Epigenetics of Drug Addiction

Abstract

Substance use disorders (SUDs) are chronic brain diseases characterized by transitions from recreational to compulsive drug use and aberrant drug craving that persists for months to years after abstinence is achieved. The transition to compulsive drug use implies that plasticity is occurring, altering the physiology of the brain to precipitate addicted states. Epigenetic phenomena represent a varied orchestra of transcriptional tuning mechanisms that, in response to environmental stimuli, create and maintain gene expression–mediated physiological outcomes. Therefore, epigenetic mechanisms represent a convergent regulatory framework through which the plasticity required to achieve an addicted state can arise and then persist long after drug use has ended. In the first section, we will introduce basic concepts in epigenetics, such as chromatin architecture, histones and their posttranslational modifications, DNA methylation, noncoding RNAs, and transcription factors, along with methods for their investigation. We will then examine the implications of these mechanisms in SUDs, with a particular focus on cocaine-mediated neuroepigenetic plasticity across multiple behavioral models of addiction.

EPIGENETIC MECHANISMS

Despite containing virtually the same DNA content, the various cell types that constitute an individual eukaryotic organism can display markedly distinct morphologies and functions, resulting in the generation of distinct tissue types and organ systems. Such diversity is accomplished, in part, through transcriptional regulation of genes contained within each cell, a process influenced by a vast network of environmental cues (and integration thereof), paracrine signals, and additional factors. These mechanisms are controlled via a highly complex and varied orchestra of histone proteins (which wrap around DNA to form spools known as nucleosomes, the fundamental repeating units of transcription)—along with their variants and posttranslational modifications (PTMs)—chromatin-remodeling enzymes/complexes, direct modifications on DNA/RNA, and the expression of noncoding RNAs, all of which affect chromatin accessibility to potentiate/attenuate the recruitment of specific transcription factors (TFs). This regulatory system allows cells to appropriately fine-tune their physiologies in response to environmental signals. Repeated exposures to environmental insults, such as drugs of abuse, can be considered a powerful disturbance to the internal state of a given organism and, as such, can result in aberrant plasticity. Given the persistence of cellular and behavioral phenotypes observed in addicted states, chromatin regulation provides a possible mechanism through which long-lasting changes in gene expression, physiology, and behavior are established in response to drugs of abuse in the absence of clear genetic predispositions.

Regulation of Chromatin Structure by Epigenetic Mechanisms

In eukaryotic organisms, DNA is located within the nucleus of each cell, packaged tightly into chromatin, consisting of nucleosomal repeats of ∼147 base pairs of DNA wrapped around a core octamer of histone protein (Kornberg 1974; Luger et al. 1997). These octameric histones exist as two copies each of four major subtypes, referred to as H2A, H2B, H3, and H4 (note that another class, H1, exists outside of the nucleosome and its members are considered to be “linker” proteins). In addition, each class of histone protein, with the exception of H4, has multiple isoforms and sequence variants (for a comprehensive discussion of identified histone variants, see Maze et al. 2014). Although histones were initially considered to act simply as structural components for DNA packaging, it is now clear that these proteins play critical roles in transcription through a complex system of coordinated regulation.

One of the major determinants of gene expression is the structure and accessibility of chromatin. Tightly packaged nucleosomal units are often referred to as heterochromatic structures, which are largely inactive, remaining in a transcriptionally repressive state owing to a lack of DNA accessibility for the transcriptional machinery (e.g., the RNA polymerase II transcriptional complex [RNAPII]). Chromatin can also exist in a more “open” state, termed euchromatin, in which DNA remains accessible for TF recruitment and permissive gene expression (Berger 2007). These opposing chromatin states are regulated, in part, by a special class of proteins known as chromatin remodelers, which in brain includes the neuronal Brg1/hBrm-associated factor (nBAF) complex. These remodelers are ATP-dependent and impact the structure of chromatin by physically moving histone proteins along the DNA template to alter chromatin topography (Owen-Hughes and Workman 1996; Yudkovsky et al. 1999; Olave et al. 2002). Such proteins can also impact chromatin structure through regulation of chromatin looping events (i.e., long-range chromatin architectural configurations), histone variant exchange, and nucleosome eviction. Chromosomal looping, a phenomenon that has only recently gained attention in the field, facilitates transcriptional regulation at distal loci from where a given TF may bind. As such, the 3D structure and arrangement of chromatin itself can serve as an additional layer of transcriptional control in multicellular organisms (Rajarajan et al. 2018, 2019). To probe the accessibility of chromatin itself, researchers can employ a technique termed assays for transposase-accessible chromatin coupled to sequencing (ATAC-seq), which relies on mutated transposase enzymes to incorporate artificial DNA adaptor sequences within accessible regions of the genome, thus allowing for quantification of the “openness” of a given region of chromatin (Buenrostro et al. 2015). This approach allows broad profiling of nucleosome spacing, thereby highlighting fundamental differences in chromatin architecture under different physiological conditions. Although incredibly powerful, ATAC-seq does not provide information about the 3D conformation of chromatin. Such assessments can be accomplished through the use of HiC-seq, in which DNA can be cross-linked, sheared, and then religated with a biotinylated nucleotide analog (Lieberman-Aiden et al. 2009). This allows for the selective immunoprecipitation (IP) of ligated fragments of DNA that can then be amplified, sequenced, and mapped onto genomes of interest to identify fragments of DNA that are close in physical space.

One interesting example of a newly emerging approach to investigating epigenetic regulation is the study of how chromatin is segregated into phase-separated structures. These occur within the nucleus of cells as either polymer–polymer phase separation (PPPS) units, which include condensed chromatin regions such as those bound to heterochromatin protein 1 (HP1), or liquid–liquid phase separation (LLPS), which occurs when high concentrations of self-associating chromatin entities create thermodynamically favorable globules as an emulsion in aqueous solution (Gibson et al. 2019; Sanulli et al. 2019). Importantly, PPPS requires a chromatin backbone to form, whereas LLPS droplets can form in the absence of chromatin. Although the physiological importance of these droplets is not yet fully understood, recent evidence suggests that the formation of liquid phase boundaries might serve to reduce rates of aqueous molecules transitioning into droplets, thereby either stabilizing condensed inactive chromatin states or preserving active transcription.

