Background
Autism spectrum disorders (ASD) are a group of neurodevelopmental disorders defined by substantial impairments in social interaction and communication, as well as patterns of rigid or repetitive behavior, with onset in the first three years of life [
1]. Social impairment is a central feature of ASD and is a primary target for neurobehavioral experimental studies. Much of this work has been pursued in the context of investigating differences in social perception and cognition, including the brain's detection and encoding of social information, attention to social stimuli, face recognition, and discrimination of social cues such as facial expression or gaze direction (for review, see [
2]). Considerably less attention has been given to investigating the neural basis of possible differences in social motivation in ASD.
It has been suggested, however, that the social impairments seen in ASD may result from aberrant limbic mediation of the reward that typically drives social interaction. The rewarding nature of social attachment and social interactions [
3] has led to speculation that neural reward mechanisms that typically reinforce and perpetuate social behavior are either dampened in ASD or recruited by non-social stimuli such as the objects of circumscribed interests or idiosyncratic sensory stimuli to which individuals with ASD may show intense attraction. However, it remains to be tested whether the affective basis of social deficits in ASD reflects aversion or simply lack of motivation (Thompson, B.L., personal communication), which may then implicate distinct but overlapping limbic circuits for avoidance (fear, disgust) or approach (reward). If the latter, it is unknown whether diminished motivation in ASD is limited to the reward of social stimuli or is a more generalized trait [
4‐
6].
The hedonic experience of pleasure depends on endogenous opioid signaling in the ventral tegmental area (VTA) of the brain [
7], which sends dopaminergic projections to the nucleus accumbens (NAc). The role of the NAc is to mediate the performance or work involved in reward seeking and anticipation [
7‐
10]. These subcortical areas project reciprocally to the ventromedial prefrontal cortex (VMPFC) and orbitofrontal cortex (OFC) [
11], which form associations between the sensory features of the reward stimulus and its hedonic value [
12] through inputs from sensory cortices of every sensory modality to OFC [
13]. In addition, the VMPFC and OFC regions compute expected reward versus reward outcomes to shape future behavior (Grabenhorst and Rolls, [
14]). The insula is important for monitoring and evaluating the impact of external stimuli on internal states [
15,
16] and the amygdala is involved in evaluating emotional stimuli for their novelty [
17], affective significance [
18,
19], and biological or behavioral relevance [
20,
21]. Separate but overlapping circuits and neurotransmitter systems mediate the hedonic ('liking') and the anticipatory ('wanting, craving') experiences of reward [
9,
22‐
24].
Palatable food is a potent stimulus for the reward system [
23,
25‐
27], as are food cues such as images of food [
28,
29]. The neural reward network hemodynamic response to food images is tightly correlated with reward sensitivity [
30], and increases with the caloric content of the foods pictured [
31] and with hunger motivational state (fasting versus satiated) [
32]. Goldstone
et al. [
33]) noted an interaction between these two variables, such that the heightened response to high calorie versus low calorie foods was greater when fasting and concluded that hunger biases the neural reward system toward high calorie foods. Behavioral evidence corroborates this, as healthy adults under fasting conditions exhibit increased gaze duration to food images [
34] and increased attentional capture by food images, resulting in decreased performance on a target detection task despite monetary incentives for accuracy [
35]. These studies converge to suggest that images of high-calorie, palatable foods under fasting conditions constitute an effective stimulus that elicits response from neural reward networks.
Studies of the neural basis of reward in ASD have focused on contrasting social versus non-social (monetary) rewards, which have been found to have highly overlapping neural substrates [
36]. Studies comparing ASD to typical control groups largely find diminished response to both social and monetary rewards [
37,
38]. Scott-Van Zeeland and colleagues noted significantly diminished response of the ventral striatum, anterior cingulate, and ventral prefrontal cortex, especially for social reward. Reported differences are generally stronger for social rewards [
37,
39] than for monetary ones. Using only monetary reward, without a contrast to social reward, Schmitz
et al. [
40] demonstrated an elevated blood oxygenation level dependent (BOLD) signal in the anterior cingulate in response to reward feedback in ASD. These discrepant results could be influenced by several variables that differed between studies, including the developmental stage (children versus adults) of the participants.
