Homeostatic Regulation of Reward via Synaptic Insertion of Calcium-Permeable AMPA Receptors in Nucleus Accumbens

Abstract

The incentive effects of food and related cues are determined by stimulus properties and the internal state of the organism. Enhanced hedonic reactivity and incentive motivation in energy deficient subjects have been demonstrated in animal models and humans. Defining the neurobiological underpinnings of these state-based modulatory effects could illuminate fundamental mechanisms of adaptive behavior, as well as provide insight into maladaptive consequences of weight loss dieting and the relationship between disturbed eating behavior and substance abuse. This article summarizes research of our laboratory aimed at identifying neuroadaptations induced by chronic food restriction (FR) that increase the reward magnitude of drugs and associated cues. The main findings are that FR decreases basal dopamine (DA) transmission, upregulates signaling downstream of the D1 DA receptor (D1R), and triggers synaptic incorporation of calcium-permeable AMPA receptors (CP-AMPARs) in the nucleus accumbens (NAc). Selective antagonism of CP-AMPARs decreases excitatory postsynaptic currents in NAc medium spiny neurons of FR rats and blocks the enhanced rewarding effects of d-amphetamine and a D1R, but not a D2R, agonist. These results suggest that FR drives CP-AMPARs into the synaptic membrane of D1R-expressing MSNs, possibly as a homeostatic response to reward loss. FR subjects also display diminished aversion for contexts associated with LiCl treatment and centrally infused cocaine. An encompassing, though speculative, hypothesis is that NAc synaptic incorporation of CP-AMPARs in response to food scarcity and other forms of sustained reward loss adaptively increases incentive effects of reward stimuli and, at the same time, diminishes responsiveness to aversive stimuli that have potential to interfere with goal pursuit.

Introduction

Motivational theory that preceded and influenced recent advances in behavioral neuroscience maintained that incentive properties of environmental stimuli are sufficient to drive behavior and that their potency is determined, in part, by the internal state of the organism [16]. An example of this joint regulation by internal and external factors is positive alliesthesia, in which the hedonic and incentive motivating effects of food and related cues are increased by energy deficit [23]. In animal models, sweet taste reactivity is enhanced by food deprivation [13,73], and the reinforcing potency of food measured in progressive ratio protocols of instrumental responding displays a pattern of increase that parallels body weight loss [60,88,89]. Negative energy balance therefore augments both orosensory reward and the incentive motivation that drives food seeking behavior. Both effects have been confirmed in human studies, where energy deprivation increases the “liking” [26,180,212] and “wanting” of food [54,158]. The neurocircuits that underlie hedonic reactivity and incentive motivation have been differentiated, with the former including a key opioid component, and the latter being largely dopamine mediated [36,99]. These separate but interacting [15,137,169] neurocircuits may be co-regulated when there are shifts in energy balance, though they can be differentially regulated as well, as is the case in addictive disorders where hedonic reactivity declines and incentive motivation escalates [164,165].

The discoveries of endocrine adiposity hormones and feeding-related neuropeptides with receptor populations in both the hypothalamus and mesolimbic pathway point to potential mechanistic bases for regulation of incentive motivation as a fundamental feature of homeostatic behavior [e.g., 50,93,120]. Thus, the presence of receptors for leptin, insulin, ghrelin, orexin, melanocortins, GLP-1 and other metabolic signaling peptides in the ventral tegmental area or nucleus accumbens point to multiple candidate mediators of the link between energy balance and reward. For most of these signaling molecules there is supporting, albeit incomplete and sometimes inconsistent, evidence of involvement in reward modulation [for reviews see: 6,104,120,139,156,213].

