Introduction

Nearly a third of the US population meets the clinical criteria (body mass index (BMI) > 30 kg/m2) for obesity (Ogden et al. 2006), a condition that carries an increased risk of diabetes, cardiovascular disease, and certain cancers leading to increased morbidity and elevated health care costs (Flegal 2005; Flegal et al. 1998). Despite a culture that demands ever larger portion sizes and relies on fast food franchises for the family dinner, not all individuals become obese. One factor that may confer an increased obesity risk is eating in response to external food cues, rather than internal hunger signals (Carnell and Wardle 2008).

Neuroimaging techniques, such as functional magnetic resonance imaging (fMRI), provide a means to determine how obesity might be related to alterations in the brain’s response to such external cues. Specifically, sensory stimuli that routinely accompany food ingestion (i.e., the sight, smell, and taste of food) acquire Pavlovian properties that can elicit both a motivational drive to eat (Berridge 2007; Berridge et al. 2010), as well as initiate digestive processes (i.e., the “cephalic phase” of digestion) (Bellisle et al. 1985). Given the brain regions that support such motivational behavior, some have speculated about a significant overlap between obesity and drug addiction (Volkow and Wise 2005). That is, sensory stimuli associated with these high reward value foods may trigger the same reward system activated in drug-seeking behavior, eliciting increased activation in known reward substrates such as the ventral striatum, amygdala, orbitofrontal cortex (OFC), medial prefrontal cortex (mPFC), hippocampus (Berthoud 2006; Mela 2006), and dorsolateral prefrontal cortex (Hare et al. 2009). Differential activation of these substrates may serve to draw attention to food sources, and govern behaviors involved in food seeking and satiety.

Numerous studies have been conducted to evaluate neuronal activation following the presentation of visual food cues (i.e., photographs of appetizing foods). Rothemund et al. (2007) observed increased neuronal activation within the dorsal striatum, anterior and posterior cingulate cortices, and lateral OFC following the presentation of high-calorie food images, suggesting an increased role for brain areas associated with reward anticipation and motivation in obese women. A similar study in fasted women shown images of high-calorie foods reported increased activation in the medial and lateral OFC, as well as the mPFC, amygdala, anterior cingulate cortex (ACC), striatum, and hippocampus (Stoeckel et al. 2008). Martin et al. (2010) also used food images to elicit activation prior to and following a standardized meal. This study demonstrated greater pre-meal activation of the mPFC and ACC within the obese individuals. Activation of the mPFC remained significantly higher in obese subjects compared with their normal weight counterparts following food intake while the differences in ACC activation dissipated (Martin et al. 2010). The above studies all show a strong role of the brain's reward system in the evaluation of visual food cues, with greater activation typically observed in obese subjects.

While food images can be compelling, food aromas may be particularly effective cues given their prominent role in flavor perception (via retronasal olfaction), and thus their role in the sensory reinforcement that helps increase consumption (de Wijk et al. 2004; Fedoroff et al. 2003, 1997; Jansen et al. 2003; Yeomans 2006). Ventromedial and medial prefrontal cortices are regions where neuronal activity varies as a function of the perceived reward value of stimuli (Hare et al. 2009, 2008; Kable and Glimcher 2007). Our laboratory has accordingly demonstrated that both alcoholic drink (Bragulat et al. 2008; Kareken et al. 2010a, b) and food aromas (Bragulat et al. 2010) robustly activate these medial frontal regions.

In the data that we report here from an ongoing study, we examine how maladaptive eating patterns may be related to medial prefrontal responses to food aromas in states of both hunger and satiety. In particular, we hypothesized that individuals who are high in “externality,” those with strong responses to external food cues (van Strien et al. 1986), would have stronger blood oxygenation level dependent (BOLD) responses to food odors in brain reward regions.

Materials and Methods

Participants

Eighteen women from the community participated, 11 of whom were normal weight (BMI, 22.2 ± 2.1; age, 23.5 ± 4.4; mean ±SD) and seven of whom were obese (BMI, 36.3 ± 4.2; age, 29.5 ± 6.0; Table 1). For the purposes of this study, the groups were collapsed and no BMI-based differences were examined. None of the participants had evidence of Axis-I psychiatric disorders or known disorders of the brain. All were nonsmokers and performed within broad normal limits on a 20-item Smell Identification Test (Doty 1995), with a range of 16–20 and a mean of 18.0. Individuals were excluded if they were pregnant or breast feeding within the past 6 months, had a history of diabetes or a fasting blood glucose level of 126 mg/dl or greater. Individuals whose food preferences were inconsistent with the food aromas used or the meals provided were also excluded. All participants voluntarily signed informed consent statements approved by the institutional review board at Indiana University.

