The central nervous norepinephrine network links a diminished sense of emotional well-being to an increased body weight

Abstract

OBJECTIVES

The neurobiological mechanisms linking obesity to emotional distress remain largely undiscovered.

METHODS

In this pilot study, we combined positron emission tomography, using the norepinephrine transporter (NET) tracer [¹¹C]-O-methylreboxetine, with functional connectivity magnetic resonance imaging, the Beck depression inventory (BDI), and the impact of weight on quality of life-Lite questionnaire (IWQOL–Lite), to investigate the role of norepinephrine in the severity of depression (BDI), as well as in the loss of emotional well-being with body weight (IWQOL–Lite).

RESULTS

In a small group of lean-to-morbidly obese individuals (n = 20), we show that an increased body mass index (BMI) is related to a lowered NET availability within the hypothalamus, known as the brain’s homeostatic control site. The hypothalamus displayed a strengthened connectivity in relation to the individual hypothalamic NET availability to the anterior insula/frontal operculum, as well as the medial orbitofrontal cortex, assumed to host the primary and secondary gustatory cortex, respectively (n = 19). The resting-state activity in these two regions was correlated positively to the BMI and IWQOL–Lite scores, but not to the BDI, suggesting that the higher the resting-state activity in these regions, and hence the higher the BMI, the stronger the negative impact of the body weight on the individual’s emotional well-being was.

CONCLUSIONS

This pilot study suggests that the loss in emotional well-being with weight is embedded within the central norepinephrine network.

INTRODUCTION

Obesity is a major health burden worldwide. The number of obese people has nearly doubled since the 1980 s and incidence rates are further rising (World Health Organization; http://www.who.int/mediacentre/factsheets/fs311/en/). Its comorbidities are not only restricted to diseases affecting the endocrine and cardiovascular system. Obesity also has influences on the brain’s function¹ and structure,² which may explain its associations to neuropsychiatric disorders ranging from attention-deficit hyperactivity disorder,³ and mild cognitive impairment,⁴ to Alzheimer’s disease and other forms of dementia.⁵ Obesity also disturbs the quality of life in other domains of health, including self-esteem and social relationships. This explains why obese individuals often present symptoms of emotional distress,⁶ such as depression.⁷ Improving weight-related emotional well-being, therefore, is a critical goal of any obesity treatment. However, central nervous mechanisms linking a diminished quality of life due to emotional distress associated with obesity and weight remain largely unexplored.

Although research into obesity has focused on dopaminergic dysfunction, recent data also suggest the involvement of the central norepinephrine system.^8,9 This system originates in norepinephrine cell groups in the brainstem including the locus coeruleus, which in turn sends projections to most areas of the neuroaxis, including metabolic and emotional brain sites. Projections connecting the locus coeruleus and other norepinephrine cell groups in the brainstem to the hypothalamus represent a crucial homeostatic-regulatory circuit, which orchestrates eating behavior according to metabolic demands.^1,10 The inhibition of the reuptake of norepinephrine from the synaptic cleft by pharmacological agents, such as the combined norepinephrine and serotonin reuptake inhibitor Sibutramine,¹¹ proved anorectic effects,¹² indicating an association between central norepinephrine, energy consumption and weight loss.

Besides its role in metabolic processes, the central norepinephrine system also has crucial influences on emotions. In the early 1960s, a deficient central norepinephrine activity has been discussed to partly contribute to the pathophysiology of mood disorders, such as major depression.¹³ In the 1990s, quantitative autoradiography of [³H]nisoxetine binding to the norepinephrine transporter (NET) in the locus coeruleus in postmortem brains of patients with major depressive disorder implied that this pathology is associated with a reduced expression of the NET on noradrenergic neurons within the locus coeruleus.¹⁴

With these functional implementations, the central norepinephrine network seems to be at the crossroad between processes controlling eating behavior through homeostatic feedback from the hypothalamus, and emotional well-being associated with body weight.¹⁵ In this pilot study, we investigated a small group of lean-to-obese individuals with a highly selective NET radiotracer for positron emission tomography (PET), called [¹¹C]-O-methylreboxetine (MRB).^8,16–18 NET is a monoamine transporter and responsible for the sodium chloride-dependent reuptake of extracellular norepinephrine. This reuptake is essential in regulating norepinephrine’s concentration within the synaptic cleft. What remains unclear is how NET availability relates to norepinephrine’s concentration within the synaptic cleft. A lower NET availability may relate to a lowered norepinephrine reuptake, resulting in higher concentrations within the synaptic cleft. Alternatively, NET may be lowered because of a lowered norepinephrine concentration within the synaptic cleft. Thus, a lower NET availability could relate to either a higher or a lower concentration of norepinephrine in the synaptic cleft. The inverse hypothesis holds true for higher NET availability.

