AUD is diagnosed as a pattern of excessive alcohol consumption with repeated cycles of abstinence followed by resumption of heavy drinking, which in turn leads to negative psychological, physical, and social consequences [1]. This phenotype has been captured in a rhesus monkey model of alcohol self-administration using repeated cycles of chronic voluntary ethanol drinking and prolonged abstinence [2]. In this protocol, subjects are first induced to drink water and then ethanol (4% w/v) under a schedule-induced polydipsia (SIP) procedure. After induction, the monkeys given “open access” conditions for 22 h/day to choose between 45 ethanol. Given the voluntary nature of self-administration, the monkeys show a wide spectrum of individual differences in their daily intake ranging from an average of 0.3 g/kg/day up to 4 g/kg/day [3]. Based on mathematical modeling procedures and using data from six cohorts of monkeys with daily 22 h access to ethanol, four categorical levels of drinking were identified: light drinker (LD), binge drinker (BD), heavy drinker (HD) and very heavy drinker (VHD) [3, 4]. When the drinking protocol was extended to include three cycles of approximately 4–5 weeks of alcohol abstinence followed by 12–14 weeks of post-abstinence open-access conditions, alcohol intake generally increased following the reintroduction of alcohol (i.e., modeling resumption of heavy drinking) and gradually returned to pre-abstinence levels in all but a few monkeys [2]. The biological and behavioral features of this repeated abstinence protocol are described elsewhere [2, 5–7]. Overall, this repeated abstinence protocol has translational value for understanding the effects of cyclical chronic drinking and abstinence on subsequent drinking patterns.

The dorsolateral prefrontal cortex (DLPFC) is implicated in the return to heavy drinking in subjects with AUD, particularly in terms of volumetric analyses with magnetic resonance imaging (MRI) predicting sustained abstinence [8]. Indeed, subjects who returned to heavy drinking after 4 weeks of abstinence showed recovery of frontal cortical volume compared to subjects who maintained abstinence [8]. This suggests that atrophy of the DLPFC seen in relapsing AUD is due to underlying aberrant neuroadaptation to chronic heavy drinking. To address the molecular adaptations of the DLPFC in abstinence and just prior to resumption of heavy drinking, the present study utilized a biopsy of PFC area 46 before and following the protocol described above that ended with a final abstinence period (i.e., at necropsy). Rodent, nonhuman primate (NHP) and human postmortem studies have all clearly established the alignment of the brain transcriptome with ethanol consumption [9–13]. Rodent studies have been helpful in untangling risk genes and gene networks from those associated with the consequences of excessive ethanol exposure, as well as short-term abstinence (72 h) [14–18]. However, relatively little attention has been focused on the effects of protracted and/or repeated abstinence following chronic consumption. In addition, the rodent prefrontal cortex is underdeveloped compared to the primate brain. Thus, the current study offers a unique dataset addressing primate brain cortical adaptations and remodeling in abstinence from chronic alcohol consumption and just prior to resumption of heavy drinking. Finally, the data presented here offer a novel opportunity to examine how chronic ethanol exposure and repeated abstinence affect the NHP cortical transcriptome over a critical period of synaptic maturation, i.e., from late adolescence to young adulthood.