Table 1 Subject characteristics
Full size table

Procedures

All participants completed two non-contiguous study days: one “fasting day” when a lunch meal was withheld until after imaging, and one “fed day” when subjects ate a lunch meal to satiety before imaging. The order of the fasting and fed days was randomized, as was the lunch type (pasta with meat sauce or beef and noodles) consumed. On both study days, participants reported to the Indiana University Clinical Research Center (CRC) at approximately 6:30 a.m., whereupon weight and height were recorded, vital signs taken, and a urine screen for pregnancy and drug use was performed. At approximately 7:40 a.m., all participants consumed a standardized breakfast of turkey sausage, French toast (with margarine and syrup) and a fruit cup, with a choice of coffee, tea, diet soda, or water provided, per the subject’s usual habit. The quantities of the breakfast were adjusted to provide 20% of the participant’s daily caloric intake based on height, weight, age, and activity level.

Fasting Day

On fasting days, the participants remained in the CRC until ∼1:00 p.m., at which time they were escorted to the imaging suite, with lunch provided upon the subject's return to the CRC immediately after imaging (∼3:00–4:00 p.m.). Lunches consisted of an ad libitum amount of the predetermined meal type (pasta or beef and noodles). Participants were provided 30 min in the CRC to finish their lunch meal before being released.

Fed Day

On fed days, participants remained in the CRC following breakfast until lunch was provided at 12:15 p.m. Between breakfast and lunch, participants could engage in normal activities (reading, watching television, and computer use) except for sleeping. Lunches consisted of an ad libitum amount of the predetermined meal type (pasta or beef and noodles), with 30 min provided to finish lunch before leaving the CRC at 1:00 p.m. for imaging.

DEBQ

Subjects completed the Dutch Eating Behavior Questionnaire (DEBQ; Van Strien et al. 1986) on one of their two study days after breakfast and prior to imaging. The DEBQ consists of 33 items across three scales reflecting: emotional (psychosomatic) eating (eating in response to emotional distress), externality (eating in response to external food cues, such as food sight and smell), and restraint (overeating in the wake of attempted restraint).

Olfactory Stimuli

All odorants were delivered at a constant rate of 2.0 l/min using a computer-controlled, eight-channel air-dilution olfactometer as previously described (Bragulat et al. 2010; Bragulat et al. 2008; Kareken et al. 2004). In this study, two classes of odorants (International Flavors and Fragrances, Union Beach, NJ) were presented to all participants: (1) food-related odors of pasta and roast beef and (2) a non-food, Douglas fir odor. Odorant concentrations were prepared by diluting the concentrated odorant in 1,2-propanediol (Sigma-Aldrich, St. Louis, MO). The pasta odorant was prepared to a 20% solution, the roast beef odorant to a 5% solution, and Douglas fir odorant was presented undiluted. A Porex® polyethylene disc (0.475” diameter; 0.250” thickness Interstate Specialty Products, Sutton, MA) saturated with odorant was placed in the bottom of a glass vial, over which the olfactometer airstream was passed before being delivered to the participant via a small polytetrafluoroethylene tube.

