Introduction

The estimated percentage of US residents who are current users of cocaine has remained more or less constant since 2007, which unfortunately translates to a steady state of nearly 2 million Americans who use cocaine at least monthly, or around 0.7% of the population aged 12 or older. It is estimated that over 850,000 Americans meet the DSM-IV criteria for cocaine dependence [1]. The cocaine abuse epidemic continues to plague each generation of US citizens. Indeed, young adults have the highest percentage of current users and an estimated 29,000 US adolescents are cocaine dependent [1]. Clearly cocaine addiction is and remains a serious public health issue in the US and worldwide.

Despite decades of intense research effort, there remain no effective treatments for cocaine addiction. This is particularly frustrating since there are at least somewhat effective therapeutic options for opioid, alcohol, and nicotine addiction. A broad array of agents acting in the central nervous system (CNS), some with stronger rationales than others, have been assessed as potential cocaine addiction therapeutics [2, 3]. All have failed. Even substitution therapy, that is using an orally administered psychostimulant such as amphetamine or methylphenidate to minimize cocaine withdrawal and craving, has had mixed success [4,5,6,7]. Moreover, the doses of psychostimulants tested to treat cocaine dependence produce relatively high rates of adverse events including sleeping problems, agitation, irritability and increased heart rate [4].

It has long been appreciated that environmental and genetic factors influence the propensity to become addicted to drugs of abuse, including cocaine. More recently, changes in epigenetic modulation of gene expression have emerged as another factor contributing to the development of addiction. Some of these epigenetic modifications are heritable, leading to changes in descendent physiology and behavior. This review focuses on environmental, genetic and epigenetic factors (including trans-generational effects) that contribute to risk of cocaine addiction. The emphasis is on the identification of physiological targets based on this research that may serve as platforms for the development of cocaine addiction therapeutics.

Environment

An intuitive prediction would be that easy access to cocaine is more likely to lead to cocaine use escalating to addiction relative to an environment in which cocaine is not present. This common-sense assumption is supported by human twin studies indicating the importance of shared environmental experiences on addiction risk [8, 9], which is a factor for addiction to all classes of abused substances rather than being drug specific [8, 9]. As an example, exposure to parental cocaine use increases the risk of substance abuse generally but not necessarily cocaine addiction specifically. Known environmental risk factors for addiction include drug availability, childhood abuse, parental attitude regarding drug use, household drug use, peer drug use and lack of participation in social activities; somewhat surprisingly parental supervision/monitoring does not appear to be a risk factor for the initiation of cocaine use [10].

The two major twin studies of addictions [8, 9] reported broadly similar results with one major exception. The earlier study indicated that, unlike other illicit drugs of abuse, most of the genetic effects on heroin addiction were unique to this substance [8]. Interestingly, this article focused exclusively on twins who served in the military during the Vietnam war era [8]; those soldiers serving in southeast Asia were exposed to an environment in which heroin was readily available. Subjects in the subsequent study in which heroin-specific heritability was not observed, in contrast, were derived from the Virginia Twin Registry [9]. Rates of heroin exposure were lower in the Virginia sample and heroin-specific heritability was not observed [9]. Although this work did not focus on cocaine, it is an excellent example of the general influence of drug availability on drug use and dependence, which skewed the results of the report from Tsuang and colleagues.

Numerous animal studies also have documented the effect of environment on subsequent drug self-administration. These studies primarily focus on environmental enrichment, which consists of group housing in larger enclosures with running wheels, toys, and other features that rodents find appealing. Although this is the accepted nomenclature, there is general acknowledgement in the field that these “enriched” environments are actually closer to a natural environment, whereas the comparison standard environment could be characterized as impoverished [11]. Numerous studies have examined the effects of rodent enriched environment on stimulant-associated behaviors. An early study showed that environmental enrichment enhanced cocaine self-administration in rats using a two-bottle choice paradigm [12]. Consistent with this finding, environmental enrichment increased amphetamine-induced conditioned reward in rats [13]. However, the same group that published this amphetamine finding subsequently reported that environmental enrichment decreased i.v. amphetamine self-administration in male and female rats [14, 15]. It also has been shown that environmental enrichment makes rats less sensitive to the discriminative stimulus effects of cocaine and amphetamine [16] and, in mice, environmental enrichment decreased cocaine conditioned reward [17, 18]. On balance, these results indicate that environmental enrichment blunts the reinforcing effects of stimulants. This conclusion fits well within the framework established by the clinical literature, whereby participation in social activities are highly predictive of reduced susceptibility to develop substance use disorder [19]. These findings were also the basis of the long-held idea that the etiology of addiction emanates from complex interactions between environmental and genetic factors.

Genetics

Drug dependence is highly familial. Thus, rates of cocaine, opioids, alcohol, marijuana, and nicotine dependence are increased in siblings and relatives of drug dependent individuals [20, 21]. Two major population based twin studies examined the major classes of addictive drugs, revealing that genetics is a substantial risk factor in every instance [8, 9]. Moreover, as noted above, abuse of drugs in one class is highly correlated with drug dependence in other categories, suggesting a common vulnerability [8]. In terms of the stimulants, relatively little genetic variance is stimulant specific [8]. These results indicate that genetic factors predisposing individuals to drug abuse are common across classes of addictive substances [9, 22]. Importantly, these conclusions apply to both males and females [23]. Interestingly, as a side note, the genetic risk factors for abuse of illicit drugs appear to differ from alcohol and nicotine [23]. The finding that the genetics of addiction are similar across illicit substances indicates that genetic variation in the primary site of action of various drugs of abuse is not likely to contribute substantially to increased risk of substance use disorder [22].

