, Mannheim, Germany.

Authors DH and WW-F contributed equally to this work.Authors GE and WHS contributed equally to this work.Address correspondence to Wolfgang H. Sommer, M.D., Ph.D., Depart-

ment of Psychopharmacology, Central Institute for Mental Health,Square J5, 68159, Mannheim, Germany; E-mail: [email protected].

teceived Apr 14, 2011; revised Jul 26, 2011; accepted Jul 27, 2011.

0006-3223/$36.00doi:10.1016/j.biopsych.2011.07.034

ty remains unclear, it would provide a strong rationale for revisitinghe biology and genetics of acute withdrawal symptoms if such aink exists (9).

Only a few approaches are currently available that allow a directomparison of alcohol-related effects on brain physiology betweenumans and experimental animals. One of those investigative toolsllowing for direct human–animal comparison are in vivo neuroim-ging-based methods that should facilitate the translational pro-ess (1,10). Noninvasive measurement of Glu concentrations withindistinct volume of the brain can be achieved by 1H magnetic

esonance spectroscopy (MRS). Recent advances in scanner techniquend data analysis allow reliable detection of a distinct Glu signal with aesolution that is sufficient to achieve measurements from compara-le brain regions of humans and small laboratory animals. Severaltudies suggest the relevance of MRS-derived Glu measures for neuro-lasticity. For example, deficits in prefrontocortical Glu were consis-

ently found in depressed patients, and these deficits improved withffective electroconvulsive therapy (11–14). Furthermore, Glu levels inhe perigenual subregion of the anterior cingulate cortex (ACC) wereositively correlated with the degree of activation in the subgenualCC during an empathy task (15) and with impulsivity measures in

ubjects with borderline personality disorder (16).Here, we aimed to provide translational evidence for increased

lutamatergic transmission during acute alcohol withdrawal andnto abstinence by assessing cerebral Glu levels using MRS in theuman ACC, a brain region critically affected by alcohol and opiateependence (17, 18), and in the corresponding medial prefrontalortex (mPFC) of the rat.

ethods and Materials

articipantsTreatment-seeking alcohol-dependent patients (n � 47) admit-

ed to inpatient detoxification were included. All patients had their

BIOL PSYCHIATRY 2011;xx:xxx© 2011 Society of Biological Psychiatry

epsuwnacHp

1

n3asr

SA

4) (w

2 BIOL PSYCHIATRY 2011;xx:xxx D. Hermann et al.

last alcoholic drink on the day of admission or the day before.Patients underwent two MRS assessments (on Day 1 and Day 14 ofadmission, MRS-d1 and MRS-d14, respectively). MRS-d1 was sched-uled at the beginning of clinical withdrawal signs (blood alcoholconcentration�1.0 mg/g) and before the application of benzodiaz-epines (diazepam or lorazepam) became necessary (see Supple-ment 1 for further details). Only clonidine (agonist at adrenergicalpha-2 receptors) for treatment of hypertension or tachycardiawas allowed and applied in 12 patients before MRS-d1. Over the90- to 120-min MRS session, withdrawal symptoms were increas-ing according to the Clinical Institute Withdrawal Assessment ofAlcohol Scale (19) (before MRS: 5.7 � 3.8, after: 7.0 � 4.1; systolic/diastolic blood pressure before: 142 � 31/89 � 13 mm Hg, after154 � 23/92 � 10 mm Hg), and mean breath alcohol levels fell from.72 � .42 g/L to .39 � .40 g/L in 35 patients. The remaining patientshad undetectable alcohol levels before MRS. An age- and sex-matched comparison group of healthy control subjects (n � 57)was recruited through newspaper advertisement. Exclusion criteriafor all participants were current or history of substance dependence

Table 1. Clinical Characteristics of Alcohol-Dependent Patients and Health

Patients (n � 47

ex (female/male) 9/38ge (years) 46.3 � 1.5

OCDS 14.6 � 1.3ADS 15.2 � 1.1AUDIT 26.4 � 1.0Duration of Alcohol Dependence (years) 15.1 � 1.6Alcohol Intake in the Previous 90 Days (g/day) 205.1 �1 7.6Breath Alcohol at Admission (g/L, n � 41) 1.4 � .12Gamma-Glutamyl Transferase (U/l) 377.1 � 97.4Alanine Transaminase (U/l) 57.8 � 8.4Depressiveness (BDI) 10.2 � .4Anxiety (STAI State) 36.4 � 1.6Current Smokers 39Cigarettes per Day (n) in Smokers 24.4 � 1.8FTND Score in Smokers 6.0 � .3

ADS, Alcohol Dependence Scale (40); AUDIT, Alcohol Use Disorders IdentNicotine Dependence (43); OCDS, Obsessive Compulsive Drinking Scale (4weeks); STAI, State-Trait Anxiety Inventory (45).

www.sobp.org/journal

xcept nicotine (and except alcoholism in the patients group), anysychotropic medication in the previous 3 months, common exclu-ion criteria for magnetic resonance imaging (e.g., metal), positiverine drug screen, history of brain injury indicated by head injuryith unconsciousness of more than 2 min, other psychiatric oreurologic brain diseases, hepatic encephalopathy, liver cirrhosis,nd severe medical illnesses. The study was approved by the ethicsommittee of the Mannheim medical faculty of the University ofeidelberg, and informed written consent was obtained from allarticipants. Participants’ characteristics are summarized in Table 1.

