Keywords

1 Alcohol Addiction: An Area of Major Unmet Medical Needs

Alcohol use accounts for a substantial proportion of global health burden. In industrialized countries, about 10 % of all disability-adjusted life years (DALYs) lost are due to the consumption of alcohol (Whiteford et al. 2013). An expert evaluation in the United Kingdom concluded that in aggregate, the harm to self and others inflicted by alcohol exceeds that caused by heroin or cocaine (Nutt et al. 2010). Reducing alcohol-related harm is therefore a major public health priority. Policies that determine availability, price, and attitudes interact with individual susceptibility factors to result in harmful consumption. In the end, alcohol-related harm is closely correlated with the individual level of consumption (Rehm et al. 2003). Accordingly, both preventative measures aimed at reducing overall levels of alcohol use in society and treatments that can help individuals with established alcohol problems are important elements in a tool-kit of evidence-based strategies to reduce alcohol-related harm (Anderson et al. 2009).

A modern, disease-oriented view of alcohol problems emerged only with the beginnings of the Alcoholics Anonymous (AA) movement in the USA, in large part as a reaction to views of drinking as a moral failing. The construct of an “alcohol dependence syndrome” was introduced almost half a century later (Edwards and Gross 1976). Epidemiological studies found that “alcohol dependence” as subsequently defined by the criteria of the fourth edition of the American Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association 2000) affected about 12 % of Americans at some point in their lives (Grant et al. 2004). Although alcoholics lose more than 20 years in average life expectancy compared to the normal population (John et al. 2013; Westman et al. 2014), only a minority of individuals identified by these criteria ever seek treatment.

Alcohol dependence is highly heterogeneous, and this also applies to the large population of people in the community who meet formal criteria for this condition but do not seek treatment. Many of these individuals do not seek or receive treatment despite suffering obvious negative consequences of their alcohol use. They are caught in a vicious cycle of addictive behavior, do not have easy access to treatments they find appealing, and have learned that for a few hours, resuming alcohol use is the easiest way to escape the misery of alcohol addiction. The resulting treatment gap is greater in alcoholism than in other psychiatric conditions such as depression, where arrival of well-tolerated therapies resulted in progressively decreased stigma and more patients being reached by treatment. Developing treatments that can attract and effectively treat the large population of alcohol-dependent people who could benefit from them is clearly a major public health priority.

Others, on the other hand, may fulfill diagnostic criteria on an interview, but do not seem to experience measurable negative consequences. The diagnostic constructs enshrined in the DMS criteria may not be particularly useful when attempting to delineate these populations, or identify patients responsive to specific treatments. For instance, when a life-time diagnosis of DSM-IV of “alcohol dependence” was established in people in the general population, and these people were reassessed five years later, two-thirds of the women and one-third of the men no longer received the same diagnosis. People with clinically significant alcohol problems seem however to know something about themselves that the diagnostic criteria do not capture. In the same study, among individuals who had ever sought treatment, diagnosis was highly stable (Culverhouse et al. 2005). The “alcohol use disorder” construct recently introduced in the fifth edition of the DSM has unfortunately further diluted the diagnostic category. Recent epidemiological studies have shown that “moderate” to “severe” alcohol use disorder corresponds to what was previously captured by “alcohol dependence”; the addition of a “mild” alcohol use disorder category represents an expansion into unknown territory (Compton et al. 2013).

In seeking to develop novel treatments, it may therefore be helpful to look beyond the ever-changing diagnostic labels. A key characteristic of people with excessive alcohol use who seek treatment is that their consumption continues despite knowledge of adverse consequences. This pattern reflects what is increasingly held to be a core feature of addictive disorders (Piazza and Deroche-Gamonet 2013). Our discussion will therefore focus on people with this type of “aversion-resistant alcohol use,” and interchangeably refer to their condition as “alcohol addiction” or simply “alcoholism,” the term preferred by the largest client organization, AA. Even within this population, different pathophysiological mechanisms are at play in different individuals, or in the same individual at different time points of their addictive career. Translational medications development efforts must recognize this heterogeneity to succeed. A key determinant of success will be the ability to inform the selection of individuals and outcomes in early clinical trials, which allows these studies to enrich efficacy signals that may be present in more homogenous subpopulations of patients, but wash out when studies enroll all-comers.