As previous MRB PET studies revealed lowered NET availability in obese as compared with lean participants,⁸ we hypothesized an obesity-related lowered NET availability in the hypothalamus. Next to MRB PET, we acquired functional connectivity magnetic resonance imaging (fMRI) scans from the same participants to investigate the connectivity of the hypothalamus in relation to its NET availability throughout the entire brain. We hypothesized a strengthened connectivity between brain sites associated with the sensory properties of food (that is, orbitofrontal cortex (OFC) and insular cortex) and food-related emotions (that is, amygdala and hippocampus). Futher, we assumed that resting-state activity within these regions scales with the body mass index (BMI), as well as the scores of the widely used questionnaire called ‘the impact of weight on quality of life–Lite’ (IWQOL–Lite). Apart from the IWQOL–Lite, addressing weight-associated emotional well-being, we assumed that the resting-state activity of brain sites being connected to the hypothalamus likewise scales with the severity of depression, assessed with the most widely used questionnaire for measuring depression, known as the ‘Beck Depression Inventory’ (BDI).

MATERIALS AND METHODS

Participants, assessment of the severity of depression using the BDI and the diminished quality of life associated with weight using the IWQOL–Lite questionnaire

Twenty right-handed participants gave written informed consent before participating and underwent the PET session (8 female, 12 male, BMI 33.2 ± 9.7 kg m⁻², aged 34 ± 9.1 years; see Figure 1a for BMI, BDI and IWQOL–Lite distribution; see Table 1 for individual anthropometric data). One subject canceled the following fMRI session due to claustrophobia during data acquisition and was excluded from further statistical analyses of the MRI data (8 female, 11 male, BMI 32.5 ± 9.6 kg m⁻², aged 33 ± 9.1 years). The local ethics committee of the medical faculty of the University of Leipzig and the German Federal Office for Radiation Protection (that is, Bundesamt für Strahlenschutz) approved the study protocol. All participants underwent physical examination before participating. Criteria for exclusion were any known contraindications for MRI, hypertension, diabetes, use of any central-acting drugs, participation in any weight loss programs during the last 6 months, pregnancy or breast feeding. Following the BMI assessment, each subject filled out the BDI¹⁹ and the IWQOL–Lite questionnaires.²⁰

Figure 1.

Distribution of body mass index (BMI), the Beck Depression Inventory (BDI), the impact of weight on quality of life (IWQOL–Lite) and the BMI versus IWQOL–Lite correlation analyses. (a) The first three scatter plots from left to right in the top row display the distribution of BMI, the BDI and the IWQOL–Lite for all 20 participants. The upper and lower hinges of the boxplot correspond to the first and third quartiles, the midline to the median. The upper and lower whisker extends from the hinge to the highest and lowest value that is within 1.5 × inter-quartile range of the hinge. (b) The scatter plots show the correlation between BMI and the IWQOL–Lite total score (P ≤ 0.05 indicated significance; rank-transformed data are presented due to Spearman’s Rho correlation analysis) as well as the post hoc correlations between BMI and the five IWQOL–Lite subscores (i.e., physical function, self-esteem, intimate relations, social embarrassment and life at work; P ≤ 0.01 indicated significance; rank-transformed data are presented). The BMI positively correlated with the IWQOL–Lite total score suggesting the IWQOL–Lite’s superior sensitvity in assessing emotional well-being in relation to an increased body weight. Furthermore, we also found significant correlations with the BMI for three out of the five IWQOL–Lite subscores (i.e., physical functions, self-esteem and social embarrassment). NS, non-significant; sig., significant; trend, strong statistical trend.

Table 1.

Anthropometric data of the study sample (n = 20) including gender, age, BMI, the BDI, the IWQOL–Lite questionnaire and its subscores, as well as the hypothalamic binding potential

Subject	Gender (0 = female, 1 = male)	Age (years)	BMI (kg m−2)	BDI-II (max. 63)	IWQOL–Lite total (max. 155)	IWQOL–Lite physical function (max. 55)	IWQOL–Lite self-esteem(max. 35)	IWQOL–Lite intimate relations (max. 20)	IWQOL–Lite social embarrassment (max. 25)	IWQOL–Lite life at work (max. 20)	Hypothalamic BPND
1	1	22	21.7	6	31	11	7	4	5	4	0.472888
2	1	40	39.6	11	72	34	13	6	11	8	0.307269
3	0	21	22.9	22	55	15	20	9	5	6	0.374778
4	0	37	22.2	1	31	11	7	4	5	4	0.572076
5	1	27	21.9	3	32	11	8	4	5	4	0.607529
6	0	25	28.7	0	31	11	7	4	5	4	0.490288
7a	1	42	45.2	4	72	41	11	4	10	6	0.115768
8	1	44	24.1	0	31	11	7	4	5	4	0.315173
9	1	36	42.6	26	78	34	14	6	13	11	0.128989
10	0	35	20.9	6	31	11	7	4	5	4	0.388802
11	1	39	26.5	0	31	11	7	4	5	4	0.522314
12	1	52	24.8	3	33	12	8	4	5	4	0.185633
13	1	31	25.2	9	32	12	7	4	5	4	0.694777
14	0	22	45.7	3	67	25	13	5	15	9	0.206971
15	1	32	41.7	0	52	26	8	4	10	4	0.473791
16	1	50	39.9	13	75	33	20	9	9	4	0.308928
17	0	37	35.7	6	31	11	7	4	5	4	0.126645
18	0	37	40.6	1	39	19	7	4	5	4	0.519554
19	0	26	47.8	9	53	15	16	6	12	4	0.471701
20	1	22	45.4	32	134	40	33	20	24	17	0.222574

Abbreviations: BDI, Beck depression inventory; BMI, body mass index; BP_ND, binding potential; fMRI, functional connectivity magnetic resonance imaging;

IWQOL–Lite, impact of weight on quality of life-Lite questionnaire.

