Obesity has been associated with prominent changes in dopamine transmission [1,2,3] and cognitive domains that are crucial for adaptive behaviour, such as motivation, decision-making, reinforcement learning and working memory [4,5,6,7,8,9]. Importantly, food-related but also general non-food-related cognitive differences have been recently highlighted in obesity [10,11,12,13,14,15,16]. However, animal studies contributed to the understanding that obesogenic diet, rather than adiposity itself, actually causes observed differences in dopamine transmission [17, 18••, 19,20,21].

Excellent obesity-related reviews have focused on either DA transmission [22], or cognition with little or no emphasis on the relation with dopamine [7, 11, 15]. Others have argued for dopamine-mediated cognitive changes in obesity [1,2,3, 8, 9, 23,24,25,26,27] paralleling findings from addiction research, with controversial opinions towards the existence of food addiction [22, 23, 28,29,30,31]. Here, we argue for a more detailed assessment on the relationship between dopamine and the variety of possibly dopamine-mediated cognitive changes in obesity.

Some reviews suggested a major role of reward function in obesity, which resonates well with abundant evidence for striatal dopamine alterations in both animals and humans. Others have argued for a deficit that is predominantly mediated by the prefrontal cortex (PFC) (e.g. [10, 32]), which is not well investigated regarding a direct link to the dopaminergic system yet. Both perspectives might mirror different angles of the same changes as dense anatomical connections exist between frontal and striatal regions. These connections are organised in functionally relevant frontostriatal loops that are strongly modulated by dopamine. This makes it important to examine the cognitive literature in obesity in the light of these frontostriatal loops.

Here, we will address the following open questions: First, do findings from animal and human studies on dopamine changes converge, given the different methodologies available to study the dopaminergic system in these species? Second, does a comprehensive picture emerge regarding major obesity-related cognitive differences and their possible association with frontostriatal loops? Third, can these differences be regarded as global, i.e. affecting several cognitive domains, or are they specific to, e.g. the food context? Fourth, to what extent are diet-induced dopamine differences responsible for the observed differences in cognitive domains? And fifth, what are candidate mechanistic links between diet and central dopamine?

In this review, we will summarise recent findings of obesity and diet-related differences in dopamine transmission, in particular in the striatum, from human and animal studies. We will then describe the different frontostriatal loops, followed by an evaluation of obesity-related differences in the sensory input to this circuit. Finally, we will discuss the neurocognitive profile of obesity within the theoretical framework of frontostriatal loops. We will point out major gaps in the literature, as well as challenges that need to be overcome in order to get at the heart of the role of dopamine in diet-induced obesity.

Obesity and Diet-Related Dopamine Differences

In humans, structural changes in the dopamine system can be imaged most directly with positron emission tomography (PET) using radioactively labeled ligands that bind to a specific substrate. Such studies have revealed obesity-related differences, in particular related to D2-receptor (D2R) (Appendix Table 1). D2R binding has been found to correlate positively to BMI in normal-weight to obese individuals in several studies [49, 52, 94], but not in others [57, 59, 71]. In the latter study, dopamine release did correlate positively with BMI in putamen and substantia nigra [59]. Another PET study revealed a positive relationship with BMI in the dorsal and lateral striatum, whereas a negative relationship was observed in the ventral striatum [53]. These, together with other inconsistent results concerning D2R binding in human obesity, have been proposed to reflect a quadratic relationship between severity of obesity and striatal D2R availability, or rather a U-shaped relationship with dopamine tone [2]. The idea of lower tone in overweight to mildly obese individuals is supported by a recent [18F] DOPA PET study [62•]. However, striatal dopamine transporter (DAT) binding as measured with single-photon emission computed tomography (SPECT) did not relate to BMI in a sample of normal-weight to severely obese participants [95], nor to self-reported ad libitum food intake in normal-weight participants [90•]. Importantly, due to the cross-sectional design of most human studies, it is debated whether or not the observed differences in dopamine transmission in humans are cause or consequence of obesity.