Whatever the label and defining criteria used, unmet medical needs related to alcohol addiction are enormous. For many countries, reliable epidemiological data to assess these unmet needs are outdated or missing. It will be an important task for epidemiologists to provide those data for better policy making. However, even among people who do seek and receive help for their alcoholism, few receive evidence-based treatments (Hester and Miller 1995). The opioid antagonists naltrexone and nalmefene, as well acamprosate, a functional glutamate antagonist (see also Spanagel et al. 2014), are approved for clinical use and have documented efficacy, but their effect sizes are limited, treatment responses are highly heterogeneous (Heilig and Egli 2006; Heilig et al. 2011; Jonas et al. 2014), and prescription rates are low (Mark et al. 2003, 2009). It is therefore clear that expanding the range of options for alcoholism treatment and developing personalized approaches to treatment should continue to be a major public health priority for a long time to come.

2 The Aspirations of Translational Research for Alcohol Addiction

In the early years of the millennium, there was considerable enthusiasm that advances in the neuroscience of alcohol addiction would soon translate into mechanistically novel alcoholism therapies, expanding the range of treatment options, and addressing unmet medical needs of patients (see, e.g., Heilig and Egli 2006). In particular, there was hope that cleverly designed experimental paradigms in rodents and humans would make it possible to accelerate the process of translation from preclinical target discovery and validation into clinical development. Since then, some advances have indeed occurred. For instance, nalmefene given as needed (rather than continuously) was recently brought to the market (Mann et al. 2013), and some medications already approved for other indications may be possible to repurpose for treatment of alcohol addiction. However, several novel mechanisms that appeared to hold considerable promise based on preclinical data failed to be translated to the human condition.

Perhaps the greatest and most surprising disappointment has been the failure of corticotropin-releasing hormone (CRH1) receptor antagonism (Kwako et al. 2014). The target validation that provided the rationale for clinical trials with CRH1 antagonists in alcoholism came from several animal models of excessive drinking (Heilig and Koob 2007; Zorrilla et al. 2013). In hindsight, there were also indications that CRH1 antagonism may not consistently reduce excessive alcohol consumption. In fact, global CRH1 receptor deletion was found to result in a prolonged increase (rather than decrease) of alcohol consumption following stress exposure, and it appears that central and pituitary CRH1 receptors have opposing influences on stress-induced alcohol drinking (Sillaber et al. 2002; Molander et al. 2012).

The failure of CRH1 antagonism in alcoholism follows failures of several CRH1 blockers in other stress-related indications, such as major depression and generalized anxiety disorder (Binneman et al. 2008; Coric et al. 2010). It has been shown that blockade of CRH1 receptors can result in opposing effects on stress responses depending on the neuronal population targeted (Refojo et al. 2011). Because there are major species differences in CRH1 receptor expression, this may result in major differences in the net effect of a global CRH1 blockade. The emerging clinical observations that global CRH1 receptor blockade is not effective may in retrospect be less surprising than they appear. Nevertheless, they have had a discouraging effect on drug development efforts. Conceptually, strategies selectively targeting specific populations of CRH1 receptor expressing neurons could potentially overcome these challenges, but it is unlikely that this mechanism will be revisited by drug companies any time soon.

The loss of confidence in commonly used preclinical target validation assays for behavioral disorders is shared between disease areas. We continue to believe that current preclinical approaches have better chances of success in the area of addiction than other areas of psychiatry. In general however, it is clear that enthusiasm for psychiatric drug development has waned. While the cost of clinical development in behavioral disorders has skyrocketed, mechanism after mechanism that appeared promising in preclinical studies has simply failed to show clinical efficacy and been abandoned, leading to an exodus of major pharmaceutical companies (Miller 2010). Rightly or wrongly, addictive disorders have also been swept up in this wave. Dr. Steven Hyman, the former director of the National Institute on Mental Health, summed up the experience in a provocative piece that stated (Hyman 2012):

current animal-based assays have failed to identify efficacious drugs with new molecular mechanisms, and given scant understanding of the pathophysiology of common psychiatric disorders, it is difficult to develop better models. Furthermore, objective diagnostic tests and treatment responsive biomarkers are lacking. Without the latter, clinical trials of psychiatric treatments are dependent on disease definitions grounded in the descriptive psychiatry of the 1960s and 1970s

Although this statement primarily referred to therapeutic areas within general psychiatry, such as schizophrenia or major depression, giving it some thought is important for addiction neuroscience as well. To positively impact the lives of patients with alcoholism, tools and strategies are needed that can persuade pharmaceutical companies to pursue discovery of novel, innovative treatments. Given the exceedingly high cost of drug development, pharmaceutical companies will not make that investment unless they believe preclinical tools exist that can help predict clinical success, and inform the design of initial clinical studies.