^aExcluded from resting-state fMRI analysis.

To assess the severity of depression, we used the most recent version of the BDI (BDI-II), which is a 21-question multiple-choice self-report inventory.¹⁹ Higher total scores indicate more severe depressive symptoms. The standardized cutoffs used are: 0–13: minimal depression, 14–19: mild depression, 20–28: moderate depression, 29–63: severe depression.

The IWQOL–Lite is a commonly used 31-item self-report instrument with high test-retest reliability and five subscales covering most relevant areas of everyday life negatively affected by obesity: physical function, self-esteem, intimate relations, social embarrassment and life at work. The subscales are reported to correlate with the total score, which we also validated for our study sample. To assess the total score, the five subscale scores are summed up so that the higher the total score, was higher the weight-related influence on the individual’s emotional well-being.²⁰

First, we applied the Kolmogorov-Smirnov test to assess whether the scores across participants significantly differed from normal distribution. In face of the small sample size and the not normally distributed IWQOL–Lite scores, we applied non-parametric Spearman’s Rho instead of parametric Pearson’s correlation analyses. Next, both scores (BDI and IWQOL–Lite) were correlated with the BMI, using one-sided Spearman’s Rho correlation analyses, to assess whether a higher BMI was associated to an increased severity of depression, as well as an increased impact of weight on the emotional well-being. Using one-sided Spearman’s Rho correlation analysis, we also tested for a positive correlation between the BDI and the IWQOL–Lite to assess whether the severity of depression scaled with the impact of weight on the emotional well-being. In case of significance, we furthermore correlated the BMI and the BDI with the five subscores of the IWQOL–Lite. We corrected the P-value of post hoc analyses to the five subscales of the IWQOL–Lite (that is, P ≤ 0.01).

Radiotracer synthesis, PET imaging and PET data analyses

Radiotracer synthesis and PET imaging was carried out in accordance with previous work by Ding and colleagues.^8,21 In brief, the preparation of MRB standard as well as the precursor was the basis for synthesizing [¹¹C]MRB with [¹¹C]methyliodide ([¹¹C]MeI). Radiolabeling was carried out by using the TRACERLab FXC automated synthesis module (GE Healthcare, USA). [¹¹C]methyliodide was produced from [¹¹C]CO2. The final formulated product was 98% radiochemically pure and the average individual injected mass was 0.026 ± 0.023 μg kg⁻¹.

At the time of intravenous bolus injection of 358.9 ± 11 MBq [¹¹C]MRB, participants were enrolled in dynamic PET scans using the ECACAT EXACT HR+ scanner in three-dimensional acquisition mode (Siemens, Erlangen, Germany; intrinsic resolution at the center: 4.3 mm (full-width at half-maximum), axial resolution: 5–6 mm and field of view: 15.5 cm). We acquired 26 time frames over the 120 min emission scan (4 × 0.25, 4 × 1, 5 × 2, 5 × 5, 8 × 10 min). Prior to the emission scan, all participants underwent a 10 min transmission scan (from three ⁶³Ga sources) for attenuation correction and iterative reconstruction (10 iterations, 16 subsets) in transverse image series (63 slices, 128 × 128 matrix, voxel size 2.6 × 2.6 × 2.4 mm³) with Hannfilter (cutoff 4.9 mm) for post-processing.

PET data were realigned using the SPM8 software (Statistical Parametric Mapping; Wellcome Trust Centre for Neuroimaging, London, UK), then co-registered with the individual MRI image using the software PMOD version 3.3 (PMOD Technologies, Zurich, Switzerland).

Following this, volumes of interest of the occipital cortex (low NET availability) and the thalamus (high NET availability)⁸ were drawn on the individual MRI images. The volumes of interests were applied to the PET data to derive the [¹¹C]MRB time–activity curves for both regions. We used the multi-linear reference tissue model with two parameters and the occipital volumes of interest as the reference to generate parametric images of the binding potential (BP_ND).²² To derive the reference tissue clearance rate constant k2′ (here k2′ = 0.0238 min⁻¹) necessary for multi-linear reference tissue model with two parameters, we applied the multi-linear reference tissue model with three parameters to the individual time–activity curves of the thalamus and the occipital cortex. This model applied to the [¹¹C]MRB data became linear at t^* = 0.^{(ref. 23)} For this reason, we sampled all data across the 26 frames over the 120 min emission scan.