Probing reward modulation using drugs as proxies for food

A hypothesis that has been well supported by basic and translational studies maintains that drugs with abuse liability target a final common pathway for the positive reinforcing effects of natural rewards [90,108,209]. Much of the supporting research has focused on the relationship between mechanisms regulating ingestive behavior and drug abuse [28,49,101,204], with the most studied link between the two being the mesoaccumbens dopamine (DA) pathway [7,8,14,59,80,102,141,146,153,157,178,187, 210]. Moreover, just as deprivation increases the rewarding effects of food, food deprivation increases the reward magnitude of abused drugs. In animal models, chronic food restriction (FR) has been shown to increase sensitivity to drug rewarding effects in self-administration [35], conditioned place preference [9,121,186,215], and electrical brain stimulation reward protocols [24,25]. FR also increases the incentive effects of contexts and cues that had been paired with the subjective effects of drugs during prior ad libitum fed states [45,215,217]. The enhancing effect of FR on drug reward magnitude, as assayed in a learning-free curve-shift protocol of intracranial self-stimulation, results from a regimen that leads to a maintained 20% decrease in body weight [24]. The effect is present whether subjects are tested before or after their single daily meal [217], and the return to baseline following restoration of ad libitum feeding occurs gradually over a three week period [25]. These results match the observation that maintenance of a substantially reduced body weight increases drug self-administration, whereas an acute withholding of food does not [33,34]. Most studies conducted to assess metabolic hormone involvement in these effects have not yielded positive results [82,83,127,172,217], though results relating to glucocorticoid involvement suggest differential involvement based on the drug under study and sex of the subjects [27,128,172].

An assumption undergirding this line of research is that just as drugs of abuse exert reinforcing effects by “hijacking” the neural substrate for natural reward, the potentiating effect of FR on drug abuse is based on a “hijacking” of neuroadaptations that normally increase the incentive motivating effects of food and related cues under the condition of negative energy balance. These neuroadaptations may contribute to the increased risk of binge eating and drug abuse among individuals who engage in weight loss dieting [2,38,64,116,131,152,181,203,205]. Relatively few basic research studies have been conducted to model the pathogenic potential of food restriction. However, it has been shown that several weeks of alternating 12-h food deprivation with 12-h sugar access, or combining longer duration food restriction with acute foot shock stress induces binge eating [43]. In the case of drug abuse, food restriction with maintenance of reduced body weight (75–90%) is sufficient to increase self-administration [33–35], reinstatement of extinguished drug-seeking [176], and cue-induced reinstatement of drug-seeking following abstinence [46].

Dopamine-related changes induced by food restriction

The food restriction protocol used in our behavioral studies (i.e, maintenance at 80% of pre-restriction body weight for a minimum of 3–4 weeks), or one very similar to it [154,155], induces a number of changes in the mesoaccumbens DA pathway. Glutamate currents in putative ventral tegmental DA neurons [147], basal DA synthetic activity in NAc, [148], and evoked DA release in NAc slices are all decreased [182]. Basal extracellular NAc DA concentrations in vivo are also decreased [154,155], as are levels of preprodynorphin and preprotachykinin mRNA contained in D1 receptor expressing medium spiny neurons (MSNs) [79]. This set of findings is suggestive of DA conservation in the mesoaccumbens pathway during FR, which is consistent with the adaptive downregulation of energy expenditure [1]. However, in response to challenge with an agonist for D1-like (D1/D5) DA receptors, MAP kinase signaling, CaMKII signaling, gene expression in NAc, and elicited behavioral responses are increased [29–32, 77–79]. This latter combination of findings suggests compensatory upregulation of postsynaptic responsiveness. Intracerebral microinjection studies with d-amphetamine and DA receptor agonists point to the D1R and NAc shell as the key DA receptor type and NAc subregion underlying the enhancing effect of FR [29,30,145,150]. Unlike the D2 receptor, D1 exists in a low affinity state and is not activated until a DA surge increases extracellular concentrations [161]. Thus, low basal DA tone with upregulated signaling downstream of D1R may tune the system to respond selectively and strongly to phasic DA-releasing stimuli (e.g., food, cues, drugs of abuse). When MSNs receive glutamate input, D1R stimulation facilitates the transition from a hyperpolarized downstate to the upstate where membrane potential is near spike threshold [189]. The underlying mechanism is not clear but depends at least in part on protein kinase A (PKA) and glutamate AMPA receptors [196].