Pre- and Post-MRI Hunger and Odorant Perception Assessment

Prior to entering the magnetic resonance imaging (MRI) scanner room, as well as following the imaging session, participants subjectively rated their hunger and the odorants’ psychophysical characteristics. Participants smelled each odorant in a random order through the olfactometer while simultaneously viewing representative images on a computer monitor (e.g., pasta odorant presented with an appealing photograph of a plate of pasta with meat sauce). Immediately following each odorant, participants rated general hunger, as well as hunger specific to each of the food items, on a vertically oriented, 100-mm labeled magnitude scale (LMS) (Cardello et al. 2005), with “greatest imaginable fullness” at the top, “greatest imaginable hunger on the bottom and “neither full nor hungry” in the center (see Fig 1). Following hunger assessment, participants again sampled each odorant through the olfactometer in a random order and rated perceived intensity, pleasantness, and “representativeness” (the extent to which an odor smelled like its intended target). Odorant intensity was rated on Green’s LMS, modified to be horizontally oriented with “Barely Detectable” on the left-hand end and “Strongest Imaginable” on the right-hand end (Green et al. 1996). Odorant pleasantness and representativeness were assessed on visual analog scales (VAS) of 1–9 in half point increments, with “very unpleasant” and “not at all representative” on the left-hand end and “very pleasant” and “very representative” on the right-hand end, respectively. Odorant perception prior to fMRI session was reassessed until each odorant was perceived by the participant as being within two intensity increments of the other odorants' perceived intensity, with adjustments in the air-dilution of the odorant made between each assessment (generally this required ∼2 adjustments).

Fig. 1
figure 1

Mean hunger scores for hunger type (general, pasta, and roast beef) assessed by Cardello’s LMS (Cardello et al. 2005) under both the fed (blue bars) and fast (red bars) session. The y-axis values indicate the labeled magnitude scale (100 mm in length) presented to the participants when rating hunger. All between-session comparisons resulted in significant (p < 0.001) differences, with greater hunger experienced in the fast session than in the fed session

Full size image

Intra-MRI Hunger Assessment

Subjective hunger was assessed at three time points during imaging; (1) prior to the first functional scan, (2) following the second functional scan, and (3) again after the fourth (last) functional scan. As above, participants were asked to rate their hunger in general, as well as their hunger specifically for pasta and beef and noodles on the same VAS presented outside the MRI. However, these intra-MRI assessments were not preceded by odorant presentation.

Activation Paradigm

Participants underwent four functional imaging scans of olfactory stimulation, with each odorant presented six times per scan in a mixed-event design that also included six presentations of the odorless control, for a total of 24 presentations per event type. Four distinct sequences of event presentations were generated by OptSeq2 (http://surfer.nmr.mgh.harvard.edu/optseq/) resulting in 8–18-s inter-stimulus intervals within 6:27 min long scans. Each sequence was run once per MRI session in a randomized order. To evaluate compliance, participants were prompted (via a 750 ms long, 750 Hz tone) to report an odorant’s presence (left button) or absence (right button) on a MRI compatible HHSC-TRK-1 trackball (Current Designs, Philadelphia, PA).

Image Acquisition

Imaging was performed on a Siemens (Erlangen, Germany) 3T Magnetom Trio-Tim scanner with a 12-channel head coil array. To facilitate anatomic localization of the functional data, a high resolution anatomic image was acquired using a 3D magnetization prepared rapid gradient echo (MPRAGE) MRI sequence (160 sagittal slices, 1.0 × 1.0 × 1.2 mm3 voxels, field of view (FOV) 256 × 256 mm, repetition time (TR) 2300 ms, echo time (TE) 2.91 ms, flip angle (FA) 9°, duration 9:14). BOLD-sensitive volumes were acquired with an echo-planar imaging pulse sequence (gradient echo, TR 2250 ms, TE 29 ms, FA 78°, FOV 220 × 220 mm, 39 interleaved 3-mm thick slices, 2.5 × 2.5 × 3.0 mm3 voxels, generalized auto-calibrating partially parallel acquisition factor of 2). 3D-PACE algorithm (Thesen et al. 2000) was enabled to adjust image acquisition in real time, accounting for head movement and significantly reducing motion-related artifacts. In addition, a deformable foam cushion stabilized head position. Participants were instructed to keep their eyes closed for the duration of each functional scan.

Image Processing and Analysis

Imaging data were preprocessed using standard SPM8 (Welcome Department of Imaging Neuroscience, University College, London, UK) procedures (slice-time acquisition correction, realignment, co-registration, and segmentation/normalization) after verification for artifacts and excessive motion. Each subject’s MPRAGE image was segmented into tissue classes and parameters of this non-linear transformation were utilized to convert participant’s structural MRI and realigned, co-registered functional image volumes into the Montreal Neurological Institute stereotactic space. The resulting normalized functional image volumes were interpolated to 2 mm per side isotropic voxels and smoothed by a 6 mm full-width at half-maximum isotropic Gaussian kernel.