Histones themselves can also be replaced throughout the genome via histone turnover to rapidly allow for activity-dependent gene expression. Disruptions in histone turnover in brain result in numerous dysregulated neuronal processes, such as aberrant synaptic connectivity and deficient activity-dependent gene expression, events that ultimately result in behavioral abnormalities (Maze et al. 2015). This can be probed utilizing stable isotope labeling by amino acids in culture/mice (SILAC/M), whereby nonradioactive, isotopically labeled amino acids can be supplemented exclusively into cell culture medium or an animal's diet. This allows for mass spectroscopy (MS)-based assessments, which represent a highly sensitive method to differentiate chemical species based on mass and charge, and can be performed on biological samples to detect the amount of newly formed proteins (i.e., those containing heavily labeled amino acids) versus proteins that existed before the labeling began (Ong et al. 2002).

Histone PTMs

In addition to aforementioned structural/architectural roles for histones, these proteins also impact transcription via the addition or removal of covalent modifications on their amino- and carboxy-terminal unstructured “tails” (note that a smaller number of modification sites within the globular domains of histones have also been identified) (Cheung et al. 2000; Barski et al. 2007; Berger 2007). These unstructured regions are highly prone to chemical modifications, owing to their accessibility for recruitment of chromatin-modifying enzymes. Different modification types display specificity at the level of amino acid sequence and can be added by enzymes that are referred to as “writers.” Most, if not all, of these modifications can similarly be removed to allow for alterations in the transcriptional state (either via in cis or in trans mechanisms) and are dictated by enzymes termed “erasers.” Additionally, many of these modifications are recognized by various “reader” proteins, which have specialized domain structures that allow for site-specific modification binding, a process that aids in the recruitment of chromatin remodelers and TFs to confer transcriptional outputs. Generally speaking, PTMs such as lysine (K) acetylation, which function in cis to impact electrostatic interactions between histones and DNA, are associated with more “open” chromatin states, leading to greater access to the transcriptional machinery. Similarly, histone phosphorylation events can generally be thought of as “on/off” switches for the recruitment of histone regulators, in some circumstances recruiting histone acetyltransferases (HATs) to increase acetylation, and also acting to kick off adjacent lysine methyl (see below) “readers” (mostly associated with active transcription). Lysine and arginine methylation (me) states, however, are more complex and can be associated with both permissive and repressive transcription depending on the site and valence state (i.e., me1, me2, or me3) of the modification being deposited. For example, H3K4 and H3K36 trimethylation are associated with permissive transcription, whereas H3K9 and H3K27 mono-, di-, and trimethylation are more heavily associated with repressive gene expression owing to their recruitment of corepressive proteins. To probe these interactions in vivo, numerous methods have been developed to examine, in an unbiased fashion, chromatin regulatory events in conjunction with their relationships to gene expression. For example, chromatin IP coupled to next-generation sequencing (chromatin IP sequencing [ChIP-seq]) represents a powerful tool that utilizes antibodies against specific histone modifications, chromatin regulators, TFs, etc., to selectively capture these factors in a chromatin-bound state, thereby allowing for investigations of their enrichment throughout the genome (Barski et al. 2007). These data can then be coupled to gene expression profiling (e.g., RNA-seq), allowing researchers to overlay their data and correlate chromatin factor enrichment with gene expression profiles in their systems of interest.

Adding an additional layer of complexity, most histone marks can exist in combinatorial configurations (i.e., the existence of multiple PTMs on the same tail or across histone tails within the same nucleosome) or even in “bivalent” states, where both active and repressive marks exist on the same histone tail at a given locus (Bernstein et al. 2006). This can be probed via MS by examining specific shifts in mass, corresponding to various chemical modifications to allow for unbiased examinations of all modifications occurring on specific histones.

Given the intricacy of histone regulation, little can be generalized about the specific effects of any given mark in isolation, as they most often exist within the context of numerous additional modifications and/or interacting proteins. As such, it has been proposed that these combinatorial readouts may constitute a so-called “histone code,” resulting in a complex network of transcriptional control. Indeed, as the putative “histone code” continues to be deciphered, deeper investigations into roles for combinatorial histone PTMs and their associated interactions with “readers” will require attention in the context of studies associated with psychiatric disorders, including drug addiction.

DNA Methylation

In addition to regulation through histone-associated processes, gene expression is also dictated by chemical modifications on DNA itself. Methylation of DNA is a process through which methyl groups can be covalently added to the 5′ position on cytosine nucleotides, resulting in the creation of 5-methylcytosine (5mC). In mammals, methylation of 5′-CpG-3′ sequences is the most heavily characterized modification and is known to be required for normal development, genetic imprinting, and X-chromosome inactivation (Tribioli et al. 1992; Li et al. 1993). CpG sequences also occur in clusters throughout the genome, and these regions are termed CpG islands. CpG islands tend to occur in promoters of genes and are typically less methylated versus CpG sequences occurring outside of these islands (Jones 2012; Zeng et al. 2014). Non-CpG methylation has also been identified and is particularly enriched in adult brain tissues and stem cells (Barrès et al. 2009). Like CpG methylation, non-CpG methylation appears to regulate gene expression. DNA methylation (DNAm) is catalyzed by a class of enzymes referred to as DNA methyltransferases (DNMTs). Demethylation remains less well-understood but involves DNA damage repair enzymes and multiple intermediates, including 5-hydroxymethylcytosine (5-hmC) (Tahiliani et al. 2009). Changes in DNAm states have been observed across a large number of experimental disease models, indicating a likely role for methylation and demethylation intermediates in disease-related transcriptional plasticity.

Functionally, DNAm has been observed to generally repress transcription by recruiting corepressor complexes or by sterically hindering recruitment of the transcriptional machinery, thereby promoting persistent patterns of repression at marked loci. However, the importance of site specificity extends to DNAm as well, as one study found that DNAm was only inversely correlated with expression when enriched in promoter regions, a correlation that is lost when examined in gene bodies (Baribault et al. 2018). Repressor complexes that recognize DNAm do so through DNA methyl-binding domain (MBD)-containing proteins, which are also required for normal growth and development. For example, mutations in these MBD proteins often result in neurodevelopmental disorders, such as Rett's syndrome (Amir et al. 1999).

