Persistent effects of obesity: a neuroplasticity hypothesis

Abstract

The obesity epidemic is a leading cause of health problems in the United States, increasing the risk of cardiovascular, endocrine, and psychiatric diseases. Though many people lose weight through changes in diet and lifestyle, keeping the weight off remains a challenge. Here, we discuss a hypothesis that seeks to explain why obesity is so persistent. There is a great degree of overlap in the circuits implicated in substance use disorder and obesity, and neural plasticity of these circuits in response to drugs of abuse is well documented. We hypothesize that obesity is also associated with neural plasticity in these circuits, and this may underlie persistent changes in behavior, energy balance, and body weight. Here, we discuss how obesity-associated reductions in motivation and physical activity may be rooted in neurophysiological alterations in these circuits. Such plasticity may alter how humans and animals use, expend, and store energy, even after weight loss.

Graphical abstract

The obesity epidemic is a leading cause of health problems in the United States, increasing the risk of cardiovascular, endocrine, and psychiatric diseases. Here, we discuss how obesity-associated reductions in motivation and physical activity may be rooted in neurophysiological alterations in these circuits. Such plasticity may alter how humans and animals use, expend, and store energy, even after weight loss.

Introduction

Despite widespread knowledge of the health risks of obesity, there has been little progress in reversing obesity in America. The obesity rate in the United States has increased approximately 30% since 1980, including a rise of about 10% since obesity was declared a national epidemic in 1999.¹ Studies show that obesity is a leading cause of death in the United States, second only to tobacco-related deaths.^2,3 Although dozens of weight loss theories and solutions have been put forward, effective approaches for combatting obesity still elude most Americans. A common perception is that reducing exposure to factors that contribute to weight gain will lead to weight loss. However, evidence suggests that this premise is flawed. While reducing caloric intake often results in immediate weight loss, it is difficult to maintain over the long term.⁴ This also occurs in animals. For example, it has been known since the 1980s that obese rodents maintain a higher average body weight than those that were never obese, even when maintained on identical diets.^5,6 Similar associations are seen in humans, supporting the concept of obesity changing a body weight “set point.”⁷ Here, we propose that obesity engages persistent changes in neural circuits that cause animals to defend an elevated body weight, even after weight loss.

The biological mechanisms that underlie such a shift in set point remain unclear and likely involve more than one system. Hypothalamic pro-opiomelanocortin (POMC)-expressing neurons in the arcuate nucleus control energy homeostasis and food intake and may contribute to the determination of body-weight set point.⁸ Basal metabolic rates can slow after weight loss, a phenomenon known as metabolic adaptation.⁹ Behavioral changes, such as reductions in physical activity, can persist following weight loss,¹⁰ contributing to a reduction in energy expenditure via physical activity. At the same time, craving for high-fat foods can increase after weight loss in animals¹¹, which may engage similar mechanisms in people to drive weight re-gain.¹² To the extent that the mechanisms underlying these changes persist, they may contribute to the persistence of obesity. Here, we discuss the hypothesis that altered plasticity in basal ganglia circuitry may underlie persistent changes in energy balance and weight.^13,14 We believe that identifying and reversing such plasticity will be necessary to achieve long-term weight loss and meaningful progress in reducing obesity rates.

Common circuits, behaviors, and underlying mechanisms between obesity and addictive disorders

Despite differences, obesity shares some similarities with substance use disorder. Obesity is associated with difficulty in controlling food intake, repeated desires and cravings for food, unsuccessful attempts to curb excessive eating, and continued overeating despite negative consequences.^15,16 Attempts to lose weight through dieting can also be reminiscent of attempts to quit drugs of abuse; both share a pattern wherein individuals abstain from the substance for a period of time followed by relapse.¹⁷ And while there is disagreement over whether obesity should be classified as an addictive state,¹⁸ there are common neural features of both substance use disorders and obesity.¹⁹ Consumption of high-fat food activates the mesolimbic dopaminergic system,^20,21 the same circuit that is activated by drugs of abuse.^22,23 Animals exposed to palatable foods can also become resistant to outcome devaluation, suggesting that palatable foods can facilitate habitual behaviors.^24–27 Here, we seek to outline a framework of circuit-based changes in substance use disorders and discuss how similar circuit adaptations may underlie the persistent behavioral changes seen in obesity.

Glutamatergic plasticity in substance use disorders

Substance abuse disorders have been linked to synaptic plasticity in the striatum, a subcortical brain region that serves as the primary input nucleus of the basal ganglia. Medium spiny neurons (MSNs) are the principal striatal cells, accounting for over 95% of rodent striatal neurons.^28,29 The striatum receives excitatory projections from the cortex, thalamus, and amygdala, as well as dopaminergic innervation from the midbrain.^30,31 The dorsal striatum is involved in regulation of motor functions, decision making, and the development of habits.^32,33 In contrast, the ventral striatum contains the nucleus accumbens (NAc) and mediates goal-directed reward and motivation, particularly under conditions that require the animal to exert effort.^34–36

At the cellular level, connections between brain regions are mediated by billions of synapses, and plasticity in these synapses regulates the flow of information. Drugs of abuse alter synaptic plasticity in mesolimbic nuclei, including the NAc and ventral tegmental area (VTA),³⁷ and dopaminergic and glutamatergic changes in the VTA underlie behavioral adaptations in response to drugs of abuse. Cocaine modifies VTA plasticity––a single in vivo cocaine exposure is sufficient to increase α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/N-methyl-D-aspartic acid (NMDA) ratios at VTA synapses, and synapses in the VTA of animals treated repeatedly with cocaine appear to be potentiated to the point of saturation³⁸ or exhibit aberrant susceptibility to long-term potentiation (LTP).³⁹ Additional work elucidated a form of cocaine-induced plasticity dependent on a distinct mechanism that appeared to involve the insertion of calcium-permeable, GluA2-lacking AMPA receptors.⁴⁰ Such cocaine-evoked plasticity in the VTA is gated by metabotropic glutamate receptors, which also control plasticity in the NAc.⁴¹ Together, this shows that synaptic plasticity in the mesolimbic circuits can be altered by acute and repeated cocaine exposure.

