Cortical γ-Aminobutyric Acid Type A–Benzodiazepine Receptors in Recovery From Alcohol Dependence: Relationship to Features of Alcohol Dependence and Cigarette Smoking

Context  Adaptations in γ-aminobutyric acid type A (GABA_A)–benzodiazepine receptors contribute to the neurobiology of human alcohol dependence and withdrawal.

Objective  To study GABA_A-benzodiazepine receptor adaptations in subjects with alcohol dependence over the first month of sobriety.

Design  Inpatients who were not receiving medication, were either smokers or nonsmokers, and had alcohol dependence completed 2 iodine I 123–labeled iomazenil single-photon emission computed tomographic scans: 1 scan at a mean ± SD of 4.9 ± 2.5 days of sobriety (n = 23) and 1 scan at a mean ± SD of 29.8 ± 7.6 days of sobriety (n = 20). Participants in a matched group of healthy subjects (n = 15) completed 1 single-photon emission computed tomographic scan.

Participants  Men with alcohol dependence (n = 27) and a matched healthy comparison group (n = 15).

Main Outcome Measures  ¹²³I-iomazenil single-photon emission computed tomographic images were converted to units of distribution volume (regional activity/free ¹²³I-iomazenil) and were analyzed using voxel-based statistical parametric mapping and regions of interest analyses. The relationships between ¹²³I-iomazenil distribution volume, clinical features of alcohol dependence, and smoking status were evaluated.

Results  ¹²³I-iomazenil uptake was elevated in several cortical regions, with a more prominent increase in nonsmokers with alcohol dependence as compared with smokers with alcohol dependence at 1 week of abstinence from alcohol. No significant differences were observed at 4 weeks of abstinence. At 1 week of abstinence, frontal ¹²³I-iomazenil uptake correlated with the severity of alcohol withdrawal and the number of days since the last alcoholic drink was consumed. No significant associations were observed for smokers with alcohol dependence.

Conclusions  These data demonstrate time-dependent regulation of cortical GABA_A-benzodiazepine receptors associated with the recovery from alcohol dependence. Higher GABA_A-benzodiazepine receptor levels during acute withdrawal may reflect a compensation for reduced receptor function, which is thought to contribute to alcohol tolerance and withdrawal. The subsequent decline may reflect “normalization” of GABA_A receptor function with sobriety. Smoking may attenuate GABA_A receptor adaptations associated with alcohol dependence and may contribute to the comorbidity between alcoholism and smoking.

Several lines of evidence implicate γ-aminobutyric acid type A (GABA_A) receptors in the neurobiology of alcohol tolerance and dependence, including (1) facilitatory effects of ethanol on γ_2L-containing GABA_A receptors,¹ (2) generalization between the discriminative stimulus effects of ethanol and GABA_A receptor agonists, (3) cross-tolerance between ethanol and benzodiazepines (BZs), and (4) the ability of BZs and other GABA_A receptor agonists to suppress alcohol withdrawal.^2,3 In fact, BZs are the principal pharmacotherapy for alcohol withdrawal symptoms.^4-6 While there is controversy regarding whether ethanol has primarily direct or indirect facilitatory effects on GABA_A receptors,^2,7,8 recent evidence of potent direct ethanol effects on GABA_A receptor subgroups and the mapping of molecular domains of GABA_A receptors through which ethanol acts⁹ highlight the importance of GABA_A receptors for ethanol action. Ethanol and BZ enhance GABAergic function through distinct interactions with the GABA_A receptor complex.¹⁰

Neuroimaging studies of patients with alcohol dependence who have been sober for at least 1 month generally suggest that GABA_A receptors are decreased. All^11-14 but one¹⁵ of these studies demonstrated decreased binding of the BZ antagonists carbon 11–labeled flumazenil and iodine I 123–labeled iomazenil in frontal, parietal, and temporal cortices. These studies may be consistent with cerebral metabolic data suggesting that the response to lorazepam is decreased in patients who have been sober for 6 to 32 days¹⁶ and continues to decrease over 2 months of abstinence.¹⁷ However, it is possible that both the receptor-based and metabolic neuroimaging findings could reflect neurotoxic consequences of alcoholism or GABA-related disturbances that preceded the onset of alcohol dependence. Human postmortem studies report unchanged,^18,19 increased,²⁰ and decreased^21,22 binding of BZ agonists in the frontal cortex, hippocampus, and cerebellum. In a small study of recently detoxified patients with alcohol dependence, 2 subjects demonstrated increased GABA_A-BZ receptor availability whereas the other 3 subjects exhibited unchanged or slightly decreased GABA_A-BZ receptor availability.²³ When reviewed collectively with the earlier in vivo studies in subjects with alcoholism who have been abstinent for greater than 1 month, these studies suggest that levels of GABA_A-BZ receptors are elevated or similar to control levels during acute abstinence and decreased with prolonged abstinence, and they also suggest that factors other than the length of abstinence may alter the levels of GABA_A-BZ receptors during alcohol withdrawal.