DNAm can be profiled using methods such as bisulfite sequencing (bisulfite-seq), in which pretreatment of DNA with the chemical bisulfite converts all unmethylated cytosine residues into uracils, allowing for the detection of cytosines, which are preferentially methylated versus unmethylated. However, this method is unable to distinguish between 5-mC and its oxidized intermediates, such as 5-hmC, limiting interpretation of results, especially in tissue types (such as brain) in which these intermediates may be enriched. Alternative methods have been developed that allow for detection of 5-hmC versus 5-mC, such as oxidative bisulfite sequencing (oxBS-seq) and Tet-assisted bisulfite seq (TAB-seq), both of which can be employed alongside of bisulfite-seq to examine both methylation and hydroxymethylation states.

Noncoding RNAs

Noncoding RNAs (ncRNAs) have also recently been highlighted for their importance in numerous biological processes. ncRNAs can largely be categorized into two groups: long ncRNAs (lncRNAs) and short ncRNAs (sncRNAs). lncRNAs represent the most abundant form of ncRNAs and can bind to chromatin-remodeling proteins, helping to recruit them to specific genetic loci to modify chromatin states and gene expression. However, they have also been shown to be important for a variety of posttranscriptional processes. A widely studied example of lncRNA function relates to the process of X-chromosome inactivation. The X-inactivation-specific transcript Xist is transcribed throughout cellular differentiation and then blankets the soon-to-be inactive X chromosome, thereby mediating recruitment of chromatin remodelers and HMTs to inactivate the chromosome (Clemson et al. 1996; Cerase et al. 2015).

In comparison, sncRNAs have received considerably more attention in the field of addiction. sncRNAs can be similarly categorized into multiple classes, including micro RNAs (miRNAs) and short interfering RNAs (siRNAs). miRNAs and siRNAs both function by binding to messenger RNA (mRNA) transcripts to block translation and promote degradation. miRNAs are transcribed by RNAPII to create hairpin-structured pre-miRNAs, which are then exported to the cytoplasm to undergo additional processing to become mature miRNAs, which then bind to RNA-induced silencing complexes (RISCs) (Bartel 2004). It is important to note, however, that many mRNAs exist in such silenced states without undergoing degradation, an important difference that exists between miRNAs and siRNAs. siRNAs are generated as double-stranded RNA molecules that undergo processing and then bind to similar RISC complexes. However, siRNAs only bind to one specific cognate mRNA and cause endonucleolytic degradation of associated transcripts. Additionally, siRNAs have been shown to induce heterochromatin formation via an RNA-induced transcriptional silencing (RITS) complex, which has been observed to increase repressive H3K9 methylation and condenses chromatin (Hamilton et al. 2002).

Transcription Factors

The canonical mechanism for regulation of gene expression is through the recruitment of TFs—both general and sequence-specific—to DNA. TFs represent a class of proteins that, in response to cellular signaling, bind to specific DNA sequences (typically in promoter and/or enhancer loci) to mediate the expression or repression of cognate genes by aiding in, or attenuating, the recruitment of the RNAPII complex (Latchman 1997). Maladaptive changes in the expression and/or regulation of these factors represent a critical mechanism controlling the establishment and/or maintenance of persistent transcriptional programs and, in turn, cellular functions. Although TFs are indeed critical mediators of gene expression, they require that the DNA itself be accessible for binding and subsequent regulation. If the DNA remains inaccessible to these factors, they cannot be recruited, indicating that a fine balance must exist between TFs and chromatin structural organization to appropriately dictate gene expression outcomes. It should be noted, however, that a class of TFs does exist that bind to relatively inaccessible regions of the genome and are capable of recruiting specific chromatin-remodeling complexes to mediate transcription in those regions. These proteins are referred to as pioneer factors (Zaret and Carroll 2011). To characterize TFs, similar methodologies to those employed for studying histone marks are typically used. For example, ChIP-seq can be employed to investigate the binding of specific TFs, which when coupled with RNA-seq allows one to correlate increased binding events with either increased or decreased gene expression. Similarly, ATAC-seq can be employed in conjunction with ChIP-seq to compare alterations in chromatin accessibility as a result of TF binding.

EPIGENETICS AND SUBSTANCE USE DISORDER

As discussed in the previous section, neuroepigenetic mechanisms regulate coordinated transcriptional responses to transduce environmental stimuli into functional changes in cellular output. The standard usage of epigenetics is the collection of mechanisms that leads to stably heritable functional changes in gene expression caused by modulation of gene expression, rather than alterations to the DNA sequence (Berger et al. 2009). This definition includes those effects that may transmit between generations or across multiple generations of offspring. For example, in the case of cocaine, several studies have confirmed that paternal cocaine exposure influences reward-based behavior in offspring through epigenetic processes, including increased susceptibility to drug-related motivated behaviors, memory/cognition, and affective measures such as mood or anxiety (Vassoler et al. 2013; White et al. 2016; Wimmer et al. 2017). Although these cross-generational effects of drug use represent an important avenue of study for understanding how susceptibility to addiction is determined, this review will focus on the mechanics of epigenetic changes on a shorter timescale, from initial drug exposure to the persistent effects that drive relapse behaviors after long periods of abstinence. The following section examines evidence from the literature that demonstrates how these processes play critical roles in establishing and maintaining the homeostatic mechanisms that regulate neuronal adaptations in response to drugs of abuse.

Substance use disorder (SUD) is a chronic, relapsing neuropsychiatric disease characterized by the abuse of psychoactive substances despite negative consequences, including decline in mental, physical, or social health. Even within the subset of affected individuals that are able to achieve abstinence, ∼40%–60% of those recovering from SUDs will relapse within 27 months (McLellan et al. 2000; Kassani et al. 2015). These statistics highlight the persistent nature of SUD, which can last the length of a lifetime. Importantly, not every individual that uses addictive substances will develop a dependence. From early twin studies and family linkage reports, the heritable component of SUD vulnerability is estimated at ∼50% (Kendler et al. 2003). The remainder of the risk for SUD is determined by a complex interaction of genetic makeup and environmental factors. The transition from recreational drug use to chronic abuse of psychoactive substances is driven by neuroplastic adaptations in the neurocircuitry of the brain's reward-processing system.