Interestingly, these plasticity mechanisms may interact with stress. This may be relevant to the development of substance use disorder, given the strong associations between stress and drug use in both humans and animals.^42–44 Accumbal MSNs in mice susceptible to stress-induced depressive-like phenotypes incorporated more synaptic calcium-permeable, glutamate ionotropic receptor AMPA type subunit 2 (GluA2)-lacking AMPARs relative to MSNs from control or resilient mice.^45,46 This suggests that related plasticity mechanisms in mesolimbic circuitry may underlie stress susceptibility and drug abuse. However, recent work examining the effects of cocaine versus chronic social defeat on evoked currents at established mushroom spines in the NAc suggests that shared mechanisms may be nuanced and cell type or stress-paradigm dependent; a week-long cocaine exposure and a chronic social-defeat paradigm produced opposite changes in corticostriatal plasticity.⁴⁷

The effects of cocaine on synaptic plasticity can be long lasting. Exposure to and withdrawal from cocaine increased accumbal expression of functionally silent synapses and promoted silent-to-functional synaptic conversion involving insertion of calcium-permeable AMPARs, respectively.^48,49 Silent synapses are important for metaplastic priming of brain regions, including the NAc, rendering them susceptible to more permanent changes in plasticity.^48,50,51 Cocaine has been shown to increase extracellular glutamate in the NAc^52,53 and potentiate accumbal glutamatergic synapses.⁵⁴ Surprisingly, in light of these synaptic changes that predict enhancements in glutamatergic synapses in the NAc, in vivo firing rates of NAc neurons are depressed following long-term cocaine exposure.^55,56 This indicates that changes in excitatory transmission may not translate directly into enhancements in accumbal firing, but may alter how accumbal neurons translate glutamatergic input into spiking. Intriguingly, reversing glutamatergic plasticity onto NAc neurons can reverse locomotor sensitization to cocaine, suggesting that this plasticity is a necessary feature of this behavioral change.^57,58 Thus, the modification of glutamatergic corticostriatal synapses by drugs of abuse can alter both striatal function and drug-adapted changes in behavior.

Cocaine can also alter LTP in the hippocampus, one of the glutamatergic inputs to the NAc. Self-administration of cocaine in rats enhanced in vitro hippocampal LTP in cornu ammonis area 1 (CA1),⁵⁹ which was sustained across 1 week of withdrawal.^59,60 In contrast, LTP was suppressed in CA1 following 3 months of withdrawal.⁶⁰ Enhanced synaptic transmission (as measured by input–output curves in CA1) was seen in rats 3–5 weeks following abstinence from cocaine self-administration.⁶¹ This suggests that cocaine alters hippocampal synaptic plasticity and that these changes can persist over long periods of time, although the specifics of these synaptic changes may depend on the dose and time course of cocaine exposure.

Finally, cocaine and other drugs of abuse are associated with dysfunction of frontal brain regions,⁶² which also send projections to the dorsal striatum and NAc. Prefrontal activity in chronic cocaine users is reduced during withdrawal, leading to the hypofrontality hypothesis of substance use disorders.⁶³ Men with alcohol use disorder exhibited lower prefrontal cortical activity (measured by functional magnetic resonance imaging (fMRI)) during a visuospatial task relative to healthy controls.⁶⁴ Furthermore, hypofrontality may persist after drug use ceases––relative glucose metabolism in frontal brain regions, as measured by positron emission tomography (PET) with [¹⁸F]-fluorodeoxyglucose, was lower in cocaine abusers 3–4 months after drug use than in healthy controls.⁶⁵ Similar effects were seen in people with alcohol use disorder over a week after alcohol withdrawal.⁶⁵ Together with alterations in corticostriatal plasticity, a persistent decrease in frontal activity may contribute to long-lasting behavioral changes in people with substance use disorder.⁶⁶ Specifically, these include cognitive changes and impulsivity that can persist after cessation of drug use.^67,68

Glutamatergic plasticity in obesity

Compared with the vast literature linking drugs of abuse to corticostriatal plasticity, relatively little is known about how these synapses change in obesity. Despite a relatively small number of studies, high-fat diet (HFD) exposure has been linked to disruptions in synaptic plasticity in the NAc (Fig. 1A). An in vitro study found that MSNs from obesity-prone rats fed HFD exhibited a mild synaptic potentiation in response to a stimulation protocol that normally would induce a synaptic depression, while MSNs from obesity-resistant rats fed HFD exhibited only a modest depression.⁶⁹ This suggests that exposure to a diet with higher fat content can alter NAc plasticity.⁶⁹ Accumbal MSNs from obesity-prone rats fed a junk food diet also exhibited an increased prevalence of calcium-permeable AMPA receptors in the membrane,⁷⁰ similar to what is seen in the NAc after exposure to cocaine.

Figure 1.