Drug addiction is among the most heritable of psychiatric disorders, with the heritability of cocaine estimated at 65% for females [24] and 79% in males [25]. Although other cocaine heritability estimates are somewhat lower than these values [23, 26], genetic influences consistently explain the majority of risk for cocaine addiction [27]. There is not a clear pattern of Mendelian transmission, suggesting that addiction heritability is likely polygenic [28]. The monozygotic:dizygotic twin concordance ratios for the inheritance of addiction are low for all drugs of abuse with the exception of cocaine. A ratio of 4:1 for cocaine suggests that multiple alleles functioning in combination may underlie the heritability of cocaine addiction [27]. Candidate gene-driven approaches have revealed a number of single nucleotide polymorphisms (SNPs) associated with aspects of cocaine addiction [29] in proximity to genes including dopamine beta hydroxylase [30], neuron-specific vesicular protein (as known as calcyon) [31], catechol-O-methyltransferase [32, 33], neuronal calcium sensor 1 [34], l-type calcium channel CACNA1D [35], cannabinoid receptor 1 [36, 37], delta opioid receptor [38, 39] tryptophan hydroxylase-2 [40], and Homer 1 [41]. Although these findings are compelling, several of these results have not yet been replicated and it is not clear that any of these genes, with the possible exception of delta opioid receptors and l-type calcium channels, represent clearly viable targets for novel drug development.

As noted above, the available evidence indicates that common genetic factors underlie the addiction liability of all illicit drugs of abuse [8, 9, 22]. The missense polymorphism rs16969968:G > A in CHRNA5, which encodes an amino acid change (D398N) that leads to hypofunctionality of (α4β2)2α5 nicotinic acetylcholine receptors (nAChRs) [42], is associated with risk for nicotine addiction [43,44,45,46]. This association is likely the most replicated finding in the human psychiatric genetics literature [47]. It was reasonable, therefore, to posit that rs16969968:G > A might also play a role in cocaine addiction. This hypothesis was tested through a genetic association analysis, which revealed a significant association between the CHRNA5 variant rs16969968 and cocaine addiction, but the direction was opposite that of nicotine addiction [48]. That is, expression of the A relative to the G allele of rs16969968 in CHRNA5 is a risk factor for nicotine addiction but is protective for cocaine addiction [48,49,50]. The physiological basis of this distinction is not completely clear but animal studies show that nAChR antagonists decrease the reinforcing effect of cocaine [51, 52], whereas mice with a null mutation of Chrna5 increase their intake of higher doses of nicotine [53].

To date, only one genome wide association study (GWAS) specifically focused on cocaine dependence has been published [54]. Using a symptom count-based analytic approach, one genome-wide significant association with cocaine dependence at rs2629540 was identified. This SNP maps to an intron of the FAM53B (family with sequence similarity 53, member B) gene on chromosome 10 [55]. FAM53B has not been studied extensively, such that its function is not clear. A gene homolog in fish is associated with cell proliferation [56] and axonal extension during development [57]. Changes in several other SNPs were identified with near genome-wide significance, which were associated with genes encoding voltage-gated sodium and potassium channels as well as cyclin-dependent kinase 1 [55]. A systems genetics study that assessed cocaine self-administration in 39 strains of recombinant inbred BDX mice revealed that Fam53b was one of six genes that exhibited midbrain expression levels that mapped to the cocaine self-administration behavioral quantitative trait loci [58].

It has been difficult to identify common germline alleles that explain even a small amount of the risk (e.g., odds ratios >1.3) for cocaine addiction [55]. Part of the reason for this is the fact that human GWASs require subject pools numbering in the tens of thousands to achieve appropriate statistical power [59]. This is a situation where animal studies represent an alternative, cost-effective means to identify alleles associated with relative levels of cocaine use. Unfortunately, few insights have emerged from the preclinical literature, which may be partially due to the fact that self-administration experiments in mice, the non-human species of choice for genetic studies, are methodologically difficult. However, a few studies in mice have addressed this issue. For example, B x D mice strains with poor reversal learning (a measure of inhibitory control over impulsive responding) acquired cocaine self-administration more readily and maintained higher rates of cocaine intake relative to strains with good reversal learning [60]. These results suggest that heritable differences in inhibitory control affect cocaine self-administration [60]. Consistent with these findings in rodents, fMRI studies of cocaine-addicted individuals show impaired functional connectivity within and between frontostriatal circuits that is associated with compulsive drug use and trait impulsivity [61]. Although experiments designed to assess the genetic underpinnings of this phenomena have not been undertaken, there is evidence that B x D strains that display poor reversal learning have decreased D2 dopamine receptor expression in the brain [62] similar to the observation in human cocaine addicts [63].