Clinically, SUD can be understood as an iterative cycle of behavior that progresses from intoxication to withdrawal, and later to preoccupation and compulsive activity. These recurring stages are driven by neuroplastic changes in specific brain circuits that underlie different aspects of reward-based learning and motivation. The cycle initiates at the taking of psychoactive drugs and is driven by the acutely reinforcing effects of these substances at their primary sites of action, causing changes in dopamine transmission in the basal ganglia, including the ventral tegmental area (VTA), the main dopaminergic nucleus of the brain, and nucleus accumbens (NAc), which receives dopaminergic input from the VTA and integrates these signals with information coming from the cortex to mediate goal-directed behaviors. Over time, maladaptations in these key brain regions can promote escalation of drug use to compulsive abuse and dependence, recruiting other cortical and limbic brain structures to precipitate a cycle of negative affect during withdrawal, continued preoccupation, and drug-seeking behaviors that can result in relapse. The ultimate aim of addiction research is thus to identify and characterize the molecular drivers of these functional alterations in reward circuitries to better leverage these phenomena for targeted SUD therapeutics in clinical settings.

Animal Models of SUD

Although it is impossible to fully recapitulate the complexity of human SUD in an animal model, preclinical studies are indispensable to parsing the molecular substrates of addiction phenotypes. Because the following section will closely examine examples of investigators using animal models of SUD to examine specific aspects of epigenetic regulatory processes in addiction, it is important to first describe the various behavioral paradigms that researchers use to model distinct aspects of addictive phenotypes.

Much of the early work in this field centered on experimenter-administered models of substance use, such as locomotor sensitization and/or conditioned place preference (CPP). Locomotor sensitization represents a phenomenon through which repeated exposures to drugs of abuse result in increased locomotion of an animal upon subsequent drug exposures. CPP, on the other hand, is a reward-related paradigm, involving pairing one chamber of a two-chambered apparatus with a rewarding stimulus (e.g., an injection of cocaine or morphine) and the other with a nonrewarding stimulus (e.g., an injection of saline). Animals are then allowed access to both sides of the chamber in the absence of either stimulus, and their time spent in the paired versus unpaired chambers is recorded as a measure of how rewarding a given substance may be. It is important to keep in mind that experimenter-administered models are indirect measures of drug-associated motivation and may not fully recapitulate etiologically relevant phenomena, despite their usefulness in characterizing the important epigenetic mechanisms in SUD.

More recently, the addiction neuroepigenetics field has begun turning its collective attention to operant models of drug abuse, such as self-administration (SA), in which animals learn to perform an action (e.g., a nose poke or lever press) to receive a rewarding/reinforcing stimulus (e.g., an intravenous infusion of cocaine). Because these paradigms involve an aspect of internal motivation, they allow for better modeling of the behavioral endophenotypes associated with SUD, such as acquisition of drug taking, escalation of drug intake following chronic exposures (a measure related to compulsivity), and abstinence-mediated incubation of drug “craving” (which can be used to model certain aspects of relapse vulnerability). Similarly, other internal states can be more concretely measured, such as motivation for drug, utilizing a progressive ratio paradigm in which an animal must perform an increasing number of learned actions to receive the same reinforcing stimulus. Sensitivity to drugs or their rewarding properties can also be probed utilizing dose–response paradigms, in which animals respond for differing doses of drug over the course of multiple sessions.

Still, even within the realm of operant models of drug abuse, variations in access to the drug can produce distinct endophenotypes with differing levels of behavioral and molecular plasticity (Kawa et al. 2019). For example, restricting drug access to only 1 hour each day results in animals that will not escalate in their drug-taking behavior over the course of a 10-day experimental timeline. On the other hand, animals that are given extended access (6 hours) to drugs will reliably increase their intake over multiple sessions (Ahmed and Koob 1998). In addition, researchers can model drug binge states, in which animals are given access to drugs following abstinence for much longer periods (usually between 12 and 24 hours of access), allowing for evaluation of behaviors that may more closely approximate human substance abuse.

Regardless of classification, the use of operant models of drug administration further allows for modeling of clinically relevant interventional states. For example, researchers can examine animal models of relapse through assessments of cue-induced seeking, in which animals are presented with drug-associated cues (e.g., a light or sound) but do not receive the drug. The number of times an animal performs the operant behavior can then be used as a measure of drug-seeking motivation. Another example of a more complex drug-associated behavior that can be examined using operant paradigms is extinction, in which animals are allowed to perform learned actions without the presence of reward-associated cues or the reward itself. Animal will eventually “extinguish” their behavior and seek the reward at a much lower level. Reinstatement of drug-seeking behaviors after extinction are another useful measure of drug-induced behavioral plasticity, in which an animal is exposed postextinction to a drug-associated cue or drug priming before being allowed to perform learned actions without receiving associated rewards. These paradigms offer researchers a standardized index of behaviors against which to measure the effects of manipulations and potential interventional strategies at clinically relevant stages of addiction.

Histones and Their PTMs in SUD

As described in the previous section, histone PTMs represent a flexible and dynamic mechanism through which gene expression programs can be regulated by incoming environmental signals. During drug exposures, changes in dopaminergic transmission to the NAc can initiate a complex cascade of regulatory enzymes to establish highly complex combinations of histone modifications. These patterns are then “read” by specialized effector proteins that recognize specific modification states and exert influence on chromatin structure or transcriptional machinery. The dynamic nature of these marks makes them well-positioned to act as a major driver of the transcriptional changes observed in reward circuitry during SUD.

The case of histone acetylation is one of the more well-characterized neuroepigenetic regulatory phenomena in addiction and provides an illustrative example of the complexity involved in these processes. Functionally, histone acetylation is generally associated with increased accessibility of chromatin structure and the opening of surrounding DNA to facilitate binding of cis-regulatory factors or transcriptional machinery. In both early candidate gene studies and genome-wide profiling, it was shown that exposures to cocaine (and other drugs of abuse) increase the acetylation state of histones H3 and H4 (Brami-Cherrier et al. 2005; Kumar et al. 2005; Renthal et al. 2009a) in mouse NAc. Importantly, these studies found that H3 and H4 acetylations at the promoters of key plasticity-associated genes (e.g., Bdnf, Cdk5, Fos, Creb) increase at different time points following cocaine exposures, from initial acute exposures to a week after abstinence from chronic injections in a mouse CPP paradigm. These PTM-specific effects may play important roles in maintaining a state that is permissive to cocaine-induced motivated behaviors in humans and animals.

Changing global levels of histone acetylation across the genome is the result of a constantly shifting balance between a large family of “writer” and “eraser” proteins that control the deposition of these marks at specific histone residues. Through a balance between HAT and HDAC activities, levels of different histone acetylation marks are controlled in a highly dynamic manner that represents a regulatory framework for the integration of environmental signals to gene expression. In this context HATs and HDACs represent intriguing candidates for pharmacological strategies that target the interaction between environmental stimuli and gene expression.

