INTRODUCTION

Alcohol use and abuse account for a tremendous burden of disease and injury and economic costs worldwide. Excessive alcohol drinking is a leading risk factor for chronic non-communicable diseases and in fact, is linked to more than 60 diseases, including cancers, cardiovascular diseases, liver cirrhosis, neuropsychiatric disorders, and fetal alcohol syndrome (Rehm et al, 2009). Consequently, alcohol use and abuse bring considerable costs to society; on a global scale, the annual costs are estimated to be 760 billion Euros. One further consequence of excessive alcohol use is that 76 million adults worldwide are alcohol dependent (Rehm et al, 2009). Currently, there are four medications approved by the U.S. Food and Drug Administration (FDA) to treat alcohol dependence: disulfiram, oral naltrexone, a long-lasting injectable naltrexone, and acamprosate (Litten et al, 2012). In Europe, nalmefene has been also approved in 2013 (Mann et al, 2013).

Almost 30 years ago, the inhibiting effect of acamprosate (calcium-bis(N-acetylhomotaurinate)—in the entire text abbreviated as Ca-AOTA) on alcohol consumption in laboratory animals was described (Boismare et al, 1984). This initial observation led to the clinical development of acamprosate (Campral) and nowadays this drug is currently used in many countries for relapse prevention in abstinent alcohol-dependent patients. Acamprosate is a safe and well-tolerated drug that does not affect craving (Umhau et al, 2011) but the risk to relapse. In a recent Cochrane Review Rösner et al (2010) summarized 24 randomized controlled trials (RCTs) on acamprosate, and concluded that acamprosate significantly reduces the risk of any drinking with a relative risk (RR) of 0.86. A RR of 1 means that there is no difference between placebo and treatment, whereas a RR<1 means that relapse occurs less frequently in the treatment group. Acamprosate has been used to treat alcohol dependence in over 1.5 million patients since its introduction in Europe in 1989 and is currently available in most European and Latin American countries, Australia, parts of Asia, and Africa (Mason and Heyser, 2010). In 2004, it was approved by the FDA for the maintenance of abstinence from alcohol in detoxified alcohol-dependent patients. A recent survey found that acamprosate is now the most widely prescribed medication for the treatment of alcoholism in the United States of America (Mark et al, 2009).

In the past three decades, huge research efforts sought to elucidate the molecular mode of action of acamprosate. Many molecular candidate targets have been described (Spanagel and Vengeliene, 2013; De Witte et al, 2005; Mann et al, 2008) but the only accumulated evidence was found with respect to an interaction with the glutamate system. In concert, these studies suggest that acamprosate attenuates hyperglutamatergic states that occur during early and protracted abstinence, possibly involving N-methyl-D-aspartate (NMDA) receptors and metabotropic glutamate receptor 5 (mGluR5) (Spanagel and Vengeliene, 2013; De Witte et al, 2005; Mann et al, 2008; Rammes et al, 2001; Harris et al, 2002). A prominent theory in the alcohol research field posits that chronic alcohol consumption leads to glutamatergic dysfunction. As a consequence, exaggerated glutamate activity is observed during alcohol withdrawal and conditioned withdrawal responses. This hyperglutamatergic state may then drive alcohol-seeking and relapse behavior (Tsai et al 1995; Spanagel and Kiefer 2008). Acamprosate dampens hyperglutamatergic activity in excessively ethanol drinking mice, thereby reducing alcohol intake (Mann et al 2008; Spanagel et al, 2005). In a recent double-blind, placebo-controlled study, which applied magnetic resonance spectroscopy acamprosate also reduced glutamate levels in the brains of detoxified alcohol-dependent patients (Umhau et al, 2010). Although the effects of acamprosate on glutamate levels are well documented, the molecular mode of action of this drug and the putative interaction between acamprosate and glutamate receptors remains unclear.