H MR Spectroscopy at 3TAll participants were scanned in a clinical 3T whole-body scan-

er (Siemens, Erlangen, Germany). The position of the voxel (15 �0 � 12 mm3, Figure 1A) in the ACC was defined on the basis ofnatomic images from isotropic 1 mm3 T1-weighted three-dimen-ional data. Spectra (Figure 1C and 1D) were acquired with a point-esolved spectroscopy sequence using the following parameters:

trols

Controls (n � 57) Test Statistics p Values

13/44 �2(1) � .2 . 64945.1 � 1.5 t(102) � .54 .588

.8 � .2 t(46.9) � 10.7 �.001

.6 � .2 t(47.2) � 13.0 �.0013.3 � .3 t(52.8) � 22.3 �.001

—23.6 � 2.7 t(48.1) � 10.2 �.001

—28.1 � 1.8 t(46.0) � 3.6 �.00129.4 � 2.2 t(52.3) � 5.0 �.001

2.3 � 1.3 t(48.1) � 5.8 �.00130.8 � .7 t(56.6) � 3.2 �.01

17 �2(1) � 28.5 �.00114.6 � 2.9 t(54) � 2.9 �.01

3.6 � .6 t(47) � –3.9 �.001

on Test (41); BDI, Beck Depression Inventory (42); FTND, Fagerström Test forithout items 7 and 8, which assess amount of alcohol drunk in previous 2

Figure 1. Increased glutamate in the anterior cingulatecortex (ACC) of alcoholic patients during early with-drawal. Patients (n � 47) were assessed by magnetic res-onance spectroscopy twice, at admission and after 2weeks of controlled abstinence. Control subjects (n � 57)were scanned once. (A) Voxel localization. (B) Data repre-sent cerebrospinal fluid corrected Glu levels (mean �SEM) in the ACC of healthy controls (HC) as well as alco-holic patients measured in early withdrawal (EW) and af-ter a short period (2 weeks) of abstinence (SA). ���p �.001 EW vs. HC; #p � .05 EW vs. SA. For detailed statistics,see Methods and Materials. (C) Typical point-resolvedspectroscopy spectrum acquired from human ACC at 3T(5.4 mL, echo time � 80 msec, repetition time � 3 sec, 100averages) and overlaid by the LCModel fit curve. Peakresonances are N-acetylaspartate (NAA), total creatine(Cr), total choline (Cho), myo-inositol (mI), glutamate(Glu), glutamine (Gln), Glu � Gln (Glx). (D) Spectrum froman intoxicated patient. ETH-CH2 and ETH-CH3, peak reso-nances for ethanol. See also Figure S1 in Supplement 1 forcomparison with rat spectra.

y Con

)

ificati

p.

f

A

mucfpsAt

E

edi

p(se4isasNttsbc(.t3p

1

siwaswitm

cdF

S

alcsloianmws

R

GA

hw[vvr

D. Hermann et al. BIOL PSYCHIATRY 2011;xx:xxx 3

echo time � 80 msec (20), repetition time � 3000 msec, bandwidth �2400 Hz, 2048 data points, and 100 averages.

Quantification of spectra was based on LCModel spectral fitting(21). Ethanol was included in the LCModel basis set. The meanCramer-Rao lower bounds for N-acetylaspartate (NAA), creatine �phosphocreatine (tCr), and choline-containing compounds (tCho)were all below 4%. Glu fits were accepted when the Cramer-Raolower bounds of the fit were less than 20%. Concentration valueswere referenced to an unsuppressed water signal acquired fromthe same voxel (repetition time � 10 sec). For further details, seeSupplement 1 (Methods) and Figure S1.

Statistical Analysis of Human DataStatistical Package for the Social Sciences was used for analysis

(version 18, SPSS, Chicago, Illinois). First, we analyzed the influenceof several confounding variables on metabolite levels. Within thecontrol group, we tested whether age, number of cigarettes perday, Fagerström Test for Nicotine Dependence (FTND), or the graymatter (GM)/(GM� white matter [WM]) ratio was associated withany of the MR metabolites using Spearman’s correlation. Because ofthe high number of nonsmokers in control participants, cigarettesper day and FTND were additionally correlated with MR metabolitesin alcohol-dependent patients. Sex differences in control subjectswere tested using Student t tests. Variables revealing a significantinfluence were included in further analyses using corrected MRmetabolite values (residuals, see Results). No metabolite was asso-ciated with the GM/(GM�WM) ratio in control subjects, nor withcigarettes per day or smoking status in alcoholic or control subjects.Although FTND scores were positively correlated with Glu (r2 � .29,

� .045) in control but not alcoholic subjects (MRS-d1: r2 � .02, p �48; MRS-d14: r2 � .08, p � .14), neither smoking status nor numberof cigarettes showed significant effects in the general linear model(GLM) analysis, and thus unadjusted GLM results for Glu were pre-sented. Because of the uneven distribution of smokers betweenpatients and control subjects, the power of the GLM to detect aneffect of smoking was low (1-� � .20 for smoking status and 1-� �.06 for number of cigarettes smoked).

In a second step, MR metabolites of three groups (MRS-d1, MRS-d14, control subjects) using all available data, which passed spectralquality criteria (Cramer-Rao lower bound �20 and visual inspectionby GE) at that time point, were included in a GLM. Post hoc tests ofmultiple comparisons provided information regarding which of thethree groups differed significantly from the others.