Responses to discrete, 2-s periods of odorant (or sham) valve events in the post-processed image time series were convolved with a standard hemodynamic response function and its time and dispersion derivatives in a within-subject “first-level” analysis. The six movement parameters from realignment were included as regressors to account for residual movement-induced effects. A high-pass filter with a cut-off of 1/128 Hz was applied to each voxel’s time series to remove low frequency noise. This model estimated within-subject activation effects, with each food (F) or non-food (NF) odorant set contrasted against sniffing of an odorless control event (CO; i.e., sham valve opening that shunted the same air source without odorant delivery). This resulted in the following BOLD response contrasts: (F > CO) and (NF > CO), respectively. The comparison of food to non-food odors also included the odorless sniffing baseline (i.e., (F > NF) = (F > CO) > (NF > CO)). The MarsBar toolbox (Brett et al. 2002) was used to define functional ROIs, and to extract mean contrast values (activation) for each of the participants for the purposes of plotting activation as a function of externality.

Results

Odor Perception and Hunger

Odor Ratings

No significant group differences in odor perception were observed between normal-weight and obese participants (p > 0.05); therefore, all perception data were collapsed across groups. All 18 participants perceived the intensity of the pasta (16.6 ± 9.8), roast beef (17.3 ± 9.6), and Douglas fir (17.5 ± 9.8) odors to be similar. Participants also found the odors to be insignificantly different in pleasantness (pasta, 7.0 ± 1.0; roast beef, 7.2 ± 0.9; Douglas fir, 7.0 ± 1.5) and representativeness (pasta, 7.5 ± 1.1, roast beef, 8.1 ± 0.5 and Douglas fir, 8.1 ± 0.7).

Hunger Ratings

Hunger ratings were insignificantly different as a function of either BMI group (p > 0.10) or the type of odor present when making hunger ratings (p > 0.30). Therefore, hunger score data were collapsed across groups and all food odors for session analyses. Across all subjects, marked between-session differences in assessed hunger emerged for hunger in general, as well as for expressed hunger for a given food (pasta and roast beef, after smelling each), with significantly (p < 0.001) greater hunger in the fast session than the fed session (Fig. 1).

Food Consumption

Participants consumed an average of 592.3 g of food, with no significant difference in consumption between lean and obese participants (p = 0.47) and no significant difference in food consumed between fed (eating at a standard lunch time) and fasting days (eating 3–4 h after missing lunch; p = 0.40).

BOLD Activation

BOLD Activation of F Compared with NF

In this sample, brain activation did not differ as a function of either BMI group or session, assessed either from individual odor classes, or as a difference between odor classes. Pooled over both sessions, both food and non-food odors elicited significant activation in a number of limbic and reward-related areas (Table 2; Fig. 2a, b). These areas included the mPFC and ACC. Areas receiving olfactory bulb projections (piriform cortex and amygdala) also showed a prominent BOLD response, as did the hippocampus. When contrasted against non-food odors, food odors produced greater activation in the medial prefrontal cortex (Fig. 2c; peak, p = 0.001, at (−4, 40, 2), and in the left lateral orbital cortex, extending into the inferior insula (Fig. 2d; peak, p = 0.001 at (−40, 20, −18).

Table 2 Odorant effects across all subjects
Full size table
Fig. 2
figure 2

BOLD activation induced by sniffing odors in a mixed sample of 18 obese and normal-weight women under both fed and fast sessions. a Food (F) odors compared with sniffing odorless control events (CO; F > CO) contrast. b Non-food (NF) odors compared with (CO; NF > CO) contrast. c, d Direct comparison of food and non-food odors (F > NF) contrast. a–d Magenta box, medial prefrontal cortex; green box, ventromedial prefrontal cortex; blue ovals, left lateral orbitofrontal cortex; red circles, left and right amydala; black boxes, left and right hippocampus; purple circle, 6-mm sphere encompassing peak of significant medial prefrontal activation reported in (Bragulat et al. 2010). Results in (a) and (b) displayed at cluster-level significance, q clust(FWE) < 0.005, voxel-wise height threshold, p < 0.001 (uncorrected). Results in (c) and (d) displayed at p < 0.01 (uncorrected), cluster size, k > 25. Color bar indicates t statistic values

Full size image

DEBQ and Correlations

The sample's mean DEBQ score was 55.3 ± 17.3, with no significant difference observed between the normal weight and obese participants (p = 0.09). Mean scores on the restraint (21.4 ± 1.2) and externality (18.9 ± 2.7) sub-scales were not different between groups (p> 0.53), although a significant group difference was present in the psychosomatic sub-scale (obese = 25.0 ± 3.3, normal weight = 15.0 ± 2.8; t[16] = 2.30, p = 0.04).