(A) Schematic showing how obesity induces neural mechanisms that lead to behavioral changes in both humans and rodents. (B) Flowchart of proposed biomolecular changes in the striatum during obesity or after HFD that lead to altered synaptic function. (C) Cartoon illustrating the cyclic nature of obesity. Obese states lead to changes in synaptic plasticity that can persist despite weight loss. This contributes to physiological and behavioral factors contributing to relapse and ultimately the cyclic nature of weight gain and loss.

Electrophysiological studies that suggest HFD affects synaptic plasticity in the NAc are bolstered by investigations of biomolecular changes in response to HFD. For instance, brain-derived neurotrophic factor (BDNF) plays an important role in structural synaptic changes⁷² and susceptibility to compulsive drug seeking.^73,74 Consumption of HFD increased BDNF in NAc tissue lysates as measured via western blotting.¹¹ This suggests that biochemical aspects of reward signaling pathways are also altered after HFD consumption. Rats fed saturated fat had increased protein levels of phosphorylated accumbal AMPA receptors, as measured by western blot, relative to rats fed a diet rich in monounsaturated fat or low-fat controls.⁷⁵ The insertion and phosphorylation of calcium-permeable AMPARs are thought to contribute to the induction of LTP,^76,77 further suggesting that HFD can engage aberrant plasticity mechanisms in the NAc (Fig. 1C). Interestingly, the rats in this study did not become obese on these diets, demonstrating that some effects of HFD can occur independent of obesity. These studies demonstrate that NAc adaptations can occur independent of weight gain.

There is also evidence of alterations in hippocampal plasticity following HFD exposure. In vivo⁷⁸ and slice⁷⁹ electrophysiological recordings from rodents show that high-frequency stimulation (HFS) of perforant pathway induces significantly lower field excitatory postsynaptic potentials (EPSPs) in dentate gyrus granular cells in overweight animals fed HFD relative to controls. Recordings in CA1 in response to Schaeffer collateral stimulation showed similar disruptions; field EPSP slopes from overweight mice fed HFD were decreased relative to controls in response to Schaeffer collateral theta burst stimulation⁸⁰ or HFS.⁸¹ However, these effects may be transient; one study that showed decreased hippocampal LTP in obese mice fed HFD saw that this phenotype was reversed after switching to a diet lower in fat.⁷⁹ In addition, it is difficult to reconcile why experiments with cocaine reported an enhanced ability to induce synaptic plasticity in the hippocampus,^59–61 while studies with HFD found the opposite.

Finally, obesity has been linked to alterations in prefrontal cortical (PFC) activity. In general, there appears to be a reduction of activity in frontal cortical areas in people with obesity, consistent with the hypofrontality hypothesis of substance use disorders.^63,82 In humans, PET scans showed women with obesity had lower cerebral blood flow in dorsolateral PFC in response to a meal (following a 36-h fast), compared with lean or formerly obese women.⁸³ This suggests that obesity may blunt frontal PFC activity in response to salient food stimuli. In the absence of a food stimulus, PET and single-photon emission-computed tomography (sPECT) scans revealed an inverse relationship between glucose metabolism and body mass index (BMI)⁸⁴ or cerebral blood flow and BMI⁸⁵ in prefrontal cortical areas, respectively. However, these studies were conducted with relatively few subjects (n = 8–12 people per lean, obese, or formerly obese group,⁸³, n = 21 subjects,⁸⁴ n = 36 subjects total⁸⁵) or only found differences in PFC activation in females.⁸³ Furthermore, other imaging studies in humans with obesity have reported increased prefrontal cortical activity relative to controls after consuming a satiating meal⁸⁶ or in response to images of highly satiating food.⁸⁷ PFC activity in children or adolescents with obesity has also been inconsistent. One fMRI study reported decreased prefrontal cortical blood flow in response to consumption of a glucose-rich beverage in adolescents with obesity relative to lean adolescent subjects.⁸⁸ Another reported higher PFC activation in response to pre-meal pictures of food in adolescents with obesity relative to controls.⁸⁹ Together, these studies suggest alterations in PFC metabolism in obesity, but it may be heterogeneous and not present in all individuals.

Dopamergic changes in substance use disorders

Dopaminergic circuits encode or incentivize reward or reward-learning,⁹⁰ and a rich body of literature has shown that substance abuse alters dopaminergic neurotransmission.^91–94 Exposure to drugs of abuse, including morphine, ethanol, cocaine, or methamphetamine induces morphological,^95–98 electrophysiological,^38,99 biomolecular,^100,101 and plasticity^98,102–104 changes in midbrain VTA dopaminergic neurons and their projections to the ventral striatum.

Drugs of abuse have many effects on the dopaminergic system and the accumbens.^105–107 Cocaine and other drugs of abuse increase extracellular dopamine in the NAc,^108–111 which is central addictive behavior in animal models.^109,112 Cocaine inhibits accumbal dopamine reuptake, the mechanism by which it increases extracellular dopamine levels.^22,113 Cocaine withdrawal is associated with decreased dopamine receptor availability.⁸² A seminal PET study showed decreased striatal dopamine receptor type 2 (D2R) availability in cocaine abusers after 1 week of detoxification,¹¹⁴ which returned to baseline after a month. A larger cohort of cocaine users showed that decreased D2R binding persisted as long as 4 months.¹¹⁵ Decreased D2R availability was also observed in individuals with alcohol use disorder.¹¹⁶ Animal work largely supports these human imaging studies. After cocaine self-administration, rodents displayed decreased D2R messenger ribonucleic acid (mRNA) levels, as measured by in situ hybridization¹¹⁷ and decreased D2R protein in the NAc.¹¹⁸ Similar results were observed via autoradiography in monkeys.¹¹⁹ Together, there is evidence across multiple species that cocaine use reduces the availability of striatal D2Rs.