Some rodent studies, although not all [64], indicate that the initial behavioral response induced by psychostimulants is predictive of subsequent higher levels of self-administration [65]. Cocaine-induced locomotor activity differed significantly across 45 inbred mouse strains, suggesting a likely genetic basis for these phenotypic differences [66]. A number of studies in rats have assessed various psychostimulant self-administration behaviors and have reported differences among (primarily inbred) strains [67,68,69,70,71]. Although the individual studies reveal interesting information, they are collectively problematic in that no one study has systematically examined a broad range of rat strains using consistent methodology and focusing on a specific psychostimulant. Thus, the extant literature suggests that there are differences in psychostimulant self-administration among rat strains but it is impossible to state with certainty the extent and magnitude of these disparities.

In a successful example of the use of animal models to investigate the genetic factors underlying cocaine addiction, a missense mutation in the Cyfip2 gene was shown to be the basis for differential acute cocaine locomotor response as well as behavioral sensitization in two substrains of the C57BL/6 model mouse [72]. C57BL/6 N mice maintained by the NIH for the International Knockout Mouse Consortium (IKMC) project show lower acute response and sensitization to cocaine at several doses than their C57BL/6 J counterparts, which are maintained at the Jackson Laboratory. An intercross quantitative trait loci (QTL) analysis between these two strains for cocaine response yielded a single QTL on chromosome 11. Further sequencing of this QTL revealed a single nonsynonymous SNP leading to a missense S968F mutation in the Cyfip2 gene, causing rapid degradation of the CYFIP2 proteins. Animals exhibiting the mutant allele have lower numbers of striatal medium spiny neuron spines and fewer mESPCs, which could account for their decreased cocaine sensitivity. Knockout of the Cyfip2 mutant allele partially restores cocaine sensitivity, thus strongly implicating Cyfip2 as being at least partially responsible for the differential cocaine behavioral phenotypes [72]. Thus, exploiting the vast amount of information regarding mouse genetics is a viable path to identifying novel targets for the development of therapeutics for cocaine addiction.

Epigenetics

Even though the identity of the specific genes underlying the heritability of cocaine addiction remains elusive, there is abundant evidence that repeated exposure to cocaine results in pronounced changes in gene expression in various nuclei, particularly in the limbic system. Candidate gene, gene array, and genome-wide approaches have delineated changes in gene expression in the nucleus accumbens, the hub of the limbic system that modulates numerous cocaine-mediated behaviors [73,74,75,76]. Changes in gene expression are regulated by complex interactions between transcription factors, chromatin, and epigenetic processes. Here, we will focus on epigenetics since the role of transcription factors in addiction have been reviewed in depth elsewhere [77, 78]. The identification of epigenetic mechanisms underlying cocaine-induced alterations in gene expression has accelerated in recent years and represents a promising new avenue for therapeutic development.

Although the definition of epigenetics is somewhat contentious (as reviewed in more detail in the next section focusing on cross-generation effects), in the most general sense epigenetics is the study of molecular mechanisms that influence gene transcription without changing the DNA sequence. Examples include DNA methylation and post-translational modifications (PTMs) of histone N-terminal tails, including acetylation and methylation. Epigenetic processes regulate the accessibility of the nearly two meters of DNA packaged into the nucleus of the vast majority of eukaryotic cells. In the nucleus, DNA is so tightly wrapped around octamers of histone proteins (two each of H2A, H2B, H3, and H4) that gene transcription cannot take place unless the DNA is first unwound. Each histone octamer is encircled with around 147 base pairs of DNA, a unit known as the nucleosome, which is somewhat analogous to thread (DNA) wrapped around a spool (histone). All nucleosomes together make up chromatin, which is further organized into chromosomes. The transition between open and closed chromatin regulates access by the transcriptional machinery to neighboring genes, thereby promoting or repressing gene transcription, respectively. Histone PTMs, which are added and removed by writer and eraser proteins, respectively, make up a “histone code”, which is decrypted by reader molecules that recruit activating or repressing complexes and influence the compaction of chromatin.

During development, many genes in a given cell are silenced permanently in the process of cell fate determination. Generally, regions of the genome harboring more highly methylated areas of the genome are less transcriptionally active [79]. Thus, irrelevant genes in neurons, for example those associated with liver function, can be silenced during development through DNA methylation. Like histone PTMs, DNA methylation can and does change across the lifespan. Indeed, there are numerous examples of environment-mediated changes in DNA methylation and histone epigenetic modifications across the lifespan that alter phenotypes [80, 81]. A growing literature focuses on epigenetic changes induced by drugs of abuse, including cocaine. Here, we review particularly salient examples of functionally relevant cocaine-induced histone PTMs or DNA methylation in the nucleus accumbens.

Histone PTMs

Although the list of histone PTMs continues to expand, neuroscience research has focused primarily on acetylation and methylation. Histone acetylation tends to result in open chromatin, which is permissive for gene transcription. The influence of histone methylation is more complex in that transcriptional enhancement or repression depends on the residue modified as well as the number of methyl marks added, which can vary from 1 to 3. An early study examined genome-wide histone acetylation and methylation the day after seven days of experimenter-delivered cocaine in mice. Their results showed that cocaine-induced increases in the expression of genes including Arc, Cart, Cdk5 NFκB, and Period 1 were associated with increased H3 and/or H4 acetylation and either unchanged or decreased H3 methylation [82]. Increased methylation at H3 following repeated cocaine injections was associated with decreased expression of Kv8.2, microtubule associated protein 2 and cGMP dependent protein kinase in the nucleus accumbens [82].