Although monetary reward paradigms are well established in their ability to recruit reward circuitry in typical adults, they may not be as ideal for individuals with ASD, who often do not manage their own money [
41,
42] and may have differences in abstract or symbolic representation even at the higher end of the spectrum [
43]. If this is the case, it may be that the 'generalized' reward system differences seen in these studies were due to the choice of non-social reward, rather than a truly generalized deficit in reward system functioning in ASD.
A recent study by Dichter
et al. [
44] provides more information with which to address the question of alternative reward stimuli in ASD by contrasting monetary rewards with non-social objects as rewards. In this study, objects were selected to have a high likelihood of representing restricted interests in ASD (that is, images related to commonly held interests such as electronics or trains). Thus, this study was an important step in modifying reward paradigms to include stimuli that are known to be visually salient and behaviorally rewarding for individuals with ASD [
45,
46]. Results revealed decreased BOLD response in reward regions in response to monetary incentives, corroborating the findings of Scott-Van Zeeland
et al. [
37] and Kohls
et al. [
38]; however, for object images, individuals with ASD showed increased reward system BOLD responses relative to controls. These findings provide support for a model of a 're-directed' neural reward response, that is, a neural response to reward that is intact but responds to different stimuli than in typically developing individuals, rather than a generalized reward deficit in ASD.
The use of a monetary reward as a comparison condition in each of these studies, however, imposes a limitation on their interpretation. Specifically, diminished response to monetary incentives in ASD may reflect generalized, intrinsic differences in neural response to reward, or it may reflect differences in the perceived reward value of money in this population. Monetary reward-specific differences could result from a diminished ability of people with ASD to attribute value to an abstract symbolic representation [
43] or even a lack of financial autonomy [
41,
42] that could impact the perceived value of monetary winnings. In the current study, response to primary reward (food) cues is investigated to address this potential confound and provide more clarity on the responsiveness of the reward system in ASD to nonsocial cues known to be rewarding in typical adults. Because monetary reward studies have shown relative sparing of nonsocial reward compared to social reward and because reduced response in these paradigms may at least partially reflect other cognitive or economic factors, we hypothesized that individuals with ASD would show similar patterns of BOLD response in brain reward regions to a comparison group of typically developing controls in response to images of palatable foods, reflecting intact reward processing for a nonsocial primary reward.
Discussion
Little is known about the neural basis of response to primary reward in ASD. As a first step, we probed the reward system using images of appetizing foods for children under conditions of mild fasting, a paradigm which has been demonstrated previously to recruit neural reward networks [
32,
59]. Our findings demonstrate that neural reward system response to food cues is not only intact, but may even be enhanced in children with ASD. This was found despite the well-known elevation of food selectivity in children with ASD [
60‐
62] and diminished gustatory discrimination ability in ASD [
63]. The foods we chose to depict were specifically targeted to be palatable to children, and were exclusively high-calorie foods, with images representing both sweet and savory flavors. High calorie foods have been demonstrated to be potent activators of neural reward circuitry [
64,
65]. Foods that were represented heavily in this stimulus set (for example, starchy foods, chicken nuggets, chocolate, pizza) were consistent with parent reports of food preferences for children in our sample, supporting the notion that our food images were appealing to children across groups.
Although both groups showed increased BOLD response to food images in a similar network of regions known to mediate reward, when comparing the ASD and TD groups directly, we found greater response in the ASD group in the insula and anterior cingulate cortex (ACC), known for their roles in assessing interoceptive states [
15,
66], and evaluating and preparing for response based on the motivational significance of these states [
67], respectively. These two regions are frequently co-activated in fMRI studies, and have been found to constitute a resting state network (the 'salience' network) [
68,
69]. The ACC has been demonstrated to be hyperactive in previous neuroimaging studies of reward in ASD [
40,
44]. The degree of connectivity between the insula and ACC at rest has been shown to be related to autistic traits in the general population [
70].