Rodent models allow for the investigation of more causal links between obesity and dopamine transmission and have begun to disentangle the effects of an obesogenic diet from adiposity. Short-term and chronic high fat diets (HFDs) as well as diet-induced obesity were shown to reduce D2R-mRNA and protein expression levels ([96,97,98,99], but see [100]) (Table 2). An elegant study suggests that diet-induced obesity may be the cause rather than the result of reduced D2R availability [18••]. The authors found decreased striatal D2R binding after a chronic HFD, despite unchanged D2R-mRNA or protein levels, which could be explained by receptor internalisation. D2R binding also did not predict weight gain, nor did deletion of D2Rs in striatal neurons increase risk for obesity. Overexpression of striatal D2Rs in development has been shown to causally relate to diet-induced obesity through its effect on energy expenditure and thermogenesis [103]. Effects of sugar on the dopamine system have also been observed, although the findings diverge. One study showed increased D2R-mRNA expression and decreased D2R-protein levels in the nucleus accumbens [104], whereas others found the opposite for the striatum as a whole [100]. This may be due to the specific striatal regions under study or the diet composition (for more details on diet composition and duration, see Appendix Table 2). Moreover, chronic HFD may also lead to reduced D1R-mRNA in the rat striatum ([97, 101], but see [18••, 105]) depending on diet composition [106]. One study showed reduced D1R signalling when a diet high in saturated, but not monounsaturated fats (palm oil vs. olive oil), was administered [101]. Finally, diet-related changes in dopamine synthesis [108], release [59, 102, 105] and uptake (DAT) [98, 99] have been observed.

In sum, excessive weight and chronic exposure to an obesogenic diet has been associated with changes in striatal dopamine transmission in humans and rodents, in particular related to D2Rs. The wide variety of observed diet-induced dopamine changes described above suggests a complex interplay between obesogenic diet and the different parts of the dopamine pathway. Future research is required to get a clearer picture of diet-induced dopamine changes in both obese and non-obese individuals and to unravel the mechanisms through which an obesogenic diet exerts its action on the dopamine system (see Box 1 for potential mechanistic links). To translate findings from animal studies to human studies, it is crucial to increase convergence between the two fields. First, human studies with dietary interventions and dopamine measurements are needed. Second, diets in animal intervention studies should be more carefully designed in terms of composition and duration to mimic human obesogenic diets. Finally, a recent review emphasises the importance of the element of dietary choice in modelling the characteristics of human diet-induced obesity [109].

figure a

Frontostriatal Loops

The striatum shares major connections with regions in the frontal cortex that are strongly modulated by dopamine and subserve adaptive behaviour. Across species, frontostriatal connections are organised in anatomically and functionally segregated loops [139,140,141,142] (Fig. 1). These loops are often grouped into three functionally relevant categories: (1) the affective loop between the ventromedial prefrontal cortex/orbitofrontal cortex (vmPFC/OFC) and nucleus accumbens (NAc) in the ventral striatum is important for motivational control, (2) the cognitive loop between the dorsolateral prefrontal cortex (dlPFC) and caudate nucleus (CN) and anterior putamen (aPUT) in the dorsomedial striatum is important for cognitive control and (3) the motor loop between the (pre) motor cortex (PMC/M1) and posterior putamen (pPUT) in the dorsolateral striatum is important for motor control. The loops also share substantial connections to enable information transfer from ventromedial to dorsolateral loops [141,142,143]. The different striatal regions may, in addition, receive input from prefrontal areas outside of their loop as was recently shown in a primate study [143], enabling cross-talk between all loops.

Fig. 1
figure 1

Schematic overview of the frontostriatal circuit and their relationship with external and internal signals as discussed in the main text. Striato-nigro-striatal connections that can serve as the interface between the loops are also displayed. The affective loop (red) consists of connections between vmPFC/OFC and NAc and is predominantly modulated by dopamine projections from VTA, the cognitive loop (yellow) of connections between dlPFC and CN/aPUT and is modulated by dopamine projections from VTA and SN, and the motor loop of connections between PMC/M1 and pPUT and is predominantly modulated by dopamine projections from SN. The color gradient between the loops reflects the shared connections that enable cross-talk between all loops. Dopamine projections are highlighted by thick grey arrows. External signals that affect the frontostriatal circuit in the context of food include visual, olfactory and gustatory sensory signals (left panel, in blue). Adaptations in sensory sensitivity (left white arrows) and subsequent processing, or cue reactivity (right white arrows), have been associated with obesity. Important internal signals that can affect the circuit include nutrients, inflammatory factors, and hormones such as leptin, insulin and ghrelin (right panel, in red). Adaptations in leptin, insulin and active ghrelin (in fasting state) have been observed in obesity, as well as in central sensitivity to these hormones (white arrows). Leptin, insulin and ghrelin can modulate the frontostriatal circuit through action on receptors in striatum, in VTA directly, or indirectly via hypothalamic control of VTA

Dopamine-dependent plasticity is thought to underlie reward-based learning in all loops [141], although dopamine has also been proposed as the interface between the loops through the striato-nigro-striatal connections [142]. As reviewed above, obesity- and diet-related changes in dopamine have been consistently shown in the ventral and dorsal striatum. However, dopamine is suggested to not only modulate learning and motivated behaviour through its effect on striatal output, but also by its effect on input coming from the PFC or sensory regions, as discussed next. Evidence of obesity-related dopamine changes in PFC is scarce, due to difficulty imaging prefrontal dopamine (but see a new PET development by [144, 145•]) and due to the ongoing debate on what are the rodent homologues of the prefrontal cortex [146].