In addition to the acute effects of dopamine on circuit function, dopamine regulates plasticity mechanisms of glutamatergic synapses.¹²⁰ Therefore, these drug-induced changes in dopamine transmission can affect subsequent glutamatergic plasticity. Some of the earliest work studying the effects of cocaine on midbrain synapses showed that a single in vivo exposure of cocaine was sufficient to alter AMPA channels at excitatory synapses in the VTA.³⁸ Furthermore, cocaine increases dopaminergic terminal density and striatal spinogenesis,^121,122 and cocaine withdrawal increases surface-level AMPA receptors in the ventral striatum.¹²³ Related changes in dopamine transmission have also been reported following obesity and HFD exposure.

Dopaminergic changes in obesity and following HFD

It is less clear how dopamine neurotransmission is disrupted in humans with obesity. The DRD2 Taq1A allele has been linked to reduced D2R availability in humans and increased likelihood of obesity.^124–126 Consistent with these findings, early evidence linked obesity to decreased D2 receptor availability.^127–129 Wang et al.²⁷ explored the striatum using PET scans for [C-11]raclopride. Later, it was reported that individuals with severe obesity (average BMI of 51.5) also exhibited lower D2R binding in prefrontal cortical regions.¹²⁸ Additionally, when D2R binding was examined in 33 subjects (BMIs ranging from 19 to 35) weak negative correlations between D2R and BMI were seen in the left caudate nucleus, right ventral striatum, and amygdala.¹²⁹ This suggested that the severity of decreased D2R may depend on the degree of obesity. A reduction in striatal D2R availability was also reported in a sample of women with obesity, relative to controls,^130,131 yet here without evidence of negative correlations between BMI and D2R availability. However, more recently, this point has been disputed. While Guo et al.¹³² reported a negative correlation between D2R binding and BMI in the ventromedial striatum, they found significant positive correlations between BMI and D2R binding in other striatal areas. This is supported by several other studies reporting positive or nonexistent correlations between BMI and striatal D2R availability.^133–136 PET scans measuring D2R binding with [¹⁸F]fallypride in women showed a positive correlation between BMI and D2R binding in the caudate.¹³³ Similarly, combined D2R and D3R binding was positively correlated with BMI in the right dorsal caudate of 12 individuals as measured by PET scans with [¹¹C]propyl-hexahydro-naphtho-oxazin ([¹¹C]PHNO).¹³⁴ D2R binding as measured by [¹¹C]n-methyl-benperidol ([¹¹C]NMB) PET scans did not correlate with the BMI of 30 individuals,¹³⁵ nor did D2R binding seen in the PET scans of 27 women using the radioligand [¹¹C]raclopride.¹³⁶ Together, this indicates that the link between D2Rs and obesity is complex and may occur differently across varying striatal compartments or different subject populations. Of interest is the idea that the relationship between D2Rs and obesity may be nonlinear, such that reductions in D2R are only seen in people with severe obesity.¹³⁷ Alternatively, the relationship may depend on subsets of people with obesity, such as people with different levels of emotional¹³⁸ or opportunistic eating.¹³² These subsets may not have been equally represented across studies.

Other changes in dopamine neurotransmission have been linked to obesity in humans. Postmortem studies of human striatum revealed an inverse correlation between BMI and dopamine transporter (DAT) binding, as well as decreased DAT and tyrosine hydroxylase (TH) mRNA in the substantia nigra of people with obesity.¹³⁹ Another study of 50 subjects supported this point, reporting highly significant negative correlations between striatal DAT binding and BMI.¹⁴⁰ However, these findings have been contradicted by others, who showed no correlation between BMI and striatal DAT binding.^141–143 Thus, while the specifics of dopamine receptor availability and dopamine transporter surface expression differs between studies, a large body of evidence supports the conclusion that dopaminergic circuits are altered in humans with obesity (Fig. 1).

Experiments conducted using animal models also suggest that diet-induced obesity alters striatal dopamine signaling.¹⁴ Compared to the complexity seen in the human literature, animal studies have consistently reported decreases in dopamine D2Rs in obesity. For instance, D2R binding^144–147 and expression²⁵ were reduced in obese rodents fed HFD^144,147 and other rodent models of obesity.^145,146 Adolescent rats maintained on a high-sugar diet also had lower dopamine receptor type 1 (D1R) and D2R expression in the NAc.¹⁴⁸ Dopaminergic receptor gene expression levels (D1R, D2R) were reported to be decreased in the VTA, NAc, and prefrontal cortex of overweight mice fed HFD from adolescence to adulthood,¹⁴⁹ an effect that recovered in VTA and prefrontal cortex, but not NAc, following 4 weeks of withdrawal from HFD.¹⁴⁹ Further work that examined D2R binding at an earlier time point (3 weeks after HFD) showed an increase in D2R striatal binding in overweight mice fed HFD,¹⁵⁰ suggesting a time-dependent effect of HFD on dopamine receptor expression. Protein homogenates of the NAc of rats fed a diet high in saturated fat for 8 weeks during adulthood showed increased D1R levels via western blot quantification, despite no difference in weight gain relative to controls,⁷⁵ suggesting that D1R translation may be differentially regulated depending on specific fat components of the diet, the age of exposure to HFD, or both. Intriguingly, some changes in ventral striatal dopamine transmission appear to occur even in the absence of weight gain. For example, in the absence of weight gain, rats exhibited decreased D2R protein levels and dopamine reuptake in the ventral striatum after HFD.^151,152 This indicates that these changes may not be secondary to the increased storage of fat. Furthermore, this highlights an important issue in the literature, namely that studies rarely parse the effect of HFD exposure versus. the effects from accumulated fat and changes in circulating factors. This would require studies to pair-feed control groups limited amounts of HFD, such that they receive all of their calories from HFD but do not gain weight.