Walter et al. [35] were the first to use the within-subject biopsy/necropsy design to examine the effects of chronic open-access ethanol consumption on Area 46 gene expression. That study used cynomolgus macaques (6 controls and 17 drinkers), aged between 5 and 7 years, and the duration of the open-access phase was approximately 28 weeks, with no imposed abstinence. The key findings may be summarized as follows. Compared to the controls, a total of 675 genes were significantly downregulated following EtOH consumption; these genes were functionally enriched for immune response, cell adhesion, plasma membrane, and extracellular matrix. A total of 567 genes were significantly upregulated following EtOH consumption, were enriched in microRNA target sites, and included target sites associated with Toll-like receptor pathways. The differentially expressed genes were also significantly enriched in transcription factor binding sites. Discussion of these data emphasized the effects on the immune response genes, which were both pro- and anti-inflammatory, e.g., Il1r1 and Cd74 vs. Il6r and Il10ra. The overall inflammatory status of area 46 was not known, but the data suggested that in association with chronic ethanol consumption, there was a reset between the pro- and anti-inflammatory pathways. There is robust evidence of relationships between neuroimmune signaling and alcohol use disorder (e.g., [36–38]). Neuroinflammatory mechanisms are associated with both the risk for and individual variation in excessive EtOH consumption (e.g., [39, 40]). Furthermore, there is ample evidence that chronic EtOH consumption has a marked effect on the expression of cytokine ligands and receptors [41–45]. It should be noted that in Walter et al. [35] and the current study, the biopsy/necropsy samples were not perfused. Thus, the possibility must be considered that some of the gene expression signals, especially immune-related signals, are associated with blood cells. For example, Sureshchandra et al. [46] and Barr et al. [47], using gene expression profiling, found that in HD macaques, there was a disruption of innate immunity. However, the profiles of the blood genes affected included an enrichment in genes associated with wound healing and blood coagulation that was not seen in the brain data.

The possibility cannot be easily dismissed that the Area 46, which is part of the cognitive circuit involved in the control of alcohol consumption, does not fit the generally accepted model of the interaction between the neuroimmune system and alcohol consumption. However, we should also note that positron emission (PET) studies using ¹⁸F-deoxyglucose (FDG) during initial abstinence have revealed in humans that the prefrontal cortex is hypometabolic (Volkow et al. 1992). We assume, although unproven, that similar data would be obtained in the NHP model. It is possible that the neuroimmune genes are particularly sensitive to this metabolic malaise. The duration of the metabolic effect under the conditions of the present experiment is unknown.

The current study used the within-subject biopsy necropsy design, along with repeated cycles of abstinence, to examine transcriptional changes associated with ethanol intake and ADE. The study differs from Walter et al. [35] in several important aspects. 1) Male rhesus macaques rather than cynomolgus macaques were the study subjects; however, the age range was similar. 2) The initial open-access ethanol drinking phase was significantly longer (60 weeks vs. 28 weeks). An additional 52 weeks of open-access consumption (two post-abstinence open access periods of approximately 26 weeks each). 3) The necropsy sample was obtained after 34–38 days of withdrawal. Despite these differences, both studies detected an “immune response” signal. As noted above, in Walter et al. [35], the signal was detected in genes downregulated in drinkers but not controls. In the current study, the signal was detected in those subjects where the ADE moved the HD group to VHD status. The question naturally arises as to whether the immune response genes were similar. This is a hypothetical question that cannot be addressed satisfactorily given the differences in study design and given that the number of immune response genes detected in Walter et al. [35] was markedly larger (92 vs. 34) than the number detected in the current study. However, 13 of the genes were in common: C3, Csf1r, Ctss, Fgr, Ifi27, Il16, Irf8, Lyn, Ncf2, Pycard, Tyrobp, Vav1 and Was. With the exception of Lyn tyrosine kinase (Lyn), this core gene grouping would be considered proinflammatory. The colony stimulating factor 1 receptor (CSF1R) and its ligand (CSF1) are key to the maintenance and proliferation of microglia, the resident macrophages in the brain. Given the role(s) of immune systems in excessive ethanol consumption noted above, microglia are likely to have some role in this process. Fortunately, this hypothesis can be directly tested since CSF1R antagonists deplete brain microglia. Warden et al. [48, 49] examined this issue from two perspectives. They found that PLX5622-induced depletion of microglia did not affect the hypnotic or sedative effects of acute ethanol intoxication and did not affect the maintenance or escalation of voluntary ethanol consumption. However, microglial depletion blocked the poly (I:C)-induced escalation of ethanol consumption, suggesting that microglia have a role in ethanol consumption in the context of sufficient immune activation [48]. When they used PLX5622 to examine the role of microglia in chronic intermittent ethanol (CIE) vapor-induced escalation of ethanol consumption (2-bottle choice), microglia depletion blocked CIE-induced increases in ethanol consumption and blocked associated CIE-induced changes in gene expression [49]. The question naturally arises as to how to interpret the decrease in Csf1r expression observed in Walter et al. [35] and the current study. One can speculate that in macaques, the decreased expression is part of a compensatory mechanism, perhaps involving all of the genes noted above, in response to high levels of ethanol consumption.

