Alcoholism is one of the most expensive (more than $185 billion/year in the US) and damaging chronic diseases (Bouchery et al., 2011), yet there are only three FDA-approved drugs for alcohol dependence (disulfiram, naltrexone, and acamprosate) (Johnson et al., 2007). Disulfiram blocks the enzyme acetaldehyde dehydrogenase, preventing formation of acetic acid from acetaldehyde (Heilig and Egli, 2006). Naltrexone blocks mu opioid receptors thought to be responsible for the rewarding effects of alcohol, thereby decreasing alcohol craving (Bouza et al., 2004); however, naltrexone was only effective in a subpopulation of recovering alcoholics. Acamprosate is thought to act by decreasing glutamatergic activity, reducing the negative effects associated with alcohol withdrawal. Treatment options are varied but limited, and there is a high rate of relapse for all existing treatments.

Current evidence suggests that brain immune or pro-inflammatory signaling is linked to alcohol action (Crews, 2012; Mayfield et al., 2013). Long-lasting increases in levels of several cytokines in mouse brain, including monocyte chemoattractant protein-1, MCP-1 (also known as chemokine C-C motif ligand 2, CCL2), were observed after pretreatment with high doses of ethanol followed by injection of lipopolysaccharide (LPS) (Qin et al., 2008). Interestingly, a similar increase in MCP-1 (CCL2) was found in brains of human alcoholics (He and Crews, 2008), and deletion of Ccl2 or its receptor decreased alcohol consumption in mice (Blednov et al., 2005). Knockout of immune/inflammatory genes in mice, which were selected based on meta-analyses of genes and functional pathways involved in regulation of alcohol drinking in mice (Mulligan et al., 2006), rats (Kimpel et al., 2007), or humans (Liu et al., 2006, 2007; Flatscher-Bader et al., 2008), reduced ethanol intake and preference, further supporting immune pathways in regulating alcohol-mediated behaviors (Blednov et al., 2012). Our research suggests that cytokines (perhaps via endotoxins) promote persistent and excessive alcohol consumption, which may in turn promote further inflammatory responses, producing a positive feedback loop leading to excessive alcohol consumption (Blednov et al., 2011). Furthermore, even a single injection of LPS induces a long-lasting decrease in dopamine neuron firing in the ventral tegmental area, suggesting that cytokine signaling can regulate reward circuitry (Blednov et al., 2011). It is also noteworthy that chronic alcohol consumption and LPS treatment can produce similar changes in mouse brain transcriptomes (Osterndorff-Kahanek et al., 2013).

Phosphodiesterases (PDEs) are a superfamily of enzymes catalyzing the hydrolysis of 3',5'-cyclic adenosine monophosphate (cAMP) and 3',5'-cyclic guanosine monophosphate (cGMP) to their inactive 5'-AMP and 5'-GMP forms, respectively. Increased cAMP and/or cGMP, resulting from inhibition of PDE, reduces inflammatory signaling (Page and Spina, 2011; Jin et al., 2012). Cyclic nucleotides are known to play pivotal roles in a number of cellular functions, including immune and inflammatory responses, cardiac activities, smooth muscle relaxation, depression and cognition. The anti-inflammatory actions of PDE inhibitors are beneficial in treating chronic obstructive pulmonary disease and asthma (Jin et al., 2012; Keravis and Lugnier, 2012). PDEs are encoded by 21 genes, grouped into 11 families, according to their structural similarity. Each gene encodes multiple protein products generated by alternative splicing and/or multiple promoters, resulting in more than 50 different PDE proteins that may be produced in mammalian cells (Bender and Beavo, 2006).

Rolipram, an inhibitor of PDE4, markedly reduced alcohol intake and preference in mice (Hu et al., 2011) and reduced alcohol seeking and consumption in alcohol-preferring fawn-hooded rats (Wen et al., 2012). Rolipram also reduced cocaine conditioned place preference and self-administration (Knapp et al., 1999; Thompson et al., 2004) and inhibited neuroimmune signaling (Zhu et al., 2001). Rolipram and other PDE4 inhibitors have been investigated as potential therapeutics in diverse CNS disease models, including depression, anxiety, schizophrenia, Parkinson's, and Alzheimer's disease (Halene and Siegel, 2007; Kanes et al., 2007; Smith et al., 2009). In this study, we compared the effects of various PDE inhibitors on alcohol consumption and preference in mice. We studied nine PDE inhibitors with different selectivity profiles (Shahid et al., 1991; Raeburn and Karlsson, 1993; Ashton et al., 1994; Meskini et al., 1994; Sugioka et al., 1994; Perry et al., 1998; Dinter et al., 2000): propentofylline (nonspecific); vinpocetine (PDE1); olprinone, milrinone (PDE3); zaprinast (PDE5); rolipram, mesopram, piclamilast, and CDP840 (PDE4) and examined their effects in two different ethanol consumption tests.