In a third step, paired Student t tests were performed usingpatients (n � 27) with Glu data passing quality criteria availablerom both MRS sessions.

nimalsSixteen male Wistar rats (Charles River, Sulzfeld, Germany), 3

onths of age at the beginning of the experiment were housednder standard conditions (2 per cage, 12-h artificial light– darkycle, lights on 7:00, temperature: 22° � 1° C, humidity: 55% � 5%,ood and water ad libitum). All experimental procedures were ap-roved by the Committee on Animal Care and Use (Regierungsprä-idium Karlsruhe), and carried out in accordance with the localnimal Welfare Act and the European Communities Council Direc-

ive of 24 November 1986 (86/609/EEC).

thanol Vapor ExposureTo induce dependence, rats were exposed to daily intermittent

xposure cycles to alcohol vapor intoxication and withdrawal, a para-igm that allows a high degree of control over brain alcohol levels and

nduces behavioral and molecular changes relevant for the patho- M

hysiology of alcoholism (7,22,23). After the first imaging sessionMRS1 at prealcohol baseline), animals were weight-matched and as-igned into the two experimental groups (n � 8 per condition) forxposure with either ethanol vapor or normal air (weight exposed50.6 � 10.4 g, control subjects 455.6 � 6.2 g) using a rodent alcohol

nhalation system (La Jolla Alcohol Research, La Jolla, California). Expo-ure was similar to that described previously (7) (Supplement 1). Ingreement with previously published data (7,22), the treatment re-ulted in daily oscillating blood alcohol levels from 2.0 to 3.5 g/L to zero.ear the end of the 45-cycle exposure period, signs of withdrawal in

he form of tail stiffness and piloerection were visible toward the end ofhe alcohol off phase. Weight gain during the exposure period wasignificantly less in the alcohol-exposed group (3% of their baselineody weight) compared with control rats (22.5%), leading to signifi-ant differences in body weights at the end of the exposure period464.4�15.4 g and 557.8�7.7 g, exposed vs. control, respectively, p �001). Rats recovered quickly after cessation of the alcohol exposure,hus body weights did not differ significantly between groups after the-week abstinence period (529.3 � 11.8 g and 562.6 � 7.9 g, ns, ex-osed vs. control, respectively).

H MRS at 9.4TRats were scanned repeatedly in a 9.4T horizontal bore animal

canner (Bruker, Ettlingen, Germany) according to the design outlinedn Figure 2A. Anaesthesia was by a gas mixture of O2: 50% and air: 50%

ith approximately 2.1% isoflurane. Respiration rate was monitored,nd body temperature was maintained at 37°C throughout the ses-ion. The position of the voxel (2 � 2 � 3 mm3, Figure 2B) in the mPFCas defined based on anatomic images from a T1-weighted fast imag-

ng with steady precession (FISP) three-dimensional sequence (echoime � 4 msec, repetition time � 8 msec, alpha � 20°, four averages;

atrix � 256 � 256 � 128; field of view � 3 cm).Spectra (Figure 2C) were acquired using a point-resolved spectros-

opy spectrum sequence (repetition time � 4 s). Quantification wasone similar to the human analysis with LCModel spectral fitting (21).or further details, see Methods and Figure S1 in Supplement 1.

tatistical Analysis of Animal DataAnimal weights over the time course of the experiments were

nalyzed by repeated-measures analysis of variance (ANOVA) fol-owed by Tukey post hoc test. Group differences in metaboliteoncentrations were tested using a mixed model, repeated-mea-ures analysis (Proc Mixed module, SAS9.2, SAS, Cary, North Caro-ina) including weight as a covariate. Because the control group wasnly measured three times, values for the missing time points were

nterpolated. Post hoc analysis compared alcohol exposure effectst the various time points against the respective time point of theaive control group through analysis of covariance of the individualetabolite changes over time with factor group and covariateeight. Glu/glutamine (Gln) ratios were analyzed by repeated-mea-

ures ANOVA followed by Tukey post hoc test.

esults

lutamate Levels in the Human ACC Are Increased Duringcute Withdrawal

The GLM revealed significantly increased Glu in the ACC of alco-ol-dependent patients in early withdrawal (MRS-d1) comparedith 2 weeks of abstinence (MRS-d14) and to control subjects

F (2,120) � 8.8, p � .001, adjusted r2 � .11; post hoc tests: MRS-d1s. control subjects: p � .001, MRS-d1 vs. MRS-d14: p � .05, MRS-d14s. control subjects: p � .43; Figure1B, Table 2]. If the analysis isepeated omitting the 12 patients who received clonidine before

RS-d1, results still showed a significant model [F(2,112) � 5.84,

www.sobp.org/journal

vbpsD

bh

pd

(a

MHP

iem2lrF

4 BIOL PSYCHIATRY 2011;xx:xxx D. Hermann et al.

p � .01, post hoc tests: MRS-d1 vs. control subjects: p � .01, MRS-d1s. MRS-d14: p � .104, MRS-d14 vs. control subjects: p � .43]. On theasis of our stringency criteria for metabolite quantification, only 27atients had two valid Glu measurements. A paired t test in thisubgroup confirmed the significant Glu reduction from Day 1 toay 14 (t � 2.2, df � 26, p � .05).

In patients, a positive correlation of Glu levels at MRS-d1 toreath alcohol levels during the MRS scan (average of breath alco-ol level before and after MRS-d1) was found (r2 � .34, p � .001).