Dopamine Modulation of Sensory Inputs

Dopamine modulates sensory perception and processing [43••, 91, 92]. Since sensory perception, in particular visual, gustatory and olfactory perception, influences when, how much and what we eat [147, 148], dopamine might play a key role in influencing food choice via this route as well.

There is solid evidence that obesity is accompanied with alterations in especially the gustatory and olfactory systems. However, as the results are contradictory, the direction and interpretation of this relationship remains unclear. For instance, some researchers show higher gustatory sensitivity [38, 149]; others lower gustatory sensitivity or no difference in obese when compared to normal-weight individuals [39, 150, 151]. Similarly, it has been shown that obese participants had higher [152] or lower olfactory sensitivity [36, 37].

Of note, the olfactory and dopamine systems are highly interconnected and are therefore of particular interest in the context of eating behaviour and obesity. It has been shown that food odour perception activates dopaminergic brain regions in a human fMRI study [93]. Moreover, dopamine neurons have been detected in the olfactory bulb [153] and a reduction of dopamine neurons induced olfactory impairment in an animal model [154]. Greater availability of DATs in the caudate nucleus and putamen, as measured with SPECT in healthy human individuals, has been associated with higher olfactory performance [92]. Interestingly, one study found evidence of decreased dopamine uptake in the caudate nucleus in Parkinson’s disease (PD) patients with and without olfactory impairments [89]. Although PD patients constitute a particular clinical group, the results support the link between dopamine function and olfactory perception.

Beyond affecting perceptual aspects in sensory systems, dopamine modulation of sensory signals may also serve as adjusting the sensory input for the affective frontostriatal loop. As such, dopamine can affect higher-order processing of sensory input, such as hedonic value of, and cue reactivity to visual, gustatory and olfactory input. It has been shown that obese compared to normal-weight individuals perceive food odours as more pleasant [152] and have a higher food cue reactivity in response to food pictures [155] and odours [44]. There is solid evidence that increased physiological (e.g. salivation, skin conductance, neural activation) and craving responses to food-related stimuli are associated with both food consumption and body weight [138]. Such food stimuli also potently activate the brain’s reward system, i.e. the striatum and PFC value areas [23, 25, 26, 93].

In conclusion, dopamine can modulate sensory input and the subsequent processing of sensory information. While enhanced reactivity of the limbic system to visual and olfactory food cues, which is likely mediated by dopamine, is quite well established in obesity, general differences in sensory sensitivity require further investigation. Whether a diet high in fat and sugar leads to similar changes through its effects on dopamine, independent of obesity or only as a result of excessive adiposity and the associated metabolic changes (Box 1), is an open question for future investigation.

Neurocognitive Profile of Obesity

Apart from distinct reactions to food and food cues, obesity has been associated with a wide variety of higher-order neurocognitive differences that crucially rely on dopamine action in all parts of the frontostriatal loops [1, 9] (Table 1). In specific tasks, impairments in executive skills including attention, (working) memory and learning [10,11,12,13,14,15,16] are consistently shown in obesity. In addition, reductions in cognitive flexibility [7] and increases in several types of impulsivity [156,157,158,159] seem to characterise obesity. Such general cognitive features are likely the roots that feed maladaptive decision-making in a food context.

Food Reward Responses and General Reinforcement Learning

When it comes to reward processing of food and non-food stimuli, brain regions associated with the affective loop, i.e. NAc and mPFC/OFC, are particularly involved [160], and activation patterns in these regions can depend on metabolic state [161]. As discussed above, enhanced reactivity to sensory food cues is typically observed in obesity. In simple reaction tasks where (often hypothetical) food rewards are anticipated, enhanced activation in the affective loop is observed in obesity [70], which may be mediated by decreases in leptin and insulin sensitivity (see also Box 1). When a food reward is received, however, hypoactivation is often reported (e.g. [8, 25]). Interestingly, Kroemer and Small [8] elegantly show how obesity-related hypoactivation of reward regions in response to reward receipt may be explained by impairments in general reinforcement learning, similar as proposed for substance addiction [162].