The effect of histone acetylation on cocaine-mediated behaviors has been assessed with histone deacetylase (HDAC) inhibitors, which attenuate the removal of acetyl groups from histone tail lysine residues thereby facilitating gene transcription. Most studies of this sort have focused on environmental conditioning associated with the subjective effects of cocaine in rodents as measured by the conditioned place preference (CPP) paradigm. Systemic administration of a general HDAC inhibitor facilitated the extinction of cocaine-mediated CPP and attenuated the subsequent reinstatement response [83]. HDACs are subdivided into subclasses. Class II HDACs include HDAC4 and HDAC5, which shuttle between the cytoplasm and nucleus in response to various cell signals. Over-expression of HDAC4 in the nucleus accumbens attenuated cocaine CPP [84]. Using the cocaine self-administration paradigm, daily HDAC inhibitor treatment into the nucleus accumbens increased the motivation of rats to consume cocaine, an effect that was reversed by virally mediated over-expression of HDAC4 in the accumbens shell [85]. Cocaine-mediated CPP was enhanced in HDAC5 knockout mice, an effect that was reversed by increasing HDAC5 expression in the nucleus accumbens [86]. Repeated cocaine dephosphorylates HDAC5 resulting in its accumulation in the nucleus [87]. Accumbens expression of a mutant form of HDAC5 that cannot be phosphorylated at S279 attenuated the development of cocaine CPP [87]. There also is evidence that the sirtuins, which are class III HDACs, are involved in cocaine-mediated behavioral plasticity. Thus, repeated cocaine exposure increased H3 acetylation associated with Sirt1 and Sirt2 as well as Sirt1 and Sirt2 catalytic activity in the nucleus accumbens [82]. Administration of a sirtuin inhibitor directly into the nucleus accumbens attenuated both cocaine CPP and self-administration [82]. Collectively, these results indicate that cocaine-induced acetylation of histones increases the expression of various genes in the nucleus accumbens [88], which combine to promote cocaine conditioned reward as well as increase the reinforcing efficacy of cocaine.

In contrast to acetylation, histone methylation can either repress or activate gene transcription depending upon which histone residue is methylated and the number of methyl groups added. Early work on repressive marks modified by repeated cocaine focused on trimethylation of lysine 9 on histone H3 (H3K9me3). Repeated experimenter-delivered cocaine injections persistently reduced H3K9me3 abundance in the nucleus accumbens, which resulted in the unsilencing of LINE-1 retrotransposable elements, among other effects [89]. These modifications may reflect global changes in genomic stabilization in the nucleus accumbens following repeated cocaine [89]. A selectively bred rat line that displays persistent cocaine seeking was shown to have increased H3K9me3 association to D2 dopamine receptor (D2DR) promoter and decreased D2DR expression in the nucleus accumbens, with both normalizing following cocaine self-administration [90]. Repeated cocaine also decreased levels of the repressive mark, H3K9me2, in the nucleus accumbens, which was due to the repression of the lysine dimethyltransferase, G9a [91]. Viral-mediated reduction in accumbal G9a increased cocaine-mediated CPP and enhanced the expression of a number of genes including several associated with cocaine-mediated dendritic changes in the nucleus accumbens [91]. Consistent with these results, over-expression of G9a specifically in D1DR-expressing, but not D2DR-containing, accumbens neurons reduced cocaine CPP [92]. Taken together, these results indicate that changes in both activating and repressing histone methylation in the nucleus accumbens contribute to cocaine-induced neuronal and behavioral plasticity.

A recent genome-wide examination of the effects of repeated cocaine exposure focused on several repressive (H3K9me2, H3K9me3, and H3K27me3) and activating (H3K4me1, H3K4me3, and H3K36me3) marks in the mouse nucleus accumbens. Their results indicated that cocaine produced substantial modifications in each of these marks at numerous genes and non-genic loci [74]. Somewhat surprisingly, repeated cocaine produced dramatic increases in alternative splicing in the accumbens. Indeed, repeated cocaine only regulated total transcript levels of around 100 genes, whereas over 1000 genes displayed altered splicing [74]. Further analysis of the neuron-specific splicing factor A2BP1 revealed that repeated cocaine resulted in translocation of A2BP1 to the nucleus where it associated with H3K4me3 and activated several hundred genes (including differential expression and alternative splicing). A functional role of A2BP1 was confirmed in that conditional knockdown of A2BP1 in the nucleus accumbens attenuated cocaine CPP [74]. Interestingly, H3K4me3 may also play an important role in mediating conditioned reward of other psychostimulants including methamphetamine. For example, viral-mediated knockdown of the H3K4me3 methyltransferase Mll1 in the nucleus accumbens attenuated CPP induced by methamphetamine [93].