Rolipram was previously shown to reduce ethanol intake and preference in mice (Hu et al., 2011) and rats (Wen et al., 2012) in 24-h two-bottle choice tests, which is consistent with our results in mice. We further demonstrate that other selective PDE4 inhibitors (but not other classes of PDE inhibitors) reduce intake and preference in the 24-h two-bottle choice paradigm. Like rolipram, CDP840 and piclamilast transiently reduce intake, whereas mesopram produces long-lasting decreases. Hu et al. (2011) did not report transient effects of rolipram, but they injected it i.p. twice daily, whereas we used a once daily oral administration. CDP840 also reduces ethanol intake and preference in the limited access test, whereas the other PDE4 inhibitors do not alter preference. Reduction of ethanol and total fluid intake, without changes in preference, in the limited access test may be due to the sedative activity of PDE4 inhibitors (Hu et al., 2011). Any sedative effects would be likely to interfere with drinking in this limited 3-h time period. However, mice in limited access or “binge” models may also display a different sensitivity to pharmacological intervention compared to mice undergoing continuous access (Crabbe et al., 2011). Furthermore, these two treatment paradigms induce different transcriptome signatures in mouse brain (Osterndorff-Kahanek et al., 2013). Other classes of inhibitors (nonspecific or PDE1, 3, or 5 inhibitors) had no effect on alcohol consumption in the continuous access test. Similar to our results with propentofylline, ibudilast (another nonspecific PDE inhibitor) did not change alcohol intake in non-dependent mice; however, ibudilast reduced drinking in alcohol-dependent mice and in alcohol-preferring P and HAD1 rats (Bell et al., 2013). Also, mice with genetic deletion of both adenylyl cyclase 1 (AC1) and AC8 exhibited decreased ethanol consumption and enhanced sensitivity to ethanol sedation (Maas et al., 2005). In contrast, mice lacking AC5 showed increased ethanol consumption and preference and reduced sensitivity to ethanol sedation (Kim et al., 2011).

It is important to note that the PDE inhibitors which did not alter alcohol consumption are nevertheless capable of exerting other CNS-dependent effects on behavior and neurophysiology. For example, olprinone and milrinone provided significant neuroprotection after cerebral ischemia/reperfusion injury in rats (Saklani et al., 2010; Genovese et al., 2011). Zaprinast and other PDE5 inhibitors improved memory (Prickaerts et al., 2004), and propentofylline improved learning and memory deficits induced by beta-amyloid protein (Yamada et al., 1998). Also, vinpocetine ameliorated hyperactivity in a mouse model of fetal alcohol spectrum disorder (Nunes et al., 2011).

Despite a long history of availability (since the late 1980's), PDE inhibitors are of limited clinical use because of lack of efficacy and/or gastroenterological side effects such as nausea and vomiting (Giembycz, 2008). One approach to mitigate the adverse effects is seeking agents with selectivity for PDE4 isoforms devoid of these effects. PDE4 isoenzymes are encoded by four genes (A, B, C, D), and it is believed that inhibition of PDE4D promotes emesis. Selective inhibition of PDE4A and/or PDE4B in pro-inflammatory and immune cells is believed to produce desired anti-inflammatory effects (Giembycz, 2008; Jin et al., 2012). PDE4B is of particular interest because it is localized in brain reward pathways (Cherry and Davis, 1999), and its expression is upregulated in mice with a genetic predisposition for high alcohol consumption (Mulligan et al., 2006). Rolipram inhibits PDE4 catalytic activity with micromolar potency but also binds to a high affinity (nanomolar) site on PDE4 (Schneider et al., 1986). Duplantier et al. (1996) noted a relationship between the high affinity site (in rats) and emesis (in ferrets), and there is a correlation between high affinity binding of denbufylline, rolipram, piclamilast, CDP840, and KF19514 and their emetic activity in the house musk shrew (Hirose et al., 2007). Thus, inhibitors with lower selectivity for PDE4D and lower binding to the high affinity site may reduce the side effects of these drugs.

Given the proposed link between neuroimmune signaling and alcohol consumption and dependence (Harris and Blednov, 2013; Mayfield et al., 2013), PDE inhibitors could potentially decrease alcohol consumption by inhibiting inflammatory signaling. Indeed, rolipram dose-dependently reduced LPS-induced production of TNFα in rat brain and microglia (Buttini et al., 1997; Yoshikawa et al., 1999). Furthermore, LPS, pro-inflammatory cytokines, and ethanol all increased the activity of PDE4 in several types of cells (microglia, astrocytes, bronchial epithelial), and these effects were reversed by rolipram (Buttini et al., 1997; Forget et al., 2003; Ghosh et al., 2012).

Our work represents the first comparison of different classes of PDE inhibitors in different ethanol drinking tests in mice with a genetic predisposition for high alcohol consumption. We provide strong evidence for a selective role of PDE4 inhibitors in reducing alcohol consumption and preference. PDE4 inhibitors have also been investigated in animal models of other CNS disorders, and the availability of more selective isoforms devoid of side effects may prove beneficial in the treatment of alcohol abuse and other inflammatory diseases.

In addition to behavioral evidence for specific PDEs in alcohol drinking in rodents, there is evidence for altered expression of PDE genes in mice following chronic ethanol. For example, Pde10 and Pde1b were upregulated in mouse prefrontal cortex after chronic ethanol intake (Osterndorff-Kahanek et al., 2013), and Pde4b was upregulated while Pde10a was downregulated in brains of mice showing a genetic predisposition to high alcohol consumption (Mulligan et al., 2006). Furthermore, the Collaborative Study on the Genetics of Alcoholism (COGA) identified a candidate single nucleotide polymorphism near PDE11A based on differing allele frequencies between alcohol-dependent and control individuals (Johnson et al., 2006). While both mouse and human studies nominate PDE genes, confirmation using multiple datasets combined with behavioral validation will ultimately be important to link them to addiction vulnerability.

Yuri A. Blednov designed and performed experiments, analyzed data, and wrote the paper; Jillian M. Benavidez and Mendy Black performed experiments; R. Adron Harris designed experiments and wrote the paper.