PET images were then normalized. To this end, we applied the parameters of the normalization procedure of the MRI data (see FMRI and MRI data analyses for further details) to the PET images. After this, PET images were smoothed with an 8 mm full-width at half-maximum Gaussian filter using SPM8. We used the WFU PickAtlas to create a region of interest mask (that is, 24 voxels; voxel size 2 × 2 × 2 mm³) covering the hypothalamic territory²⁴ and extracted the first Eigenvariate of the BP_ND across all voxels within the hypothalamic region of each individual PET data set within SPM8. These data were applied to one-sided Spearman’s Rho correlation analyses to test for a negative correlation with individual BMI, the IWQOL–Lite total score and the BDI using the SPSS 19.0 statistic software (SPSS Inc., Chicago, IL, USA). The P-value was adjusted to the number of correlations using Bonferroni correction (P ≤ 0.017).

FMRI and MRI data analyses

For resting state, or fMRI, we visually presented a fixation cross to the participants after placing them in a supine position in the MRI scanner. Participants were told to keep their eyes open and to fixate on the cross during the entire scanning session. Although 15 participants were investigated using a Siemens Verio 3 T-Scanner (Erlangen, Germany, 300 acquired echo planar imaging volumes, voxel size 2 × 2 × 2 mm³, repetition time 2000 ms, echo time 30 ms and slice thickness 3 mm), another 4 participants were investigated with a Siemens TimTrio 3 T-Scanner (120 acquired echo planar imaging volumes, voxel size 2 × 2 × 2 mm³, repetition time 2000 ms, echo time 30 ms and slice thickness 3 mm).

FMRI data were pre-processed using SPM8 and LIPSIA Software (Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany). Within SPM8, images were realigned, normalized to the Montreal Neurological Institute template and finally smoothed using an 8 mm full-width at half-maximum Gaussian filter. With LIPSIA, we additionally band-pass filtered (cutoff frequency: 1/80 Hz) and detrended the images.

On the single-subject level, the images were applied to the general linear model, with one regressor representing the scans acquired over time. Next, we applied the hypothalamic region of interest (as used for extracting the mean BP_ND from the NET–PET data) to extract the first Eigenvariate of the beta values across all voxels within the hypothalamic region. This individual hypothalamic time series was next implemented within the same single-subject general linear model, as an additional non-interacting regressor, and analyzed to test for a positive correlation (that is, strengthened connectivity) of the hypothalamic seed region throughout the entire brain.

The individual statistical maps, representing the connectivity of each individual hypothalamic seed, were next entered into a random effects group-level one-sample t-test, together with the corresponding individual BP_ND values from the same hypothalamic region (that is, within the same region of interest), to assess the functional connectivity of the hypothalamus in relation to the individual hypothalamic BP_ND (that is, interaction between fMRI and BP_ND). To account for gender effects, as well as the two different MRI scanners together with the slightly different scanning protocols, we added two non-interacting regressors to the model to account for the corresponding nuisance variance. Following our a priori hypotheses on brain sites being connected to the hypothalamic seed region (that is, brain sites associated with the sensory properties of food and food-related emotions, see Introduction section), we applied small volume correction as implemented in SPM8. To this end, we used a sphere with a diameter of 15 mm centered on Montreal Neurological Institute coordinates previously reported in the context of eating (OFC (9, 37, −26; x, y, z; in mm),²⁵ anterior insula/front. operculum (43, 18, 1),²⁶ amygdala (right hemisphere: 24, −7, −9; left: −26 −8 −14)²⁷ and hippocampus (right: 34, −20, −8; left: −34 −20 −10).²⁸ We applied a family-wise error corrected threshold at P ≤ 0.05.

Correlation between the BMI, the BDI, the IWQOL–Lite scores and resting-state activity in brain regions connected to the hypothalamus

From brain regions that were significantly connected (that is, correlated) to the hypothalamic seed, we extracted the first Eigenvariate of the resting-state activity (that is, beta values) across significantly activated clusters, which were then applied to one-sided Spearman’s Rho correlation analysis within SPSS 19 (SPSS Inc.) to test for a positive correlation with the BMI, the BDI as well as the IWQOL–Lite total score. The P-values were adjusted to the number of significantly connected brain sites using Bonferroni correction.

In case of significance, we then performed post hoc analyses and correlated each of the five subscales of the IWQOL–Lite with the first Eigenvariate of the resting-state activity (that is, beta values) across significantly activated voxels from those regions connected to the hypothalamus. To account for multiple comparisons, the P-value for these post hoc analyses were adjusted to the five subscales of the IWQOL–Lite and the number of significantly connected brain sites using Bonferroni correction.

