PET/fMRI demonstrates that bariatric surgery may reverse striatal dopaminergic dysfunction in women with obesity

Abstract

Background:

Central mechanisms may play a role in the success of bariatric surgery (BS), the treatment of choice for refractory obesity. We hypothesize that central dopaminergic receptor function in striatal brain regions is a pivotal mechanism in the success of BS.

Methods:

We conducted a cross-sectional study to investigate central dopamine type 2 and 3 receptors (D2/3 R) within striatal brain regions in successful weight loss (WL) through BS. Positron Emission Tomography was used to map nondisplaceable binding potential (BP_ND) of D2/3 R in 48 women: 19 successful responders to BS, 12 with obesity (OB), and 17 normal-weight controls. Parametric maps were compared between-groups in regions of interest and at voxel-level. We also investigated brain blood oxygenation level-dependent (BOLD) responses to food content using functional Magnetic Resonance Imaging (fMRI) and how key variables correlate with D2/3 R binding.

Results:

We find mean D2/3 R BP_ND significant differences between OB and controls in the ventral striatum (p = 0.042) and at voxel-level across striatum between OB and the other groups (p < 0.05). Food content (Food > Non-food, p = 0.05) reveals significantly higher neural activation in striatum also for OB compared to the other groups. Moreover, D2/3 R BP_ND values correlate with dysfunctional self-report measures of eating behaviors, incentive salience to food cue and high-calorie food preferences in obesity. Notably, BOLD responses (Food > Baseline) in striatum correlate positively with D2/3 R binding in ventral striatum.

Conclusions:

Striatal dopaminergic dysfunction in obesity may enhance salience to food cues, driving cravings and compulsive eating. BS may reverse the striatal molecular and functional disruptions found in obesity.

Plain language summary

Bariatric surgery is a type of surgery that helps people with severe obesity lose weight. The success of this surgery may involve changes in the brain. This study investigated how these changes impact weight loss and eating behavior, focusing on dopamine, a brain chemical known to have an effect on pleasure, food cravings and eating behaviors. We used brain imaging to examine dopamine in women with successful weight loss after bariatric surgery, women with obesity, and normal weight women. We found that obesity is associated with changes in dopamine, which may contribute to food cravings and compulsive eating. Successful weight loss after bariatric surgery appear to reverse these changes. This work may help clarify how bariatric surgery promotes weight loss.

Pais et al. use multimodal brain imaging to examine dopamine function and eating behavior. Striatal dopaminergic dysfunction in obesity may increase the incentive salience of food cues, promoting cravings and compulsive eating, with bariatric surgery appearing to reverse these molecular and functional disruptions.

Introduction

The World Health Organization defines obesity as a body mass index (BMI) equal to or greater than 30 kg/m². Despite this simple operational definition, obesity is a complex multifactorial condition¹. The prevalence and severity of this condition have been rising at an alarming rate each year². Between 1980 and 2017, the BMI distribution curve in the United States of America shifted towards severe obesity³. Moreover, severe obesity is projected to reach 257 million adults worldwide in 2025, nearly 1.5 times the 2014 estimate⁴. This trend may persist into future generations as the prevalence of obesity in children and adolescents remains high and continues to rise⁵.