H3K4me3 is in close proximity to H3R2me2a, a repressive mark that counteracts the activating effect of H3K4me3. Protein-R-methyltransferase-6 (PRMT6) is a nuclear enzyme that asymmetrically dimethylates R2 on H3. Recent evidence indicates that H3R2me2a and PRMT6 are decreased in the nucleus accumbens following repeated exposure to cocaine in mice, rats and humans [94]. Interestingly, these cocaine-induced decreases in accumbens H3R2me2a and PRMT6 occurred selectively in D2DR-containing accumbens output neurons, activation of which suppresses cocaine-mediated behavioral effects. Indeed, over-expression of PRMT6 in D2DR-containing accumbens neurons increased cocaine CPP [94]. The decrease in accumbens H3R2me2a was associated with increased H3K4me3, which promoted the transcription of a small set of genes including Srcin1, a Src repressor that dampens the behavioral effects of cocaine [94].

Given the potentially broad effects of histone PTMs on gene expression, a challenge to the field is to develop methodologies to modulate histone PTMs as selectively as possible. To this end, engineered transcription factors were used to selectively modify chromatin at specific genes. Zinc-finger proteins (ZFPs) or transcription activator-like effectors (TALEs) were designed to recognize and bind to a narrowly defined locus in the genome and deliver specific histone modifications [95]. A recent study investigating the mechanism linking chromatin dynamics to reward pathology applied ZFP-p65 or ZFP-G9a, targeting histone H3 lysine 9/14 acetylation (H3K9/14ac), a transcriptionally active mark, or H3K9me2, which is associated with transcriptional repression, respectively, to the Fosb gene [95], the transcriptional regulation of which is strongly implicated in the actions of drugs of abuse [91, 96]. These ZFPs were sufficient to modify histones at the targeted region of the Fosb promoter in nucleus accumbens and to control drug-evoked transcriptional and behavioral responses. Intriguingly, Fosb-ZFP-G9a was sufficient to block cocaine-induced Fosb activation via interference with CREB phosphorylation [95], providing direct evidence of the hierarchy between chromatin modifiers and transcription factors in gene regulation. More recent studies applied cell-type specific epigenetic editing using Fosb-ZFPs expressed in specific neuronal cell types using Cre-dependent HSV expression in mice transgenic for Cre recombinase in one of two types of accumbens principal neurons [97].

Fosb is just one of thousands of genes under epigenetic control in the context of drug exposure. There is growing evidence that cyclin-dependent kinase 5 (Cdk5) expression in accumbens influences reward-related behaviors [98]. A recent study found that HSV-mediated expression of ZFP-p65 and ZFP-G9a targeting the Cdk5 locus in NAc were sufficient to bidirectionally regulate Cdk5 gene expression via enrichment of their respective histone modifications at the Cdk5 promoter [99]. Further, Cdk5-targeted H3K9/14ac increased cocaine-induced locomotor behavior while Cdk5-targeted H3K9me2 attenuated cocaine reward. These data are especially compelling given that conventional Cdk5 overexpression or knockdown caused opposite behavioral phenotypes [98, 100, 101], demonstrating the importance of targeted epigenetic remodeling tools to invoke subtler, but physiologically-relevant, changes in gene regulation.

DNA methylation

Gene transcription also is profoundly influenced by cytosine modifications including methylation, which occurs mainly at 5′-cytosine-phosphate-guanine-3′ (CpG) sites. Regions of DNA with a high concentration of CpG sites are known as CpG islands, which tend to be concentrated at promoters located near transcription start sites. Methylation of multiple CpGs in an island, via enzymes such as DNA methyltransferases (DNMTs), can result in the stable silencing of gene transcription. In contrast, cytosine methylation within the gene body can promote gene expression [102]. Interactions with other proteins is necessary to produce DNA conformational changes. For example, methyl CpG binding protein 2 (MeCP2) binds to methylated DNA and recruits additional proteins to form a complex that can result in chromatin compaction, HDAC recruitment and transcription factor repulsion, all of which result in gene silencing. It also has been shown that in certain circumstances MeCP2 recruits the transcription factor CREB and promotes gene expression [103].

In animal models, cocaine exposure has been shown to induce global hypomethylation in the nucleus accumbens [104]. Somewhat surprisingly, rodents exposed to cocaine repeatedly also displayed persistently increased expression of methylation-inducing Dnmt3a [105] and Dnmt3b [104]. Moreover, the increases in accumbens Dnmt3a following repeated cocaine were selective to D1DR-expressing neurons [106]. These changes in DNMTs were shown to be functional in that systemic administration of a methyl donor attenuated the development of cocaine-induced behavioral sensitization and cocaine priming-induced reinstatement of cocaine seeking [104]. Consistent with these results, viral-mediated modulation of accumbens Dnmt3a resulted in opposing effects on cocaine CPP in that upregulation of accumbens Dnmt3a suppressed cocaine CPP, while down-regulation of accumbens Dnmt3a produced the opposite effect [105]. DNMT3a over-expression also reversed cocaine-induced increases in spine density in the accumbens [105].