A well-characterized example of a dysregulated epigenetic “eraser” in response to cocaine exposures is the histone deacetylase 5 (HDAC5), which is an enzyme responsible for removing acetylation marks (Chawla et al. 2003). Although cocaine exposures do not affect global expression levels of HDAC5, early studies found that chronic cocaine injections regulate its subcellular localization by inducing an increase in HDAC5 phosphorylation, which in turn leads to export of HDAC5 from the nucleus and an overall decrease in HDAC5-mediated deacetylation. Using viral-based approaches to artificially manipulate HDAC5 expression in the mouse NAc, these studies found that overexpression of HDAC5 reduces cocaine preference in a CPP paradigm compared to wild-type (WT) mice, whereas knockdown enhances CPP (Renthal et al. 2007). Importantly, HDAC5 manipulations have no effect on sucrose place preference, suggesting that this mechanism does not influence natural reward processes (Taniguchi et al. 2012). These findings indicate that cocaine exposures induce nucleocytoplasmic shuttling of HDAC5, resulting in decreased HDAC5-mediated removal of histone acetylation and facilitation of permissive, drug-induced transcription states.

However, recent investigations of HDAC5 using an SA model of cocaine use revealed additional layers of complexity. To identify which aspects of cocaine-associated behaviors are specifically mediated by HDAC5 nuclear localization, Taniguchi et al. compared the effects of overexpressing a WT version of HDAC5 to overexpression of a nonphosphorylatable mutant that is artificially sequestered to the nucleus, and thus maintains deacytylase activity during cocaine exposures (Taniguchi et al. 2017). Interestingly, little to no differences were observed in measures related to cocaine taking, progressive ratio, dose responsiveness, or extinction rates. However, in the nuclear-enriched HDAC5 mutant animals, significantly reduced levels of cue-induced reinstatement were observed. This finding suggests that the importance of HDAC5 activity may be specific to drug-seeking behaviors during abstinence, which in the clinical setting is a highly relevant phase of drug dependence for targeting by new therapeutics. The example of HDAC5 also highlights the necessity of using different behavioral paradigms to investigate distinct aspects of addictive-like behaviors when manipulating epigenetic modifiers.

Overall, the current literature suggests that the interplay between histone acetylation “writers” and “erasers” is critical for releasing the regulatory brakes on increased transcriptional plasticity during cocaine use. However, it is important to keep in mind that increases in nuclear HDAC activity do not always correlate with enhanced drug-seeking behaviors. An illustrative example of this point involves regulation of the sirtuin family of proteins—specifically the NAD-dependent deacetylases sirtuin (SIRT)1 and SIRT2—in response to cocaine. Given that global increases in acetylation are observed in response to cocaine exposures, the up-regulation of these HDACs following chronic drug administration was hypothesized to act as a compensatory, neuroprotective mechanism. However, pharmacological inhibition of SIRT1 and SIRT2 unexpectedly results in decreased preference for cocaine in the CPP paradigm. Conversely, induction of SIRT1 and SIRT2 expression is associated with increased H3 acetylation and drug place preference (Renthal et al. 2009b). This effect has also been observed following morphine exposures, indicating a possible convergent mechanism for SIRT1/2 in the precipitation of addictive-like behaviors (Ferguson et al. 2013). The case of SIRT proteins following cocaine exposures demonstrates that, depending on the gene locus in question, decreases in histone acetylation may repress the expression of a gene product that would otherwise work to control cocaine-related transcriptional programs (Renthal and Nestler 2009). Therefore, it will become increasingly important to utilize locus-specific approaches when exploring patterns of histone modifications to make inferences about the effects of drugs on gene expression.

Unlike histone acetylation, the addition of methyl groups to histone tails can have vastly different effects on chromatin regulation depending on which residue of the histone has been modified. Similar to histone acetylation, differential histone methylation states have also been implicated in the precipitation of addictive-like states. Although there are many examples of activating histone methylation marks increasing during cocaine exposures, this section will focus on repressive histone methylation to illustrate how psychostimulants can release regulatory brakes on transcription, thereby increasing neuroplasticity and driving cocaine-related behaviors in SUD. The HMT enzyme G9a (Ehmt2), which selectively catalyzes repressive mono- and dimethylation of H3K9, was shown to be decreased in its expression following chronic, but not acute, cocaine administration, leading to global reductions in the repressive H3K9me2 mark (Maze et al. 2010). It was further shown that decreases in H3K9me2 correlate with increased gene expression for transcripts known to be associated with synaptic and dendritic plasticity (e.g., ΔFosB), suggesting a mechanistic link between G9a down-regulation, loss of H3K9me2, and enhanced neuroplasticity in this region. As confirmation of this link, the authors used viral manipulations to demonstrate that overexpression of G9a in the mouse NAc results in decreased cocaine CPP (Maze et al. 2010). In a follow-up paper, Sun et al. (2012) utilized an NAc-specific conditional knockout approach to show that loss of G9a in this region significantly increases CPP in response to morphine, suggesting that regulation of this enzyme may be a generalizable mechanism of transcriptional dysfunction across different classes of abused substances.

Paradoxically, more recent investigations using operant cocaine SA paradigms in rat models have found that overexpression of G9a in the NAc not only increases cocaine taking and motivated behaviors in a progressive ratio test, but also enhances reinstatement of drug seeking after periods of cocaine abstinence (Anderson et al. 2018). Knockdown of G9a in NAc before the cocaine SA paradigm, on the other hand, results in decreased cocaine taking, decreased motivation, and reinstatement of drug seeking (Anderson et al. 2018). These findings again highlight the importance of behavioral model choice when studying SUDs, as they sometimes result in seemingly opposing behavioral outcomes. There are numerous differences between cocaine CPP and SA that may account for these discrepancies, including (1) acute cocaine exposure in CPP versus chronic exposure in SA, (2) intraperitoneal injections in CPP versus intravenous infusion in SA, (3) experimenter administration of cocaine in CPP versus volitional taking in SA, or (4) classical versus operant conditioning (Anderson et al. 2018). Any or all of these factors may affect the distinct transcriptional patterns initiated in either paradigm. The specific genes that lose or gain histone modifications may then be different in either case depending on the parameters of the drug experience.