Obesity in Europe and North America can be divided into three groups regarding severity: class I obesity (30–34.9 kg/m²), class II obesity (35–39.9 kg/m²) and class III obesity or higher: ≥40 kg/m²⁶. These three degrees of obesity are the pillars of obesity management that guide treatment recommendations in most clinical units⁶. Bariatric surgery (BS) is the treatment of choice when behavioral interventions and pharmacotherapy approaches fail to treat refractory obesity⁶. The American Society for Metabolic and Bariatric Surgery and the International Federation for the Surgery of Obesity and Metabolic Disorders recommend bariatric procedures for individuals with class II obesity or higher regardless of obesity-related comorbidities and for those with class I obesity if an associated metabolic disease is present⁶. Bariatric procedures work through multiple mechanisms, such as restriction and malabsorption^7,8. In restrictive procedures, like sleeve gastrectomy (SG), the gastric volume is conditioned to a smaller space, reducing the dietary intake⁷. On the other hand, hybrid approaches, such as Roux-en-Y Gastric Bypass (RYGB), add to the gastric volume restriction the malabsorption of nutrients by shortening the track between the small intestine and stomach^7,8. These two types of BS, SG and RYGB, are considered the actual gold standard treatment for refractory obesity due to their high efficacy in terms of weight loss (WL) and duration of effectiveness⁶. Even though BS is the most effective treatment for refractory obesity, not all patients succeed in losing weight⁹. Successful degree of WL is typically defined at 1 year after BS by evaluating the excess weight loss (EWL) or total body weight loss (TWL), $>$50% or $>$20%, respectively^10–12. While most individuals achieve a successful degree of WL, approximately 20% of individuals fail to achieve this goal^10,12,13. Moreover, for many, weight regain remains a major concern, resulting in the return of obesity-related comorbidities and a decrease in quality of life⁹. Weight regain is characterized by a progressive increase in weight following an initially successful WL¹⁴. It has been reported that 20–25% of patients experience weight regain after BS¹⁴, with ~50% regaining the weight lost within 2 years after surgery¹⁵. A prospective cohort study of 1406 RYGB patients also revealed that the weight regained within 2 years after surgery continues to increase for up to 5 years¹⁶, while the Swedish Obese Subjects study showed that weight gain persists for up to 8 years¹⁷. Failure to achieve long-term successful results is likely multi-factorial¹⁰. Understanding the physiological mechanisms of action through which BS promotes a successful WL outcome may help elucidate how this surgery promotes maintenance of WL, decreased food intake and reduced motivational drive to eat¹⁸.

The gut-brain axis has led to the suggestion that WL after BS is likely due to changes in the central nervous system (CNS) rather than only modifications in gastrointestinal malabsorption^19,20. As such, investigation the physiological mechanisms related to the success of the BS requires considering neural pathways. Among the central mechanisms implicated, the mesolimbic pathway plays a key role in eating behavior, as food stimuli activate brain regions within this pathway^21,22. Moreover, dysfunctions within this pathway are associated with obsessive-compulsive behaviors related to the rewarding effects of food consumption, such as food addiction^21,22. This link between food addiction and the mesolimbic pathway is further supported by evidence showing that individuals with food addiction, overweight, and obesity exhibit increased activation in mesolimbic brain regions, particularly in response to high-calorie food cues^23,24. Within the mesolimbic pathway, dopaminergic projections from the ventral tegmental area (VTA) to the nucleus accumbens (NAc) are thought to drive reward system activity^25,26. Notably, food addiction—like drugs of abuse—has been linked to a decrease in dopamine type 2 receptor (D2R) density due to dysfunctions within this pathway²⁷. Furthermore, blocking D2R increases appetite and results in weight gain²⁸, and treatment with dopamine receptor agonists that have a high affinity for D2R causes WL²⁹. Some hypotheses have been proposed to explain this link between central dopamine receptors and obesity^30,31. Yet, the role of these receptors in eating behavior, their impact on body weight, and their function in the context of successful WL through BS remains unknown.

In the brain, the D2R is found at high levels in striatal brain regions, which are crucial for reward processing³². The ventral striatum, in particular, plays a key role in driving immediate reward-based responses, emphasizing its central role in addiction disorders³³. The dorsal striatum also contributes to reward processing, particularly during decision-making³⁴. Given the role of striatum in the brain reward system, its involvement in addiction and obesity, we hypothesize that central dopamine receptors in striatal reward-related brain regions are critical to successful BS outcomes. Therefore, our main research question focuses on investigating the role of dopamine type 2 and 3 receptors (D2/3 R) in successful WL through BS. For that purpose, Positron Emission Tomography (PET) with [¹¹C]raclopride was used to map striatal brain D2/3 R nondisplaceable binding potential (BP_ND) of 19 women with a successful response to BS (pBS^success), 12 women with obesity (OB) and 17 normal-weight women (NW). To further address our aim, we investigated brain blood oxygenation level-dependent (BOLD) responses to food content in each group using functional Magnetic Resonance Imaging (fMRI). Moreover, we aimed to establish the link between molecular phenotypes and brain activity in regions involved in dopaminergic processing by investigating how activation within clusters of interest from visual food cue contrasts and key anthropometric and behavioral self-reported measures correlate with the binding values of the central dopamine receptors.