Reports of extracellular dopamine concentrations in rodents maintained on HFD vary. Dopamine levels in the VTA and NAc were reported to be reduced in obese mice maintained on an HFD.^153,154 Using high-performance liquid chromatography (HPLC) assays, dopamine levels were also reduced in accumbal homogenates of obese mice maintained on an HFD.¹⁵³ Obesity-prone rats fed HFD for 5 days exhibited lower dopamine levels in the NAc as measured by in vivo microdialysis and subsequent HPLC.¹⁵⁴ In obese rats fed a cafeteria diet, NAc extracellular dopamine levels and amplitude of evoked dopamine responses decreased relative to chow controls.¹⁵⁵ However, a different group reported no difference in extracellular dopamine levels between rats fed HFD and controls, yet dopamine turnover––assessed by normalizing concentrations of 3,4-dihydroxyphenylacetic acid (DOPAC), a dopamine metabolite, to extracellular dopamine levels–– decreased in the striatum.¹⁵⁶ This effect was not observed in mice maintained on an HFD for 18 weeks, after which extracellular striatal dopamine, DOPAC, and the ratio between them were no different from control mice.¹⁴⁷ Finally, other studies reported that HFD increased extracellular dopamine in the accumbens of obesity-prone rats as measured by microdialysis¹⁵⁷ and that peri-adolescent exposure to HFD increased burst firing in the VTA and subsequent extracellular dopamine levels in the NAc.¹⁴⁸ On the basis of the variance in this animal work, conclusions regarding the effect of HFD on extracellular dopamine levels in rodents remain difficult to draw.

In contrast to the variance seen in studies of extracellular dopamine, HFD appears to negatively regulate surface expression of ventral striatal dopamine transporters. Fast-scan cyclic voltammetry experiments in the NAc showed that obese mice on an HFD did not show normal insulin-enhanced dopamine clearance and thus exhibited inhibited DAT function,¹⁵⁸ while high-speed chrono-amperometry experiments showed delayed clearance of injected dopamine in the striatum of obese rats fed HFD.¹⁵⁹ Inhibited dopamine transporter insertion into the membrane appears to occur after HFD even in the absence of weight gain,¹⁵² supported by decreased DAT protein levels seen in NAc lysates following exposure to saturated fat in the absence of weight gain.⁷⁵ The apparent inhibition of DAT function after HFD is similar to what has been documented following the effects of cocaine; cocaine causes delayed presynaptic dopamine transport, and repeated exposure over time reduces the ability of cocaine to inhibit accumbal synaptic DATs.¹⁰⁶ This highlights further potential parallel mechanisms spanning drug abuse and obesity and brings to light the question of what mechanistic deficiencies underlie both states.

These effects of HFD on dopamine transmission may alter dopamine-dependent striatal glutamatergic plasticity. As discussed above, drugs of abuse, including cocaine and alcohol, increase dopamine release into the striatum, and over time result in diminished dopamine receptor availability and inhibited dopamine reuptake.^114–116 This disrupts normal synaptic plasticity in projections from the prefrontal cortex to the NAc.¹⁶⁰ The disruptions in dopamine release, binding, or reuptake seen in rodents fed HFD^{144–147,149,151} suggest a role for HFD in altering D2R-dependent plasticity mechanisms²⁹ (Fig. 1B). Such changes may underlie the changes in physical inactivity and food choices that accompany obesity (Fig. 1A).

Opioid changes in substance use disorders

Opioid receptors have been implicated in regulating intake of multiple drugs of abuse. For example, alcohol interacts with the opioid system,^161,162 and alcohol drinking in rats was bidirectionally gated by injection of μ-opioid receptor agonists or antagonists into the ventral pallidum.¹⁶³ Furthermore, intraventricular administration of a κ-opioid receptor antagonist attenuates anxiety-like phenotypes exhibited by mice following alcohol withdrawal, while κ-opioid receptor agonist administration induces anxiety-like behaviors similar to those seen in alcohol withdrawal.¹⁶⁴ Similarly, stress-induced cocaine seeking in rats was precluded with treatment with a κ-opioid receptor antagonist.¹⁶⁵ These studies suggest that opioid receptors may gate certain behavioral aspects of substance abuse.

Opioid receptors are also involved in the regulation of glutamatergic synaptic plasticity. Administration of μ-opioid receptor antagonists hindered the induction of LTP at mossy fiber–CA3 terminals in the hippocampus.¹⁶⁶. In vitro recordings in the dorsal striatum showed that endogenous opioids, including met-enkephalin and dynorphin-A, as well as μ- and δ-opioid agonists, induced long-term depression (LTD),¹⁶⁷ suggesting that opioid receptors modulate corticostriatal plasticity per se by facilitating synaptic depression. Recent work suggests a specific role for dynorphin in the control of LTP; optogenetic-facilitated release of dynorphin in striatal slices following theta-burst stimulation of synaptic potentiation impaired LTP in the dorsal medial striatum.¹⁶⁸ Studies examining the effects of μ-opioid receptor agonists in striatum suggested that μ-opioid receptors gate excitatory and inhibitory inputs to NAc MSNs in a cell type–specific manner.^169,170 κ-Opioid receptors enhance excitatory inputs from the amygdala to D1R-expressing MSNs in the NAc and inhibit excitatory input from amygdala and ventral hippocampus to accumbal D2R-expressing MSNs,¹⁷¹ suggesting that κ-opioid receptors control the balance of excitation and inhibition to the NAc in a cell-specific and projection-specific manner. These studies suggest that opioid receptors regulate glutamatergic synaptic plasticity. Furthermore, opioid receptors may control both the onset and persistence of obesity through their regulation of food intake as well as synaptic plasticity.