We hypothesized that some of the conflicts in the existing literature might reflect the confounding effects of time-dependent changes in GABA_A-receptor populations during recovery (return to baseline) from alcohol dependence as well as other clinical features, including severity of alcohol dependence and withdrawal, subtype of alcoholism (reflected in age at onset), and the frequent comorbidity of alcohol dependence and tobacco smoking. The objectives of the current study were to directly evaluate whether GABA_A-BZ–receptor levels change over time during recovery from alcohol dependence, to explore the relationships to clinical correlates of alcohol dependence, and to determine if tobacco smoking modifies GABA_A-BZ receptors during early sobriety.

Thirty-six men with alcohol dependence and 15 age-matched healthy men provided written informed consent for participation in this study. The study was approved by the Department of Psychiatry Research Committee, Human Investigation Committee, and Radiation Safety Committee, Yale University School of Medicine (New Haven, Conn), and the Human Subjects Subcommittee of the West Haven Veterans Affairs Connecticut Healthcare System (West Haven, Conn). Eligibility was evaluated with a structured interview, physical examination, laboratory blood tests, urine drug screen, and an electrocardiogram. Control subjects had no history or evidence of a serious medical or neurological illness, did not meet criteria for any psychiatric or substance abuse diagnosis as determined by the Structured Clinical Interview for DSM-IV nonpatient version, had no history of a DSM-IV Axis I psychiatric disorder in a first-degree relative, and had not used psychotropic substances other than alcohol, nicotine, or tobacco for at least 1 year preceding the study. Control subjects drank less than 8 alcoholic drinks per week as defined by the timeline follow-back (ie, 1 drink = 12 oz of beer, 5 oz of wine [10% alcohol], 1.5 oz of hard liquor, etc)²⁴ and were asked to refrain from alcoholic and caffeine-containing drinks for 2 weeks preceding the single-photon emission computed tomographic (SPECT) scan.

Participants in the alcohol-dependent group met DSM-IV criteria for alcohol dependence and currently had no other Axis I disorder other than nicotine dependence, as determined by the Structured Clinical Interview for DSM-IV.²⁵ They had no medical or neurological disorders that might compromise the interpretation of the SPECT data, had not taken any psychotropic medications in the last month, and demonstrated alcohol withdrawal symptoms of a Clinical Institute Withdrawal Assessment⁴ score of less than 10 at the initial assessment. Subjects with alcohol dependence were accepted with abnormal liver function test values for γ-glutamyltransferase, serum glutamic oxaloacetic transaminase, and serum glutamate pyruvate transaminase of up to 3 times the upper normal limit in recognition of the effect of recent drinking on liver function.

Subjects with alcohol dependence were admitted into the Connecticut Mental Health Center Clinical Neuroscience Research Unit (New Haven) for 1 month. They were monitored for alcohol withdrawal symptoms using the Clinical Institute Withdrawal Assessment every 6 hours after admission until scores were consistently 0. Patients with Clinical Institute Withdrawal Assessment ratings of 10 or higher were evaluated to determine if immediate BZ administration was necessary. The BZs were automatically given if the Clinical Institute Withdrawal Assessment score exceeded 12. Once BZs were administered, no subsequent tests were performed for at least 1 month; ie, the first imaging session was cancelled and, since long-term BZ administration does not alter GABA_A-BZ receptor levels,²⁶ subjects were encouraged to participate in the second scan. Of the 36 patients who signed consent forms, 10 subjects were excluded for the following reasons: (1) withdrew consent and terminated participation before completing an imaging session; (2) did not have a magnetic resonance image taken; (3) had abnormal magnetic resonance image results; (4) had first-degree family history of depression; (5) had cocaine use discovered after the study; or (6) there was difficulty in placing the intravenous needle and/or collecting blood samples during the SPECT scan. Three subjects with alcohol dependence participated in the second scan but did not participate in the first scan because they received BZ to treat withdrawal symptoms, the camera was not available (because of a change in location), and entrance into the study was after the first week of abstinence. Six subjects with alcohol dependence dropped out after the first scan and did not participate in the second scan. Alcohol consumption in the month preceding the SPECT scan was determined using the timeline follow-back.²⁴ Age at onset of alcohol dependence was obtained from the Structured Clinical Interview for DSM-IV. Antisocial personality disorder was determined using the Structured Clinical Interview for DSM-IV Axis II. Severity of alcohol dependence was assessed using the Alcohol Dependence Scale,²⁷ and alcohol craving was assessed on the day of the SPECT scan using the Tiffany Scale.²⁸ Subjects with alcohol dependence had SPECT images taken within 10 days and/or at 4 weeks after their last alcoholic drink.