The insula and ACC have been postulated by Craig [
71,
72] to constitute an integrated system of emotional perception and action, analagous to primary sensory and motor cortices. Included in Craig's model is the unique concentration of von Economo neurons in these two regions, which he proposes form the basis for rapid communication between them despite their physical separation. A recent neuroanatomical study reported a higher proportion of Von Economo neurons to pyramidal neurons in the insulae of their sample with ASD relative to controls [
73] and the authors theorized that this neural difference might give rise to heightened interoception. Our result of enhanced response in insula and ACC to food cues in ASD may thus suggest that children in the ASD group were more attuned to an internal state of hunger or food craving elicited by the pictures than controls.
The work of Craig and others has demonstrated a posterior-anterior gradient of interoceptive representation within the insula, with posterior regions responding to objective features of the stimuli themselves and more anterior regions to more subjective assessment of their emotional significance [
74,
75]. It is of note that our comparison of the ASD > TD contrast revealed three distinct clusters of significantly higher response in the insulae of the ASD group, distributed along this axis (Figure
1b). This suggests that they may have experienced both stronger signals of hunger or 'wanting' the food in the images, as well as a more intense emotional reaction to these interoceptive signals. The role of the insula in integrating interoceptive sensation with reward evaluation in the context of reward-motivated behavior such as drug craving is currently being actively investigated [
76,
77].
The insula is responsive to visual food cues [
29,
78] and is also the site of the primary gustatory cortex, although recent studies provide evidence that a more accurate characterization is a multimodal oral sensory region that integrates taste with other sensory features such as texture and temperature [
79]. While the primary taste cortex occupies the most anterior region of the insula in nonhuman primates [
80,
81], it is positioned further posterior in humans [
82]. The most anterior portion of the human insula has been hypothesized to have evolved more recently along with increased human capacity for self-awareness [
71,
72]. Although not statistically significant, the positive correlation of the BOLD response in the insula with parent reports of food cravings and preferences is consistent with the known function of this region. Further work is needed to explore the differences in insula response in ASD exhibited in the current study. The lack of significant correlation between BOLD response in these regions and ADI-R scores summarizing clinical severity of ASD may suggest that the enhanced response in these regions is unrelated to core features of ASD, or it may reflect a lack of power to detect a relationship, possibly due to small sample size and/or the diagnostic rather than quantitative nature of the ADI-R algorithm.
Conclusion
Despite an aberrantly enhanced response in the insula and anterior cingulate in the ASD group, the orbitofrontal cortex, nucleus accumbens, and amygdala were similarly responsive in both groups, although we noted slight differences in the laterality of response in the nucleus accumbens and OFC. Thus, all nodes in the neural reward circuit are responsive to primary reward in ASD, suggesting that social deficits are not explainable by a generalized under-responsiveness of the reward system.
This study is a first step in assessing neural response to primary rewards in ASD although much more work needs to be done to fill in remaining gaps. Although children in both groups fasted for the same minimum amount of time, subjective hunger ratings and/or hedonic ratings of food images would be an important variable for future studies to collect and report. Further, our paradigm did not allow us to separate motivational from hedonic aspects of food reward. Additional fMRI studies incorporating an anticipatory phase and actual palatable food delivery, or utilizing behavioral paradigms that confer the ability to separate 'liking' from 'wanting' (for example, [
83]), should be undertaken in the future. An important next step will also be to directly compare food reward with social and object reward cues, to provide a clearer picture of the reward system as a whole in ASD. Finally, application of neuroimaging and reward paradigms to younger children and/or at-risk sibling groups will facilitate the translation of this knowledge into new approaches for early identification and intervention in ASD. The current finding of enhanced response to primary reward advances our understanding of the similarities and differences in the brain's response to rewarding stimuli in ASD; this understanding will ultimately provide opportunities to harness the power of the reward system to optimize educational and treatment approaches in children with ASD.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
JF assisted with image analysis, participant recruitment and assessment, and helped to draft the manuscript. JH coordinated participant recruitment, scheduling, and scan acquisition. CN performed clinical assessments and provided expert clinical opinion for inclusion of participants in the ASD group. BR, MB and RC assisted with study design and interpretation. BR and AC assisted with image processing and analyses. CC conceived of the study, participated in its design, coordination, and analyses, and drafted the manuscript. All authors read and approved the final manuscript.