Continuing the theme of compensatory mechanisms is the observation that the RANK signaling pathway was significantly upregulated in the necropsy vs. biopsy samples of the subjects that converted from HD to VHD. Key protein members of the pathway include receptor activator of nuclear factor-κB (RANK), the RANK ligand (RANKL), and osteoprotegerin (OPG), forming the RRO axis. OPG is a decoy receptor for RANKL. The involvement of the RRO axis in bone remodeling and peripheral immune responses is well established [50–53]. In the brain, the RRO axis tunes the neuroinflammatory response, depending on the molecular, cellular and pathological context. RANK/RANKL signaling is neuroprotective in TLR-mediated inflammation, which has been linked to ethanol-induced neuroinflammation (see, e.g., [54–57]). To our knowledge, this is the first report linking RANK signaling to excessive ethanol consumption. Given the large number of compounds and biologics (e.g., antibodies) that have been developed to affect bone remodeling via RANK signaling, some of these may be effective CNS therapies [34]. However, we also consider that the RANK signaling changes may be protective adaptations and not especially suitable as therapeutic targets. A similar argument could be raised for the changes in MAPK signaling (see below).

Given the widespread association of MAPK signaling with numerous brain mechanisms and pathways, it cannot be surprising to have detected an association between MAPK signaling and very heavy drinking. From a genomics perspective, Mulligan et al. [58], in a meta-analysis of microarray data, detected an association between MAPK signaling and ethanol preference. Subsequent studies (e.g., [59, 60]) have confirmed the association between Ras/MAPK pathways and preference consumption. Here, we focus on Ferguson et al. [10], who examined and reviewed 17 human postmortem studies involving a total of 230 cases and 233 controls; the data were derived from both microarray and RNAseq experiments and involved multiple brain regions. Upregulated KEGG pathways in cases included focal adhesion, MAPK signaling, cytokine-cytokine receptor interaction, cell adhesion and regulation of actin cytoskeleton. Downregulated KEGG pathways included oxidative phosphorylation, Parkinson’s disease and DNA replication. (See Table 2—Ferguson et al. [10]—for a complete list of up- and downregulated KEGG pathways). In the current study, we observed that VHD is associated with the upregulation of genes that negatively impact MAPK signaling. The signal was seen in both the HD→VHD and the VHD→VHD groups but not in the BD→HD group. Furthermore, the MAPK signal was not detected in Walter et al. [35], although the downregulation of immune-related genes was observed (see above). Overall, these data suggest that ADE leads to MAPK signaling. The question we cannot answer is whether there was an upregulation of MAPK genes [10] that preceded the upregulation of genes that would blunt such an effect. Is this another example of compensation? It is of interest that a core set of genes associated with the negative regulation of MAPK signaling was the same in the HD→VHD and VHD→VHD groups. The core set included dual specificity phosphatases, members of the sprouty/spred gene family, Rgs2 and Bmp4. Here, we note that the DUSP and SPRY/SPRED proteins have been suggested to be novel targets for CNS therapeutics [61, 62].