Smokers in the control group and the alcohol-dependent group were instructed to maintain their normal smoking patterns. Inpatient smokers with alcohol dependence were limited to smoking 4 times per day. Smoking characteristics, including the ages at which the subjects began smoking, the total number of years they smoked, the number of cigarettes smoked per day, and the number of attempts to quit smoking, were documented. Nicotine dependence was evaluated using the Fagerstrom Test for Nicotine Dependence.^29,30 Plasma cotinine (the principal metabolite of nicotine) was measured on the day of the SPECT scan. All nonsmokers with alcohol dependence and 8 of the 10 control nonsmokers were “never smokers” as defined by a lifetime history of smoking less than 40 cigarettes or nicotine-related products.

Cotinine concentrations in serum were assayed using reverse-phase high-performance liquid chromatography. The procedure³¹ was modified by substitution of an aqueous microvolume back-extraction clean-up step in place of solvent evaporation. Following the addition of an internal standard (2-phenylimidazole), cotinine was extracted from alkalinized serum with a 40:60 mixture of dichloromethane and hexane. Following a microvolume back-extraction into 0.1M phosphoric acid, the aqueous phase was analyzed by reverse-phase chromatography on a C6 reverse-phase column using a mobile phase of 10% acetonitrile buffered to pH 4.8 and containing 20 mL of triethylamine and 0.6 g/L of octane-sulfonic acid. Between-day coefficients of variation, in routine use at concentrations of 20 μg/L and 200 μg/L, were 11.6% and 6.6%, respectively.

¹²³I-iomazenil was prepared as described previously.³² The average yield was a mean ± SD of 63.0% ± 11.7% (n = 57 preparations) and its radiochemical purity was a mean ± SD of 97.8% ± 1.8%. All participants were pretreated with saturated potassium iodide to reduce thyroid uptake of ¹²³I. ¹²³I-iomazenil was administered by intravenous injection (71.3 ± 2.1 MBq, 72.6 ± 1.9 MBq, and 72.5 ± 1.5 MBq for subjects with alcohol dependence at 1 week of abstinence, subjects with alcohol dependence at 4 weeks of abstinence, and control subjects, respectively) plus constant infusion (275.9 ± 7.3 MBq/h, 280.1 ± 7.7 MBq/h, and 278.2 ± 5.1 MBq/h, respectively). The duration of infusion was 6.5 hours with a mean ± SD total injected dose of 221.2 ± 3.1 MBq for subjects with alcohol dependence and 223.2 ± 1.8 MBq for control subjects. Prior to imaging, 5 external fiducial markers containing 0.037 MBq of ¹²³I were placed on the scalp to provide a common reference for coregistration with emission images. Simultaneous transmission and emission scans were acquired with a line source containing 740 MBq of cobalt 57 on a PRISM 3000XP 3-headed camera (Picker International, Cleveland, Ohio). Three consecutive 12-minute emission scans were acquired in continuous mode between 330 and 370 minutes of ¹²³I-iomazenil infusion. Three venous blood samples were collected at the midpoint of the SPECT scan (at 350 minutes) for measurement of free ¹²³I-iomazenil.³²¹²³I-iomazenil SPECT has 83% to 90% reproducibility.³³

Magnetic resonance imaging was performed on a Signa 1.5T system (General Electric Co, Milwaukee, Wis). Axial images were acquired parallel to the anteroposterior commissural line with an echo time of 5 milliseconds; repetition time of 24 milliseconds; matrix 256 × 192; number of excitations of 1; field of view of 24 cm; and 124 contiguous slices with a thickness of 1.2 mm.