Recently developed neuroepigenetic editing approaches can also facilitate our understanding of why these differences between paradigms occur, by allowing researchers to manipulate a modification at the specific gene loci. To investigate roles for G9a at specific loci in the context of repeated cocaine exposures, Heller et al. (2014) used engineered zinc finger proteins (ZFPs) fused to the catalytic domain of G9a to selectively direct repressive HMTase activity toward a specific gene locus of interest. In this case, the authors directed G9a to the ΔFosB promoter in the mouse NAc, a TF that has previously been found to mediate cocaine-related gene expression and to reduce levels of G9a itself. They found that targeting repressive G9a activity to ΔFosB results in increased enrichment of H3K9me2 at this locus and attenuates ΔFosB expression (thus opposing the endogenous actions of cocaine). Interestingly, expression of the G9a-ZFP is sufficient to attenuate drug-induced behavioral plasticity by reducing cocaine locomotor sensitization. Despite differential effects observed between CPP and SA, reports using both paradigms have overall contributed to a general mechanistic framework, which suggests that inappropriate chromatin dynamics—particularly those that potentiate transcriptionally permissive states—may prime the brain's reward-processing systems toward maladaptive addiction-like behavioral states.

Beyond acetylation and methylation, there are many other examples of posttranslational covalent modifications to histone proteins, including phosphorylation, SUMOylation, and glycosylation, all of which contribute to the complex regulatory network that controls transcription. For example, two modifications have recently been characterized, termed propionylation and butyrylation, that result from the addition of fatty acid metabolites—propionyl and butyryl molecules, respectively—to histone proteins (Chen et al. 2007). These modifications have been found to specifically mark highly expressed genes and to change in their expression in response to metabolic alterations such as fasting, suggesting that these novel PTMs may work to couple gene expression alterations to changes in the energetic state of cells (Kebede et al. 2017). These findings demonstrate the potential for novel histone modifications to act as a direct mechanistic link between external stimuli, cellular signaling cascades, and transcriptional responses.

Recently, it was discovered that biogenic monoamines (e.g., serotonin, dopamine) can also be chemically added to histone tails, indicating potential novel roles for this class of neuromodulator in the direct mediation of transcriptional plasticity associated with cocaine exposures. Although dopamine signaling has previously been implicated in chromatin remodeling (Schroeder et al. 2008), a new report identified histone H3 dopaminylation (glutamine 5 dopaminylation/H3Q5dop) as a novel epigenetic modification in brain. Following extended access to cocaine in rats, the authors detected significant increases in H3Q5dop expression in the VTA following 30 days of forced abstinence from the drug, but not earlier after final cocaine exposures, suggesting that this modification may increase during cocaine withdrawal to promote drug-seeking behaviors and relapse vulnerability. Critically, preventing accumulation of H3Q5dop in the VTA during withdrawal reverses both cocaine-mediated gene expression programs and attenuates drug-seeking behaviors. These data suggest that this modification may represent an important epigenetic mechanism directly linking alterations in dopamine signaling during/after drug exposures to potentiate aberrant drug-induced gene expression programs and maladaptive behaviors (Lepack et al. 2020).

Chromatin Structure in SUD

As chromatin structure and locus accessibility ultimately determine gene expression patterns, recent investigations have further emphasized roles for chromatin-remodeling complexes in addiction models. For example, multiple chromatin remodelers have been found to be increased in their expression following cocaine SA and prolonged periods of drug abstinence, including Brg1 and the INO80 complex ATPase subunit (INO80), a member of the SWI/SNF family of chromatin-remodeling ATPases. Viral manipulation studies for both of these enzymes have found that their overexpression potentiates cue seeking for cocaine following abstinence, implicating chromatin remodeling in the NAc as an important process underlying relapse-related behaviors (Wang et al. 2016). In the case of INO80, ChIP-seq analyses of INO80 additionally revealed regulation of multiple downstream transcriptional and synaptic targets, providing evidence that increased expression of this chromatin remodeler likely mediates large-scale changes in cellular physiology (Werner et al. 2019). Interestingly, however, a more recent study revealed that different regions within the brain's reward circuitry may display unique patterns of regulation for this epigenetic machinery. For example, in prefrontal cortex (PFC), Brg1 appears to serve neuroprotective roles following heroin SA, in which overexpression of Brg1 results in decreased motivation for heroin in a progressive ratio paradigm (Martin et al. 2018).

Moving forward, it will be critical to further characterize roles for these chromatin-remodeling complexes/events using genome-wide approaches, such as ATAC-seq, to allow for the identification of differentially accessible chromatin regions after drug exposures. For example, in human postmortem tissues of heroin users, ATAC-seq was recently used to identify open chromatin regions specific to drug dependence. Interestingly, these open regions were found to correlate with deposition patterns of the activating mark H3K27ac (typically found in enhancers), suggesting that this modification may contribute to a more accessible chromatin conformation in drug dependents (Egervari et al. 2017).

In addition to profiling the two-dimensional landscape of chromatin with ATAC-seq, newly developed chromosome conformation capture-based techniques (e.g., 3C, Hi-C) can also yield insights into the hierarchical three-dimensional structure of chromatin, and how different stimuli may regulate the spatial organization of genes in relation to their regulatory sequences. A recent report employed 3C approaches in the mouse NAc to show that cocaine exposures disrupt the interaction between two gene loci (Auts2 and Caln1) that are found more than 1500 kb apart (Engmann et al. 2017). These two loci are normally found to interact through chromosomal looping events; however, chronic cocaine injections result in the destabilization of the DNA loop that facilitates this association, as well as alterations in other epigenetic signatures at these loci, including H3K4me3 and DNA methylation. Furthermore, these alterations are neuronal cell-type-specific, affecting D2-like medium spiny neurons (MSNs) in the NAc, but not D1-like MSNs, suggesting a mechanism through which cocaine may target functional outputs of distinct cell types in the brain's reward systems. In future studies, integration of ATAC-seq and Hi-C with other epigenetic and transcriptomic profiling methods promises to further illuminate targets of chromatin-remodeling complexes during different stages of drug dependence in both preclinical models of addiction and human SUD.