AMPA receptor-related changes induced by food restriction

The DA innervation of NAc is convergent with several forebrain glutamate inputs [75] and AMPARs are co-expressed with DA receptors in NAc neurons [12,68]. Most are either GluA1/GluA2 or GluA2/GluA3 heteromers [159]. Changes in AMPA receptor abundance in the synaptic membrane mediate dynamic tuning of synaptic transmission as well as plasticity [40,103]. In cultured MSNs, D1R activation rapidly increases GluA1 surface expression in a PKA-dependent manner [37,126]. It is therefore noteworthy that FR increases phosphorylation of GluA1 at Ser845, the PKA site, in response to administration of a D1R agonist, d-amphetamine, cocaine, sucrose, and a morphine-paired environment [30,98,121,150,151]. This set of findings points to a mechanism, upregulated by FR, that is activated by natural reward, drugs of abuse, and an environment associated with subjective effects of a drug. Phosphorylation at Ser845 stabilizes GluA1 in the membrane and facilitates synaptic insertion [53,55,85,86,125,144,166,177]. Consistent with pSer845 involvement in AMPAR trafficking, it was found that sucrose and d-amphetamine each increased levels of GluA1 and GluA2 in the NAc postsynaptic density (PSD), with a greater effect in FR than in ad libitum fed (AL) rats [150,151,197]. Given that most GluA1 in NAc are associated with GluA2, and most GluA2 not associated with GluA1 are partially assembled receptors [159], it is likely that reward stimuli during FR increase insertion of GluA1/GluA2 heteromers.

In the course of investigating effects of reward stimuli on AMPAR trafficking, it was discovered that FR, itself, increases NAc synaptic abundance of GluA1. Using a BS³ cross-linking method [18], it was shown that FR increases surface expression of GluA1 but not GluA2 [145]. Subcellular fractionation revealed that FR specifically increases GluA1 protein levels in the NAc PSD [145]. These findings suggested that FR increases synaptic insertion of GluA2-lacking, calcium-permeable AMPA receptors (CP-AMPARs). This conclusion was substantiated by electrophysiological findings in which Naspm, a selective antagonist of CP-AMPARs, decreased the amplitude of evoked excitatory postsynaptic currents (EPSCs) in NAc shell MSNs of FR, but not AL, rats [145]. Further, behavioral experiments indicated that CP-AMPARs mediate the enhanced responses of FR subjects to drug challenge. Microinjection of Naspm into NAc shell reversed the enhanced rewarding effects of SKF-82958 (D1/D5R agonist) [30] and d-amphetamine [150], and reversed the enhanced locomotor-activating effect of SKF-82958 in FR subjects [145]. Naspm had no effect on the rewarding or locomotor-activating effects of the D2-like (D2/D3R) agonist, quinpirole [30,145]. The significance of a synaptic population of CP-AMPARs lies in distinct properties they possess relative to other AMPAR types, including larger single channel conductance, faster kinetics, and triggering of postsynaptic signaling cascades that depend on Ca²⁺ [115].

FR does not increase density or affinity of D1 binding sites in NAc [77] nor does it increase D1R-stimulated adenylyl cyclase activity [32], which activates canonical signaling leading to phosphorylation of GluA1 at Ser845 via activation of PKA. However, recent findings point to an alternative mechanistic basis for increased DA-dependent pSer845-GluA1 and AMPAR surface expression in the NAc of FR rats. In cultured striatal MSNs, a cooperative relationship was shown between CP-AMPARs and D1Rs in regulating phosphorylation and trafficking of AMPARs [198]. Briefly, D1R stimulation increased phosphorylation of GluA1 at Ser845, via PKA, thereby increasing extrasynaptic accumulation of GluA1-containing AMPARs. Concurrent stimulation of CP-AMPARs triggered intracellular Ca²⁺ signaling, activating cGMP and cGKII, which augmented the phosphorylation of GluA1 at Ser845, and was followed by PKC-mediated synaptic insertion of GluA1 and GluA2. According to this model, synaptic insertion of CP-AMPARs may be the adaptive response to FR that enables amplification of cellular and behavioral effects downstream of D1R stimulation.