Here we describe a series of experiments and come to the surprising conclusion that N-acetylhomotaurinate by itself is not an active psychotropic molecule. Instead, calcium is the active moiety of acamprosate! In the first set of experiments, we tested the putative interactions of acamprosate with NMDA receptors and mGluR5 (Mann et al, 2008; Harris et al, 2002; Madamba et al, 1996). From these experiments, we have to conclude that the proposed glutamate receptor interactions of acamprosate cannot sufficiently explain the anti-relapse action of this drug. In comparative experiments, we then studied the effects of Ca-AOTA vs sodium-N-acetylhomotaurinate (Na-AOTA) in three preclinical models of either excessive alcohol drinking, alcohol-seeking, or relapse-like drinking behavior. The rationale for doing these comparative experiments stems from US patent 4,355,043 (1982) where the initial investigators of acamprosate described that various salts of N-acetylhomotaurinate produce their effects according to the nature of the counter ion. This assertion has so far not been tested. From these animal experiments, we conclude that calcium exhibits anti-relapse effects and seems to be the major active ingredient of acamprosate. We further translated these findings at the clinical level. Using a clinical sample of placebo vs acamprosate-treated abstinent alcohol-dependent patients, we measured calcium plasma concentrations and show that patients with high plasma calcium levels due to acamprosate treatment exhibit better primary efficacy parameters.

MATERIALS AND METHODS

Six different studies were performed. All information is provided in the Supplementary Information.

Study 1: Screening panel for the mode of action of acamprosate.

Study 2: Testing different salt formulations of acamprosate in the ADE model.

Study 3: Testing different salt forms of acamprosate in alcohol-seeking rats in the cue-induced reinstatement model.

Study 4: Testing different salt forms of acamprosate in alcohol-preferring iP-rats.

Study 5: Pharmacokinetic (PK) profiles of different salt forms of acamprosate.

Study 6: Calcium plasma levels in placebo and acamprosate-treated patients.

RESULTS

Acamprosate (Ca-AOTA) does not Interact with NMDA Receptors or mGluR1/5

Several possible modes of action of acamprosate have been described (Mann et al, 2008); however, a body of evidence suggests that acamprosate interacts with NMDA receptors and/or mGluR5 (Mann et al, 2008; Harris et al, 2002; Madamba et al, 1996). To test these putative interactions, we applied an extensive screening panel. By expressing human NR1 and NR2B subunits in Xenopus laevis oocytes, we first tested agonist activity of acamprosate on the glycine and glutamate-binding site of the NMDA receptor. The two subunit types expressed for our purposes (hNR1A and hNR2B) provide a combination that is thought to be predominantly present in the human forebrain (Scherzer et al, 1998; Kosinski et al, 1998). To determine NMDA receptor agonist activity, glycine and glutamate were substituted with Ca-AOTA. Multiple concentrations of Ca-AOTA were tested up to a maximal concentration of 1 mM. Elicited currents were measured and compared with currents seen with saturating concentrations of glycine and glutamate. Even at a maximal concentration, no activation of the receptor was seen indicating that acamprosate does not appear to have any effect at the glycine agonist binding site nor at the glutamate-binding site (Figure 1a and b). To determine NMDA receptor antagonist activity, standard Schild-plot analyses were performed to assess whether or not acamprosate could shift the affinity of glycine and glutamate. Concentration response curves of glycine and glutamate were again prepared with varying concentrations of Ca-AOTA (100, 500, and 1000 μM). No shift in affinity was seen in concentration response curves for glycine or glutamate, indicating no apparent antagonist activity in our assay (Figure 1c and d). In addition, using acute brain slices from rats, electrically evoked NMDA receptor dependent excitatory post-synaptic potentials (EPSPs) were measured in the nucleus accumbens. Modest changes in EPSP amplitude were observed in ∼50% of the cells following addition of Ca-AOTA (Supplementary Figure 1) suggesting that acamprosate may have some effect on the quantum release of presynaptic glutamate.