Correlations between food odor-induced activation and the scores on the Externality subscale of the DEBQ were examined. Of particular interest were correlations with the (F > CO) BOLD contrast under both fed and fast conditions. The (F > CO) response correlated with externality only under the Fed condition (Fig. 3b), and in the ventral aspect of the mPFC cluster (vmPFC) activated by food odors (i.e., under the (F > CO) contrast in the fed condition; Fig. 3a). To illustrate the distribution of the effects, we extracted mean (F > CO) contrast values from the vmPFC cluster of significant correlation, as well as the contrast values from the same region in the (NF > CO) contrast. The positive correlation between (F > CO) and Externality observed in the Fed session (Fig. 4a) was absent in the fast session (Fig. 4b). There were no significant voxel-wise correlations between (NF > CO) and externality under either the fed or fast sessions (p > 0.05; Fig. 4c, d), and no significant voxel-wise correlations between either (F > CO) or (NF > CO) in other activating reward areas. Adding BMI as a covariate did not change the results.

Fig. 3
figure 3

a (F > CO) BOLD contrast in a mixed sample of 18 obese and normal-weight women under the fed session. Voxel-wise height threshold, p < 0.001 (uncorrected), cluster size, k > 50. b Positive correlation between DEBQ externality sub-scale scores and (F > CO) response under the fed session. Display threshold, p < 0.001 (uncorrected), cluster size, k > 10

Full size image
Fig. 4
figure 4

a Illustrative plot of correlation between externality and (F > CO) response, as extracted from the cluster of significant, positive correlation (threshold, p < 0.001) under the fed session. However, non-significant relationships between externality scores were observed with b (F > CO) response under the fast session, c (NF > CO) response under the fed session, and d (NF > CO) response under the fast session

Full size image

Discussion

As olfaction is a primary foraging mechanism across nearly all species, the aromas of ingested primary rewards should be particularly salient reward cues. Moreover, food aromas are routinely present during consumption and critical to flavor perception through the means of retronasal olfaction (Mozell et al. 1969; Murphy et al. 1977; Shepherd 2006). Thus, via Pavlovian conditioning, they become closely associated with food availability, foods’ inherently pleasurable sensations, and the rewarding sensations (or negative reinforcement) from hunger relief and satiety. In the present study, both food and non-food aromas led to BOLD activation in a number of brain regions associated with both reward and olfaction. Medial frontal cortex (including some ACC) activation was, however, largest to food odors. The location of this BOLD response is consistent with others’ observations of regional activation that correlates with the perceived value of a reward (Bray et al. 2010; Hare et al. 2009, 2011, 2008; Kable and Glimcher 2007). The OFC, also known to play a role in coding appetitive goal values (Arana et al. 2003; Chib et al. 2009; Hare et al. 2010; Hare et al. 2008; Paulus and Frank 2003; Plassmann et al. 2010), similarly showed a significant BOLD response to food odors above and beyond that of non-food odors in a location previously reported to display increased activity in obese participants following the presentation of high-calorie, food images (Stoeckel et al. 2008). Although normally studied in the context of anterograde learning and memory, the hippocampus has been shown to regulate food intake (Davidson et al. 2007; Tracy et al. 2001), distinguish between introceptive signals of hunger and satiety (Hebben et al. 1985; Davidson and Jarrard 1993; Hock and Bunsey; 1998), and activate in the presence of food related stimuli and food cravings (Bragulat et al. 2010; Cornier et al. 2009; Gautier et al. 1999; LaBar et al. 2001; Pelchat et al. 2004). Although the hippocampus did activate to food odors in our study, the activation was not significantly different from that of non-food odors, at least in this mixed sample.