Target validation in alcoholism has typically relied on behavioral assays centered around drug-seeking and self-administration, often thought to reflect the clinical construct of craving (Egli 2005). But craving is not uniformly experienced by patients. Other behavioral domains, such as impulsivity or attention bias, may also offer strategies for preclinical target validation. These domains will, however, need to be further deconstructed, mapped to clinical symptoms, and evaluated for predictive validity. For example, two drugs that both have documented clinical efficacy to reduce impulsivity in patients with ADHD have very different signatures in rodents: Amphetamine improves performance in delay discounting tests but worsens premature responding in a stop signal task, while atomoxetine has the opposite effects (Broos et al. 2012). Furthermore, when impulsivity measures were improved in alcoholics as a result of modafinil treatment, clinical drinking outcomes were dependent on baseline patient characteristics: Patients with initially poor inhibitory control showed improvement in abstinence and drinking rates, but those with better baseline impulsivity scores worsened in response to the same treatment (Joos et al. 2013).

We continue to work toward and hope for increasingly refined behavioral assays for target validation and translation, but also note that simpler, biochemical measures may have advantageous measurement properties and translational utility. In the remainder of this chapter, we discuss some of these more biochemically oriented translational strategies.

3 The Promise of Translational Biomarkers

In search of strategies to better guide medications development in alcoholism, some useful lessons may be learned from advances in the genetics of complex traits. Despite well-established heritability of many complex behavioral traits, attempts to identify their genetic underpinnings have over the years been fraught with frustration and seemingly spectacular findings that have then failed to replicate. One successful strategy to address this challenge has been to expand sample sizes far beyond what was once thought necessary. A recent showcase project in the field of schizophrenia recently obtained 128 genome-wide significant findings in a multistage schizophrenia GWAS including 36,989 cases and 113,075 controls (Schizophrenia Working Group of the Psychiatric Genomics 2014). These kinds of sample sizes are, however, not realistic in clinical development.

Another, complementary strategy from complex behavioral genetics may have a greater appeal. The influential concept of endophenotypes as empirically tractable constructs with advantageous measurement properties was originally introduced largely on theoretical grounds (Gottesman and Gould 2003), but its practical value has since been proven in numerous instances (see, e.g., Munafo et al. 2008), where it has helped inform findings obtained with more complex, distal phenotypes such as depression (Caspi et al. 2010). The basic contention of this approach is that relatively simple traits that are not visible to the unaided eye and are closer to the biology sometimes offer advantages over attempts to use the distal, complex traits themselves as readouts.

An important example of this concept are the Research Domain Criteria (RDoC). These attempt to describe psychiatric conditions within a matrix of functional dimensions such as positive or negative valence systems (e.g., reward learning or acute/potential threat, respectively) or cognitive control (e.g., inhibition/suppression) that are investigated across a number of variables ranging from genetics and circuit activity to psychology and behavior (Insel and Cuthbert 2009; Cuthbert 2014). The RDoC approach studies dysfunction in basic systems aiming to understand sets of symptoms that may cut across multiple disorders. RDoC provide a multilevel framework and database for integrating a large variety of studies from many sites.

Capturing targetable feature of alcohol addiction through a set of intermediate phenotypes directly implies that these responses may be utilized as translational biomarkers for medication development. Along these lines, “a translational biomarker strategy” can be applied to medication development in alcohol addiction. In this approach, preclinical target validation does not have to rely on attempts to model complex behavioral phenotypes thought to be more or less homologous to behavioral elements of alcohol addiction. Instead, a biomarker strategy that focuses on simpler biological measures is chosen because these have some mechanistic relation to alcohol effects on the nervous system, have advantageous measurement properties, can be obtained both in experimental animals and in humans, and are predictably responsive to pharmacological interventions. While the interpretation of any behavioral response in terms of underlying mental processes is difficult in both humans and animals, a more direct approach is provided by neuroimaging methods that allow to objectively ascertaining brain responses at various levels, including molecular, structural, and functional levels, and thus may comprise a translational tool for the proposed biomarker strategy. The remainder of this chapter will focus on examples that illustrate this strategy, and discuss our experience with these approaches to date.