Glutamate inputs to nucleus accumbens

Many distinct components of reward-related behavior are encoded by NAc neurons [170], and strong evidence indicates that orosensory reward and consummatory behavior are initiated and sustained by neuronal inhibition [5,57,100,110,124,160,167,168,183,191,206]. Recently, O’Connor and coworkers identified D1R-expressing MSNs in NAc medial shell that project to lateral hypothalamus as the specific sub-population whose inhibition drives consummatory behavior [142]. In contrast, incentive motivation, as reflected in the initiation and vigor of approach triggered by reward cues, correlates with NAc neuronal firing rate [44,84,91,133], and is dependent on NAc neuronal excitation [42]. Taking a direct approach to assessing functional consequences of NAc neuronal excitation, optogenetic studies have shown that activation of glutamate terminals in NAc is positively reinforcing [19,188], as is direct activation of NAc medial shell neurons. These effects, and the potentiating effect of NAc neuronal activation on cocaine reward, have been attributed to D1R-expressing MSNs [19,111,122,199]. A caveat, however, is in recent findings that, depending on context and task, activation of D2R MSNs can also facilitate positively motivated behavior [179]. Nevertheless, based on the aforementioned pharmacological and biochemical findings, it may be predicted that FR increases synaptic insertion of CP-AMPARs in D1R MSNs of NAc medial shell. This prediction remains to be tested.

An important gap in mechanistic understanding of FR effects is the source(s) of glutamate that acts upon altered AMPAR mechanisms to drive behavior, and the question of whether any glutamate input is differentially involved in FR relative to AL rats. Glutamate inputs to NAc include those from medial prefrontal cortex (mPFC), ventral subiculum (vSub), basolateral amygdala (BLA) and paraventricular thalamus (PVT) [11,19,65,66,175]. Most addiction research has been focused on the former three inputs, which display a high degree of convergence upon individual MSNs [19,62,65,66,75,132], and whose terminals in NAc can be stimulated to increase locomotor activity [4] and reinforce intracranial self-stimulation [19,185,188]. The fact that 95% of MSNs are excited by activation of any one of these three input pathways [19], has led to the suggestion that the amount of glutamate release in NAc may be of greater functional importance than the source [19,199]. Based on such findings, it might be expected that all glutamate inputs to NAc have the potential to interact with a behaviorally significant population of CP-AMPARs. Yet, a source of glutamate input to NAc shell that is well-suited to signal in a manner that is dependent on diet and metabolic status, is PVT [11,20,118]. PVT neurons are excited by food deprivation [195], glucoprivation [51], orexigenic peptide pathways [130], and are activated when FR subjects are anticipating a meal [138]. Importantly, drugs of abuse also activate PVT neurons [48,63,67]. Selective elimination of PVT neurons that innervate NAc decreases acquisition of cocaine self-administration [140], and inactivation of PVT blocks cocaine prime-induced reinstatement [95], CPP expression [21], and meal-anticipatory locomotor activity [138]. However, not unlike the sub-population of NAc shell neurons whose inhibition promotes consummatory behavior, experimental inhibition of PVT neurons increases consumption [184,214]. PVT contains a nearly pure population of glutamate cells [61], most of which innervate NAc shell [52,92,119], where they form synaptic connections with MSNs [119] and converge with DA afferents [10]. There has been a recent test of PVT involvement in drug-seeking behavior of rats on a mild FR regimen in which body weight was maintained at 90% of the pre-FR value. Chemogenetic inhibition of PVT had no effect on heroin seeking or its augmentation by FR in subjects that were withdrawn from heroin [39]. However, chemogenetic excitation preferentially altered the behavior of FR relative to AL rats, but the effect was to inhibit their heroin-seeking. While this result is consistent with the existence of an upregulated response to glutamate of PVT origin, it is not known which PVT target area(s) is involved, nor is the status of NAc AMPAR mechanisms under this protocol known. That is, it is not known whether a mild FR protocol induces the same neuroadaptations as the protocol described above, and opioid withdrawal, itself, induces excitatory plasticity in NAc [123].

Synaptic insertion of CP-AMPARs and motivational valence

In addition to increasing the unconditioned rewarding effects of abused drugs, FR increases the magnitude and persistence of a cocaine conditioned place preference (CPP) acquired during a prior ad libitum (AL) fed state [215–217]. The same effect is seen in morphine CPP [98]. Further, in both cases, CPP expression was associated with increased NAc pSer845-GluA1, which correlated with magnitude of CPP. However, when a LiCl place aversion was conditioned, and subjects were switched to FR for testing, the CPA weakened and extinguished more rapidly than in AL subjects. The opposing effects of FR on cocaine/morphine CPP and LiCl CPA suggest that the effects on CPP expression are not due to a general enhancement of recall or attentiveness but rather, increased incentive effects of the drug-paired context.