The reinstatement of cue-induced cocaine seeking behavior, an animal model of craving, progressively increases, or “incubates”, over the first several weeks following the cessation of cocaine self-administration [107]. This is particularly interesting given that 1 day following cocaine self-administration Dnmt3a was decreased but Dnmt3a expression increased after 28 days of forced cocaine abstinence [105]. Consistent with this observation, the extent of DNA methylation in the nucleus accumbens was enhanced following 30 days of cocaine abstinence relative to the first day. These cocaine-induced increases in accumbens DNA methylation were at least partly negatively correlated with gene expression [108]. Intra-accumbens administration of a DNMT inhibitor attenuated the incubation of cocaine reinstatement, whereas a methyl donor enhanced cocaine cue-mediated reinstatement of drug seeking [108]. Administration of the DNMT inhibitor prior to the reinstatement session influenced a number of genes including demethylation of the estrogen receptor 1 (Esr1) promoter and activation of mRNA expression. A subsequent experiment revealed that an ESR1 agonist attenuated the incubation of cocaine reinstatement [108]. This work demonstrates how examination of patterns of DNA methylation following cocaine self-administration can lead to the identification of novel potential therapeutic targets for cocaine addiction.

The transcriptional silencing effect of MeCP2 is mediated, in part, by its phosphorylation at Ser421, which is enhanced by exposure to amphetamine in a D2DR-dependent manner [109]. MeCP2 also appears to play a role in the conditioned rewarding effect of amphetamine in that viral-mediated knockdown of MeCP2 in the nucleus accumbens enhanced amphetamine-induced CPP, whereas over-expression of accumbens MeCP2 produced the opposite effect [109]. Cocaine self-administration increases MeCP2 expression in the dorsal striatum and viral-mediated knockdown of MeCP2 in this brain region reduces cocaine self-administration [110]. Decreased dorsal striatal MeCP2 also potentiates the ability of cocaine to increase expression of the short non-coding RNA, miR-212 [110]. Viral-mediated increases in striatal miR-212 decreases the expression of BDNF, a neurotrophin that promotes enhanced cocaine intake [110]. It is important to note that microRNA-induced RNA silencing is a post-translational process that regulates gene expression and is, therefore, epigenetic [111]. While we do not include it in the scope of this review, an emerging literature focuses on the role of microRNAs in cocaine-induced neuronal and behavioral plasticity [112, 113].

Oxidation of 5-methylcytosine to products including 5-hydroxymethylcytosine (5hmC) has been identified as an intermediate step to DNA demethylation, but recent evidence suggests 5hmC itself may function as an epigenetic mark to regulate gene expression. Repeated administration of cocaine increased 5hmC at gene body and intergenic regions associated with enhancer elements and alternative splicing sites [114]. Enrichment of 5hmC was associated with upregulation of genes involved in synaptic plasticity and implicated in addiction including Adcy1 and Nrtk2, a receptor for BDNF. Moreover, increased 5hmC concomitant with higher Adcy1 and Nrtk2 gene expression persisted for one month following cocaine administration [114]. Interestingly, repeated cocaine delivery also downregulated expression of ten-eleven translocation 1 (Tet1), an enzyme known to catalyze the conversion of 5-methylcytosine to 5hmC [114]. Virally-mediated knockdown of Tet1 increased cocaine CPP and, counterintuitively, mirrored the effect of cocaine to enrich 5hmC in the same subset of genes [114]. This study highlights the importance of locus in determining the impact of epigenetic marks like methylation or hydroxymethylation as well as the effect of factors known to regulate their expression. Elucidating the role of novel epigenetic marks like 5hmC in the cellular and behavioral response to cocaine may be critical to fully understanding the epigenetic landscape of addiction.

Cross-generational epigenetics

One of the more interesting aspects of epigenetic regulation is the influence of parental life experiences on subsequent generations through stable epigenetic alterations in the germline. Unfortunately, the terminology in the general area of heritable epigenetics is not always standardized, often confusing and sometimes controversial. The very definition of the term epigenetics has been a source of debate, with each faction of related fields favoring one version over the other. The word was coined by Waddington [115] and literally means “above” or in “addition to” genetics. The standard definition of an epigenetic trait is a stably heritable phenotype resulting from changes in a chromosome without alterations in the DNA sequence [116] that are transmitted either to progeny through germline modifications or to daughter cells directly. In contrast, we have used the term epigenetics to describe all functional changes in DNA methylation and histone PTMs in neurons, the vast majority of which are post-mitotic. Semantics aside, this section focuses on influences of exposure to cocaine on the behavior and physiology of descendants, with an emphasis on studies with an epigenetic mechanistic component.

In this context, another important distinction to make is the one between intergenerational and transgenerational effects of drug exposure. When the effects of environmental perturbations such as drug exposure are transmitted across generations and direct effects on fetuses or gametes cannot be ruled out, this is known as intergenerational heritability. Fetal alcohol syndrome is an example of intergenerational inheritance. A pregnant woman who drinks alcohol heavily puts her baby at risk of CNS damage among other signs and symptoms. Thus, maternal behavior can influence the phenotype of the offspring, which may or may not result from changes in the baby’s DNA sequence or epigenetic marks. If the fetus is female, three generations are nested one inside of the other: the mother, the fetus and the primary oocytes that will ultimately produce the grandchildren. Paternal environmental insults also can produce intergenerational effects by influencing the sperm epigenome, which can reorient early developmental programming and impact progeny. Therefore, any changes in biology or behavior in the children (maternal and paternal lineage) or grandchildren (maternal lineage) could completely or partly result from direct exposure to the environmental insult making it difficult to tease apart the mechanisms underlying physiological changes in descendants.