Opioid receptors can also regulate drug-induced synaptic changes. Alcohol-induced LTD between fast-spiking interneurons and MSNs in the dorsolateral striatum is dependent on δ-opioid receptors and was attenuated in the presence of the antagonist naloxone.¹⁷² κ-Opioid receptor antagonists reversed the effects of stress on LTP of γ-aminobutyric acid (GABA)-ergic synapses in the VTA and prevented reinstatement of cocaine seeking.¹⁷³ Finally, stress-induced changes in hippocampal metaplasticity were prevented by administration of a κ-opioid receptor antagonist.¹⁷⁴ Synaptic plasticity and various drug-seeking behaviors are differentially regulated by opioid receptors, raising the question as to whether these receptors also gate circuit and behavioral changes associated with obesity.

Opioid changes in obesity

Opioids have a complex interaction with food intake, and specifically the intake of palatable diets. It has been known for decades that stimulation of endogenous μ-opioid receptors in the NAc promotes feeding.^175,176 A genome-wide association study also revealed that adolescent humans carrying the minor OPRM1 allele were more likely to eat less fat and have a lower body fat percentage.¹⁷⁷ Supporting this, systemic administration of naltrexone, a nonselective opiate receptor antagonist, reduced the consumption of palatable foods in rats without altering intake of chow.¹⁷⁸ Consistently, gene knockout mice lacking μ-opioid receptors exhibited normal feeding when presented with a low-fat diet, yet were resistant to obesity when given access to an HFD.¹⁷⁹ Together, these results suggest that μ-opioid receptors are necessary for fat appetite. However, men¹⁸⁰ and women¹³⁶ with obesity have reportedly lower levels of accumbal μ-opioid receptors, which is difficult to reconcile with this point. In addition, others have reported that μ-opioid receptor gene knockout mice gained weight on a chow diet, relative to control mice,^181,182 indicating that μ-opioid receptors may have effects on food intake independent from fat content. There has been less work investigating how the other classes of opioid receptor relate to feeding, but evidence suggests that κ-opiates can also drive intake of fat. Injecting a κ-opioid receptor agonist into the lateral ventricle of rats caused them to increase intake of high-fat but not low-fat food.¹⁸³ Consistently, κ-opioid receptor gene knockout mice had lower body weights than controls after being maintained on an HFD.¹⁸⁴

While the role of opioid receptors on feeding and obesity has been well documented, the effects of HFD on opioid receptors are not well understood. Obese rodents fed HFD or cafeteria diet exhibited reduced μ-opioid mRNA in the accumbens and in the VTA and prefrontal cortex,^185–188 possibly due to increased methylation or heterochromatic location of the μ-opioid receptor.^185,188 A study in rats showed that one week of HFD exposure increased the binding of naloxone, an opioid receptor antagonist, in tissue homogenates of the cortex and midbrain.¹⁸⁹ Interestingly, Alsio et al.¹⁹⁰ showed that access to a palatable, high-fat/high-sugar diet in the absence of weight gain did not alter NAc μ-opioid receptor mRNA, but that ad libitum access to such a diet (which caused weight gain) decreased NAc μ-opioid receptor mRNA. Other studies reported no difference in μ-opioid receptor binding in the NAc of rats who gained weight on a diet including 33% condensed milk.¹⁹¹ Further work aimed at elucidating the mechanisms underlying opioid receptor control of feeding and HFD, and importantly of the role of HFD in modulating opioid receptors, is needed to understand potential opioid contributions to obesity.

Changes in dopamine persist after weight loss

Multiple studies suggest that an HFD has persistent actions on dopamine release and transmission in striatum, even after weight loss. Rodent exposure to HFD increased body weight and depleted striatal dopamine levels, which persisted despite weight loss¹⁹² or decreased further^11,153 after the rodents were switched back to chow. HFD-induced obesity decreased dopaminergic gene expression in the NAc.¹⁴⁹ This effect appeared to persist partially, with D1R expression levels recovering to baseline after diet was switched back to chow, and D2R gene expression remaining low.¹⁴⁹ Mice on an HFD had lower TH protein levels in the NAc relative to mice fed a low-fat control diet; when these HFD mice were switched to a normal chow diet, the low accumbal TH protein levels persisted.¹¹ While more studies are needed to understand the time course and persistence of these changes, there are several indications that dopaminergic changes can persist following withdrawal from HFD and subsequent weight loss. To the extent that these changes drive behavior and energetic regulation, they may contribute to the persistence of obesity.

Many researchers have also looked at the persistence of behavioral adaptations following withdrawal from obesogenic diets. Female rats given a “junk -food” high-carbohydrate, low-protein diet had lower break points in an operant task for sucrose reward that persisted for 9 days after the diet was withdrawn.¹⁹³ Mice maintained on an HFD spent less time in the open arm of an elevated plus maze (interpreted as a high anxiety score), and this was exacerbated after withdrawal from HFD.¹¹ However, rats put on a long-term monounsaturated or saturated fat diet did not show increased anxiety-like behavior,¹⁹⁴ and these rats did not gain weight. Together, this suggests that HFD-induced obese states contribute to anxiety-like behavior, but the consumption of fat alone is not anxiogenic.