One determinant of vulnerability to drug abuse is the balance between rewarding and aversive drug effects [163,201,202]. For example, a substantial percentage of psychostimulant users report adverse effects (e.g., depression, anxiety, irritability), and this often leads to moderation of use [208]. Moreover, a hallmark feature of addiction is drug-seeking despite adverse conditions [47,200], which may be homologous with the suppression of aversion that accompanies strong food-motivated behavior [22,107]. Consequently, organismic states that include a negative gating of aversive stimuli may enhance use of drugs with mixed effects and promote addictive behavior. It was therefore considered whether the attenuation of LiCl CPA might be reflective of an aversion-attenuating effect, possibly related to the anxiolytic and analgesic effects that can be induced by FR [81,94,117,134,211]. To assess whether FR may have a moderating effect on the aversive subjective effect of cocaine, we adapted a protocol in which cocaine reinforces a CPA if pairing of drug and context are delayed by 15-min [56]. Using i.c.v. administration, delayed pairing reinforced a modest but significant cocaine CPA. When testing resumed three weeks later, AL subjects continued to display avoidance of the cocaine-paired compartment, but subjects that had been switched to FR displayed preference [Figure 1]. This reversal in motivational valence could result from the known reward-enhancing effect of FR. That is, if cocaine has mixed effects that net out to aversion in the AL fed subject, and those mixed effects are elicited by the paired context as well, FR may selectively enhance the rewarding component and switch valence. Yet, in the LiCl experiment, neither the unconditioned nor conditioned stimulus has a rewarding component available for augmentation.

Figure 1.

In eight consecutive daily conditioning sessions, rats (male and female) were injected with cocaine (200 μg, i.c.v.) or saline vehicle in alternation. Fifteen minutes after injection, rats were confined (15-min) to corresponding compartments of a place conditioning apparatus. Acquisition of a cocaine conditioned place aversion was confirmed (Test 2; n=25 total; p<.05). Rats were then either maintained AL (top) or switched to FR (bottom) for 3 weeks prior to resumption of testing. Mean time (sec) on the saline- and cocaine-paired sides are displayed. Time not spent on either side was spent in the center compartment. Analysis of difference scores (tests 3–6) indicates an effect of diet (F_1,23=7.45, p=.01) [Lee and Carr, unpublished].

An aversion-attenuating mechanism is in line with interest in NAc as a target for modulating pain and negative affect [136], and basic science studies point to a possible mechanistic overlap between these effects and the reward potentiating effect of FR. Specifically, chronic pain increases NAc synaptic abundance of CP-AMPARs, and these receptors mediate a negative feedback mechanism which attenuates the accompanying aversive motivational-affective state [71,114]. Current understanding of the way in which reward and aversion are coded in NAc does not readily lend itself to a model in which separate glutamate inputs acting through CP-AMPARs facilitate reward and inhibit aversion. Optogenetic studies generally support the view that activation of D1R-expressing MSNs has a pro-reward effect and, though somewhat less well supported, activation of D2R-expressing MSNs has an anti-reward effect [179]. In this simple model of valence coding, synaptic insertion of CP-AMPARs in D1R-expressing MSNs might be expected to have the dual effect of enhancing reward and diminishing aversion by shifting the balance of excitability between functionally opponent cell types. It is also possible that reward-enhancing and aversion-attenuating populations of CP-AMPARs are anatomically separable. Optogenetic activation of glutamate terminals from prelimbic cortex in NAc core induces analgesia, inhibits the aversive motivational state, and restores sucrose preference in the chronic pain model [114]. The research on CP-AMPAR involvement in the reward-enhancing effect of FR has, on the other hand, implicated the NAc shell, while the core has not been similarly examined.