3.1 Alcohol-Induced Dopamine (DA) Activation as Translational Biomarker

An ability to activate mesolimbic alcohol-induced dopamine (DA) transmission is a shared feature of most addictive drugs and has long been thought to play a key role in addictive processes (Wise and Bozarth 1987; Di Chiara and Imperato 1988). It has since become clear that this is only one of numerous processes involved in the development and maintenance of addiction (e.g., Koob and Le Moal 2008), but drug-induced DA-activation fulfills several criteria outlined above: When it occurs, it is simpler than the complex processes of drug-seeking and can be measured with reasonable reliability and precision in both experimental animals and humans. Drug-induced mesolimbic DA-release was originally established using animal models, pharmacological manipulations, and brain microdialysis, but was subsequently found to translate to the human situation (reviewed, e.g., in Volkow et al. 2004). Similar to other addictive drugs, mesolimbic DA-activation is also seen in response to alcohol, although alcohol-induced activation is less pronounced and more variable than that resulting from, e.g., cocaine or stimulants. Alcohol-induced DA-release was originally demonstrated in experimental animals using passive, non-contingent alcohol administration (Di Chiara and Imperato 1988; Tanda and Di Chiara 1998), but was subsequently also observed as a result of oral alcohol self-administration in rats (Weiss et al. 1993). An appealing feature of the latter observation was that DA-activation in Wistar-derived rats with high-alcohol preference as a result of genetic selection was markedly higher than that of non-selected Wistar rats. In a recent meta-analysis of in vivo microdialysis DA datasets, derived from more than 7400 rats, ethanol dose-dependently, globally, and independently of brain site increased DA basal concentrations up to 270 % (Brand et al. 2013).

Detection of alcohol-induced mesolimbic DA activity in humans must necessarily rely on more indirect measures than microdialysis in animals. Despite this challenge, mesolimbic DA-activation by alcohol has been demonstrated. Key observations have been made using PET, and alcohol-induced reduction in binding potential, or ΔBP for short, for the D2/3 ligand [11C]-raclopride (Boileau et al. 2003; Martinez et al. 2005; Urban et al. 2010; Ramchandani et al. 2011). Although this PET-based approach to measuring human DA-responses to addictive drugs is indirect, experiments in non-human primates have established that it reliably detects the effects of DA-releasing reference drugs (Dewey et al. 1993) and yields measures that have an excellent correlation with those obtained directly using microdialysis (Endres et al. 1997).

Drug-induced changes in raclopride BP are commonly called “displacement,” implying a competitive mechanism for the reduction in BP, but it may be worth noting that non-competitive mechanisms, such as ligand-induced receptor internalization, could also be involved. These processes may be affected by disease pathophysiology in some conditions, such as schizophrenia, where they would make interpretation of data more complex (Ginovart 2005). No data suggest, however, that changes in receptor internalization would confound PET measures of DA-release in alcohol research. Drug-induced ΔBP for raclopride is a neurochemically specific measure of resulting endogenous DA-release, and a gold standard in human studies. While fundamentally reflecting the same approach, a more recent DA-D2/3 ligand, [11C]-PHNO, may have potential to offer an improvement over the modest sensitivity offered by raclopride, because PHNO is an agonist that selectively labels DA-receptors in their high-affinity state (Willeit et al. 2008; Shotbolt et al. 2012).

PET-based measures of alcohol-induced DA-activation are appealing, because they have high neurochemical specificity, strong validation against direct measures of DA-release, and excellent measurement properties, while advances in ligand development may offer further improvements in sensitivity. However, PET-based measures also have distinct disadvantages as drug development tools. Most importantly, they are costly to obtain, require on-site radiosynthesis capacity, and have a low temporal resolution. Functional magnetic imaging (fMRI) lacks the neurochemical specificity of PET and is considerably noisier. However, the BOLD fMRI signal from the ventral striatum largely reflects DA-transmission in this structure (Knutson and Gibbs 2007), has small marginal cost, and allows a temporal resolution that is superior to that offered by PET. Using this approach, brain alcohol exposure closely controlled for pharmacokinetic variation has been shown to generate a robust signal from the ventral striatum of social drinkers, and the magnitude of this signal correlated strongly with subjective intoxication (Gilman et al. 2008). These data suggest that fMRI-based measures of alcohol-induced ventro-striatal activity may offer a useful complement to PET-based measures, although they are likely to require larger group sizes because of their inherently higher variance.

A consistent line of evidence supports the notion that alcohol-induced mesolimbic DA-activation can serve as a treatment responsive translational biomarker in alcoholism studies. The opioid antagonist naltrexone, an approved and effective alcoholism therapy (Jonas et al. 2014), was originally discovered in the absence of in-depth mechanistic understanding (Altshuler et al. 1980; O’Malley et al. 1992; Volpicelli et al. 1992). However, subsequent work showed that naltrexone and other, more mu-selective opioid antagonists, suppress alcohol-induced DA-activation in experimental animals (Gonzales and Weiss 1998; Tanda and Di Chiara 1998). This predicted that naltrexone would block alcohol-induced DA-release in humans. Surprisingly, this prediction has to our knowledge not been directly evaluated to date. However, data from a natural experiment provide strong indirect support for this notion.