RESULTS

Correlation between the BMI, the BDI and the IWQOL–Lite

We tested whether the IWQOL–Lite total score scaled with its five subscores using one-sided Spearman’s Rho correlations together with a Bonferroni-corrected P ≤ 0.01. Also, the post hoc correlations between the BMI and the five IWQOL–Lite subscores were Bonferroni corrected (that is, P ≤ 0.01). The analyses confirmed a positive correlation with BMI for the IWQOL–Lite total score, indicating that a higher score in the questionnaire related to an increased BMI (IWQOL–Lite total: r = 0.701, P<0.001; Figure 1b).^29,30 We also found significant correlations with BMI for three out of the five IWQOL–Lite subscores (physical function: r = 0.754, P<0.001; self-esteem: r = 0.560, P = 0.005; social embarrassment: r = 0.830, P<0.001), whereas the intimate relations subscale and the life at work subscale missed significance (BMI intimate relations: r = 0.479, P = 0.016; BMI work: r = 0.498, P = 0.013; Figure 1b). Furthermore, the IWQOL–Lite subscores correlated with the total score (r in the range between 0.753 for life at work and 0.951 for physical function; P<0.001 for all correlations).²⁹

Contrarily to the IWQOL–Lite, we found no significant correlation between an increased BMI and increased BDI (that is, n = 20; one-sided Spearman’s Rho correlation r = 0.208, P = 0.190; Supplementary Figure 1A),^31–33 although the BDI was directly correlated with the IWQOL–Lite total score (r = 0.634, P = 0.001), as well as with three out of the five IWQOL–Lite subscores (physical function: r = 0.509, P = 0.011; self-esteem: r = 0.668, P = 0.001; intimate relations: r = 0.765, P<0.001; social embarrassment: r = 0.450, P = 0.023; life at work: r = 0.558, P = 0.005; Supplementary Figure 2).

Correlation between hypothalamic NET availability and the BMI, the BDI and the IWQOL–Lite

In line with our a priori hypothesis (see Introduction section), the individual NET availability (BP_ND 0.3753 ± 0.1687) in the hypothalamus was negatively correlated with the IWQOL–Lite (r = −0.501, P = 0.012; Figure 2), whereas the correlation with BMI (that is, n = 20; one-sided Spearman’s Rho correlation r = −0.439, P = 0.026) and BDI (r = −0.402, P = 0.039) did not reach significance (adjusted P ≤ 0.017; Supplementary Figure 1B). In other words, the higher the IWQOL–Lite score, the lower the NET availability in the hypothalamus was. Addressing age and gender as putative confounding covariates did not affect this relationships (that is, the r- and P-value remained the same for six positions after decimal point).

Figure 2.

Hypothalamic norepinephrine transporter (NET) availability and its correlation to the impact of weight on quality of life-Lite questionnaire (IWQOL–Lite). (a) The brain slices display the NET availability as revealed by [¹¹C]MRB PET. Brain slices in the top row show the mean binding potential (BP_ND) across the whole group (n = 20). Brain slices in the middle row show the BP_ND for lean and overweight participants (n = 10, body mass index (BMI)<30 kg m⁻²), whereas the bottom row shows the BP_ND for obese participants (n = 10, BMI ⩾ 30 kg m⁻²). The marker indicates the peak voxel of the hypothalamic region and R the right brain hemisphere. Note the higher tracer uptake within this region in lean and overweight as compared with obese participants. (b) Correlation between BP_ND values obtained from the hypothalamic region and the IWQOL–Lite total score (P ≤ 0.017 indicated significance, rank-transformed data are presented due to Spearman’s Rho correlation analysis). This correlation suggests that the higher the impact of the body weight on the emotional well-being (indexed by the IWQOL–Lite), the lower the hypothalamic NET availability was.

Functional connectivity of the hypothalamus in relation to individual hypothalamic NET availability

Here, we investigated the NET availability–related functional connectivity of the hypothalamus throughout the entire brain (n = 19). This interaction analysis between fMRI and PET data revealed a significantly strengthened connectivity between the hypothalamus and the insula/frontal operculum (x, y, z: 56, 15, 3 mm; T = 4.9720; P = 0.021 family-wise error corrected), as well as the medial OFC (x, y, z: −1, 45, −21 mm; T = 3.6576; P = 0.046 family-wise error corrected; Figure 3a). Neither the amygdala, nor the hippocampus search spheres revealed any significantly activated voxels. These findings suggest that the lower the hypothalamic NET availability, and hence the higher the BMI, the stronger the connection between the hypothalamus and the anterior insula/frontal operculum, and the medial OFC was.

Figure 3.

Resting-state connectivity of the hypothalamus in relation to the individual hypothalamic norepinephrine transporter (NET) availability, and correlations between resting-state activity in brain regions connected to the hypothalamus and the impact of weight on quality of life-Lite questionnaire (IWQOL–Lite). (a) Functional connectivity of the hypothalamic seed region in relation to the individual hypothalamic NET availability revealed two significantly connected regions (P ≤ 0.05 family-wise error corrected small volume corrected), namely the anterior insula/frontal operculum and the medial OFC. The thresholds for extracting the beta values from the clusters are displayed: red P ≤ 0.001 uncorrected (anterior insula/frontal operculum), yellow P ≤ 0.005 uncorrected (medial OFC). R indicates the right brain hemisphere. (b) The two scatter plots show the correlations between the resting-state activity obtained from the two regions shown in a, and the IWQOL–Lite total score (P ≤ 0.025 indicated significance; n = 19). Together, these correlations suggest that the higher the resting-state activity obtained from the anterior insula/frontal operculum and the medial OFC, the higher the BMI and the impact of body weight on emotional sense of well-being was. The scatter plots show the rank-transformed data due to Spearman’s Rho correlation analysis. sig., significant.