Human studies also support a role for persistent changes in dopamine function and physical activity after weight loss. Two PET studies showed lack of recovery of D2R binding after gastric bypass and associated weight loss.^195,196 However, a report from another group demonstrated a partial recovery of D2R binding 6 weeks after bariatric surgery.¹⁹⁷ Similarly, objective assessments of vigorous physical activity 6–12 months after bariatric surgery show that, despite weight loss, physical activity levels remained low in these individuals.^10,198,199 Animal studies have supported these observations. Loss of adiposity correlated with further decreases physical activity in nonhuman primates²⁰⁰ and dogs.^201,202 Further work is needed to investigate the circuit changes that may underlie these persistent behavioral effects, but it is possible they reflect persistent changes in dopamine function.

Lastly, persistent effects seen in offspring exposed to HFD in utero or in those born to mothers with obesity also support a lasting role for these––at times transient––conditions. Experiments in nonhuman primates showed Japanese macaques exposed to HFD in utero had lower D1R and D2R protein levels and a decrease in TH positive terminal fibers in the prefrontal cortex, as measured by immunohistochemistry,²⁰³ and that maternal obesity (but not maternal HFD) resulted in increased offspring body weight 7 months after weaning.²⁰³ Rats exposed to HFD in utero exhibited impaired adaptive dopamine signals in the NAc in response to repeated pinch stress²⁰⁴ and decreased levels of NAc dopamine in response to a palatable food reward compared with rats born to lean mothers.²⁰⁵ Rats exposed to a high-fat, high-sugar diet in utero exhibited increased lever presses to achieve a palatable reward relative to those born to chow-controlled mothers,²⁰⁶ Together, these studies suggest that exposure to HFD during development can alter dopamine transmission later in life, which may increase the vulnerability of the offspring to obesity. While exposure to HFD during fetal development is much different than exposure during adolescence or adulthood, it underscores how changes caused by short-term exposure to HFD or obesity may have long-lasting impacts on neural circuits and behavior.

Potential mechanisms of persistent effects of obesity

There are multiple possible contributing factors involved in the sustained effects of obesity or obesogenic diets. Factors such as triglycerides and free fatty acids are elevated following HFD exposure, and fat itself can alter synaptic plasticity.²⁰⁷ Evidence of this is found in multiple brain regions. The presence of acutely applied triglycerides disrupted the maintenance of in vitro hippocampal LTP,²⁰⁸ suggesting that acute exposure to fat can disrupt synaptic potentiation. Recordings in rodent hippocampus after exposure to HFD and subsequent obesity showed disruptions in LTD²⁰⁹ and reductions in synaptic efficiency.^80,209 Recent work showed that bath application of monounsaturated fatty acid decreased both neuronal firing of VTA dopaminergic neurons in slice recordings and action potential–independent excitatory release onto those neurons,²¹⁰ suggesting that fat itself may affect VTA projections to the NAc. Together, this supports the hypothesis that fat, or components of fat including fatty acids, can alter synaptic plasticity. Subcellular mechanisms underlying changes in plasticity may include disruptions in pathways involving the holoenzyme protein kinase A (PKA), which is important for LTP^211,212 and LTD regulation.²¹³ During LTP, AMPA channels are driven into the membrane,²¹⁴ in part due to activity-dependent PKA phosphorylation of glutamate ionotropic receptor AMPA subunit 1 (GluA1) at serine 845.²¹⁵ Rats fed saturated fat diets in the absence of weight gain showed increases in accumbal protein markers known to be targets of PKA: phospho-dopamine and cAMP–regulated phosphoprotein 32 (DARPP32) and phospho-serine845-GluA1.⁷⁵ Mice fed HFD became obese and exhibited increased PKA activity in the hypothalamus.²¹⁶ Together, these studies suggest that HFD and obese states alter PKA-based molecular pathways in the brain, which may be particularly important for synaptic plasticity in the striatum.

Persistent effects of HFD may also include epigenetic or gene expression changes, which have been implicated in altered synaptic plasticity following exposure to drugs of abuse.²¹⁷ Mice fed HFD exhibited altered gene expression in the cortex,²¹⁸ and RNA-sequencing data from an Alzheimer’s mouse model following exposure to HFD showed an upregulation of immune-linked genes and a downregulation of genes ontologically linked to synaptic transmission in the cortex.²¹⁹ A study examining DNA methylation after HFD in mice showed that the μ-opioid receptor (MOR) gene had enriched methylation in tissue from the VTA, NAc, and PFC of mice fed HFD and was more likely to be located within inactive chromatin, suggesting that HFD acted to epigenetically suppress MOR transcription.¹⁸⁵ This may relate to human findings of reductions in MOR binding in striatum of subjects with obesity.^136,180 Furthermore, protein levels of the canonical transcription factors phosphorylated-cAMP response element binding protein (CREB) and δ-FBJ murine osteosarcoma viral oncogene homolog B (deltaFosB) were increased in the amygdala and NAc of mice fed HFD, respectively.¹¹ CREB has also been implicated in the influence of drugs of abuse over cellular plasticity.¹⁰⁵ Overexpression of CREB in the NAc induced aversion to cocaine²²⁰ and reduced preference for morphine²²¹ in rats in a conditioned place- preference task, while expression of a dominant-negative mutant CREB enhanced preference for both cocaine and morphine.^220,221 This suggests that regulation of transcription factors, such as CREB, may underlie subcellular changes common to exposure to drugs of abuse and obesogenic diets. Recent work showed that, in the absence of weight gain, rats fed monounsaturated or saturated fats exhibited altered transcription profiles of stress-related genes, including corticotropin releasing hormone (CRH) and its associated receptor CRH receptor 1 (CRH-R1), in the paraventricular nucleus (PVN) of the hypothalamus as well as amygdalar nuclei.¹⁹⁴ PVN CRH transcription levels positively correlate with susceptibility to social stress.²²² Together, these findings suggest a mechanism for persistent changes in gene expression of multiple brain circuits following HFD.