To obtain these data, we capitalized on naturally occurring functional genetic variation at the locus encoding the target for naltrexone, an OPRM1 A118G single nucleotide polymorphism (SNP) that encodes an amino acid substitution in the N-terminal extracellular loop of the receptor, and mutates out a glycosylation site (Bond et al. 1998). Using PET and raclopride displacement, we found that the mu-opioid receptor variant encoded by the major 118A allele at this locus is associated with markedly lower mesolimbic DA-response to alcohol than the minor G-allele, essentially mimicking the functional consequences of mu-opioid receptor antagonism. As a reverse-translational tool, we then generated two humanized OPRM1 mouse lines, identical throughout the genome with the exception of the OPRM1 polymorphism. In agreement with the human PET findings, alcohol-induced DA-release in the ventral striatum of mice homozygous for the A-allele was dramatically lower than that found in GG-mice (Ramchandani et al. 2011).

Subsequent experiments in these humanized mice have provided data supporting the potential of alcohol-induced DA-release as a treatment responsive translational biomarker. Using a classical model of drug reward, suppression of intracranial self-administration (ICSS) thresholds, we found robust rewarding effects of alcohol in the GG-mice, while these effects were markedly attenuated or absent in the AA-mice. ICSS measures of alcohol reward were blocked in the GG-mice by naltrexone, while no effects of naltrexone were seen in AA-mice. Finally, in agreement with these data, naltrexone as well as another opioid antagonist, nalmefene, was markedly more effective in its ability to suppress alcohol self-administration in GG-mice compared to AA-animals (Bilbao et al. 2014). The latter data are in agreement with the original proposition, based on a secondary analysis of clinical trial data, that alcohol-addicted patients carrying the OPRM1 118G allele are particularly responsive to naltrexone treatment (Oslin et al. 2003). More recent secondary analyses, based on larger patient samples, have provided further support for this observation (Chamorro et al. 2012, Garbutt et al. 2014). A subsequent, prospectively genotyped study failed to detect a moderating effect of OPRM1 A118G variation on naltrexone efficacy (Oslin et al. in press), but that study also failed to detect overall efficacy of naltrexone as such, and can therefore not really inform the question of genetic moderation. It is simply a failed trial, as are indeed for unknown reasons about half of psychiatric trials that include a medication with documented efficacy (Khin et al. 2011).

Alcohol-induced DA-activation is not uniformly seen in all individuals. We have already noted that genetic variation at the OPRM1 locus is a potent determinant of this response, but other factors that influence this response have also been established. In a provocative PET-study, it was shown that the mesolimbic DA-response to alcohol in females is negligible compared to males (Urban et al. 2010). This observation is consistent with our findings in non-human primates, where only males responded with psychomotor stimulation to an alcohol challenge (Barr et al. 2007). Furthermore, several studies suggest that activation of mesolimbic DA by alcohol declines with progression into alcoholism. This has been observed in the ventral striatum both using [11C]-raclopride displacement (Martinez et al. 2005) and fMRI (Gilman et al. 2012; Spagnolo et al. 2014). Rather than reflecting limitations of the biomarker, these data highlight its strength. If the signature of a novel therapeutic mechanism is to attenuate alcohol-induced DA-release in experimental animals, then the absence or low magnitude of a DA-response to alcohol will help identify patients who are less likely to respond to therapeutics targeting that mechanism. Such individuals should then not be included in clinical efficacy trials, where they would only dilute efficacy signals from responsive patients. Accordingly, although available data may have some methodological limitations, being male appears to be a predictor of naltrexone response (Garbutt et al. 2014).

In summary, both PET and fMRI-based measures of alcohol-induced mesolimbic DA-activation should have considerable potential as treatment responsive translational biomarkers in developing novel alcoholism pharmacotherapies. It is important to point out that a narrow use of DA-responses as translational biomarkers of drug effects does not rest on any specific hypotheses about the role of the DA system in addiction pathophysiology. Several novel mechanisms of potential interest as alcoholism therapies have shown a signature that includes an ability to inhibit alcohol-induced mesolimbic DA-activation in preclinical studies. Among these, blockade of receptors for the appetite regulating hormone ghrelin appears particularly promising (Jerlhag et al. 2006, 2009; Landgren et al. 2012). Initial translation of this mechanism is now underway by Leggio and colleagues at the National Institutes of Health using the non-peptide ghrelin-1a receptor antagonist PF-05190457 (NCT02039349). If determined to be safe, this mechanism will be able to benefit from the biomarker strategy described here. Other candidate mechanisms with an ability to inhibit measures of alcohol-induced DA-activation are also in the pipeline, such as antagonism of the appetite regulating neuropeptides melanin-concentrating hormone type-1 receptors (MCH1-R) (Cippitelli et al. 2010). Successful translation of these promising preclinical findings will benefit from the application of a biomarker strategy outlined above.