In contrast, if transmission of the phenotype remains stable across multiple generations in the absence of fetus or oocyte exposure to the environmental perturbation, that is an example of transgenerational inheritance. Several instances of transgenerational heritability in humans have been reported. For example, around the turn of the twentieth century, the remote community of Överkalix, Sweden routinely experienced periods of food abundance and famine. Retrospective studies indicated that if the paternal grandfather experienced a period of food excess during the slow growth phase of adolescence, his grandsons had increased risk of diabetes-related mortality [117, 118]. The mechanism underlying this transgenerational phenomenon may be conserved germline epigenetic marks. This is a controversial concept since, traditionally, it was thought that all epigenetic marks were erased soon after fertilization. Moreover, sperm histones are replaced by protamines, which allows for higher chromatin compaction. Following fertilization, paternal protamines are replaced by histones of maternal origin [119]. Despite these steps to erase any current epigenetic information early on in development, it is now clear that some parental DNA methylation is conserved [120] and a limited number of paternal histones and their epigenetic marks are retained and influence development following fertilization [121]. Findings like these provide mechanistic bases for transgenerational epigenetic inheritance. What follows are several recent examples of heritable effects of exposure to cocaine. This section will focus on paternal manipulations, which circumvent some of the potentially confounding factors associated with studies of maternal cocaine exposure including changes in maternal behaviors. In rodent models, the sole contribution of sires to their progeny comes from the genetic and epigenetic material in their sperm. Hence studies of paternal cocaine exposure provide an opportunity to systematically delineate the possible modes of transmission from fathers to their descendants. This is by no means a comprehensive survey of this literature. Recent reviews provide a broader examination of the cross-generational influences of exposure to drugs of abuse [122,123,124].

In order to assess the effects of paternal cocaine exposure on offspring behavior, investigators have used different rodent models of addiction. Generally, male rodents were either allowed to self-administer cocaine or received non-contingent injections of cocaine for over two months, which covers the duration of spermatogenesis in mice and rats. During the latter stages of spermatogenesis, chromatin becomes highly compacted obviating modifications in epigenetic marks. Thus, there is a relatively short window of sensitivity when cocaine can plausibly interact with the epigenetic and transcriptional machineries in the germline to reprogram heritable epigenetic marks that can later influence neurodevelopment. Following the cocaine regimen, sires are mated with drug naïve females and the resulting first generation (F1) progeny were behaviorally assessed as adults. As the number of paternal cocaine exposure studies has grown, so too has the variability in the observed phenotypes on the next generation of animals. The unequivocal conclusions that can be reached from this research is that paternal cocaine exposure affects several behavioral modalities including propensity to consume drugs of abuse, reward processing, measures of mood and anxiety as well as cognitive processes [125,126,127,128,129,130,131].

Most publications have reported that paternal cocaine exposure has a protective effect on addiction-like behavior using either drug self-administration [126, 131] or conditioned place preference [132]. In two of these examples, the effects were specific to male offspring and cocaine-reward [126, 131]. In sharp contrast, a multigenerational mouse model suggested that paternal cocaine affected both non-drug (sucrose) and cocaine-rewards in both male and female offspring [132]. These discrepancies highlight the need to assess both male and female progeny in these studies and underscore the complexity of how information can be carried across generations. One possibility is that the species used in the studies (mice vs. rat) can lead to divergent results, which is not surprising and consistent with a large body of research on this subject. The mechanisms underlying the sex specificity of many of the transgenerational effects of cocaine remain unclear and represent a major challenge for the field.

The jury is still out regarding changes in anxiety-like behavior, with reports noting an increase in anxiety-like behavior in the progeny of cocaine-exposed sires [127, 132] versus two other articles observing no change in baseline anxiety [130, 131]. Given the fact that anxiety-like behavior is notoriously difficult to assess in rodents and the added variability in methodologies to generate the F1 generation, these discrepancies are perhaps not surprising. Similarly, about half of the publications reports deficits in cognitive capacities in progeny of cocaine-exposed sires [128, 129], while others observed no impact of paternal cocaine on memory in male or female offspring [130, 132].

Only a few articles have explored the neuroepigenetic mechanisms underlying the behavioral phenotypes derived from paternal cocaine exposure. The consensus is that BDNF mRNA and protein expression in the PFC of drug naïve, adult cocaine-sired rats are increased [126, 131] and that these changes are functionally relevant for reducing the reinforcing efficacy of cocaine in the male F1 cocaine-sired progeny [126]. Overexpression of BDNF was likely driven by increased acetylation of histone H3 in the PFC. Intriguingly, the association of acetylated histone H3 with Bdnf promoters was also increased in the sperm of sires that self-administered cocaine [126]. These findings indicated that paternal cocaine self-administration reprograms the germline resulting in enhanced BDNF expression in the mPFC, which blunted the reinforcing efficacy of cocaine only in the male progeny. Another potential mode of transmission between sires and progeny is through methylation of DNA in sperm [131]. Some of the differences in DNA methylation elicited by paternal cocaine taking were maintained from the sires to the F1 generation, particularly near the transcription start sites of genes. These stable alterations to the germline epigenome could explain the transgenerational reduction in cocaine self-administration reported in both F1 and F2 male offspring [131].