Emission data from SPECT scans were reconstructed from counts acquired in the ¹²³I photopeak (159 keV) with a 20% symmetric window using a Butterworth filter (power factor = 10, cutoff = 0.24 cm). An attenuation map was reconstructed from the transmission and flood data and was used for nonuniform attenuation correction.³⁴ The magnetic resonance image was coregistered to the emission image and reoriented to the intercommissural plane. Standardized 2-dimensional region of interest templates were placed on the medial frontal cortex (12.3 cm²), anterior cingulate (3.1 cm²), hippocampal-amygdala area (5.4 cm²), and cerebellum (25 cm²). Three-dimensional volumes of interest were transferred to coregistered SPECT images. Regional activities from right and left hemispheres were averaged, decay corrected, and expressed as kilobecquerels per cubic centimeter using a calibration factor of 337.44 Bq/cpm derived from a ¹²³I-distributed source phantom. Regional activities (kilobecquerels per cubic centimeter) were normalized to free ¹²³I-iomazenil (kilobecquerels per milliliter) plasma levels.³⁵

A voxel-based analysis was conducted using statistical parametric mapping (SPM99; FIL, Wellcome Department of Imaging Neuroscience, University College of London, London, England). For each subject, a mean image was made from the 3 ¹²³I-iomazenil emission scans scaled to the total volume of distribution (V_T; milliliters per cubic centimeter). A mean iomazenil V_T template was made from all study subjects and was spatially normalized to the SPECT blood flow template in SPM99 using a mask to occlude uncommon brain areas. Thereafter, each mean V_T map from each subject was spatially normalized to the mean iomazenil SPECT V_T template and smoothed using a gaussian kernel of 12 mm × 12 mm × 12 mm. Statistical analyses between comparison groups were conducted without global normalization. Overall effects were determined using an F threshold of 300 between 5 groups, including controls, smokers with alcohol dependence who were abstinent from alcohol for less than 1 week, nonsmokers with alcohol dependence who were abstinent from alcohol for less than 1 week, smokers with alcohol dependence who were abstinent from alcohol for less than 4 weeks, and nonsmokers with alcohol dependence who were abstinent from alcohol for less than 4 weeks. The t statistic images were thresholded to a minimum cluster size of 42 voxels, calculated based on the nominal resolution, ie, the full width at half maximum of a point source in water in the Picker Prism 3000XP SPECT camera. Between-group analyses were conducted comparing subjects in the control and alcohol-dependent groups. In the alcohol-dependent groups, the relationship between ¹²³I-iomazenil V_T and various clinical measures of alcohol dependence were explored. The anatomic location of the most significant voxel in clusters demonstrating statistical significance after corrections for multiple comparisons was determined after conversion of Montreal Neurological Institute stereotaxic coordinates to Talairach coordinates and submission of the results to Talairach Daemon.^36,37

Region of interest data were analyzed using SAS version 8.2 (SAS Institute Inc, Cary, NC). Data were examined for normality using normal probability plots and Kolmogorov-Smirnov test statistics. Differences in V_T between subjects in the control and alcohol-dependent groups and between smokers and nonsmokers were assessed using a mixed-effects regression model with ¹²³I-iomazenil V_T as the dependent variable and with group (subjects with alcohol dependence or control subjects), smoking status (yes or no), region of interest, and interactions as fixed factors and unstructured variance-covariance matrix across regions. Nonsignificant interactions were dropped from the model for parsimony. Age was considered as a covariate but was not significant and was dropped from the model. Correlations with clinical variables for the patients were also assessed using mixed models with ¹²³I-iomazenil V_T as the dependent variable and smoking, region of interest, 1 clinical variable at a time, and all possible interactions as independent variables. Different slopes for smokers and nonsmokers were estimated to illustrate the relationship between significant clinical variables and the dependent variable. Statistical significance was considered at P = .05. Bonferroni correction was used to correct for multiple comparisons for post hoc region of interest analyses.

Twenty-three men with alcohol dependence, including 15 smokers and 8 nonsmokers, and 15 control men, including 5 smokers and 10 nonsmokers, were included in the clinical population (Table 1). Smokers in the alcohol-dependent and control groups demonstrated similar Fagerstrom Test for Nicotine Dependence scores. Smokers with alcohol dependence tended to have lower plasma cotinine levels compared with control smokers (t₁₇ = 2, P = .06). Subjects with alcohol dependence reported smoking, on average, 1 pack of cigarettes per day, which was similar to control smokers; however, once they became inpatients where they had up to 4 escorted “smoking breaks” each day, they were not able to match their outpatient smoking frequency, as demonstrated by differences in plasma cotinine levels and the severity of nicotine withdrawal (Table 1). More than half of the men with alcohol dependence exhibited a paternal history of alcohol dependence, with a higher proportion of this family history in nonsmokers as compared with smokers. The average age at onset, the severity of alcohol dependence, and alcohol craving were similar between smokers and nonsmokers. An insignificant trend toward greater severity of alcohol withdrawal symptoms was noted in the nonsmokers as compared with smokers (Table 2). Smokers with alcohol dependence drank more alcohol than nonsmokers with alcohol dependence in the past month.