Figure 1
figure 1

(a) Glycine-site agonist screen. Glycine was substituted with acamprosate in the presence of excess co-agonist glutamate. No activation of the receptor was seen indicating that acamprosate (Ca-AOTA) does not appear to have any effect at the glycine agonist binding site up to a 1 mM concentration. (b) Glutamate-site agonist screen. Glutamate was substituted with acamprosate in the presence of excess co-agonist glycine. No activation of the receptor was seen indicating that acamprosate does not appear to have any effect at the glutamate agonist binding site up to a 1 mM concentration. (c) Glycine-site antagonist screen of acamprosate. Concentration response curves of glycine show no shift in affinity when increasing concentrations of acamprosate is introduced, indicating no apparent glycine-site antagonist activity. (d) Glutamate-site antagonist screen of acamprosate. Concentration response curves of glutamate show no shift in affinity when increasing concentrations of acamprosate is introduced, indicating no apparent glutamate-site antagonist activity. (e) Functional Ca2+ flux assays illustrate dose-dependent mGluR1/5 activation by quisqualate and antagonism of this response by AIDA and MPEP, respectively. (f) Shows the lack of effect of Ca-AOTA (acamprosate) on the response to quisqualate.

PowerPoint slide

Next, calcium flux assays were used to study acamprosate’s interaction with mGluR1/5. For these experiments, human mGluR1 and 5 were expressed in HEK 293 cells. The mGluR1/5 agonist quisqualate produced a robust dose-dependent increase in the fluorescence signal that was completely inhibited by the mGluR1/5 antagonist AIDA as well as by MPEP (Figure 1e). Ca-AOTA had no effect in this assay (Figure 1f). Then we performed patch clamp electrophysiology to measure effects of mGluR1/5 on resting membrane potential and spike number. Again, brain slices were prepared from 6–10-week-old male rats. Using whole-cell patch clamp, the effects of the mGluR1/5 agonist DHPG on resting membrane potential and spike frequency were measured. DHPG-induced changes to the membrane potential were inhibited by the mGluR1 antagonist LY367385 whereas DHPG-induced increases in action potentials were blunted by addition of the mGluR1 and 5 antagonists, LY367385 and MPEP, respectively. In contrast, Ca-AOTA had no effect under these conditions on membrane potential or the number of action potentials (Figure 2). From these studies, we conclude that the proposed glutamate receptor interactions of acamprosate, especially with NMDA- and mGluR5 receptors cannot sufficiently explain the anti-relapse action of this drug.

Figure 2
figure 2

(a) Effects of DHPG on resting membrane potential (Vm change) in CA1 neurons of the hippocampal region in the presence of mGluR1/5 antagonists or acamprosate. Application of 100 μM DHPG (mGluR1/5 agonist) depolarized cells by ∼1.8 mV (n=33). Addition of 300 μM LY367385 (mGluR 1 antagonist) significantly reduced DHPG-induced depolarization (n=11) (P<0.05). DHPG+10 μM MPEP (mGluR 5 antagonist) resulted in a modest but not significant decrease in the depolarization (n=9). Acamprosate (Ca-AOTA; 300 μM) had no effect on DHPG-induced depolarization (n=13) (b) Effects of DHPG on spike number change in the presence of mGluR1/5 antagonists or acamprosate. DHPG alone (n=33) increased the spike number by more than 3. Addition of LY367385 (n=10) or MPEP (n=9) significantly reduced the spike number in the presence of DHPG (P<0.05). Ca-AOTA (n=12) had no effect on the number of spikes. The data are presented as means±SEM * Indicates significant differences from vehicle control group, P<0.05. Note: In the condition DHPG+LY367385 and DHPG+acamprosate we lost in each experiment one recording during spike induction.