Additionally, obesity-linked alterations in insulin and leptin resistance may underlie molecular mechanisms contributing to sustained changes in the dopaminergic system. Obesity causes prolonged increased secretion and deficient clearance of insulin, which ultimately leads to insulin resistance.^223,224 Through phosphoinositide 3-kinase (PI3K) and renin–angiotensin system (Ras) pathway activation, insulin binding in the brain is involved in DAT surface expression,²²⁵ suggesting that sustained insulin exposure due to insulin resistance may disrupt DAT expression. Reduced DAT expression has been linked to obesity in humans^139,140 and HFD in rodents,⁷⁵ though there have been conflicting reports on this point.^141–143 Indeed, obese rats fed HFD for a month exhibited elevated plasma insulin levels indicative of insulin resistance and also showed reduced striatal DAT expression,¹⁵⁹ similar to the observed insulin resistance, impaired blood glucose clearance, and reduced dopamine clearance prevalent in mice fed HFD for 6 weeks.¹⁵⁸ Inhibition of protein tyrosine phosphtase 1B––and subsequent promotion of phosphorylation of insulin receptor tyrosine residues––promoted the insulin receptor signaling cascade and effectively rescued dopamine reuptake,¹⁵⁸ supporting the link between insulin resistance and changes observed in the dopamine system. Even in the absence of obesity, mice fed high-fructose corn syrup (HFCS) for several months had impaired blood glucose clearance and impaired dopamine release in the dorsal striatum.²²⁶

Finally, a large literature indicates that obesity is associated with inflammation in the hippocampus²²⁷ and impaired cognition or hippocampal-dependent memory.^228,229 Broadly, sustained inflammation alters synaptic plasticity and disrupts neural homeostasis,²³⁰ and recent work suggests that HFD-induced obesity caused blood–brain barrier leakage²³¹ and may increase markers of immune response in the brain,²¹⁹ even acutely in the absence of weight gain.²³² Elevated levels of the inflammatory molecules interleukin 6 (IL-6) or tumor necrosis factor α (TNF-α) inhibit LTP in rodent hippocampus.^233,234 Persistent inflammation is associated with depression,²³⁵ a comorbidity of obesity that involves the ventral striatum and prefrontal cortex. Obesity is also associated with inflammation in the hypothalamus,^236,237 which projects to the VTA, further implicating the circuitry that is involved in reward and depression. Though hypothalamic inflammation appears to recede with weight loss,²³⁸ increased inflammation during obesity may affect hypothalamic to VTA plasticity and in turn alter VTA to NAc projections.²³⁷ Persistent effects of inflammation may be another mechanism whereby obesity disrupts neural circuits, even after weight loss.

Conclusions and ongoing questions

Understanding how synaptic plasticity is linked to obesity may reveal new insights into the challenges of both weight loss and weight maintenance. However, many outstanding questions remain. For instance, given the heterogeneity of human food consumption, it is difficult to attribute changes in weight to the presence or absence of any particular food substance. A similar problem exists in rodent studies, where experiments are often conducted using different formulations of HFD. Thus, the question of what elements of a diet contribute most strongly to weight gain or loss remains controversial. With regard to plasticity, it is possible that certain types of fat are more likely to alter synaptic plasticity mechanisms than others. Finally, fat itself is an endocrine organ, and the accumulation of fat during obesity alters the concentration of many circulating factors in the blood. Therefore, there is a need to decouple the effects of HFD and obesity to identify whether specific components of the diet are likely to alter synaptic plasticity mechanisms. This can be done with pair feeding that allows animals to gain all of their calories from an HFD without becoming obese.

While there is widespread acknowledgement that obesity is associated with reductions in physical activity, the mechanism underlying the link between the two remains unclear. We have reviewed evidence that HFD alters dopamine binding, which in turn controls synaptic potentiation and depression. This leads to the question of whether overconsumption of fat disrupts striatal dopamine binding or transmission and whether this is a main contributor to decreases in physical activity that accompany obesity. Physical activity is one of the best predictors of human health, and unraveling the mechanisms that cause inactivity may be necessary to increase activity in people with obesity.

Perhaps the most pertinent question on this topic is whether persistent changes observed in people with obesity can be reversed. There is some evidence of D2R recovery after food restriction in rats,¹⁴⁶ as well as recovery of D2R binding in human striatum following gastric bypass.¹⁹⁷ This suggests that dramatic changes in diet may be one option for treating persistent effects of obesity. Additionally, a study in mice demonstrated increased D2R availability in striatum after chronic wheel running,²³⁹ suggesting that high-intensity physical activity may aid in the recovery of striatal dopamine dysfunction after obesity. However, this has not yet been examined by other researchers or in humans.

Changes in basal ganglia circuitry in obesity may be like the circuit alterations caused by drugs of abuse and may lead to some behavioral adaptations that resemble substance use disorders. Such changes may persist in people who have lost weight and may underlie susceptibility to relapse for compulsive eating and weight gain. Understanding these circuit alterations may lead to new treatments for obesity and to better understanding of why obesity is so difficult to reverse.