There were no between-group differences in the injected dose, bolus-infusion ratio, and time of scan (data not shown). There was a trend difference in mean ± SD levels of total ¹²³I-iomazenil: control nonsmokers, 0.18 ± 0.04 kBq/mL; control smokers, 0.16 ± 0.03 kBq/mL; nonsmokers with alcohol dependence who were abstinent from alcohol for less than 1 week, 0.15 ± 0.03 kBq/mL; smokers with alcohol dependence who were abstinent from alcohol for less than 1 week, 0.14 ± 0.03 kBq/mL; nonsmokers with alcohol dependence who were abstinent from alcohol for less than 4 weeks, 0.16 ± 0.05 kBq/mL; and smokers with alcohol dependence who were abstinent from alcohol for less than 4 weeks, 0.14 ± 0.04 kBq/mL; where F_2,31 = 3.06 and P = .06. There was also a trend difference in mean ± SD levels of free ¹²³I-iomazenil: control nonsmokers, 0.07 ± 0.02 kBq/mL; control smokers, 0.05 ± 0.01 kBq/mL; nonsmokers with alcohol dependence who were abstinent from alcohol for less than 1 week, 0.05 ± 0.01 kBq/mL; smokers with alcohol dependence who were abstinent from alcohol for less than 1 week, 0.05 ± 0.01 kBq/mL; nonsmokers with alcohol dependence who were abstinent from alcohol for less than 4 weeks, 0.05 ± 0.01 kBq/mL; and smokers with alcohol dependence who were abstinent from alcohol for less than 4 weeks, 0.05 ± 0.01 kBq/mL; where F_2,31 = 5.30 and P = .01, with lower levels in subjects with alcohol dependence compared with levels in control subjects. The difference in ¹²³I-iomazenil clearance is corrected for in the outcome measure, V_T, which is equal to the total regional activity divided by the free ¹²³I-iomazenil in blood. Plasma ¹²³I-iomazenil levels were similar between smokers and nonsmokers.

¹²³I-iomazenil uptake was higher throughout the brain for subjects with alcohol dependence, but the degree of difference and significance varied in a region-dependent manner. A voxel-based analysis designed to determine if there was any group × time interaction in a 1-way analysis of variance of alterations in ¹²³I-iomazenil uptake between (1) control nonsmokers, n = 10, plus control smokers, n = 5 (¹²³I-iomazenil V_T uptake did not differ between control smokers and nonsmokers, and n = 5 is too small for voxel-based analyses; therefore, control smokers and nonsmokers were pooled); (2) smokers with alcohol dependence who were abstinent from alcohol for less than 1 week (n = 15); (3) nonsmokers with alcohol dependence who were abstinent from alcohol for less than 1 week (n = 8); (4) smokers with alcohol dependence who were abstinent from alcohol for less than 4 weeks; and (5) nonsmokers with alcohol dependence who were abstinent from alcohol for less than 4 weeks. This analysis demonstrated significant differences in left frontal inferior and middle gyrus, Brodmann areas (BAs) 11 and 47; right occipital lobe, BA 19; left and right temporal lobe, BA 21; and the cingulate gyrus, BA 31. The most significant cluster (with the highest F value) was noted in the frontal cortex. A plot of this cluster demonstrated higher ¹²³I-iomazenil uptake in nonsmokers with alcohol dependence who were abstinent from alcohol for less than 1 week compared with smokers with alcohol dependence who were abstinent from alcohol for 1 week and control subjects. This effect appeared to normalize by 4 weeks of abstinence (Figure 1).