However, breath alcohol levels during MRS-d1 were also correlatedto indices of severity of alcoholism like the amount of alcohol con-sumption within the previous 3 months (r2 � .14, p � .05), percent-age of heavy drinking days within the previous 3 months (r2 � .19,

� .01), and the dosage of benzodiazepines needed to treat with-rawal symptoms during the consecutive detoxification period

Table 2. Concentration of Magnetic Resonance Spectro

Glu

Healthy Control Subjects 9.25 � .17Alcohol-Dependent PatientsDay 1 Withdrawal 10.65 � .35a

Day 14 Abstinence 9.66 � .23c

Values are corrected for cerebrospinal fluid contentCho, choline; Cr, creatine; Glu, glutamate; NAA, N-acap � .001.bp � .01 vs. control subjects.c

www.sobp.org/journal

r2 � .18, p � .01). Finally, no correlation was found between Glund measures of withdrawal severity (Supplement 1, Results).

RS Analysis in mPFC of Alcohol-Dependent Rats Parallelsuman Findings and Identifies the Glu/Gln Ratio as aotential Biomarker

Mixed model analysis with weight as a covariate revealed signif-cant differences in Glu levels between treatments for the factorsthanol exposure [F(1,55) � 6.41, p � .01], time point of measure-ent [F(4,55) � 3.38, p � .05], but not for their interaction [F(4,55) �

.03, ns]. Post hoc tests demonstrated significantly increased Gluevels 12 h into withdrawal (F test: p � .01, dependent vs. controlats; F test: p � .05, 12-h withdrawal vs. baseline and vs. intoxication;igure 2D, Table 3).

Figure 2. Alterations in medial prefrontal cortex (mPFC)glutamate (Glu) and glutamine levels in alcohol depen-dent rats. (A) Design of the animal study. The ethanol-exposed group (n � 8) was scanned by magnetic reso-nance spectroscopy (MRS) five times, at baseline(MRS1), at the end of the chronic intermittent ethanolexposure (wavy line) during intoxication in exposurecycle 42/43 (MRS2), 12 h (MRS3) and 60 h (MRS3) afterthe last intoxication cycle, and following a 3-week pe-riod of abstinence (MRS5). Because we had no reason toassume shifting glutamate levels in unchallenged ani-mals during the brief period from MRS2 to MRS4 (5days), the controls (n � 8) went through only three MRSsessions placed within 2 weeks of the time points MRS1, MRS 2– 4, and MRS 5. (B) Typical point-resolved spec-troscopy spectrum acquired from rat mPFC at 9.4T (12�L, echo time � 10 msec, repetition time � 4 s, 256averages) and overlaid by the LCModel fit curve. Peakresonances are as in Figure 1 with the addition of tau-rine (Tau), -aminobutyric acid (GABA). (C) Voxel local-ization. (D) Data represent Glu levels and (E) Glu/Glnratios (mean � SEM) observed during the five sessionsMRS1-5 in the ethanol exposed (open circles) and threesessions in the control groups (closed circles). �p � .05,���p � .001 exposed rats vs. controls measured atsecond time point. Within-subject analysis of varianceof Glu levels in the control group showed no significantresult. EtOH, ethanol.

ic Metabolites in the Human Anterior Cingulate Cortex

NAA Cho Cr

11.91 � .13 2.65 � .05 11.0 � .17

11.05 � .20b 2.70 � .05 11.47 � .1711.88 � .14d 2.63 � .06 11.27 � .17

iven in mmol/L (mean � SEM).partate.

scop

and getylas

f[pdti

wniac.[i4c

Na

(2ieCab

mitqesssd

bsoCatoMrfdw

Fd(ehc

D. Hermann et al. BIOL PSYCHIATRY 2011;xx:xxx 5

The analysis for mPFC Gln (Table 3) levels showed no main effectsor ethanol exposure [F (1,55) � .01, ns], but highly significant timeF (4,55) � 10.52, p � .01] and interaction effects [F (4,55) � 13.09,� .01]. Post hoc analysis revealed significantly decreased Gln in

ependent rats 12 and 60 h into withdrawal (F test: p � .05, at bothime points vs. control subjects; F test: p � .01, 12-h withdrawal vs.ntoxication).

Given the increased Glu and decreased Gln levels during alcoholithdrawal and that these two metabolites are intricately con-ected through the glutamate-glutamine cycle involved in replen-

shing synaptic Glu stores (24), we tested individual Glu/Gln ratioss a potential indicator of increased Glu release. We found signifi-ant effects for the factors ethanol exposure [F (1,55) � 11.75, p �

01], time point [F (4,55) � 9.78, p � .001] and their interactionF (4,55) � 9.79, p � .001]. Tukey post hoc test showed highly signif-cantly increased Glu/Gln ratios during acute withdrawal (4.9 � .51,.8 � .22, and 3.0 � .17, p � .001, 12-h and 60-h withdrawal vs.ontrol subjects, respectively, Figure 2E).

-Acteylaspartate, Choline, and Total Creatine in Humansnd Rats

The effects of alcohol dependence on other major metabolitesi.e., NAA, Cho, tCr) in the humans and rat brain are shown in Tablesand 3. For statistical analysis, see Supplement 1, Results. Compar-

ng data from MRS-d1 (human) and MRS3 (rat), we found consistentffects of early withdrawal on Glu and NAA levels, whereas tCr andho behaved differently between the two species (Figure 3). Alllterations were transient and returned to about normal levels inoth species within a few weeks of abstinence.