PowerPoint slide

Na-AOTA Is Ineffective in Preclinical Animal Models but Calcium Salts Produce Acamprosate-Like Effects

In US patent 4,355,043 (1982), the initial investigators of acamprosate described that various salts of N-acetylhomotaurinate produce their effects according to the nature of the counter ion. This assertion has so far not been tested. Therefore, in a comparative experiment, we studied the effects of Ca-AOTA vs Na-AOTA in the alcohol deprivation effect (ADE) model. This is a standard preclinical rat model that measures relapse-like drinking behavior by monitoring the ADE (Spanagel, 2009). This model provides excellent face and construct validity (Vengeliene et al, 2009), and has shown predictive validity as well (Spanagel and Kiefer, 2008). In particular, acamprosate was found in previous studies to reduce the ADE (Spanagel et al, 1996a; Heyser et al, 1998; Lidö et al, 2012). As described in these previous studies, the calcium salt of N-acetylhomotaurinate (acamprosate) decreased the ADE in the present experiment. Hence, a two-way repeated measures ANOVA displayed a significantly different alcohol intake during ADE days between vehicle and Ca-AOTA-treated animal groups (F(5,250)=10.327, P<0.0001). In contrast, an equimolar concentration of the corresponding sodium salt had no effect on the ADE (P=0.950) (Figure 3a and b). This surprising finding suggests that calcium is a critical component for the efficacy of acamprosate, and we hypothesized that calcium salts should produce acamprosate-like effects on the ADE. Indeed, equimolar concentrations of calcium derived either from calcium chloride or calcium gluconate reduced the ADE in a similar way as acamprosate two-way repeated measures ANOVA revealed significantly lower alcohol intake during ADE days in both calcium chloride- and calcium gluconate-treated animal groups when compared with the control animal group (F(10,105)=2.860, P<0.003) (Figure 3c). We then tested a mixture of Na-AOTA and calcium chloride, which restored the effect of acamprosate on the ADE (F(5,60)=3.685, P<0.006) (Figure 3d). Locomotor activity was monitored throughout the ADE measurements by the home cage E-motion system. Ca-AOTA, as observed in previous experiments by us, (Spanagel et al, 1996b) reduced home cage activity following the first injection. This effect was, however, absent in NA-AOTA (Supplementary Figure 2). This set of experiments suggests that N-acetylhomotaurinate is an inactive molecule and that calcium is the active moiety of acamprosate.

Figure 3
figure 3

Total ethanol intake (g/kg/day) before and after an alcohol deprivation period of 3 weeks in male Wistar rats. The average of the last 3 pre-abstinence days measurement of ethanol intake is given as baseline drinking ‘B’. Arrows indicate the i.p. administration of either saline (n=7–26) or compounds: 200 mg/kg of Ca-AOTA (n=26) (a), 200 mg/kg of Na-AOTA (n=9) (b), 73.4 mg/kg of CaCl2 × 2 H2O (n=8) and 215 mg/kg of Ca-gluconate (n=7) (c), and 200 mg/kg of Na-AOTA combined with 73.4 mg/kg of CaCl2 (n=7) (d). Note 1: 73.4 mg/kg of CaCl2, 215 mg/kg of Ca-gluconate and 200 mg/kg of Ca-AOTA contain equivalent amounts of Ca2+ ions (0.499 mmol/kg). Note 2: 200 mg/kg Na-AOTA and 200 mg/kg Ca-AOTA contain almost equal amounts of N-acetylhomotaurinate (0.984 mmol/kg vs 0.999 mmol/kg; corresponding to 1.5% difference). All data are presented as means±SEM * indicates significant differences from vehicle control group, P<0.05.