DNA Methylation in SUD

As mentioned previously, modifications to histone proteins are not the only chemical alterations that can influence chromatin conformation. DNA cytosine residues can be directly modified by methyl groups to serve as binding platforms and recognition motifs for epigenetic “reader” proteins to affect downstream gene transcription. Earlier studies of DNAm suggested that this modification is relatively stable and mediates long-lasting gene silencing, particularly during development and tissue specification (Li et al. 1993). However, more recent work has revealed that DNAm is itself reversible (Miller and Sweatt 2007) through the actions of DNA demethylases, such as the TET family of proteins, which can convert 5mC to 5hmC (Tahiliani et al. 2009). These findings have highlighted the possibility of DNAm potentially acting as a dynamic epigenetic regulatory mechanism in normal brain function, one that when perturbed may also contribute to neuropsychiatric disorders. Subsequent work has shown DNAm plays an active role in transcriptional plasticity associated with addictive disorders (Feng et al. 2015), illustrated through reports of alterations in global levels of DNAm and enrichment at specific genomic loci (Vaillancourt et al. 2017).

In general, DNA methylation works to silence transcription in one of two ways: (1) by blocking the binding of TFs to gene promoters, and/or (2) by recruiting methyl-binding proteins to modify surrounding chromatin toward a repressed state. In the case of cocaine, one study examining DNAm following SA identified differential patterns of DNAm using a candidate gene approach by measuring methylation at more than 40 genes of interest in a microarray-based design. In this study, the authors detected a correlation between increased DNAm at specific loci (e.g., CREB) and reduced expression of those genes in response to drug exposures (Massart et al. 2015). Furthermore, patterns of DNAm enrichment are specific to distinct phases of cocaine SA, with nonoverlapping sets of genes displaying differential methylation immediately after 10 days of cocaine access versus 21 days of abstinence (Baker-Andresen et al. 2015). These data suggest that DNAm profiles may reflect the specific regulatory processes underlying transcriptional alterations at different time points following drug exposures.

The enzymes that catalyze DNAm have also been shown to play a role in mediating plasticity changes following drug exposures. Early work by LaPlant et al. (2010) showed that repeated experimenter-administered cocaine injections acutely decrease levels of the methyltransferase DNMT3a (a de novo DNA methyltransferase), followed by sustained increases in this enzyme after cocaine abstinence. These findings raised the possibility that an increase in DNMT3a may be an epigenetic mechanism that is specifically important to the abstinence stage of addiction. Interestingly, when DNMT3a is knocked down prior to abstinence, the authors observed an increase in cocaine preference during CPP, suggesting that increases in DNMT3a during abstinence may be a neuroprotective mechanism that works to counteract addiction-related plasticity. However, more recently, roles for DNMT3a during drug abstinence and relapse were investigated utilizing an operant cocaine SA paradigm. In this study, the authors did not identify immediate changes in DNMT3a expression following cocaine SA, but when animals were put through forced abstinence, significant increases in DNMT3a expression were observed beginning 2 days after the last cocaine infusion. These findings raise the possibility that increased DNMT3a activity during withdrawal may serve as an epigenetic mechanism important for potentiating relapse vulnerability. Through viral manipulations studies, knockdown of DNMT3a in the NAc was found to significantly reduce rates of cue-induced reinstatement (Cannella et al. 2018), suggesting that DNMT3a overexpression may indeed be a key driver of drug-seeking behaviors.

Alterations in epigenetic “readers” that recognize and bind DNAm have also been implicated in animal models of drug abuse and are typically conceptualized as enzymes that work to silence genes near their cognate DNAm binding sites (Chen et al. 2003). Specifically, the MBD-containing protein MeCP2 has been shown to be increased in the dorsal striatum (including the caudate nucleus and putamen) following extended access to cocaine SA (Im et al. 2010). Furthermore, viral knockdown of MeCP2 decreases cocaine taking through a mechanism mediated by the ncRNA miR-212 (Im et al. 2010). Another example of miRNA-associated regulation of MeCP2 comes from heroin SA studies, in which Yan et al. (2017) observed that miR-218, which targets MeCP2 directly, is down-regulated following heroin SA. Lentiviral-mediated rescue of miR-218 attenuates MeCP2 levels and inhibits heroin-induced reinforcement in CPP and SA paradigms. These examples of miR targeting of MECP2 to influence transcriptional repression further highlight the complex interplay between epigenetic regulatory systems—that is, these processes do not work in isolation, but rather function in concert to exert coordinated physiological changes observed in addictive disorders.

ncRNAs in SUD

Roles for lcNRAs and miRNAs have not yet been extensively studied in the addiction field; however, there are several salient examples of recent work examining ncRNAs in operant paradigms of drug use. The most notable example is miR-212, which has been shown to increase in its expression during cocaine SA, having a seemingly neuroprotective effect by increasing pCREB signaling and acting as a compensatory mechanism against transcriptional plasticity driving cocaine taking (Hollander et al. 2010). In manipulation experiments, overexpression of miR-212 was found to decrease cocaine SA, whereas blockade of miR-212 potentiates cocaine SA behaviors (Hollander et al. 2010). The ncRNA miR-495 provides an additional example of miRNA regulation of relapse-related behaviors. miR-495 was initially found to be rapidly down-regulated by acute injections of cocaine (Bastle et al. 2018). Using an operant model of cocaine SA, it was later shown that miR-495 directly targets the 3′ UTR of multiple genes known to be increased in SUD, such as Bdnf, Camk2a, and Arc. Importantly, viral rescue of miR-495 was found to decrease motivation for cocaine under progressive ratio conditions (Bastle et al. 2018).

In studies of morphine dependence, mostly using experimenter-administered models, dynamic regulation of miR expression has also been observed. An interesting example of this is the role of miR-219-5p in regulating morphine analgesic tolerance, in which this miRNA was found to attenuate expression of γCamKII, which in turn alleviates morphine tolerance via up-regulation of the NMDA subunit NR1. Indeed, many miRs have been shown to interact with a variety of proteins, such as CXCR4 and β-arrestin, to impact morphine behaviors, indicating important roles for these molecules in alleviating opiate tolerance, an important clinical target when considering the development/use of opiate-based analgesics (Wang et al. 2017).

Transcription Factors in SUD

Although many TFs have been implicated in the pathophysiology of SUDs, this section will focus on the particularly well-characterized examples of ΔFosB and cAMP-responsive element-binding protein (CREB) to illustrate the broad impact of TF regulation in addictive behaviors. Both of these TFs represent convergent end points for multiple complex intracellular signaling cascades, while also acting as primary regulators of activity-dependent transcriptional programs. Because these TFs are positioned at the interface between external stimuli and the amplification of gene expression responses, they have been the focus of many studies aimed at characterizing how drug exposures are transduced into functional outputs in the brain's reward-processing systems.