We found altered higher D2/3 R BP_ND values and increased brain activation in response to food-related cues contrasts (Food > Non-food, p = 0.05) within striatal brain regions in obesity compared to the other groups. Since no significant differences were found in striatal D2/3 R BP_ND values between pBS^success and NW, this study suggests that successful WL after BS may reverse striatal dopaminergic dysfunction associated with obesity.

Methods

The protocol for this cross-sectional multimodal neuroimaging study in women was previously published in BMJ Open journal³⁵.

Study Design and Ethics

This study was conducted under the Declaration of Helsinki. We obtained ethical approval from the local medical ethics committee of the Faculty of Medicine of the University of Coimbra (FMUC – approval ID, CE_Proc. CE-088/202) and Coimbra Hospital and University Centre (CHUC, approval ID OBS.SF.143/2021). Participants signed written informed consent, approved by the ethics committee, after a detailed explanation of the study procedure.

Dataset

To address the proposed goals of this study, a total of 48 women were included in the final dataset: 19 pBS^success, 12 OB and 17 NW. The pBS^success group was recruited from the Portuguese association of patients who underwent BS— Associação Portuguesa dos Bariátricos (APOBARI) and public advertisements, while the NW and OB groups were recruited from our volunteer database and public advertisements. We included only two types of BS, SG and RYGB. Extra study-specific inclusion criteria included signed informed consent, women aged between 18 and 65, BMI $\ge$30 kg/m² for OB and 18.5–24.9 kg/m² for NW. There was no BMI range restriction for participants in the pBS^success group as long as they demonstrated a successful response to BS. Our definition of success was based on the following criteria: >50% EWL or >20% TWL at 1 year after surgery^10–12. Since we included participants from 1−5 years postoperative: (1) we used a retrospective approach to verify these criteria at 1 year after surgery, considering the weight participants had at that time; and (2) ensured that, at the time of data acquisition, participants still met the criteria defined for 1 year after surgery (>50% EWL or 20% TWL). The exclusion criteria of this study followed the European Association of Nuclear Medicine (EANM) guidelines for brain neurotransmission PET imaging with D2R ligands³⁶ and were customized to our study design. These criteria include previous BS, less than 1 year or more than 5 years postoperative from BS, current treatment involving medication interacting with [¹¹C]raclopride uptake, active neoplastic or inflammatory disease, history of neurological disease, traumatic brain injury or psychiatric disorder, previous or current alcohol or other substance abuse and conditions that preclude Magnetic Resonance Imaging (MRI) or PET, such as pregnancy and claustrophobia. All participants performed clinical, anthropometric and behavior self-report measures assessment, MRI and PET acquisitions at the Institute of Nuclear Sciences Applied to Health (ICNAS). Although two participants in this dataset did not undergo MRI, all participants underwent PET with [¹¹C]raclopride and clinical, anthropometric and self-report behavior measures assessment. Therefore, aside from the results specifically derived from the MRI scan, all other analyses include these 48 participants.

Clinical, anthropometric and self-report behavior measures assessment

The key variables of the clinical assessment were collected using a clinical report form. They included age, hormonal status, list of CNS medication, metabolic diseases and physical activity, diet and smoking habits. In terms of anthropometric data, a commercial body analysis scale (Becken BAS-3057) was used to assess weight and fat percentage and a plastic tape meter was used to determine neck, waist and hip circumferences. BMI, waist/hip circumference ratio, TWL and EWL were calculated using anthropometric equations. Finally, a brief behavior evaluation using self-report measures was performed, including the Eating Disorder Examination Questionnaire (EDE-Q) of 28 items³⁷, the Three Eating Factor Questionnaire (TEFQ) of 21 items³⁸ and the Depression Anxiety and Stress Scale (DASS) of 21 items³⁹, all translated and validated for the portuguese population^40–42. EDE-Q of 28 items is also validated for the portuguese bariatric population⁴³. In 1994, Fairburn and Beglin published the EDE-Q³⁷, a widely used self-report questionnaire that reflect the severity of the psychopathology of eating disorders, where higher scores indicate higher levels of disordered eating attitudes and behaviors. On the other hand, the TFEQ is one of the most widely used instruments to measure eating behaviors, precisely cognitive restraint, uncontrolled eating and emotional eating⁴¹. Originally, TFEQ contained 51 items and was designed to assess eating behaviors in people with obesity⁴⁴. Later, a study using a large sample of individuals with and without obesity developed a revised and shorter version of TFEQ with 21 items³⁸. Lastly, DASS is a set of three self-report scales designed to measure the emotional states of depression, anxiety and stress⁴⁵.