The current results provide new insights into the transcriptional changes that occur over the period from late adolescence to early adulthood (roughly 8–10 macaque years) and provide new information on how chronic ethanol consumption affects these transcriptional changes. Detailed information on the spatiotemporal dynamics of the developing NHP brain transcriptome is available (e.g., [63, 64]). However, these studies do not cover the period examined in the current study. The data presented here are clear that there are marked transcriptional changes that occur over young adulthood. The main caveat to these data is the procedures the animals encountered over the nearly 3-year duration of the study. The animals were anesthetized and underwent surgery. The pre-alcohol biopsy sample was removed in all but four low to binge drinkers, maltodextrin was administered chronically, and the animals were individually housed. The precise impact of these manipulations and circumstances are unknown. However, we observed significant downregulation of genes associated with cell migration, signal transduction and the plasma membrane in the control animals in biopsy compared to necropsy samples. Upregulated genes were associated with translation. In the BD→HD group, these ontologies were reduced, further reduced in the HD→VHD group and absent in the VHD→VHD group. Assuming that the transcriptional changes seen in the controls are at least, in part, signatures from late cortical developmental processes, ethanol plus the ADE reverses these changes in a dose-related fashion. What is not known is the persistence of the ethanol effects. That is, the data do not address if the absence of altered gene expression ontologies seen in ethanol drinkers compared to controls is representative of a “critical window” such that brain growth trajectories will not recover.

Given our understanding of the mechanisms associated with chronic ethanol intoxication and withdrawal in highly dependent individuals, the data presented here suggest that in the NHP brain, area 46 has relatively subtle adaptations, perhaps reflecting the relative homogeneity of the prefrontal molecular/granular layers examined here. The question then arises: do other brain regions drive excessive consumption? Previous studies (e.g., [65]) have suggested that the central nucleus of the amygdala (CeA) is key in regulating ethanol preference. Iancu et al. [11] aligned gene expression in Area 32 and the CeA with the drinking patterns in 31 rhesus macaques that had completed the standard 12-month open-access ethanol consumption protocol [20]. The CeA transcriptional data, compared to the cortical Area 32 data, aligned more significantly with the drinking pattern data. Membrane and synaptic genes, e.g., Chrm3, Chrna4, Chrna7, Glra2, Grm1 and Grm2, were especially enriched in the genes positively correlated with ethanol consumption. This signal was not observed in Area 46. Overall, the data presented here are part of a larger and ongoing effort [9, 11, 35, 66] to understand how chronic ethanol consumption affects the primate brain, including expanding the analyses to both males and females.

The data presented here have translational value for understanding the effects of voluntary chronic ethanol drinking, coupled with periods of involuntary abstinence, on subsequent drinking patterns. The data illustrate that across all levels of NHP ethanol consumption, imposed abstinence increased consumption when ethanol was reintroduced. The approximately 4 weeks of abstinence was modeled after the most common inpatient treatment protocol for severe AUD [8]. Escalated drinking, as seen in the return to heavy drinking following imposed abstinence, is a hallmark symptom of the addiction cycle [67] and is commonly modeled in rodents [68]. Furthermore, for many monkeys, the effects of repeated abstinence on ethanol intake escalated to higher steady state levels, for example, in the conversion of HDs to VHDs and the increase in consumption among the VHDs. As we hypothesize that the escalated drinking after abstinence cycles represent a model of the most severe AUD cases, we focused on the transcriptional effects of repeated abstinence to suggest or support new strategies for pharmacotherapeutic development that could be incorporated into abstinence-based therapies. As noted above, the within-subject DLPFC Area 46 transcriptional changes suggest that neuroinflammation mechanisms are prominent in the state of the brain about to return to heavy drinking (i.e., in prolonged abstinence just prior to resumed access to alcohol). Although neuroinflammation may be protective in some situations [69], recent data suggest that centrally acting anti-inflammatory drugs, such as the PDE4 inhibitor apremilast, could be beneficial in the treatment of AUD [70]. The data presented here suggest that drugs that affect MAPK and RANK signaling may also be effective, especially in therapeutic situations associated with involuntary abstinence.