Discussion

We are the first to demonstrate increased Glu levels during acutewithdrawal in the ACC in humans. Importantly, this response iscomparably observed in a widely used animal model of alcoholdependence (6), thereby supporting previous inference from ani-mal studies about the neurobiological mechanisms involved in al-cohol withdrawal as a valid construct for the human condition andsupporting the feasibility of a human–animal translational imagingapproach when applied to highly homologous conditions.

Obtaining MRS data from patients in acute alcohol withdrawal ishighly challenging and has to our knowledge so far not beenachieved. The descending limb of the blood alcohol curve providesa brief window in which withdrawal symptoms are beginning to

Table 3. Concentration of Magnetic Resonance Spectro

Glu

BaselineControl 9.90 � .46 2.71EtOH 9.82 � .34 2.54

DependentControl 9.05 � .18 3.07EtOH, intoxication 9.92 � .20 3.72EtOH,12-h withdrawal 10.79 � .21b 2.31EtOH, 60-h withdrawal 10.66 � .29 2.26

PostdependentControl 8.67 � .47 2.95EtOH 9.33 � .54 2.96

Values are given in mmol/L (mean � SEM).Cho, choline; Cr, creatine; EtOH, ethanol; Gln, glutamap � .01 vs. 12-h withdrawal.bp � .05 vs. respective control group.

emerge, yet patient cooperation can still be obtained without phar-ts

acologic treatment. This time point of assessment is critical fornterpreting our findings, because at this time withdrawal symp-oms are highly variable. Indeed, the large number of spectra failinguality criteria for Glu quantification reflects the fact that patientsxperiencing withdrawal symptoms have a much harder time lyingtill during the scanning procedure. Because of these practical is-ues and, more important, for ethical reasons patients could not becanned during the peak of withdrawal symptoms. This limitation isiscussed further in Supplement 1, Results.

Furthermore, the emergence of clinical withdrawal also led to aias for patients with higher estimates of addiction and withdrawaleverity to be scanned at higher breath alcohol levels, becausetherwise, withdrawal symptoms would have become too severe.orrelation of breath alcohol levels to amount and frequency oflcohol consumption and to withdrawal severity supports the exis-ence of this bias. It is likely that this selection bias also underlies thebserved correlation between breath alcohol and Glu levels duringRS-d1. Although we cannot rule out that increased MRS Glu could

esult from the presence of alcohol, our animal data support theormer explanation as Glu levels during intoxication in ethanol-ependent rats did not differ from naive control subjects but roseithin 12 hours of withdrawal from alcohol.

ic Metabolites in the Rat Medial Prefrontal Cortex

NAA Cho Cr

9.99 � .37 1.65 � .04 7.53 � .249.80 � .36 1.65 � .05 7.78 � .15

10.40 � .11 1.71 � .04 7.75 � .17a 9.27 � .32b 2.07 � .08 7.08 � .18b 10.13 � .36 1.92 � .06 7.35 � .22b 10.59 � .15 1.87 � .03 7.56 � .17

9.57 � .52 1.67 � .13 7.86 � .319.82 � .29 1.74 � .05 7.95 � .15

lu, glutamate; NAA, N-acetylaspartate.

igure 3. Metabolic profiles in human and rats during acute alcohol with-rawal. The diagram summarizes the relative deviation (%) of glutamate

Glu), N-acetylaspartate (tNAA), total creatine (tCr), and choline (Cho) duringarly withdrawal (MRS-d1 in humans, MRS3 in rats) from their respectiveuman or rat control groups. ��p � .01, ���p � .001 MRS-d1 vs. healthyontrol subjects; #p � .05 MRS3 exposed vs. controls. Metabolite concentra-

scop

Gln

� .21� .13

� .17� .27� .16� .13

� .24� .06

ine; G

ions and detailed statistics are in Tables 2 and 3. MRS, magnetic resonancepectroscopy.

www.sobp.org/journal

viGMdcntiidnrGelehd

ts2(fadsdbps

lMirdenZidNcnturtAhes

pltaaprossici

tNwTuF

c

6 BIOL PSYCHIATRY 2011;xx:xxx D. Hermann et al.

Glu levels apparently normalize after 2 weeks of abstinence. Twoother MRS studies of human alcoholics have measured Glu (25,26).Mason et al. (25) found no effect of alcohol dependence or time ofabstinence (1 vs. 4 weeks) on central Glx (Glu � Gln) levels. Interest-ingly, this group also reported increased central -aminobutyricacid levels after 1 week of abstinence in nonsmoking, but not smok-ing, alcohol-dependent subjects compared with healthy controlsubjects. Levels normalized after 4 weeks of abstinence. Umhau etal. (26) reported a trend toward increased Glu levels in the ACC(albeit relative to tCr levels) after 4 weeks of abstinence. In this smallclinical trial, acamprosate robustly decreased Glu levels in patients.However, the first measurement was taken after acute withdrawalhad subsided, and no healthy control subjects were included,which makes a comparison with our data difficult.