The TF ΔFosB, for example, is generated from a transcript encoded by the FosB gene; however, this variant is subjected to an alternative splicing event, which causes ΔFosB to lack two key regulatory protein domains. This deletion ultimately leads to a fourfold increase in protein stability when compared to other Fos family protein members (Kelz et al. 1999). The stabilized version of this protein is further protected from degradation via a phosphorylation modification, thus allowing it to regulate particularly sustained changes in gene expression during substance use (Kelz et al. 1999). In fact, multiple reports have found that the amount of ΔFosB accumulates significantly with each subsequent exposure in a chronic drug exposure paradigm (Kelz et al. 1999). The timing of ΔFosB accumulation following drug exposures, however, indicates that it may primarily act to set into motion a transcriptional program that will, in turn, prime the reward system toward addiction-like behaviors. Larson et al. (2010) found that after 3 weeks of abstinence from cocaine SA, ΔFosB protein levels return to normal, indicating that its priming effects may indirectly function to maintain relapse-vulnerable states. In manipulation studies, overexpression of ΔFosB results in increased locomotor sensitivity to cocaine, as well as increased CPP to both cocaine and morphine (Kelz et al. 1999; Zachariou et al. 2006). However, these changes only partially translate to SA models, in which viral overexpression of ΔFosB in the orbital frontal cortex was found to increase impulsivity during abstinence, with overall drug taking remaining unaffected (Winstanley et al. 2009). Additionally, another study in which ΔFosB was overexpressed in the NAc recently revealed a protective effect for this TF, whereby decreased levels of cocaine SA and enhanced extinction of drug taking were observed (Zhang et al. 2014).

The TF CREB has also been heavily implicated in drug-induced transcriptional plasticity. CREB has been shown to be increased in its expression following experimenter-administered cocaine. Viral overexpression of CREB increases sensitivity to cocaine in CPP paradigms, such that low doses of cocaine that do not normally cause animals to form a place preference become aversive (Carlezon et al. 1998). CREB has also been probed in SA models, in which overexpression of the TF has been found to increase sensitivity to cocaine in a dose–response paradigm, with elevated breakpoints observed under progressive ratio conditions, indicating enhanced motivation for drugs of abuse; however, these effects have held true only for high doses of cocaine (Larson et al. 2011).

Regulation of CREB activity is additionally modulated by its phosphorylation at serine 133, termed phospho-CREB (pCREB). This modified pCREB represents the form of the protein that interacts with DNA to induce transcriptional activation. An interesting aspect of pCREB signaling in addiction is that it appears to be intimately linked to MSN excitability and thus may serve as an important regulator of cellular physiology during drug use. As discussed previously, MSNs are a neuronal subtype that make up the majority of the neuronal population in the NAc and therefore represent an important substrate for addiction phenotypes. The role of MSN neurotransmission and pCREB in drug-related behaviors has long been established in earlier studies demonstrating that overexpression of a constitutively active form of CREB (mimicking pCREB) increases NAc MSN excitability, whereas overexpression of a dominant-negative isoform of CREB (one that cannot be activated) exerts opposing effects, while also increasing behavioral sensitization to cocaine (Dong et al. 2006).

These findings suggest a relationship between pCREB activity and dysregulated cellular processes in the NAc; however, more recent work examining pCREB in extended access cocaine SA has revealed somewhat contradictory results, with both increases and decreases in pCREB observed after chronic cocaine taking (Hollander et al. 2010; Sun et al. 2013). These differences may be due to variations in brain region–specific responses to cocaine or other factors, such as the timing of tissue-collection post-drug exposures. To fully tease apart these phenomena will require further application of more sophisticated genomic targeting approaches, molecular profiling, and behavioral phenotyping to determine the precise roles of pCREB in cocaine dependence. Although ΔFosB and CREB are likely the most heavily studied TFs in the addiction field, numerous other TFs have also been implicated in the behavioral phenotypes induced by cocaine administration (both self- and experimenter-administered), such as early growth response 1 (EGR1) and SYR-Box TF 4 (SOX4), indicating that large numbers of TFs are likely dysregulated in addictive behaviors (Wimmer et al. 2019).

FUTURE DIRECTIONS FOR THE FIELD

The multitude of studies discussed throughout this review collectively highlight the enormous complexity of epigenetic regulation in drug addiction, with the majority of this work—mostly coming from experimenter-administered and/or operant models of drug abuse—pointing to SUD being characterized by enhanced, aberrant transcriptional plasticity within key brain reward regions. This plasticity appears, in large part, to be dictated by chromatin-mediated transcriptional priming effects that act to maintain persistent cellular and behavioral phenotypes associated with the long-lasting nature of addiction (Fig. 1). As the field moves forward, it will be important to continue utilizing the power of operant models of drug use to more fully delineate the epigenetic alterations that are indeed causal in mediating the endophenotypes of SUD. Furthermore, it will be critical that researchers continuing leveraging currently available genome-wide sequencing methods to more comprehensively profile gene expression patterns in conjunction with additional assessments of the epigenetic regulome. Importantly, these types of experiments can now also be performed on human postmortem brain tissues from drug dependents, followed by overlay assessments using data collected from animal models. These efforts promise to greatly improve the translatability of identified key regulators and aid in the identification of bona fide targets for the future development of pharmacotherapeutics aimed at treating these disorders.

Figure 1.

Epigenetic priming in substance use disorders. Represented here are four examples of observed patterns of neuroepigenetic regulation following acute versus chronic exposures to drugs of abuse. Gene A represents a repressed locus that can be activated by acute exposures, yet following periods of abstinence, becomes primed at the level of chromatin to be hyperactivated upon reintroduction to drug stimuli. Gene B, on the other hand, represents a permissive gene that becomes repressed by acute exposures and then remains inappropriately repressed during withdrawal and in response to subsequent drug priming and/or cue reexposures. Gene C represents a gene that might become activated by acute drug exposures, similar to gene A. However, following chronic drug use and withdrawal, gene C becomes inappropriately repressed, so that upon subsequent drug exposures, the locus remains in a chronically repressed state. Gene D, although not activated by acute exposures, becomes aberrantly activated by chronic drug experiences and then primed during abstinence, such that reexposure to drugs of abuse and/or their associated cues result in hyperpermissiveness of the gene product. (TF) transcription factor, (RNAPII) RNA polymerase II transcriptional complex.