On the other hand, observations that females and patients in later stages of alcohol addiction do not show robust alcohol-induced DA-activation suggest considerable heterogeneity among “alcoholics.” Clearly, a “smorgasbord” of biomarker strategies will be needed to facilitate translation across these heterogeneous populations.

3.2 Measures of Glutamate Activity as Translational Biomarkers

It is increasingly recognized that alcohol addiction is an evolving process, characterized by progressive and widespread recruitment of neuroadaptive changes in the central nervous system over time. At a behavioral level, this evolving process has been characterized as a shift from positively reinforced alcohol use to consumption that is increasingly driven by negative reinforcement, or from “reward craving” to “relief craving.” Neuroadaptive changes are initially triggered acutely during states of acute withdrawal, but ultimately persist into what can be called “protracted abstinence,” at which time they generate powerful incentives to resume alcohol-seeking and use (Heilig et al. 2010; Glockner-Rist et al. 2013; Meinhardt and Sommer 2015). Of course, in reality this process is not nearly as uniform and neat as described here. Based on the presence or absence of pre-existing genetic susceptibility factors and exposure to environmental influences, such as drug consumption itself or life stressors, people arrive at “neuroadapted” alcoholism through very different trajectories (Heilig et al. 2011). Irrespective of trajectory, however, once these neuroadaptations are in place, they are likely to offer additional translational biomarkers.

Among a multitude of neuroadaptive changes reported in alcoholism, adaptations within the glutamatergic system appear to be prominent, and offer a rich pharmacology that holds the promise of yielding novel alcoholism treatments (Spanagel and Kiefer 2008; Spanagel 2009; Holmes et al. 2013). Chronic excessive use of alcohol ultimately results in a hyperglutamatergic state, characterized by elevated extracellular glutamate and altered glutamate receptors and transporters. Pharmacologically manipulating glutamatergic neurotransmission alters a wide range of alcohol-related behaviors, such as acute intoxication and withdrawal, but also alcohol-seeking and consumption, in both rodents and humans. Accordingly, several elements of glutamatergic neurotransmission have been proposed as attractive targets for novel alcoholism treatments. For instance, blockade of NMDA and AMPA receptors reduces alcohol consumption in rats and mice. However, side effects are likely to limit the therapeutic potential of drugs that block ionotropic glutamate receptors, and experience with this strategy in stroke has not been encouraging (Gladstone et al. 2002). Targeting metabotropic glutamate receptors (mGluRs) may offer a better tolerated approach, in particular if pursued using allosteric modulators. Indeed, blocking mGluR5 potently affects various alcohol-related behaviors in rodents, and mGluR2/3 agonism or mGluR2 positive allosteric modulation also suppresses alcohol consumption (Spanagel 2010). Finally, glutamate transporter upregulation may mitigate behavioral and neurotoxic sequelae of excess glutamate caused by alcohol, and attenuate alcohol drinking.

The possibility that targeting the glutamate system holds promise for developing new alcoholism treatments makes it a priority to establish translational biomarkers responsive to glutamatergic medications. Once again, in establishing biomarkers of glutamate function, one can remain fairly agnostic regarding underlying pathophysiology or synaptic function. The measures that will be best able to aid translational efforts are those that are relatively simple, robust, and can be obtained in experimental animals as well as in humans with alcoholism, and are similarly responsive to drugs in animal and human models.

Early rat studies showed that extracellular glutamate levels, measured in rat striatum using brain microdialysis, increase during acute alcohol withdrawal (Rossetti and Carboni 1995). Benzodiazepines, the standard clinical treatment for acute alcohol withdrawal, blocked the behavioral withdrawal signs in that study, but did not prevent the glutamate surge. In contrast, the non-competitive NMDA antagonist MK-801 blocked both the behavioral and the neurochemical withdrawal symptoms. It was subsequently shown that within the ventral striatum, withdrawal-induced glutamate elevations increase with repeated cycles of withdrawal (Dahchour et al. 1998). A different pattern was found in the hippocampus, where glutamate elevations were found after a single withdrawal episode, but dissipated over subsequent cycles (Dahchour and De Witte 1999).