PET acquisition and processing

Dynamic PET scan of the entire brain in 3D mode for 60 min (24 frames: 4 x 15 sec, 4 x 30 sec, 3 x 60 sec, 2 x 120 sec, 5 × 240 sec, 6 x 300 sec) together with Computed Tomography (CT) was acquired using PET/CT scanner (Siemens Biograph Vision 600) after intravenous bolus injection of a maximum of 15 mCi of [¹¹C]raclopride. The emission data was corrected for attenuation using the CT scan. The [¹¹C]raclopride synthesis was performed in the facilities of ICNAS according to the methods previously described⁴⁶. The PET images were reconstructed using the Ordered Subset Expectation Maximization (OSEM) 3D algorithm plus time-of-flight (TOF) information (8 iterations, 5 subsets)⁴⁷, with resolution modeling into a 440 × 440 image matrix with isotropic voxel spacing of 1.65 mm.

PET data processing was performed using the Pmod software (PMOD, version 4.105; PMOD Technologies; Zurich, Switzerland) and SPM12 (Statistical Parametric Mapping, version 12, Wellcome Trust Centre for Neuroimaging, London, UK). We first visually assessed all dynamic PET images to detect head motion. In cases where motion was detected, we averaged the first frames with negligible head motion to serve as a static reference to correct the frames with motion. After that, accurate co-registration (between PET images and their anatomical MRI images) was performed and further confirmed by visual inspection across all planes. Due to conditions that preclude MRI acquisition, two participants of this dataset could not undergo MRI scans. PET images were co-registered to an averaged template of 152 T1-weighted MRI scans in those two cases. Perfusion-like data acquired in the first minutes after tracer injection reflecting cerebral blood flow were used to improve the co-registration process⁴⁸. As such, the first 13 frames (1–10 min after tracer injection) were averaged to obtain a perfusion-like image and co-registered to the corresponding structural MRI scans by applying a normalized mutual-information-based rigid registration. The resulting transformation matrix was applied to all frames of the dynamic PET image. The maximum probability atlas (Hammers N30R83)⁴⁹ was then intersected with the gray matter segmentation map and used for the region of interest (ROI) definition. As binding of [¹¹C]raclopride is negligible in the cerebellum, the simplified reference tissue model was used to compute the parametric maps of the D2/3 R BP_ND. These maps were generated using the cerebellum as reference region and MRTM2 (multilinear reference tissue model 2)⁵⁰. Finally, spatial smoothing was performed using a Gaussian kernel with a Full Width Half Maximum (FWHM) of 5 mm and PET and MRI scans were spatially normalised to the Montreal Neurological Institute (MNI) template. ROI and voxel-based analyses were carried out. For ROI-based analysis, mean regional D2/3 R BP_ND values were extracted in three striatum subregions: ventral striatum, putamen and caudate. Although observations suggest that extrastriatal [¹¹C]raclopride measurements also represent a true D2/3 R signal⁵¹, we decided to restrict our analysis to striatal brain regions due to their role in the reward-related pathways of the central dopamine system⁵² and the high-intensity of [¹¹C]raclopride signal in these regions⁵¹.