Analysis of the t statistic maps comparing nonsmokers with alcohol dependence who were abstinent from alcohol for less than 1 week vs control subjects (n = 15) indicated significantly higher ¹²³I-iomazenil uptake in the frontal (BA 47), insular (BA 13), occipital (BAs 17 and 19), temporal (BA 21), and parietal (BA 40) cortices as an overall group (P = 0; set-level) and as 6 individual clusters after corrections for multiple comparisons, with a P value range of .002 to .05 (Figure 2). The effects of smoking were also examined using a mixed-model analysis for 4 brain areas, including medial frontal cortex, anterior cingulate cortex, hippocampal-amygdala area, and cerebellum. There was significant group by smoking interaction (F_1,34 = 5.54; P = .02) with significant differences between control subjects and nonsmokers with alcohol dependence who were abstinent from alcohol for 1 week (28.4%, 29.5%, 31.6%, and 13.7% for medial frontal cortex, anterior cingulate cortex, hippocampal-amygdala area, and cerebellum, respectively; F_1,34 = 6.24; P = .02), but not between control subjects and smokers with alcohol dependence who were abstinent for 1 week (14%, 14%, 6.2%, and 5.2% for medial frontal cortex, anterior cingulate cortex, hippocampal-amygdala area, and cerebellum, respectively; F_1,34 = 0.81; P = .37) (Figure 3, Table 3). There were no significant differences between smokers in the control and alcohol-dependent groups and nonsmokers in these groups who were abstinent from alcohol for 4 weeks. There was a significant positive association between the number of days of abstinence from alcohol for nonsmokers with alcohol dependence (t₁₉ = 2.35, P = .03) but not for smokers with alcohol dependence (t₁₉ = −0.34, P = .70) (Table 4). Pearson product correlations demonstrated statistically significant correlation coefficients between days of abstinence from alcohol and ¹²³I-iomazenil uptake in medial frontal cortex and cerebellum (Figure 4), suggesting that GABA_A-BZ receptors increased over the first week of abstinence in nonsmokers with alcohol dependence but not in smokers with alcohol dependence.

Voxel-based comparisons of ¹²³I-iomazenil uptake between subjects with alcohol dependence who were abstinent from alcohol for less than 1 week and for less than 4 weeks did not yield significant differences in gray matter brain areas (data not shown). Likewise, in the mixed-model analysis, ¹²³I-iomazenil V_T was not significantly different between smokers and nonsmokers with alcohol dependence between 1 and 4 weeks of abstinence from alcohol. These findings are not consistent with the early findings in the F test analysis, most likely because of the small number of subjects in the alcohol-dependent nonsmoker groups.

The relationship between regional ¹²³I-iomazenil uptake and various clinical characteristics of the subjects with alcohol dependence were assessed (Figure 4 and Figure 5; Table 2 and Table 4). A significant positive correlation with ¹²³I-iomazenil uptake in medial frontal cortex and cerebellum and the days since the last alcoholic drink was consumed was noted at 1 week of abstinence in nonsmokers with alcohol dependence but not in smokers with alcohol dependence. Voxel-based comparisons in subjects with alcohol dependence who were abstinent from alcohol for less than 1 week demonstrated significant positive correlation with the severity of alcohol withdrawal in the occipital lobe (BAs 17, 18, and 19) and the cerebellum, which was also demonstrated in the cerebellum region of interest analysis. There were no significant correlations with the other clinical variables.

The present study evaluated adaptive changes in GABA_A-BZ–receptor availability in subjects with alcohol dependence over the first month of sobriety using SPECT and ¹²³I-iomazenil. ¹²³I-iomazenil is a potent competitive BZ antagonist with weak inverse agonist properties^38,39 that has pharmacological specificity similar to flumazenil, a well-known, structurally related BZ antagonist,^39-43 and Ro4513, a partial inverse agonist. ¹²³I-iomazenil binds to all α₁ through α₆ subtypes and thus is a measure of all GABA_A receptors in the brain. The present findings demonstrate higher ¹²³I-iomazenil binding to GABA_A-BZ receptors in the parietal, frontal, cingulate, temporal, insular, and occipital cortices of nonsmokers with alcohol dependence at 1 week of abstinence from alcohol but not of smokers with alcohol dependence at 1 week of abstinence from alcohol, as compared with the control group. The GABA_A-BZ–receptor availability correlated with the severity of alcohol withdrawal in all of the subjects with alcohol dependence and correlated positively with the number of days since the last alcoholic drink was consumed by the nonsmokers with alcohol dependence but not the smokers with alcohol dependence. The time-dependent changes in GABA_A-BZ–receptor availability are supported by an earlier study¹⁷ that demonstrated a time-dependent change in response to BZ challenge in orbitofrontal and cingulate cortices in subjects with alcoholism who were abstinent from alcohol for 2 to 3 weeks and for 6 to 8 weeks. Collectively, these findings suggest that both time since consuming the last alcoholic drink and smoking status are critical variables that need to be taken into account when studying the adaptations in GABA_A-BZ–receptor expression throughout recovery from alcohol dependence.