Medial frontal cortex has become a compelling target in the study of stimuli that are either rewarding in and of themselves (such as the secondary reinforcer of money; e.g., Kable and Glimcher 2007), or sensory stimuli that are classically conditioned with drug rewards, such as alcohol intoxication (Bragulat et al. 2008; Filbey et al. 2008; Kareken et al. 2010a, b, 2011; Myrick et al. 2008). In particular, activation in this region appears to correlate with the extent of the reward’s perceived value and with decision-making based on these valuations (Chib et al. 2009; Hare et al. 2010, 2009, 2011; Noonan et al. 2011). Most recently, Noonan et al. (2011) have shown that the mPFC specifically represents the value of anticipated outcomes. Similar work by Hare et al. suggests that mPFC activation drives choice when confronted with salient stimuli related to food reward (Hare et al. 2009, 2011). In our own hands, we have observed remarkably similar medial prefrontal activation from not only food aromas (Bragulat et al. 2010) but also the aromas of preferred alcoholic drinks (Bragulat et al. 2008)—activation which varies as a function of particular alcoholism risk factors, such as familial alcoholism and polymorphisms in the GABRA2 gene (Kareken et al. 2010a, b). Given the previously documented relationship between the mPFC response to stimuli associated with reward and perceived reward value, one might expect a significant relationship between the extent to which feeding is driven by external cues (i.e., the extent to which food cues can induce a motivated state) and food cue induced activity in this region. Consistent with this notion, we did, indeed, note a region of ventral mPFC that not only responded to food odors but also correlated with the degree to which subjects reported themselves to be high in “externality” on the DEBQ scale, although this appeared only in the post-meal, fed condition. Activation from non-food odors under both fed and fast sessions was not significantly correlated with externality scores.

The DEBQ Externality sub-scale was designed to measure the phenomenon of “external eating,” defined as eating in response to food related stimuli irrespective of internal signals of hunger or satiety (Schachter et al. 1968; van Strien et al. 1986). Those scoring high on the Externality sub-scale need not be obese, as individuals are endowed with their own range of homeostatically regulated body weight (Herman and Polivy 1980; Polivy and Herman 1983). This is true for our dataset, which shows no significant group difference in externality scores, although it is conceivable that with larger samples our small difference (p = 0.09) will become significant. The fact that a correlation emerged solely under the fed condition may not be particularly surprising, as irrespective of externality, most individuals should have a significant response to food stimuli while hungry (Castellanos et al. 2009). The strong positive correlation under the Fed condition between externality and the food odor activation suggests that when sated, vmPFC in these individuals may continue to signal continued rewarding value that is disproportionate to recent food intake. In fact, the peak voxel effect at (−4, 34, −14) in our correlation between externality and the (F > CO) response was very close to a focus (−6, 41, −14) in the most recent study conducted by Hare et al. (2011) in which mPFC activation from food images correlated with taste and health ratings of the images. Thus, we postulate that individuals high in externality may overeat, regardless of hunger, in part from inappropriate food valuation signals in the mPFC—a supposition supported by work from Gearhardt et al. (2011), in which medial OFC BOLD responses to milkshake tastes correlate positively with Yale Food Addiction Scale scores.

Although some between-group trends between normal-weight and obese subjects did begin to emerge in the data reported here, none reached statistical significance. Thus, while the activations in similar studies were all greater in obese subjects when compared with normal-weight controls, the lack of a group difference in our ongoing study is very likely due to the sample sizes. In addition, pooling over obese and normal-weight participants may also explain why we observed no differences in (F > CO) activation between Fed and Fasting states, especially over the relatively short normal-meal fasting interval between breakfast and lunch. It should also be noted that the normal-weight group participants are younger than the obese participants (23.6 ± 4.4 and 30.1 ± 5.0, respectively, p < 0.05). To date, we have also not seen sensory-specific satiety effects, such as those reported by O’Doherty et al. (2000), where activation to a food odor is reduced after the food it represents is eaten to satiety. However, a larger sample may again help determine whether such an effect exists in these subjects.

Conclusions

In summary, this study demonstrated that the peak differences between food and non-food odors arise within medial prefrontal and lateral orbitofrontal areas. The current study also revealed a positive correlation between ventral mPFC activation from food aromas and scores on the externality sub-scale of the DEBQ under sated conditions. This suggests the possibility that, in subjects whose feeding behaviors are driven strongly by external cues, the vmPFC may be involved in the inappropriate valuation of external food stimuli, even after diminished hunger.