Reward-related learning within the frontostriatal loops critically relies on dopamine-dependent plasticity [141], which may go beyond reward learning and extends to associative learning ([161], preprint). The reinterpretation of the findings in a learning framework supports the idea that obesity is related to general rather than food-specific differences, and dovetails with the widespread role of dopamine in motivation, cognition and behaviour. General reinforcement learning differences in obesity, be it impairments [5, 6, 58, 61, 74] or improvements ([4], Kube et al., under revision), are indeed suggested based on evidence from non-food and food-related reinforcement learning tasks. Difficulties with integrating negative feedback may be central to observed impairments [5, 6, 58], which could result in insensitivity to the negative consequences associated with obesity.

Food-Related Attentional Bias and Craving

What happens once food-related stimuli have been registered and led to initially enhanced responses in terms of activation of affective frontostriatal regions, invoking craving, or attracting attention? In case no food is actually available or you are trying to break a habit of giving into temptations, disengaging your attention or regulation of craving may be necessary. Obese individuals show an enhanced attentional bias to food cues across different experimental paradigms and measures ([6, 58, 61, 74], but see [163]), which may be due to difficulty disengaging from such stimuli. Food attentional bias has been linked to striatal DAT binding, although no relationship was observed between DAT binding and craving or ad libitum food intake [90•]. Furthermore, glucose intake enhanced attentional food bias in obesity [41] and intra-individual variability in a similar bias measure was stronger in obesity [64], supporting the dynamic nature of attentional bias [164•]. Food attentional bias can be attenuated by cognitive factors such as a healthy mindset [87], which emphasises the importance of cross-talk between the frontostriatal loops. Regulation of craving has been associated with differential activation in the putamen and functional connectivity between the putamen and dlPFC [50], also spanning the loops.

Self-Control and Cost–Benefit Decision-Making

What if food is available? Then self-control may be needed, which again requires cross-talk between the affective and cognitive loops. Exercising self-control in a food choice task involved dlPFC activity in dieting human participants, which correlated with vmPFC activity [86]. In a similar task, Medic and colleagues [67] found no evidence for a difference in vmPFC activity for overweight to severely obese participants, whereas vmPFC activity did predict subsequent consumption of, particularly, unhealthy foods at a buffet. At the level of the striatum, reduced NAc food cue reactivity was also associated with successful self-control of eating behaviour in daily life in dieting female students, as measured with experience sampling [165]. Experience sampling is a promising method that can be used in obesity research to link neurocognitive findings to maladaptive decisions in daily life. The right food choice always depends on your current state and situation, but also on possibly conflicting internal goals. Decreased goal-directed control of behaviour in a food context has also been associated with obesity [54, 55].

Food often comes at a cost, which requires weighing your options. Obese individuals may be less willing to, first, pay money for plain than highly palatable food items [73]; second, exert effort to obtain food or monetary rewards [66, 166]; and third, wait for a larger reward if a smaller immediate reward is offered simultaneously, as consistently observed in delay discounting tasks ([167,168,169,170], but see [51•]). Willingness to exert effort relies on regions in the affective loop and is particularly interesting because of its link to dopamine [171, 172] as well as low-grade systemic inflammation [173], which is highly prevalent in obesity (see Box 1). A recent PET study using a highly specific D2 tracer [51•] together with measures of insulin sensitivity revealed no difference in willingness to wait (nor striatal D2R -binding) in obese relative to non-obese participants. However, greater D2R availability in obese was associated with less willingness to wait and reduced insulin sensitivity. This raises the questions whether and how striatal D2R binding and metabolic factors interact to affect temporal discounting in obesity. In another study, lower willingness to wait (i.e. steeper temporal discounting) was associated with reduced dlPFC–vmPFC connectivity in obesity [169] and may thus rely on cross-talk between the affective and cognitive loop. Interestingly, thinking about how the larger later rewards can be used (i.e. future thinking) has been effective in reducing temporal discounting and food intake in obesity [174,175,176,177,178,179] and involves cognitive control areas ACC and dlPFC, as well as mPFC–hippocampus interaction [180, 181]. Of note, diet effects have been consistently shown in the hippocampus [11].

Behavioural Control and Action Inhibition

It can also occur that an action has already been initiated upon perceiving a palatable food stimulus. The decision to stop such a response can happen at the level of behavioural control in the cognitive loop, or gating of motor responses. Investigations of response inhibition, tapping into the latter, revealed small obesity-related differences [45]. However, in a resting state fMRI study, disruption was observed in motorcortico-striatal networks in obesity consistent with habit formation theories [182]. More evident results have been observed using the go/no-go task, which indicated impaired performance in obesity [183]. A behavioural intervention that trained no-go responding to high-calorie food cues led to devaluation of those items in normal-weight [184, 185] and in morbidly obese participants [34, 186], as well as impulsive food choices in normal-weight individuals [184, 185]. The authors have proposed that training acts bottom-up by creating associations between no-go food items and stopping responses and reducing valuation of no-go food items (in the affective loop) [187]. A similar mechanism may explain a reduction of food intake in uncontrolled eaters after inhibitory control training [82] and of approach bias to unhealthy food cues in obesity after training automated action tendencies [42, 188].