Increased GABA_A-BZ–receptor expression, as well as augmented GABA-mediated enhancement of BZ binding, has been observed in the superior frontal cortex of postmortem human brain from subjects with alcohol dependence who were without complications.^18,19,44 Since BZ binding is differentially sensitive to α subunits,⁴⁵ the increase may be a consequence of subunit changes that occur in response to abrupt withdrawal after chronic alcohol consumption.⁴⁶ In postmortem frontal cortex from humans with alcoholism, α₁ and β₃ messenger RNA (mRNA) is enhanced whereas α₂, α₃, and α₄ mRNA is not significantly altered.^47,48 In rats chronically treated with ethanol, levels of cortical α₁, α₂, and α₅ mRNA and α₁, α₂, and α₃ polypeptides are decreased whereas α₄ mRNA levels are increased.^49-53 During withdrawal, α₁ and α₄ levels rapidly normalize.^54,55 The decrease in α₁ subunit levels appears to be owing to ethanol-induced internalization of cortical α₁-GABA_A receptors⁵⁶ that results in decreased cell surface and increased intracellular GABA_A-BZ receptors. It is not clear whether iomazenil labels cell surface or intracellular receptors or binds to both receptor populations. If iomazenil labels both populations, then the observed increase reflects increased internalized or intracellular receptors in addition to decreased cell surface receptors, which would have an overall net effect of decreased GABA_A-BZ–receptor function during early sobriety.

The up-regulation in ¹²³I-iomazenil uptake and the correlation with days since consuming the last alcoholic drink observed in nonsmokers with alcohol dependence were attenuated in smokers, suggesting that smoking suppressed the increase in levels of GABA_A-BZ receptors during early sobriety. This finding is of great importance given the high prevalence (up to 90%) of smoking among subjects with alcohol dependence. While smokers tend to suffer from more severe alcohol dependence,^57-59 they also report feeling less intoxicated on alcohol challenge.⁶⁰ In the present study, there was a trend for smokers with alcohol dependence to report fewer alcohol withdrawal symptoms, suggesting that smoking may reduce the severity of withdrawal. In support of this, mice treated long-term with ethanol develop a greater and long-lasting response to inverse agonists, suggesting that BZ inverse agonists may antagonize ethanol intoxication.^61,62 These findings, combined with the observation for a tendency for ¹²³I-iomazenil to correlate with the reported severity of withdrawal in the nonsmokers with alcohol dependence but not in smokers with alcohol dependence, suggest that the greater the adaptive increase in GABA_A-BZ–receptor expression is, the more severe the withdrawal symptoms are. Thus, it would appear that smoking might prevent GABA_A- receptor adaptations associated with alcohol dependence and withdrawal. Furthermore, these findings predict that smokers with alcohol dependence should have greater success with abstaining from alcohol should they continue to smoke during acute withdrawal. Moderate smokers tend to increase the intensity of smoking during ethanol detoxification to suppress ethanol withdrawal symptoms.⁶³

Several constituents of tobacco smoke may play a role in the impact of smoking. Nicotine directly stimulates GABA neuronal activity and may, therefore, suppress alcohol withdrawal symptoms.⁶⁴ Cigarette smoke also contains β-carbolines that have been measured in the plasma from smokers in the control and alcohol-dependent groups.⁶⁵ The β-carbolines are well known as monoamine oxidase inhibitors but have also been suggested to be GABA_A-BZ– receptor inverse agonists and to antagonize some of the effects of ethanol. There has been some debate about whether the β-carbolines reach sufficient levels in the brain to have pharmacological effects.^66,67 They do achieve sufficient levels to inhibit monoamine oxidase enzymes,^68-71 but they have a 2- to 10-fold lower affinity for binding to GABA_A-BZ receptors as compared with monoamine oxidase.^66,67,72-74 Definitive conclusions about the role of the β-carbolines on GABA_A-BZ–receptor availability in the present study without plasma measures are speculative, and this should be addressed in future studies. Additionally, other compounds in tobacco smoke or other receptor mechanisms may be responsible for the opposing effects of tobacco smoke on alcohol withdrawal symptoms.