PowerPoint slide

To further confirm these surprising findings, we used yet another standard preclinical rat model—the reinstatement model (Shaham et al, 2003; Sanchis-Segura and Spanagel, 2006). This model refers to the resumption of extinguished lever-pressing behavior after exposure of an animal to alcohol-conditioned stimuli. Reinstatement of alcohol-seeking is used to study the neurobiological and molecular basis of craving, as there appears to be a good correspondence between the events that induce alcohol-seeking in laboratory animals and those that provoke craving in humans (Shaham et al, 2003). In a previous study, it was shown that acamprosate completely abolishes the cue-induced reinstatement response (Bachteler et al, 2005). As described in this previous report, the calcium salt of N-acetylhomotaurinate (acamprosate) abolished the reinstatement response but equimolar concentration of the corresponding sodium salt formulation of acamprosate had no effect (Figure 4). Hence administration of Ca-AOTA and Na-AOTA as well as calcium chloride significantly changed lever responding during the cue-induced reinstatement sessions (factor drug F(3,87)=12.7, P<0.0001; factor drug × lever interaction: F(6,87)=4.9, P<0.001 and factor drug × lever × session interaction: F(6,87)=2.3, P<0.05). Post hoc comparisons revealed that administration of both Ca-AOTA and calcium chloride significantly reduced responses on the ethanol-associated lever during S+/CS+ session (Figure 4a) and had no effect on lever responding during S−/CS− session (Figure 4b). No significant effect on lever responding during the cue-induced reinstatement sessions was observed in the Na-AOTA-treated group (Figure 4a and b). Responding on the inactive lever was not affected by either drug during both reinstatement sessions, indicating the absence of a nonspecific reduction in lever-pressing behavior (Figure 4a and b). In summary, these results clearly support our conclusion from the ADE model that calcium is the active moiety of acamprosate.

Figure 4
figure 4

The effect of either vehicle (n=9) or 200 mg/kg of Na-AOTA (n=8), 200 mg/kg of Ca-AOTA (n=8) and 73.4 mg/kg of CaCl2 (n=8) on cue-induced reinstatement under S+/CS+ (a) and S−/CS− (b) conditions. Data are shown as the average number of lever presses on the active lever during the last four extinction sessions (extinction) and as the number of responses on the active lever (active lever) after the presentation of a stimulus previously paired with either ethanol (S+/CS+ session) (a) or water (S−/CS− session) (b). Inactive lever (inactive lever) was present throughout the experiment and was used as a measure of a sedative effect of the treatment. All animals were injected with either of the compounds 12 h and 2 h before reinstatement sessions. Data are presented as means±SEM * indicates significant differences from the control vehicle group, P<0.05.

PowerPoint slide

In light of the major implications of this study, we initiated an independent blinded replication of our findings in another laboratory (Lawrence group at Florey Institute of Neuroscience & Mental Health, University of Melbourne). Alcohol-preferring rats have been used for decades to study excessive alcohol consumption and the efficacy of putative pharmacological interventions (Spanagel and Kiefer, 2008; Bell et al, 2006). In particular, iP rats self-administering alcohol show sensitivity to acamprosate treatment (Cowen et al, 2005). Accordingly, Ca-AOTA and Na-AOTA were tested on operant behavior under a fixed ratio schedule 3 (FR3) (Supplementary Figure 3). As reported previously, (Cowen et al, 2005) Ca-AOTA (200 mg/kg i.p.) reduced significantly the number of ethanol reinforcers and calcium chloride produced a very similar response. A repeated measure one-way ANOVA indicated a main effect of treatment (F(3,31)=11.18, P<0.001). Post hoc pairwise multiple comparison revealed for saline vs Ca-AOTA P<0.001 and for saline vs calcium chloride P<0.007. In a second group of animals, saline treatment was compared with Na-AOTA and calcium chloride. Although statistics indicated a treatment effect (F(3,28)=4.59, P=0.025) post hoc analysis indicated that Na-AOTA did not significantly reduce the numbers of ethanol rewards (saline vs Na-AOTA, P=0.083), whereas again calcium chloride produced a significant reduction in earned ethanol rewards (saline vs CaCl2, P=0.02). This independent experiment further supports our conclusion that N-acetylhomotaurinate by itself is not an active psychotropic molecule. Instead, calcium seems to be the active moiety of acamprosate.