Correlation between BMI and the resting-state activity in the insula/frontal operculum and the medial OFC

Next, we investigated whether the resting-state activity (that is, beta values) from the two brain sites with strengthened hypothalamic connectivity (that is, medial OFC, thresholded at P = 0.005 uncorrected, cluster of 40 voxels, anterior insula/frontal operculum, thresholded at P = 0.001 uncorrected, cluster of eight voxels) correlated with an increased BMI. To this end, we used a one-sided non-parametric Spearman’s Rho correlation analysis together with a Bonferroni correction adjusted to P ≤ 0.025. Both the anterior insula/frontal operculum (r = 0.628, P = 0.002) as well as the medial OFC (r = 0.582, P = 0.004) were found to correlate positively with the BMI (Supplementary Figure 1C), suggesting that the resting-state activity (that is, beta values) obtained from these regions was enhanced with an increased BMI.

Correlation between BDI and the resting-state activity in the anterior insula/frontal operculum and the medial OFC

The correlation between the BDI and MRI resting-state activity (Bonferroni-adjusted P-value = 0.025) revealed no significant correlation for the anterior insula/frontal operculum (r = 0.279, P = 0.124) or the medial OFC (r = 0.440, P = 0.030; Supplementary Figure 1C).

Correlation between IWQOL–Lite scores and the resting state activity in the anterior insula/frontal operculum and the medial OFC

Contrarily to the BDI correlations (see previous paragraph), the correlation between the IWQOL–Lite total score and fMRI resting-state activity (Bonferroni-adjusted P-value = 0.025) revealed a positive correlation for the anterior insula/frontal operculum (r = 0.550, P = 0.007) and the medial OFC (r = 0.643, P = 0.002; Figure 3b). The post hoc analyses with the five subscores of the IWQOL–Lite were computed with non-parametric Spearman’s Rho correlation together with Bonferroni correction at P = 0.01. For the medial OFC, the analysis revealed a significant correlation for the subscores on physical function (r = 0.570, P = 0.005), self-esteem (r = 0.685, P = 0.001) and social embarrassment (r = 0.539, P = 0.009), as well as a strong statistical trend for the intimate relations subscale (r = 0.515, P = 0.012; Figure 4). For the anterior insula/frontal operculum, we found a positive correlation with physical function (r = 0.559, P = 0.006) and social embarrassment (r = 0.630, P = 0.002), but a strong tendency only for the self-esteem subscale (r = 0.512, P = 0.013) and no correlation with intimate relations (r = 0.384, P = 0.052; Figure 4). Neither the OFC (r = 0.364, P = 0.063) nor the anterior insula/frontal operculum (r = 0.289, P = 0.115) was related to the life at work subscale (Figure 4). Table 2 lists the P-values, significance thresholds and coefficients of all correlation analyses presented above.

Figure 4.

Post hoc correlations between resting-state activity of the anterior insula/frontal operculum, as well as the medial orbitofrontal (OFC) and the five IWQOL–Lite (impact of weight on quality of life) subscores (i.e., physical function, self-esteem, intimate relations, social embarrassment and life at work). This figure displays the post hoc correlations between the resting-state activity in the anterior insula/frontal operculum, as well as the medial OFC (P ≤ 0.01 indicated significance; n = 19) and the five IWQOL–Lite subscores. Note the significant relationships between resting-state activity in the anterior insula/frontal operculum for two out of the five dimensions included in the IWQOL–Lite, as well as in the medial OFC for three out of the five dimensions. Self-esteem missed significance only in the anterior insula/frontal operculum, whereas intimate relations and life at work were not correlated to the resting-state activity in any of the two regions. Note the IWQOL–Lite’s low sensitivity in capturing the influence of overweightness on emotional sense of well-being (lowest possible scores in lean and overweight participants in the scatter plots). The scatter plots show the rank-transformed data due to the Spearman’s Rho correlation analysis. NS, non-significant; sig., significant; trend, strong statistical trend.

Table 2.

P-values, thresholds and coefficients of all the Spearman’s Rho correlation analyses performed in this study