The mode of drug delivery in sires (non contingent vs. self-administration) may have a profound influence on the behavioral endpoints in the progeny. In fact, a recent study suggests that the motivational state of the sires for earning infusions of cocaine is a critical component for conferring a cocaine-resistance phenotype in the F1 generation and for reprogramming of the germline methylome by cocaine [131]. These kinds of studies begin to indirectly address the elusive question of how exactly cocaine reprograms the germline. There are cocaine binding sites in the testes, but it remains unclear whether reprogramming occurs as a consequence of direct interactions in the testes and/or through a combination of direct and indirect mechanisms that also involve cocaine’s actions on the brain and/or endocrine systems. To date, no study has directly tackled this fascinating question. It is notable that the number of studies examining the impact of paternal cocaine exposure beyond the first generation offspring is extremely sparse [131].

Although the study of potential cross-generational effects of addictive drugs remains an emerging field, several intriguing effects have been delineated. It is particularly interesting that most of the influences of ancestral exposure to drugs of abuse are sex specific. As reviewed above, there is evidence that drugs of abuse can modify epigenetic marks in the germline. Sex-specific influences on progeny could result from changes in mitochondrial genes or genes found on the Y chromosome. Genomic imprinting also is a plausible mechanism that thus far remains unexplored. The impact of sex hormones during development also may be suppressed or enhanced due to complex epigenetic effects leading to changes in behavior. There are well-characterized effects of female sex hormones on drug self-administration behavior [133]. Testosterone also has profound influences during development leading to a number of sexually dimorphic characteristics. For example, the bed nucleus of the stria terminalis, which plays an important role in addiction-related behaviors, is larger in males than females [134]. A major challenge going forward is to characterize drug-induced epigenetic changes in the germline that might plausibly influence physiology and behavior across generations. This is a significant hurdle for the entire field of transgenerational epigenetics, which will likely require the development of new methods to accurately trace the influence of a specific germline modification through embryogenesis and development.

Conclusions

According to PubMed, over 39,500 scientific papers focusing on cocaine have been published to date. Although we have reviewed only a small fraction of these articles here, the question remains: how can we know so much about the physiological effects of cocaine and still have no therapeutics for cocaine addiction? Fortunately, the situation is not completely bleak. There are some promising developments that could lead to therapeutics based on limiting the access of cocaine to the brain via a cocaine vaccine [135,136,137] or cocaine hydrolase [138, 139]. Initial studies also suggest that transcranial magnetic stimulation may produce prolonged reduction of cocaine intake in human addicts [140, 141]. Preclinical [142,143,144,145,146] as well as clinical case studies [147, 148] suggest that deep brain stimulation of the nucleus accumbens may be effective in the treatment of refractory cocaine dependence. We must acknowledge, however, that these advances are based on immunology, pharmacology and systems neuroscience rather than genetics or epigenetics.

The single genome wide analysis of genes associated with cocaine dependence failed to generate promising novel targets for cocaine therapeutic development [55]. This could be due to the fact that the heritability of cocaine addiction is highly polygenic and/or because the study was under-powered. It is now clear that GWASs focusing on psychiatric disorders require subject pools numbering in the tens of thousands. Human studies assessing specific genes in addiction have produced numerous positive effects, but the targets are derived almost exclusively from the animal literature. Given the staggering cost of human GWASs, animal genetic experiments are a more cost-effective and potentially equally valid alternative [72]. Cross-generational animal studies are a potential source of targets for cocaine addiction drug development. Multiple studies have shown that cocaine reinforcement is blunted in the male offspring of cocaine-exposed sires (Vassoler et al. 2013; [131, 132]). A comprehensive survey of the changes in gene expression and epigenetic marks in these cocaine-resistant rats has yet to be assessed and is critically important.

Therapeutics targeting epigenetic mechanisms are becoming more common, particularly in the treatment of various cancers. Valproate is an anticonvulsant that also used to treat bipolar disorder and migraine. The specific mechanism of action responsible for its therapeutic effects is unclear but valproate acts as an HDAC inhibitor, among other effects [149]. As reviewed above, animal studies indicate that HDAC inhibitors attenuate cocaine-induced conditioned place preference [84, 86, 87] as well as cocaine self-administration [85]. Valproate is the subject of ongoing cocaine addiction clinical trials although published work has been mixed in terms of reducing cocaine use [150, 151] or craving [152]. A limitation of drugs such as valproate is a lack of specificity generally including its epigenetic influences. The latest epigenetic editing tools such as ZFPs and TALEs, which target specific genomic locations of epigenetic marks, have effectively modulated cocaine-mediated behaviors in animal studies [95, 99] and are likely to produce fewer side-effects than non-specific compounds such as general HDAC inhibitors. Exploiting these new developments in epigenetics in a clinical setting is clearly a distant prospect, but these kinds of advances hold the promise for new therapeutic developments for cocaine addiction, which are desperately needed.