To ensure that both salt formulations of N-acetylhomotaurinate lead to similar blood concentrations over time, we established pharmacokinetic (PK) profiles after i.p. injections and found no significant differences in assessed PK values for Na-AOTA and Ca-AOTA (Supplementary Table 1, Supplementary Figure 4). In addition, we measured PK parameters in the brain 30 min after i.p. injections of equimolar doses of 200 mg/kg Na-AOTA and Ca-AOTA. Amounts of N-acetylhomotaurinate detected in brain tissue were 3.88±0.28 μg/g for Na-AOTA and 4.08±0.58 μg/g for Ca-AOTA, respectively. We also calculated the brain:plasma ratio and found a ratio of 0.034±0.004 for Na-AOTA vs 0.026±0.004 for Ca-AOTA. We conclude that Na-AOTA and Ca-AOTA have a similar PK profile in blood and that similar amounts of each compound reach the brain. Given these similarities in PK profile and previous data demonstrating that salts of N-acetylhomotaurinate become totally dissociated in hydrophilic media (Chabenat et al, 1988), we can exclude the possibility that differences of bioavailability for these two salt forms of N-acetylhomotaurinate can account for the observed behavioral effects and therefore conclude that calcium acts as the active moiety of acamprosate in rats.

Plasma Calcium Levels Correlate with Acamprosate Response in Alcohol-Dependent Patients

We further translated these animal findings at the clinical level. Using a clinical sample of placebo vs acamprosate-treated abstinent alcohol-dependent patients, we measured calcium plasma concentrations. Before treatment, the same physiological calcium concentration of about 2.4 mmol/l were found in the placebo group and acamprosate group (Supplementary Figure 5). Additionally, blood was drawn after 1, 2, and 3 months of treatment. Although calcium levels are extremely tightly controlled by the kidneys and parathyroid hormone and are more or less constant in each individual, we found a tendency (P<0.1) of enhanced calcium plasma levels in acamprosate vs placebo treated patients (Supplementary Figure 5). When correlating the primary efficacy parameters of first drink, severe relapse, and cumulative abstinence with calcium plasma levels, we did not find any correlation in the placebo group, whereas significant correlations were found in the acamprosate group (Figure 5). Thus, patients with high plasma calcium levels due to acamprosate treatment showed better primary efficacy parameters.

Figure 5
figure 5

Plasma Ca2+ concentrations in patients receiving placebo (n=12) or acamprosate (n=19) treatment (1998 mg/day) in correlation to abstinence duration. Blood was drawn 1, 2, and 3 months after treatment begun and mean calcium concentrations were correlated with the primary efficacy parameters first drink, severe relapse, and cumulative abstinence.

PowerPoint slide

DISCUSSION

In summary, our combined in vitro, animal and human data demonstrate that N-acetylhomotaurinate by itself is not an active psychotropic molecule and therefore it is not surprising that a molecular mode of action could not be established for acamprosate. Although previous studies have suggested that acamprosate might interact with NMDA receptors and mGluR5 (Mann et al 2008; Harris et al 2002; Madamba et al, 1996) our in vitro screening panel does not support this assumption. Instead, our findings in three independent preclinical animal studies suggest that calcium produces anti-drinking, anti-alcohol-seeking and anti-relapse effects. Furthermore, when translating these findings to the human condition, we demonstrated that calcium plasma levels in acamprosate-treated alcohol-dependent patients correlate with primary efficacy parameters such as time to first drink, time to severe relapse, and cumulative abstinence duration. Collectively, these data lead to the conclusion that calcium is the active moiety of acamprosate.