Our animal data allow further insight into the nature of theobserved human imaging phenomena. Using an absolute quantifi-cation approach based on the internal water signal we obtainedhighly comparable Glu estimates in both species under normalconditions and in ethanol withdrawal as well as over the course ofabstinence. The higher field strength used for MRS in rats allowedus to estimate simultaneously Gln, a metabolite that is intricatelyrelated to the synaptic pool of Glu through the so called glutamate–glutamine shuttle between neurons and astrocytes (24,27,28). Afterrelease from synaptic vesicles, Glu is taken up by astrocytes andmetabolized into Gln, which on demand is transferred to neurons,converted back into Glu, and used for refilling the synaptic vesicles(24). Although there is no stoichiometric relationship betweenbrain Glu and Gln levels, the majority of synaptic Glu is derived fromthis cycle (27,28). In our study, we found increased Glu and de-creased Gln levels during acute withdrawal, which may point toalterations in the glutamate– glutamine cycle. Further investiga-tions into the role of Glu–Gln cycling and its implication for gluta-mate hyperactivity in alcoholism will require methods for assessingrate-based processes, for example, by 13C-MRS (28,29).

Clearly, MRS cannot differentiate between the various Glu poolsin the brain. However, it is worth noting that the observed magni-tude in Glu signal enhancement (about 10%) could indeed resultfrom the synaptic Glu pool. This pool may hold nearly half of the Glucontent of a neuron. This assumption is based on the more then20-fold higher Glu concentration within the synaptic vesicles com-pared with the cytoplasm (30,31) and data indicating that the total

olume of synaptic vesicles of a neuron may amount to about 5% ofts perikaryon (32–34). Here, the withdrawal-induced alterations inlu and Gln persisted for at least 3 days. The time course of theseRS metabolite levels is clearly different from previously observed

uration of physical withdrawal signs or extracellular Glu levels (2);onsequently, the observed changes in MRS Glu and Gln levels mayot directly reflect transmitter release, but a process closely related

o excitatory neurotransmission such as replenishing synaptic ves-cles. Further support for this conclusion is offered from the exper-mental medicine study by Umhau et al. (26), who found that MRSerived Glu measures are sensitive to modulation of glutamatergiceurotransmission by acamprosate, but are uncorrelated to Glu as

eflected by cerebrospinal fluid level. The observed alterations inlu and Gln levels during acute withdrawal and into abstinence arexpected to impact on synaptic and extrasynaptic glutamate with

ong-lasting consequences for the glutamate homeostasis. Suchnduring imbalances, particularly within corticostriatal circuits, areypothesized to underlie the difficulties for addicts in modulatingrug-seeking behavior (35).

Additional animal studies assessing central Glu levels by MRSduring alcohol withdrawal are so far lacking. However, two reports

www.sobp.org/journal

oxication paradigm, similar to the one used in our study, animalshowed progressively increasing Glu levels over a period from 16 to4 weeks of alcohol exposure compared with naive control subjects

36). In contrast, after a 4-day binge intoxication paradigm, no dif-erences in Glu levels were observed (37). Together with our data, itppears that intoxication-related rise in Glu requires prolongeduration to sufficiently high levels of alcohol. Also, the Glu/Gln ratioeemed unaffected by alcohol intoxication but increased rapidlyuring acute withdrawal and remained high even a short periodeyond withdrawal. It remains to be seen whether this responserovides information that could be used clinically as a biomarker foreverity of alcohol dependence.

Apart from Glu, several other metabolites, including NAA, cho-ine, and creatine have been repeatedly assessed in alcoholics by

RS and found to deviate from levels in healthy subjects as well asn animal models of alcohol addiction. These findings have beenecently reviewed (38). In summary and in line with our study,etoxified patients consistently show decreased NAA levels in sev-ral brain regions, including frontal brain lobes, which increase orormalize within some weeks of abstinence. The animal study ofahr et al. (37) as well as our study indicate that excessive alcohol

ntake is causally linked to decreased NAA. Both studies foundecreased NAA in intoxicated animals as well as a rapid recovery ofAA with discontinuation of alcohol exposure. Alterations in Chooncentration by alcohol dependence and during abstinence doot appear to be consistent across studies. The significance of the

emporal Cho level dynamics as a function of drinking severity isnclear. They are likely complicated by the fact that Cho levels

eflect different biological processes, including cellular membraneurnover and density, as well as myelin anabolism and catabolism.lterations in total creatine have so far not been reported in alco-olics. Studies in rats have been conducted under highly variablexperimental conditions, which make generalization of these re-ults difficult.

This study contributes importantly to our understanding of theathophysiology of alcoholism by validating a mechanism estab-

ished in animal models for the human condition, that is, hyperac-ivity of glutamatergic neurotransmission during withdrawal. Inddition, our data show a strongly increased Glu/Gln ratio duringlcohol withdrawal in the rat mPFC. Future studies including MRSrotocols suitable for assessing the Glu/Gln ratio in humans will

esolve whether this parameter is informative about componentsf the addiction process. Finally, combined animal and human MRStudies as well as other neuroimaging approaches seem to have atrong and valid translational component, essentially demonstrat-ng that a given animal model is indeed relevant to the humanondition and therefore useful for expanding mechanistic insight

nto psychiatric disorders.

This study was funded by Deutsche Forschungsgemeinschafthrough a center grant (No. SFB636) and within the frameworks ofationales Genomforschungsnetz (NGFN� No. 01GS08152, see underww.ngfn-alkohol.de and Spanagel et al. 2010) (39) and ERA-Net

ransaltional Neuroimaging in Alcoholism (TRANSALC 01EW1112, seender www.transalc.eu) by the Bundesministerium für Bildung undorschung.