	P-values	threshold	coefficient
BMI vs
BDI	0.190	≤0.017	0.208
IWQOL–Lite total score	<0.001a	≤0.017	0.701
Physical function	<0.001a	≤0.01	0.754
Self-esteem	0.005a	≤0.01	0.560
Intimate relations	0.016	≤0.01	0.479
Social embarrassment	<0.001a	≤0.01	0.830
Life at work	0.013	≤0.01	0.498
BDI vs
IWQOL–Lite total score	0.001a	≤0.017	0.634
Physical function	0.011	≤0.01	0.509
Self-esteem	0.001a	≤0.01	0.668
Intimate relations	<0.001a	≤0.01	0.765
Social embarrrassment	0.023	≤0.01	0.450
Life at work	0.005a	≤0.01	0.558
IWQOL–Lite total score vs subscores
Physical function	<0.001a	≤0.01	0.951
Self-esteem	<0.001a	≤0.01	0.899
Intimate relations	<0.001a	≤0.01	0.786
Social embarrassment	<0.001a	≤0.01	0.823
Life at work	<0.001a	≤ 0.01	0.753
Hypothalamus BPND vs
BMI	0.026	≤ 0.017	−0.439
BDI	0.039	≤ 0.017	−0.402
IWQOL–Lite total score	0.012a	≤0.017	−0.501
Resting-state activity ant.insula/front. operculum vs
BMI	0.002a	≤0.025	0.628
BDI	0.124	≤0.025	0.279
IWQOL–Lite total score	0.007a	≤0.025	0.550
Physical function	0.006a	≤0.01	0.559
Self-esteem	0.013	≤0.01	0.512
Intimate relations	0.052	≤0.01	0.384
Social embarrassment	0.002a	≤0.01	0.630
Life at work	0.115	≤0.01	0.289
Resting-state activity med. OFC vs
BMI	0.004a	≤0.025	0.582
BDI	0.030	≤0 025	0.440
IWQOL–Lite total score	0.002a	≤0.025	0.643
Physical function	0.005a	≤0.01	0.570
Self-esteem	0.001a	≤0.01	0.685
Intimate relations	0.012	≤0.01	0.515
Social embarrassment	0.009a	≤0.01	0.539
Life at work	0.063	≤0.01	0.364

Abbreviations: ant., anterior; BDI, Beck depression inventory; BMI, body mass index; BP_ND, binding potential; IWQOL–Lite, impact of weight on quality of life-Lite questionnaire; med., medial.

^aSignificant values.

DISCUSSION

Using the NET tracer [¹¹C]MRB for PET, we show an IWQOL–Lite (Impact of Weight on Quality of Life) related lowered NET availability in the hypothalamus, whereas the BMI and BDI only showed a vague statistical trend into the same direction. The hypothalamus, in turn, presented a strengthened NET availability-related functional connectivity to the anterior insula/frontal operculum, as well as the medial OFC, supposed to host the primary³⁴ and the secondary gustatory cortex,²⁶ respectively. Resting-state activity in these two regions significantly correlated with the BMI and the IWQOL–Lite questionnaire scores, suggesting that the higher their resting-state activity, and hence the higher the BMI, the stronger the impact of the body weight was on the individual’s quality of life. In line with previous assessments,^31–33 we found no correlation between the BMI and the BDI. Furthermore, the BDI was also not correlated with the resting-state activity, neither in the anterior insula/frontal operculum, nor in the medial OFC, suggesting that these regions are rather more involved in the loss of emotional well-being associated with weight (as indexed by the IWQOL–Lite) than in more general aspects of emotional well-being (as indexed by the BDI).

Diminished quality of life related to weight and the IWQOL–Lite questionnaire

The IWQOL–Lite consists of five subscales covering the five areas of daily life most seriously affected by an increased body weight.²⁰ Previous intervention studies showed that all these five areas of life improve significantly if obese individuals lose weight.^35,36 In the present study, the IWQOL–Lite and its subscores were correlated with the BDI, but only the IWQOL–Lite and not the BDI was related to the BMI. Together, these findings suggest the IWQOL–Lite’s superior sensitivity for evaluating emotional well-being in relation to an increased body weight. The five IWQOL–Lite subscales are well-known to correlate with each other, as well as with the total score, suggesting a mutual influence of the five dimensions on the individual’s weight-related quality of life.²⁹

In the present study, the resting-state activity obtained from the anterior insula/frontal operculum and the medial OFC scaled with the IWQOL–Lite’s subscores for physical function, social embarrassment and self-esteem, whereas only the medial OFC resting-state activity showed a trend for a correlation with the loss in the quality of the intimate relations. Besides these associations, neither the activity obtained from the anterior insula/frontal operculum, nor the medial OFC scaled with the loss in the quality of working life. This finding is in line with previous work suggesting that an increased body weight has a lower emotional impact on the quality of one’s working life as compared with the other four areas of daily life included in the IWQOL–Lite.³⁵

Emotions, eating behavior and associated brain sites

Emotions influence eating behavior, but eating, in turn, also influences emotional processing and the sense of well-being, not only when overeating over time causes obesity and associated negative consequences for mental health, but also during eating, when eating elicits mainly positive emotions. These mutual influences between eating and emotions are anchored in the architecture of the brain circuitries that typically orchestrate eating behavior. Homeostatic control sites within the hypothalamus are integrated within a far-reaching network that incorporates information on homeostatic demands for the hedonic control over eating behavior. During eating, exteroceptive food-related sensory signals from taste and olfactory receptor cells activate the anterior insula/frontal operculum (that is, primary gustatory cortex), where stimulus identity and intensity are merged into a stable representation, independent of the homeostatic or motivational state.^25,37,38 Guided by sensory information, key regions of the reward circuit, such as the striatum, attribute rewarding values to the food, thereby enabling the food to shape emotional responses.³⁹ Integration of interoceptive signals indexing the actual metabolic demand, with exteroceptive sensory signals elicited by the food’s look, smell, taste and texture, define its rewarding value as well as the emotional response to it.^40–42