This conclusion may be somewhat surprising; however, previous studies showed that (i) calcium sensitivity of the synapse is important for alcohol tolerance development (Lynch and Littleton, 1983, ii) calcium given intraventricularly significantly enhances alcohol intoxication in a dose-dependent manner (Erickson et al, 1978, iii) activity of calcium-dependent ion channels modulate alcohol drinking and recently small conductance calcium-activated potassium type 2 channels (SK2) were shown to modulate hyperglutamatergic activity and alcohol consumption in rats (Mulholland et al, 2011; Hopf et al, 2011), finally and most importantly, (iv) clinical studies from the early 1950s proposed calcium therapy (at that time called calmonose) for the treatment of alcoholism (O'Brien, 1952; O'Brien 1964). Our data suggest that acamprosate may also act as a calcium therapy and indeed the application of a daily recommended dose of 1998 mg of Campral leads to enhanced calcium plasma levels in alcohol-dependent patients. Furthermore, enhanced calcium plasma levels are associated with better primary efficacy parameters such as time to relapse and cumulative abstinence. What could be a putative mode of action of a calcium bolus on alcohol drinking, seeking, and relapse considering that plasma and brain calcium are tightly regulated? First, given that a calcium bolus can enhance plasma calcium levels, it may thereby also enhance brain calcium levels and affect brain function. This suggestion is supported by population-based studies, which show that calcium in the drinking water correlates with cognitive functions (Emsley et al, 2000) and that a low content of calcium in drinking water is a risk factor in Alzheimer dementia (Stutzmann, 2007). Thus, although tightly regulated exogenously applied calcium seems to affect brain function and disease. Therefore, we speculate that acamprosate treatment may enhance brain calcium levels and thereby the activity of calcium-dependent ion channels including the SK2 channels, which may reduce hyperglutamatergic activity and alcohol consumption (Mulholland et al, 2011; Hopf et al, 2011). Second, in alcohol-dependent patients, impairment in electrolyte regulation is often observed and a calcium bolus might counteract this impairment leading to well being and less alcohol consumption. Whatever the putative mode of action of a calcium bolus may be, our screening panel for the mode of action of the calcium salt of N-acetylhomotaurinate argues against a direct effect on glutamate receptors.

Indeed our electrophysiological results obtained in Xenopus oocytes where we expressed hNR1A and hNR2B subunits—a subunit composition, which is thought to be predominantly present in the human forebrain (Scherzer et al, 1998; Kosinski et al, 1998)—show that acamprosate acts not via NMDA receptors. Thus, glycine and glutamate-mediated activation of the receptor could not be substituted even by 1 mM of acamprosate. Further, no shift in affinity was noted in concentration response curves for glycine or glutamate, indicating no apparent antagonist activity in our assay. Rammes et al (2001) reported that acamprosate acts as a weak antagonist against NMDA-induced currents in Xenopus oocytes expressing NR1A/2B receptors. Testing acamprosate as the calcium salt also adds excess calcium to the preparation. Therefore, Rammes et al (2001) tested in addition the effects of equimolar calcium concentrations and found that in four out of the eight cells tested, calcium also decreased NMDA-induced currents and a pronounced effect of acamprosate was only seen in these four cells. Thus, it seems that even the effects of high concentrations of acamprosate on NMDA receptors are not due to acamprosate itself, but rather due to nonspecific effects of calcium ions added (Rammes et al, 2001) Altogether, a detailed review of published reports and our data demonstrate that acamprosate exhibits neither agonistic nor antagonistic effects on NMDA receptors. However, there is a limitation to this conclusion, namely that in our Xenopus oocyte test system, we expressed only one possible combination (hNR1A/hNR2B) and cannot exclude the possibility that other NMDA receptor subunit compositions would yield different results. Furthermore, our experiments on accumbal slices indicate that NMDA receptor dependent post-synaptic potentials seem to be affected by very high acamprosate concentrations (1 mM). Thus, acamprosate produced a modest nonsignificant decrease in EPSP amplitude in approximately half of the recorded cells. A similar observation was made in an earlier in vitro recording study in rat neocortical neurons (Zeise et al, 1993).

Harris et al (2002) showed that acamprosate has binding and functional characteristics similar to group I mGluR antagonists, in particular to mGluR5. However, in those brain-binding studies in rats, trans-ACPD was used, which is a non-selective ligand at mGluRs. In Xenopus oocytes injected with mGluR1 or mGluR5 cRNA, acamprosate did not affect receptor function directly, nor did it alter glutamate responses during co-application (Reilly et al, 2008). Here we expressed human mGluR1 and 5 in HEK 293 cells and performed functional calcium flux assays. We then performed patch clamp electrophysiology in rat brain slices to measure effects of mGluR1/5 on resting membrane potential and spike number. Acamprosate had no effect on calcium flux, membrane potential, or the number of action potentials. In summary, neither in expression systems nor in slices from rat brain did acamprosate interfere with mGluR1/5 binding or receptor function.