MRI acquisition and processing

Two participants of this dataset could not undergo MRI scans, one from the OB group due to claustrophobia and another from the pBS^success group who had a metallic heart implant incompatible with our MRI scanner. Structural and functional MRI were acquired on a 3 T imaging system (MAGNETOM Prisma, Siemens Medical Solutions) using a 64-channel head-coil for the remaining participants. First, we acquired the structural MRI data using 3D-T1 weighted sequence (192 slices; echo time (TE): 3.5 msec; repetition time (TR): 2530 msec; voxel size: 1 × 1 × 1 mm; flip angle (FA): 7°; field of view (FOV): 256 × 256 mm). After that, we acquired the block-design fMRI paradigm (6 runs of ~2 min each) using 2D-T2* weighted sequences (72 interleaved slices; TE: 37 msec; TR: 1000 msec; voxel size: 2 × 2 × 2 mm; FA: 68°; FOV: 200 × 200 mm). In each run, participants were presented with 5 images of Low-calorie Foods, High-calorie Foods and Non-food Objects from the Cross-Cultural Food Images Database (CROCUFID)⁵³. For all participants, the sequence of the fMRI task was the same, where the first block switched between High- and Low-calorie Foods, while the block of Non-food Objects was interleaved in the middle. Finally, after the fMRI acquisition, the participants performed a rating task using a horizontal, 100 mm, bidirectional scale between the extremes “I don’t like it at all” and “I like it a lot” to rate High- and Low-calorie foods from fMRI visual task in terms of preferences.

Pre-processing of MRI data was performed in BrainVoyager 22.0 and 22.4 software (Brain Innovation, Maastricht, The Netherlands). Pre-processing included 3D head motion correction (trilinear / sinc interpolation), slice scan time correction (cubic spline) and temporal filtering (high pass using a cut-off frequency of 0.0125 Hz). This cut-off frequency value was calculated based on the duration of the fMRI paradigm, allowing us to select the frequency ranges relevant to the task while minimizing noise. Geometrical distortions were corrected using the COPE plugin⁵⁴. Finally, structural and fMRI data were co-registered and transformed into the MNI space, and spatial smooth was performed using a Gaussian kernel with an FWHM of 4 mm.

Statistics and reproducibility

This study includes a sample of 48 women divided into three groups: 19 pBS^success, 12 OB and 17 NW. Levene’s and Shapiro-Wilk tests were conducted to check the homogeneity of variances and normality of distributions, respectively, for analysis of variance (ANOVA), analysis of covariance (ANCOVA) and multivariate analysis of covariance (MANCOVA). For Pearson correlation, in addition to the Shapiro-Wilk test, Quantile-Quantile (Q-Q) plots of the model residuals were examined to assess whether the assumption of linearity was valid. The non-parametric alternatives were used when the data failed to meet at least one of these assumptions.

Clinical, anthropometric and self-report measures analysis

To assess whether the clinical, anthropometric and behavior self-report measures differ between groups, we use the statistical test ANOVA and post hoc Bonferroni tests for multiple comparisons. The non-parametric alternative, the Kruskal-Wallis test, was performed when the assumptions above mentioned were violated. In the NW group, one outlier was identified in the waist/hip circumference ratio and another in the fat percentage using the interquartile range (IQR) method (IQR > 3). As such, these two values were excluded from the statistical analysis. These analyses were performed in IBM SPSS Statistics for Windows, version 25 (IBM Corp., Armonk, N.Y., USA). Key anthropometric and behavior self-report measures were further used in the correlation analysis.

PET statistical analysis

MANCOVA and post hoc Bonferroni tests for multiple comparisons were performed to assess D2/3 R BP_ND alterations between groups at the ROI level. One outlier was identified in the mean D2/3 R BP_ND values of the putamen for the OB group using the IQR method (IQR > 3). As such, it was excluded from this analysis. D2/3 R BP_ND alterations between groups were also assessed at the voxel level by performing an ANCOVA using an F-Contrast ([1 −1 0; 0 1 −1]) and the false discovery rate (FDR) correction (p < 0.05). Second, we also test three different T-Contrasts: pBS^success > NW: [1 0 −1]; OB > pBS^success: [−1 1 0]; and OB > NW: [0 1 −1] using again FDR correction (p < 0.05). Finally, to investigate the influence of BS type on the PET analyses, the rank-based non-parametric alternative to MANCOVA was used to compare the two types of BS included in this study: SG and RYGB. Since the dopaminergic neurotransmission might fluctuate across age⁵⁵ and due to the scarcity of studies regarding the influence of the menstrual cycle on this topic^56,57, the menstrual cycle and age were used as a covariate in both ROI and voxel-based PET analyses. These analyses were performed in IBM SPSS Statistics for Windows, version 25 (IBM Corp., Armonk, N.Y., USA), SPM12 (Statistical Parametric Mapping, version 12, Wellcome Trust Centre for Neuroimaging, London, UK) and SnPM13 (Statistical nonParametric Mapping, version 13, University of Warwick, Coventry, UK, URL http://warwick.ac.uk/snpm).