In a recent meta-analysis of in vivo microdialysis datasets, derived from 104 alcohol-dependent rats, consistent evidence was obtained for elevated extracellular glutamate levels in various brain sites that correlated with the intensity of the withdrawal response (Fliegel et al. 2013). Recently, Sommer and colleagues were able to detect withdrawal-induced increases in glutamate levels in rats using proton MR-spectroscopy (MRS) at high field, 9.4T (Hermann et al. 2012). These data represent a major advance because they tie together the direct, microdialysis-based measures, with those detected by MRS, showing that the latter to some extent reflect the extracellular glutamate pool that originates from synaptic transmission, and emphasizing the feasibility of a translational neuroimaging approach

Perhaps the best evidence for a potential of glutamate MRS as a translational biomarker in alcohol studies comes from data obtained with the glutamatergic modulator acamprosate, a clinically approved therapy for alcoholism. In mice with a deletion of the clock gene Per2, escalated voluntary alcohol consumption was observed by the Spanagel laboratory and was tied to decreased clearance of glutamate by the glial Glutamate Aspartate Transporter (GLAST; also Excitatory Amino Acid Transporter 1, EAAT1, or SLC1A3). Acamprosate rescued both the escalated alcohol self-administration and the striatal elevations of extracellular glutamate, measured directly by microdialysis, in the Per2 null-mutants (Spanagel et al. 2005). These data provided some of the most important support for a causal role of a hyperglutamatergic state in driving excessive alcohol consumption. Because MRS is able to tap into a signal that represents elevated overflow of synaptic glutamate, we hypothesized that MRS-based measures may offer useful biomarkers in drug development for alcoholism.

To address this hypothesis, we carried out an experimental medicine study and attempted to establish whether MRS at 3T would be sensitive and specific enough to detect a reduction in central glutamate resulting from clinical acamprosate treatment. A challenge for MRS studies at 3T is that glutamate and its precursor glutamine overlap significantly in the 1H resonance spectrum. Higher magnetic field strength makes it possible to resolve their respective resonances, but is not yet widely available for human studies. At 3T, the overlapping glutamate and glutamine peaks are frequently combined into a “GluX” peak, but because of the glutamine–glutamate cycle (Bak et al. 2006), this approach does not provide sufficiently detailed information about the functional state of glutamatergic transmission. An echo-time-averaged, point-resolved technique (TE-averaged PRESS) has been shown to detect an unobstructed glutamate signal at 3T that is resolved from glutamine at 2.35 ppm. TE-averaged PRESS therefore provides an unambiguous measurement of glutamate at 3T (Hurd et al. 2004; Srinivasan et al. 2006) and holds potential as a biomarker.

We used TE-averaged PRESS and scanned treatment-seeking alcohol-addicted patients randomized to acamprosate or placebo, as well as healthy volunteers who were scanned for comparison. Our first scan awaited steady state for acamprosate to be reached and was therefore carried out after acute withdrawal had subsided. At this point, there was no elevation of glutamate within a voxel placed in the anterior cingulate cortex (ACC) compared to controls. When these patients were rescanned three weeks later, however, the placebo-treated group showed significantly elevated glutamate levels, while levels in the acamprosate-treated group had, if anything, declined; the two groups were clearly separated at that time (Umhau et al. 2010).

In one important aspect, our data are complementary to those obtained by Sommer and colleagues (Hermann et al. 2012). Piecing together a time course from these two studies, it appears that acute alcohol withdrawal may be associated with a transient elevation of central glutamate, and that, in alcohol-dependent patients, levels start creeping up again in protracted abstinence, when relapse most frequently occurs. The suppression of that delayed elevation by acamprosate can be detected by MRS at 3T using TE-averaged PRESS technology. Of note, measures of glutamate in the cerebrospinal fluid (CSF) do not offer an alternative to the MRS-based biomarker; CSF glutamate appears to have a different origin (Umhau et al. 2010).

These approaches hold considerable potential to facilitate alcoholism therapies targeting mGluR2 receptors, a key player in controlling glutamate homeostasis identified as a promising target by converging lines of evidence. In rats, chronic intermittent alcohol intoxication results in a post-dependent state characterized by escalation of subsequent voluntary alcohol intake (Heilig et al. 2010; Meinhardt and Sommer 2015). Post-dependent animals also show long-lasting deficits in executive control, e.g., attention and response selection (Trantham-Davidson et al. 2014), functions dependent on the medial prefrontal cortex (mPFC). We have shown that the post-dependent state is associated with a persistent insult to the infralimbic mPFC, where a lasting suppression of mGluR2 receptor expression occurs (Meinhardt et al. 2013). The translational value of these findings is supported by human postmortem data showing a reduction in mGluR2 expression in the corresponding PFC region in alcoholics. A mechanistic role for the loss of infralimbic mGluR2 expression is demonstrated by the finding that viral vector-mediated rescue of mGluR2 expression in the infralimbic mPFC of post-dependent rats results in a rescue of their escalated self-administration.