The authors report no biomedical financial interests or potentialonflicts of interest.

Supplementary material cited in this article is available online.

1. Spanagel R (2009): Alcoholism: A systems approach from molecularphysiology to addictive behavior. Physiol Rev 89:649 –705.

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

4

D. Hermann et al. BIOL PSYCHIATRY 2011;xx:xxx 7

2. Rossetti ZL, Carboni S (1995): Ethanol withdrawal is associated withincreased extracellular glutamate in the rat striatum. Eur J Pharmacol283:177–183.

3. De Witte P, Pinto E, Ansseau M, Verbanck P (2003): Alcohol and with-drawal: From animal research to clinical issues. Neurosci Biobehav Rev27:189 –197.

4. Gass JT, Olive MF (2008): Glutamatergic substrates of drug addictionand alcoholism. Biochem Pharmacol 75:218 –265.

5. Krupitsky EM, Rudenko AA, Burakov AM, Slavina TY, Grinenko AA, Pitt-man B, et al. (2007): Antiglutamatergic strategies for ethanol detoxifica-tion: Comparison with placebo and diazepam. Alcohol Clin Exp Res 31:604 – 611.

6. Björk K, Hansson AC, Sommer WH (2010): Genetic variation and braingene expression in rodent models of alcoholism implications for medi-cation development. Int Rev Neurobiol 91:129 –171.

7. Rimondini R, Arlinde C, Sommer W, Heilig M (2002): Long-lasting in-crease in voluntary ethanol consumption and transcriptional regulationin the rat brain after intermittent exposure to alcohol. FASEB J 16:27–35.

8. Breese GR, Overstreet DH, Knapp DJ (2005): Conceptual framework forthe etiology of alcoholism: A “kindling”/stress hypothesis. Psychophar-macology (Berl) 178:367–380.

9. Heilig M, Egli M, Crabbe JC, Becker HC (2010): Acute withdrawal, pro-tracted abstinence and negative affect in alcoholism: Are they linked?Addict Biol 15:169 –184.

10. Heinz A, Beck A, Grusser SM, Grace AA, Wrase J (2009): Identifying theneural circuitry of alcohol craving and relapse vulnerability. Addict Biol14:108 –118.

11. Auer DP, Putz B, Kraft E, Lipinski B, Schill J, Holsboer F (2000): Reducedglutamate in the anterior cingulate cortex in depression: An in vivoproton magnetic resonance spectroscopy study. Biol Psychiatry 47:305–313.

12. Hasler G, van der Veen JW, Tumonis T, Meyers N, Shen J, Drevets WC(2007): Reduced prefrontal glutamate/glutamine and gamma-ami-nobutyric acid levels in major depression determined using protonmagnetic resonance spectroscopy. Arch Gen Psychiatry 64:193–200.

13. Portella MJ, de Diego-Adelino J, Gomez-Anson B, Morgan-Ferrando R,Vives Y, Puigdemont D, et al. (2011): Ventromedial prefrontal spectro-scopic abnormalities over the course of depression: A comparisonamong first episode, remitted recurrent and chronic patients. J PsychiatrRes 45:427– 434.

14. Pfleiderer B, Michael N, Erfurth A, Ohrmann P, Hohmann U, Wolgast M, etal. (2003): Effective electroconvulsive therapy reverses glutamate/glu-tamine deficit in the left anterior cingulum of unipolar depressed pa-tients. Psychiatry Res 122:185–192.

15. Duncan NW, Enzi B, Wiebking C, Northoff G (2011): Involvement ofglutamate in rest-stimulus interaction between perigenual and su-pragenual anterior cingulate cortex: A combined fMRI-MRS study [pub-lished online ahead of print February 8]. Hum Brain Mapp.

16. Hoerst M, Weber-Fahr W, Tunc-Skarka N, Ruf M, Bohus M, Schmahl C,Ende G (2010): Correlation of glutamate levels in the anterior cingulatecortex with self-reported impulsivity in patients with borderline person-ality disorder and healthy controls. Arch Gen Psychiatry 67:946 –954.

17. Vollstadt-Klein S, Hermann D, Rabinstein J, Wichert S, Klein O, Ende G,Mann K (2010): Increased activation of the ACC during a spatial workingmemory task in alcohol-dependence versus heavy social drinking. Alco-hol Clin Exp Res 34:771–776.

18. Hermann D, Frischknecht U, Heinrich M, Hoerst M, Vollmert C, Vollstadt-Klein S, et al. (2011): MR spectroscopy in opiate maintenance therapy:Association of glutamate with the number of previous withdrawals inthe anterior cingulate cortex [published online ahead of print February11]. Addict Biol.

19. Sullivan JT, Sykora K, Schneiderman J, Naranjo CA, Sellers EM (1989):Assessment of alcohol withdrawal: The revised clinical institute with-drawal assessment for alcohol scale (CIWA-Ar). Br J Addict 84:1353–1357.

20. Schubert F, Gallinat J, Seifert F, Rinneberg H (2004): Glutamate concen-trations in human brain using single voxel proton magnetic resonancespectroscopy at 3 Tesla. Neuroimage 21:1762–1771.

21. Provencher SW (1993): Estimation of metabolite concentrations fromlocalized in vivo proton NMR spectra. Magn Reson Med 30:672– 679.