Interpretation of the current study in light of concurrent cortical GABA level measurements made in the same subjects may shed light on the functional implications of the time-dependent changes in ¹²³I-iomazenil observed in subjects in this study.⁷⁵ The GABA levels are modulated by GABA_A-BZ receptors to maintain GABAergic neurotransmission. For example, clonazepam produces a substantial decrease in cortical GABA levels measured using ¹H-magnetic resonance spectroscopy in healthy human subjects.⁷⁶ Thus, it is possible that a shift from α₄ or α₆ BZ-insensitive, inverse agonist–sensitive GABA_A receptors to α₁ BZ-sensitive GABA_A receptors during the recovery from alcohol dependence would be marked by a decline in cortical GABA levels. In fact, nonsmokers with alcohol dependence in this study exhibited normal cortical GABA levels in the first week of abstinence from alcohol that declined over the first month of sobriety⁷⁵ and remained low over the next 5 months whereas smokers with alcohol dependence demonstrated low cortical GABA levels that did not vary between 1 and 4 weeks of abstinence.^75,77 In nondrinkers, cortical GABA levels do not vary between men who do and who do not smoke.⁷⁸ The effects of smoking on GABA_A-BZ–receptor availability have yet to be evaluated. The relationship of the subunit substitutions during alcohol dependence and withdrawal to GABA_A-receptor function is not clear. However, there is a wealth of data demonstrating decreased GABA_A- receptor function in response to chronic ethanol use that may occur owing to increased inverse agonist–sensitive α₄ subunits that increase the sensitivity to inverse agonists that would reduce GABA_A-receptor function. With sobriety, the restoration of BZ-sensitive, inverse agonist–insensitive GABA_A receptors may alleviate a GABA_A-BZ–receptor functional deficit and may contribute to the decline in cortical GABA_A binding and GABA levels associated with recovery. It is not yet clear whether the reductions in GABA levels, ¹²³I-iomazenil uptake, and lorazepam response that may persist with extended sobriety reflect a persisting consequence of alcohol dependence, alcohol dependence–related neurotoxicity, or a preexisting trait.⁷⁹

The current data suggest that GABA_A-receptor adaptations may contribute to acute ethanol withdrawal and to the recovery from alcohol dependence. While BZ treatments are the predominant detoxification strategy, a growing number of alternative approaches are being explored, including anticonvulsant medications and N-methyl-D-aspartate glutamate receptor antagonists.^80,81 The current data raise the possibility that treatments that accelerate the normalization of GABA_A-receptor populations may increase the rate of recovery whereas treatments that have ethanol-like effects on GABA_A-receptor populations may delay recovery. The effects of these detoxification strategies on alcohol-related adaptations in human GABA_A-receptor populations are currently unknown. However, there is growing interest in the possibility that these treatments might avoid the negative effects of BZ-assisted detoxification on the initiation of abstinence in patients who have completed acute detoxification.⁸⁰ In this regard, it is possible that substances in tobacco smoke, such as BZ inverse agonist β-carbolines¹ or nicotine, may provide clues to novel pharmacotherapeutic approaches to alcohol dependence that might prevent or treat acute withdrawal symptoms and promote the initiation and maintenance of sobriety.

Correspondence: Julie K. Staley, PhD, Department of Psychiatry, Yale University School of Medicine & VACHS 116A2, 950 Campbell Ave, West Haven, CT 06516 (julie.staley@yale.edu).

Submitted for Publication: July 30, 2004; final revision received January 10, 2005; accepted January 13, 2005.

Funding/Support: This work was supported by the Dana Foundation, New York, NY, the Veterans Affairs Mental Illness Research, Education, and Clinical Center for Dual Diagnosis, and the Veterans Affairs Alcohol Research Center, West Haven, Conn, and grants K01AA00288, K05 AA014715-01, RO1 AA1132, and P50 DA13334 from the National Institutes of Health, Bethesda, Md.

Previous Presentation: These findings have been presented in part at the Ninth Annual Meeting of the Society on Research on Nicotine and Tobacco; February 19-22, 2003; New Orleans, La; the Annual Meeting of the Society for Neuroscience; November 2002; Orlando, Fla; the Congreso Internacional las Neurosciencias y las Adicciones: Nuevos Desarrollos, Nuevas Esperanzas; June 2002; Mexico City, Mexico; the 25th Annual Meeting of the Research Society on Alcoholism; June 2002; San Francisco, Calif; the 26th Annual Meeting of the Research Society on Alcoholism; June 2003; Miami, Fla; and the 42nd Annual Meeting of the American College for Neuropsychopharmacology; December 2003; San Juan, Puerto Rico.

Acknowledgment: We thank Louis Amici and Nina Sheung for their expert technical assistance in the synthesis and metabolite analyses of ¹²³I-iomazenil.