MRI statistical analysis

We used the standard fMRI statistical analysis that relies on the general linear model (GLM) and two levels of analysis^58,59. In the first-level analysis, we performed a general linear model (GLM) for each participant. We obtained the predictor’s model by convolution of the time course of each condition with a two-gamma hemodynamic response function (HRF). Three predictors were distinguished: High-calorie Foods, Low-calorie Foods, and Non-food Objects. Even though none of the runs exceeded 5 mm of movement in any axis, we added motion parameters, such as spikes of motion pattern, as confounding predictors into the individual GLM model. In the second-level analysis, we performed a whole-brain Random Effects Analysis (RFX) two-factor ANOVA model to test whether the difference (Food > Non-Food Objects, p = 0.05) was significantly higher for the OB compared to the other two groups. We have used this model since the two-factor ANOVA introduced with BrainVoyager 20.4 supports unbalanced designs (groups with unequal subjects) by employing robust techniques such as Type I, II and III sums of squares, which ensures the validity of our conclusions. In addition a method for the correction of multiple comparisons was conducted by using cluster-size thresholding plugin⁶⁰ and a more recent work that extended this approach to 3D statistical maps⁶¹.

To define the fMRI activation metrics (i.e., β-values within clusters of interest from visual food cue contrasts) to be used in the correlation analysis, we applied a whole-brain RFX GLM contrasting High-calorie Foods + Low-calorie Foods (Food) >  Baseline to identify the activation clusters related to visual food cue processing. After FDR correction for multiple comparisons, clusters with a p < 0.001 for the whole striatum and p < 0.05 for smaller anatomical ROIs were considered significant. This approach accounts for the fact that smaller anatomical regions of interest typically require a more liberal threshold in localizers to ensure sufficient detection of activation. In addition, the cluster-size thresholding plugin method for multiple comparisons correction was also applied. The whole striatum was the first anatomical region used to define the activation clusters due to its role in reward-related pathways of the central dopamine system⁵². Moreover, the $\beta$-values extracted from this cluster were used in the correlation analysis with the binding values of [¹¹C]raclopride PET images, where the same region was used. In addition, we also extract the $\beta$-values from VTA, NAc and external globus pallidus (GPe). The VTA and NAc are the core structures of dopaminergic mesolimbic pathway^21,62, and the GPe, within the indirect pathways of the basal ganglia, projects via D2R and plays a critical role in reward-based, aversive learning and in the mechanisms underlying addiction^63,64. As such, we extracted the $\beta$-values for the cluster that resulted from the intersection between the anatomical boundaries of these ROIs and the whole RFX GLM analysis (Food > Baseline, p-FDR = 0.001 for the striatum and p-FDR = 0.05 for NAc and GPe) maps. See Fig. 1 and Table 1 for details. The $\beta$-values were extracted in 4 contrasts of interest: Food (High-calorie Foods + Low-calorie Foods) > Baseline; Food > Non-Food; High-calorie Foods > Low-calorie Foods; and High-calorie Foods > Non-Food.

Fig. 1. Clusters used in the fMRI analysis.

Clusters result from the intersection between the whole RFX GLM brain analysis (Food > Baseline, p-FDR = 0.001 for striatum [green] and p-FDR 0.05 for NAc [orange] and GPe [yellow]) and the anatomical boundaries of these regions. FDR false discovery rate; fMRI functional Magnetic Resonance Imaging; GLM general linear model; GPe external segment of the globus pallidum; NAc nucleus accumbens; RFX Random Effects Analysis.

Table 1.