In sum, obesity-related differences are predominantly observed in the affective and cognitive frontostriatal loops. A simple explanation could be that studies on motor gating or learning are lacking. Many of the neurocognitive constructs investigated in obesity rely on cross-talk between the loops. Studies implementing tasks that specifically investigate the interplay between the different loops may help us further. Also, more convergence in the use of experimental stimuli and task parameters in food-related neuroimaging is needed (as argued by [181]), and there is a need of inclusion of a wider BMI range in obesity studies (see Table 1). However, a more mechanistic explanation is also plausible. That is, the more ventral and medial parts of the frontostriatal circuits may be particularly vulnerable for the effects of diet and adiposity-related metabolic factors in the bloodstream. It is important to better understand the mechanism of the cross-talk between loops. That is, where in the loops can interaction occur (e.g. is the motivational signal from the affective loop to higher-order cognitive loop, or can the cognitive loop affect the state of the affective loop?) and at what point in the process can maladaptive decisions be prevented from being made?


Both diet and obesity affect dopaminergic transmission. However, site and direction of effects are inconsistent across species and studies. Non-specific changes are observed spanning all frontostriatal loops, from sensory input to motivated behaviour. Given the impact of peripheral signals on central dopaminergic signalling and the interaction between the frontostriatal loops, modulation of dopamine likely propagates through all loops and, thus, affects behaviour on various levels of complexity. In line with [112], we highlight in Box 1 that homeostatic factors have direct access to hedonic systems via dopaminergic modulation, indicating that these can be highly interdependent, going against the historical, dichotomous concept of homeostatic vs. hedonic control over eating behaviour. However, in this review, we mostly focused on the hedonic system. Interactions between the hypothalamus and the frontostriatal circuits require further investigation.

Despite the wealth of literature, it has proven difficult to evaluate the degree of convergence of findings between animal and human studies on the role of dopamine in diet-induced obesity. The main reason is the lack of studies utilising overlapping measures of dopamine and cognition in both species. Human studies are largely observational in nature and lack direct measures of dopaminergic transmission. As such, there is a great need for diet intervention studies, more longitudinal studies [13] and mechanistic studies on the relationship between dopamine and the observed neurocognitive differences, preferably linked to metabolic factors as discussed in Box 1. Further, although some attempts have been made [109], the usage of animal diets that do not closely resemble human obesogenic diets limits comparability of the effects of diet exposure, and higher-order cognition is often not studied in relation to diet-induced dopamine changes.

Due to the narrow scope of the current review, some aspects should be highlighted that were not directly addressed but are likely of high relevance for understanding diet-induced obesity. First, this review dealt with the relationship between diet-induced obesity, cognition and dopamine transmission. Although playing a central role in motivation and cognition, dopamine is not the only neurotransmitter involved. In fact, dopamine interacts closely with other systems such as the opioid, serotonin and noradrenergic system [57, 71, 189, 190]. A relationship to obesity has been demonstrated for all of them. Second, most human research is cross-sectional in nature. In addiction, the existence of subsequent behavioural phases has been proposed, leading from incentive-guided towards compulsive behaviour with accompanying central changes that indicate a transition of changes from ventromedial to dorsolateral frontostriatal loops [191]. Here, we cannot tell whether group differences similarly relate to a transition-in-progress or resemble an endpoint (although a vicious cycle model has been proposed by [11]). We do not even know whether or not overweight people are prone to obesity or represent a special “subpopulation”. This transitional aspect could be addressed in animal studies that longitudinally monitor changes following diet exposure or obesity induction. It would also be of interest to take into account the severity of obesity and the individual history of being obese in human studies. Moreover, inconsistent results in the literature could be related to not assessing important latent variables such as the genetic background or possible epigenetic edits induced by lifestyle or family history. Although common variation in dopaminergic genes seems not to have a direct relationship to obesity [24], its relation to cognition is well established. Thus, ignoring this information may lead to either false positives or negatives assigned to the obesity factor.

Finally, an intriguing open question that deserves attention in future research is whether or not the changes that are observed in diet-induced obesity are really maladaptive in nature. Whereas physiological, behavioural and neural differences are often interpreted as maladaptive, it may be that some actually reflect functional adaptations that could be beneficial either at the individual or population level. This would call for a more nuanced interpretation of any obesity-related differences.