The medial OFC, as the secondary gustatory cortex, seems to be crucially involved in the encoding of the emotional value or pleasantness of food.^43–45 Unlike the anterior insula/frontal operculum (that is, primary gustatory cortex), the medial OFC reduces its responsiveness to food that had already been eaten to satiety and its activation scales with the subjective pleasantness during eating.^46–48 Furthermore, the medial OFC integrates sensory signals with various peripheral homeostatic hunger and satiety-related signals, as received from the nucleus of the solitary tract, via thalamic and hypothalamic nuclei, which together seem to form the food’s affective value.⁴⁹ In view of the present findings, the medial OFC may therefore have a twofold complementary role in processes underpinning the generation of pleasant emotions during eating on the one hand, and mainly negative emotions in response to weight gain and the associated emotional distress on the other.

Besides the medial OFC, we also found that resting-state activity in the anterior insula/frontal operculum scaled with a diminished quality of life related to weight, suggesting that the primary gustatory cortex, assumed to rather attribute sensory information to food than the affective value, has a crucial role in the emotional consequences to being obese nonetheless.

Norepinephrine’s role in obesity and emotional processing

Norepinephrine is well-known for its involvement in emotional processing and well-being. The composition of this system is characterized by its far-reaching alerting excitatory inputs to most brain sites, where norepinephrine exerts arousal for initiating wakefulness, motivation and attention. These alerting influences can initiate chronic emotional stress and lead to negative emotional states, such as fear or anxiety.¹⁵ In obesity, such negative emotions may cause further overeating⁵⁰ and weight gain, which, in turn, further disturbs perception of body image⁵¹ and self-esteem,⁵² leading to psychological distress and depression.⁵³ Depression, in turn, may alter eating habits toward emotional and uninhibited eating^54,55 supposed to temporarily compensate for symptoms of negative emotional stress.^56–58 This suggests that overeating and the presumably enhanced positive emotional response mediated by dopamine, endorphins and many other neurotransmitters may temporarily override the lowered NET availability–dependent altered norepinephrine concentrations within the synaptic clefts of the hypothalamus together with the strengthened connectivity to sensory-emotional brain sites and hence, related negative emotions to body weight.

Limitations

Our pilot study has several limitations. First, our study sample is small and evidences are based on correlation analyses. That is why this study cannot distinguish cause from consequence regarding the putative role of norepinephrine in either obesity or emotional well-being. Furthermore, it also remains unclear how this correlation between the NET availability and the IWQOL–Lite in the hypothalamus relates to the central norepinephrine concentration, as a lower NET availability could relate to both, a higher or a lower concentration of norepinephrine within the synaptic cleft (see also Introduction section). Although participants cover the BMI range from lean to morbidly obese, the IWQOL–Lite questionnaire is not sensitive enough to capture the impact of slight overweightness on emotional well-being. That is why most of these participants showed a floor effect in the IWQOL–Lite correlation analyses. Our study sample was also too small to investigate the obese participants alone or address gender-specific effects, which have been reported in previous brain imaging studies.^59,60 Second, one previous study also used PET imaging together with the NET tracer [¹¹C]MRB in lean-to-obese participants, but found significant BMI differences in the thalamus, including the pulvinar, but not in the hypothalamus.⁸ Although we had no hypotheses regarding the involvement of the thalamus or the pulvinar in weight-related emotions, we correlated the BMI with the thalamic NET availability (using the thalamic region of the WFU PickAtlas together with the same methods as applied for assessing the hypothalamic NET availability), but found no relationship (that is, n = 20; one-sided Spearman’s Rho correlation r = −0.266, P = 0.128). The most striking difference between the present and the previous study by Li et al.⁸ is the broader BMI range in our sample (up to 47.8 kg m⁻²) that might explain those differences. Alternatively, this missing relationship may simply be another power problem. Third, the fMRI resting-state measurements were performed using two different scanners. That is why we included a non-interacting regressor in the corresponding fMRI analysis to account for the related nuisance variance. Future studies should consider these limitations to draw a more conclusive picture of norepinephrine’s role in obesity and the diminished sense of emotional well-being associated with body weight.

Summary

Our findings highlight norepinephrine as one neurotransmitter involved in emotional distress related to body weight. Thus, the present findings may explain why the modulation of norepinephrine’s reuptake from the synaptic cleft through agents such as Sibutramine¹¹ supports weight loss, and the regain of an increased quality of life related to weight.¹² Sibutramine, unfortunately, caused major side effects, such as heart attacks, which consequently led to its withdrawal from the market.⁶¹ The development of other pharmacological agents that modulate the norepinephrine system effectively with comparable anorectic effectiveness, such as Sibutramine, but without the high risk of serious cardio- or cerebrovascular side effects, seems unachievable, as the effects on the vascular system are principle effects of norepinephrine. Future molecular imaging studies should, therefore, investigate other neurotransmitter systems, such as the dopaminergic, serotonergic, GABAergic, endocannabinergic or opioid system, which may also contribute to the emotional distress related to obesity, and potentially offer new pharmacological targets for its treatment.