Convergent support for a key role of mGluR2 receptors in control of alcohol-seeking and consumption comes from a recent study in P rats, a line selected for high innate alcohol preference. This work identified a premature stop codon in Grm2, the gene encoding mGluR2 receptors, which contributes about 25 % of the increased alcohol consumption in these animals (Zhou et al. 2013) Collectively, these results suggest that mGluR2 loss in rodent and human neural circuits, which provide cortical control over deeper brain structures involved in motivational and emotional regulation, may be a major consequence of alcohol dependence and a key pathophysiological mechanism for the increased propensity to relapse. Observations that PFC control over drug-seeking behavior can be restored by rescuing or enhancing the function of metabotropic glutamate receptors (Meinhardt et al. 2013; Gass et al. 2014) provides a strong rationale for translational studies aimed at assessing these processes in alcoholic patients.

Although glutamate MRS is an appealing translational biomarker for these studies, fMRI-based approaches may once again have utility. This is illustrated by a recent animal study that used pharmacological fMRI in awake rats, and demonstrated a modulation of ketamine-induced BOLD response by an mGluR2/3 agonist (Chin et al. 2011). Several mGluR2/3 agonists and mGluR2 positive allosteric modulators (PAMs) have cleared human safety studies. Both MRS and pharmacofunctional MRI-based biomarkers can likely be used to probe the functional state of the glutamate systems and provide a window on the molecular and neuronal basis of executive function impairment seen in many alcoholics. These probes should offer attractive translational biomarkers for medication development that targets dysregulated glutamate transmission in order to improve executive control in these patients.

4 Future Directions: The Whole—More than the Sum of Its Parts?

Specific neurochemical systems contribute to alcohol addiction, but the disorder is a systems level problem. Neuroimaging can provide an unbiased view of brain function at the systems level that allows ascertainment of network activity, but so far, only a few studies have used these tools in alcoholism (Camchong et al. 2013; Muller-Oehring et al. 2014). In general, functional connectivity analysis in abstinent alcoholics has shown an overall integrity of large-scale functional networks, but has found specific pathology in distinct sub-networks, in particular expanded connectivity in attention and visual input networks, supporting the concept of network expansion as a neural mechanism for functional compensation.

Expanding this approach could help identify a “disease-network” for alcohol use disorders. Because aberrant network states can be shifted by pharmacological interventions (Schmaal et al. 2013), “disease-network” states could potentially function as biomarkers for treatment development, even in the absence of a distinct target mechanism. The biomarker would then be an ability to force network activity in the direction of a “normal” state. Imaging studies have shown that spatial and functional characteristics of intrinsic brain network architecture are conserved across species, including rodents and humans. Thus, network analysis may emerge as a translational tool for medication development. In an attempt to explore the potential of brain network analysis in rats for studying alcoholism-related brain network states, Dudek and colleagues (Dudek et al. 2014) used manganese-enhanced MRI to investigate network activity after alcohol drinking and abstinence. Many of the activated brain regions have previously been implicated in alcohol reward, but in the prelimbic cortex, ventral hippocampus, and subthalamic nucleus, activation persisted into abstinence, supporting the idea of long-term changes comprising a “relapse-prone” network state. Future studies will address to what extent “relapse-prone” networks overlap between animals and humans, and can offer translational biomarkers.

5 Conclusion

Over the past two decades, major advances have been made in the neuroscience of alcohol addiction, but few of these have directly translated into improved treatments for patients with alcoholism. During this time, the field has devoted considerable attention to developing and debating behavioral animal assays that might hold promise of predictive validity, and that might facilitate translation (Egli 2005; Heilig and Egli 2006; Litten et al. 2012). To bridge the proverbial “translational valley of death,” other types of tools may have to be deployed. Treatment responsive translational biomarkers top the list of such tools, and our present paper describes some of those already available.

Tools that follow the same principles as those described, but can index alcohol and drug effects on other systems need to be added to the toolkit. A notable example in this category would be a displaceable ligand for the endocannabinoid CB1 receptor. Successful development and deployment of translational biomarkers will to a large extent have to rely on the ability of academia and pharma to work together, in turn related to the ability of governments to provide incentives and regulatory space for that kind of efforts (Hudson and Khazragui 2013). Although the pullout of major pharmaceutical companies from the CNS area confronts us with a rather gloomy picture compared to the hopes and expectations of only a decade ago, there may be a silver lining. As major pharmaceutical companies discontinue or scale down their psychiatry and addiction programs, many interesting molecules become available for licensing by consortia of biotech and academia. Those consortia may have to rely on venture capital or other sources, because few public funding mechanisms are adequate to support the costs of clinical development. Data obtained using translational biomarkers will be an important part of any portfolio that can attract investment, and ultimately benefit patients through development of novel therapeutics.