The triggering condition for NAc neuroadaptations

The triggering condition for NAc synaptic incorporation of CP-AMPARs during FR has not been identified, though reward deficiency and the neurochemical concomitant of decreased DA transmission merit investigation. There are several other conditions in which loss of reward is followed by synaptic incorporation of CP-AMPARs in NAc. These include withdrawal from chronic cocaine [41], amphetamine [96], morphine [87], ‘junk food’ [143], and ventral tegmental DA neuronal stimulation [149]. Moreover, these CP-AMPARs have been shown to mediate the enhanced cue-triggered motivation and “craving” which characterize the withdrawn state [41,143]. Synaptic insertion of CP-AMPARs may therefore be a cellular response aimed at maintaining reward homeostasis by enhancing the incentive potency of appetitive stimuli and, inadvertently, their proxies (e.g., drugs of abuse). This line of thinking is a speculative extension of findings obtained in cultured neurons, sensory cortices, and brainstem sensory relays following deprivation of input, where homeostatic trafficking of CP-AMPARs to the synapse either restores stable activity or compensates for loss of function [69,70,76,97,105,115,171,190,192,193]. To our knowledge, the only prior evidence bearing on the possibility that deficient DA transmission is a stimulus for CP-AMPAR insertion comes from extreme cases of DA depletion; a genetic mouse model of Parkinson’s Disease increases striatal levels of GluA1 and pSer845-GluA1 [109], and 6-OHDA-induced DA depletion causes dorsal striatal synaptic incorporation of CP-AMPARs [3] and an L-DOPA-induced dyskinesia that is dependent on striatal CP-AMPARs [106]. However, there is also the recent finding that mice withdrawn from chronic optogenetic self-stimulation of ventral tegmental DA neurons show synaptic insertion of CP-AMPARs in D1R-, but not D2R-, expressing MSNs in NAc, coinciding with emergence of vigorous cue-induced relapse to seeking DA stimulation [149].

Translation and caveats

The basic science findings relating to food restriction, dopamine, AMPA receptors, and drugs of abuse were obtained in male rats. This points to a serious gap in the literature. National surveys in the US indicate that at any given time twice as many women as men are dieting to lose weight [17,162,174,207], and women consistently seek to lose a greater percentage of body weight than men [17,58,112]. Moreover, transition from dieting to an eating disorder occurs predominantly in girls and women [181]. Comorbidity of eating disorders with substance use disorders ranges from 23 to 37%, which is much higher than prevalence in the general population of 2–3% for illicit drugs and 12% for alcohol [135]. Considering the female predominance among those who diet, suffer from eating disorders, and express a comorbid substance use disorder, the use of males in the FR studies cited above provides an incomplete view of the translational relevance of findings obtained.

Should continuing research support a model in which NAc synaptic insertion of CP-AMPARs during FR increases the incentive potency of appetitive reward stimuli and concurrently diminishes aversion, the utility of such a neuroadaptation during periods of food scarcity in the wild would be clear. However, modern human societies offer opportunities for maladaptive consequences. These could include self-medication of negative affective states by abstinence from feeding, breakthrough episodes of bingeing among weight loss dieters, and increased relapse risk among those who diet to combat rebound weight gain which occurs during cessation of smoking and other psychostimulant addictions.

A moderating consideration when judging the above model as a feature of homeostatic control or a basis for the pathogenic potential of weight loss dieting is the impoverished environment occupied by rodent subjects in most of the aforementioned studies. Because subjects in these studies typically have surgical implants, and/or their food access and consumption need to be precisely regulated, they are usually housed individually without social contact or enrichment objects. The incentive value of food is accentuated under these housing conditions, as suggested by the fact that free-feeding rats housed in this manner take in more calories and are more obese than rats housed in an enriched environment [129]. Consequently, “impoverished” housing would increase the severity of reward loss when food is removed. Interestingly, transfer of rats from impoverished housing to social housing with novel objects increases NAc metabolic activity in both the short- and long-term [72,113] and specifically increases basal extracellular [DA] and potassium-induced DA release [173]. These changes are accompanied by decreased potency of other incentive stimuli. For example, just 22 hours of environmental enrichment decreased rats’ consumption and responsiveness to sucrose-paired cues [74], and three weeks of enrichment decreased cue-induced reinstatement of cocaine-seeking [194]. Consequently, CNS and behavioral effects of FR in naturalistic environments may be less extreme than those identified in the studies described, and this additional factor merits systematic investigation.

Highlights.

• Food restriction increases the reward magnitude of abused drugs and associated cues

• Behavioral effects are mediated by Ca²⁺-permeable AMPA receptors in n. accumbens

• Results may provide insight into dieting as a risk factor for binge eating and drug abuse