Taking all of these results into consideration, one can conclude that despite substantive investigation it is not possible to characterize a binding site for acamprosate in the CNS. However, recently it was reported that acamprosate may work via direct or indirect interference with the inhibitory glycine receptor (Chau et al, 2010a, 2010b). We have not tested this putative binding site in the present study but earlier electrophysiological experiments in Xenopus oocytes expressing different homomeric and heteromeric glycine receptors excluded the possibility that acamprosate is directly binding at inhibitory glycine receptors (Reilly et al, 2008). Therefore, we asked ourselves whether acamprosate was actually a biologically active molecule and so we reviewed in detail the US patent 4,355,043 (1982). In this patent, the initial investigators of acamprosate described that various salts of N-acetylhomotaurinate produced their effects according to the nature of the counter ion. This would imply that the counter ion could be the biologically active molecule. Therefore, we set out to test the sodium vs calcium salt of N-acetylhomotaurinate on relapse-like drinking behavior in rats. In rats that had long-term voluntary access to alcohol followed by deprivation for several weeks, the re-presentation of alcohol leads to relapse-like drinking, a temporal increase in alcohol intake over baseline drinking. This robust phenomenon is called the ADE. In recent years, this model has become widely used for examining the efficacy of pharmacological agents in preventing compulsive alcohol consumption and relapse (Spanagel, 2009; Vengeliene et al, 2009). Moreover, acamprosate produces a reliable reduction of the ADE (Spanagel et al, 1996a; Heyser et al, 1998; Lidö et al, 2012). Critically, this effect is only observed if the calcium salt of N-acetylhomotaurinate or any other calcium salt is applied, which strongly suggests that the calcium counter ion of the acamprosate molecule is the biologically active species. Similar results were obtained in the reinstatement test (Shaham et al, 2003; Sanchis-Segura and Spanagel, 2006) where we could show that a cue-induced alcohol-seeking response is only decreased by calcium. Our conclusion was further supported by an independent series of experiments in iP rats. This latter experiment was performed in a blind fashion in yet another laboratory (Lawrence lab); ie, the experimenters did not know which substances they applied to their animals. Again the results of these experiments further suggest that calcium loads given by i.p. bolus injections can cause a reduction in excessive alcohol consumption. From a translational perspective, we found a correlation of calcium levels and primary efficacy parameters of acamprosate treatment in alcohol-dependent patients. This analysis was, however, done in a retrospective manner from the German COMBINE study (Kiefer et al, 2003) were only a limited number of plasma samples were still available. Clearly, a prospective study should be done to confirm these findings.

Accordingly, we provide convincing evidence that N-acetylhomotaurinate is in fact a biologically inactive molecule and that the effects of acamprosate described in more than 450 published original investigations and clinical trials can be attributed to calcium. There is now a strong need for a complete independent replication of our findings before withdrawal of acamprosate from the market. In particular, a clinical investigation should be initiated to study the effects of calcium salts in terms of relapse prevention. However, a historical perspective is also important in this respect. Clinical studies from the early 1950s proposed intensive calcium therapy (at that time called calmonose) for the treatment of alcoholism (O'Brien 1952; O'Brien 1964). Although this therapy disappeared from the treatment landscape for unknown reasons, in light of our new findings, calcium supplements could be easily re-introduced into treatment programs. Furthermore, epidemiological data show that the calcium level in drinking water significantly affects cognitive functions (Emsley et al, 2000). For that reason, one could speculate that calcium levels in drinking water and food might also influence excessive alcohol drinking on a population wide scale.

FUNDING AND DISCLOSURE

RS acted as a consultant for XenoPort, Inc. All other authors declare no competing financial interests.