22. Sommer WH, Rimondini R, Hansson AC, Hipskind PA, Gehlert DR, BarrCS, Heilig M (2008): Upregulation of voluntary alcohol intake, behavioralsensitivity to stress, and amygdala Crhr1 expression following a history

3. Heilig M, Thorsell A, Sommer WH, Hansson AC, Ramchandani VA,George DT, et al. (2010): Translating the neuroscience of alcoholism intoclinical treatments: From blocking the buzz to curing the blues. NeurosciBiobehav Rev 35:332–344.

4. Broman J, Hassel B, Rinvik E, Ottersen OP (2000): Biochemistry andanatomy of transmitter glutamate. In Ottersen OP, Storm-Mathisen J.,editors. Glutamate. Amsterdam: Elsevier, 1– 44.

5. Mason GF, Petrakis IL, de Graaf RA, Gueorguieva R, Guidone E, Coric V, etal. (2006): Cortical gamma-aminobutyric acid levels and the recoveryfrom ethanol dependence: Preliminary evidence of modification bycigarette smoking. Biol Psychiatry 59:85–93.

6. Umhau JC, Momenan R, Schwandt ML, Singley E, Lifshitz M, Doty L, et al.(2010): Effect of acamprosate on magnetic resonance spectroscopymeasures of central glutamate in detoxified alcohol-dependent individ-uals: A randomized controlled experimental medicine study. Arch GenPsychiatry 67:1069 –1077.

7. Kvamme E (1998): Synthesis of glutamate and its regulation. Prog BrainRes 116:73– 85.

8. Rothman DL, Behar KL, Hyder F, Shulman RG (2003): In vivo NMR studiesof the glutamate neurotransmitter flux and neuroenergetics: Implica-tions for brain function. Annu Rev Physiol 65:401– 427.

9. Zwingmann C, Leibfritz D (2003): Regulation of glial metabolism studiedby 13C-NMR. NMR Biomed 16:370 –399.

0. Ottersen OP, Zhang N, Walberg F (1992): Metabolic compartmentationof glutamate and glutamine: morphological evidence obtained byquantitative immunocytochemistry in rat cerebellum. Neuroscience 46:519 –534.

1. Gras G, Porcheray F, Samah B, Leone C (2006): The glutamate-glutaminecycle as an inducible, protective face of macrophage activation. J LeukocBiol 80:1067–1075.

2. Pakkenberg B, Pelvig D, Marner L, Bundgaard MJ, Gundersen HJ, Nyen-gaard JR, Regeur L (2003): Aging and the human neocortex. Exp Gerontol38:95–99.

3. Ikeda K, Bekkers JM (2009): Counting the number of releasable synapticvesicles in a presynaptic terminal. Proc Natl Acad Sci U S A 106:2945–2950.

4. Qu L, Akbergenova Y, Hu Y, Schikorski T (2009): Synapse-to-synapsevariation in mean synaptic vesicle size and its relationship with synapticmorphology and function. J Comp Neurol 514:343–352.

5. Kalivas PW (2009): The glutamate homeostasis hypothesis of addiction.Nat Rev Neurosci 10:561–572.

6. Zahr NM, Mayer D, Vinco S, Orduna J, Luong R, Sullivan EV, PfefferbaumA (2009): In vivo evidence for alcohol-induced neurochemical changesin rat brain without protracted withdrawal, pronounced thiamine defi-ciency, or severe liver damage. Neuropsychopharmacology 34:1427–1442.

7. Zahr NM, Mayer D, Rohlfing T, Hasak MP, Hsu O, Vinco S, et al. (2010):Brain injury and recovery following binge ethanol: Evidence from in vivomagnetic resonance spectroscopy. Biol Psychiatry 67:846 – 854.

8. Meyerhoff DJ, Durazzo TC, Ende G (2010): Chronic alcohol consumption,abstinence and relapse: Brain proton magnetic resonance spectroscopystudies in animals and humans. In: Sommer WH, Spanagel R, editors.Behavioral Neurobiology of Alcohol Addiction. New York: Springer.

9. Spanagel R, Bartsch D, Brors B, Dahmen N, Deussing J, Eils R, et al. (2010):An integrated genome research network for studying the genetics ofalcohol addiction. Addict Biol 15:369 –379.

0. Skinner HA, Allen BA (1982): Alcohol dependence syndrome: M0000-easurement and validation. J Abnorm Psychol 91:199 –209.

1. Saunders JB, Aasland OG, Babor TF, Delafuente JR, Grant M (1993):Development of the Alcohol Use Disorders Identification Test (AUDIT):WHO Collaborative Project on Early Detection of Persons with HarmfulAlcohol Consumption—II. Addiction 88:791– 804.

2. Hautzinger M, Bailer M, Worall H, Keller F. (1994): Beck-Depression-Inven-tory (BDI). Bern: Huber.

3. Heatherton TF, Kozlowski LT, Frecker RC, Fagerstrom KO (1991): TheFagerstrom Test for Nicotine Dependence: A revision of the FagerstromTolerance Questionnaire. Br J Addict 86:1119 –1127.

4. Mann K, Ackermann K. (2000): The OCDS-G: Psychometric Characteris-tics of the German Version of the Obsessive Drinking Scale. Sucht 46:90 –100.

5. Laux L, Glanzmann P, Schaffner P, Spielberger CD (1981): The State-Trait

www.sobp.org/journal