List of regions of clusters used in fMRI analysis

Whole-brain RFX GLM (Food > Baseline)	ROIS	MNI coordinate (peak)			Nr voxels
		x	y	z
p-FDR = 0.001	Striatum	-17	-18	20	3687
p-FDR = 0.05	NAc	9	8	-14	3
	GPe	16	5	3	264

Clusters result from the intersection between the whole RFX GLM brain analysis (Food > Baseline, p-FDR = 0.001 for striatum and p-FDR = 0.05 for NAc and GPe) and the anatomical boundaries of these regions. FDR false discovery rate; fMRI functional Magnetic Resonance Imaging; GLM general linear model; GPe external segment of the globus pallidum; NAc nucleus accumbens; RFX Random Effects Analysis.

Correlation analysis

Finally, Pearson correlation (r), or the non-parametric alternative Spearman correlation (r_s) was performed to examine how fMRI metrics related to food cue processing and key anthropometric and behavior self-report measures correlate with the dopaminergic system. This analysis was performed in RStudio: Integrated Development Environment for R. Posit Software, PBC, Boston, MA. Of all the behavior self-report questionnaires and scale used, we selected the TEFQ for the correlation analysis because its subscales are closely aligned with the emotional dimensions of food rewards, making it especially suitable for examining correlations with the dopaminergic system.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Clinical, Anthropometric and Self Report Measures Analysis

ANOVA proved that the experimental groups were matched for age, showing no statistical difference between them (F(2,45) = 0.06, p = 0.94). Similarly, Fisher’s exact test (value = 7.51, p = 0.74 (two-tailed)) found no significant association between menstrual cycle phases and groups. See Table 2 for details on the clinical and anthropometric characteristics of the dataset. The self-report behavior data are summarized in Supplementary Table S1. For additional clinical details (such as CNS medication), see Supplementary Table S2.

Table 2.

Clinical and anthropometric characteristics

	Clinical and anthropometric characteristics			pBSsuccess (n = 19)	OB (n = 12)	NW (n = 17)	Statistical analyses
Clinical	Age (years)			45.44$\pm$8.72, 31.09 - 61.16	44.00$\pm$13.65, 24.46 - 62.20	44.98$\pm$11.38, 25.62 – 59.62	F(2,45) = 0.06, p = 0.94
	Type of BS		RYGB	8 (42.11%)	NA	NA	NA
			SG	11 (57.89%)	NA	NA	NA
	Follow-up time (years)			1.99$\pm$1.09, 1.07−4.86	NA	NA	NA
	Menstrual cycle phases	Progestin-only HC		4 (21.05%)	2 (16.67%)	1 (5.88%)	Fisher’s exact test (value = 7.51, p = 0.74 (two-tailed))
		Combined HC	Active	5 (26.32%)	4 (33.33%)	3 (17.65%)
			Placebo	0 (0%)	1 (8.33%)	0 (0%)
		Naturally cycling	Follicular	2 (10.53%)	0 (0%)	2 (11.76%)
			Luteal	2 (10.53%)	1(8.33%)	4 (23.53%)
		Menopause		6 (31.58%)	4 (33.33%)	7 (41.18%)
Anthropometric	BMI (kg/m2)			26.94 ± 4.07, 20.21 − 34.74	36.97 ± 3.13, 32.96 – 41.16	22.13 ± 1.66, 19.79 − 24.77	H(2) = 33.14, p = 6.36$x{10}^{-8}$
	Neck circumference (cm)			31.11$\pm$2.23, 27 − 36	34.75$\pm$2.34, 30 − 38	29.47$\pm$2.63, 25 − 36	F(2,45) = 17.25, p = 2.75$x{10}^{-6}$
	Waist/hip circumference ratio			0.85$\pm$0.05, 0.79 – 0.95	0.83$\pm$0.07, 0.74 – 0.96	0.81$\pm$ 0.04, 0.75 –0.90	H(2) = 2.72, p = 0.26
	Fat (%)			40.12$\pm$4.35, 31.30 − 47.10	47.88$\pm$1.85, 45 − 50	36.18$\pm$2,52, 32.3 − 40.50	H(2) = 29.32, p = 4.30$x{10}^{-7}$
	TWL (%)			36.16$\pm$6.52, 21.56 – 47.02	NA	NA	NA
	EWL (%)			77.89$\pm$17.82, 42.84– 116.34	NA	NA	NA

BMI body mass index; BS bariatric surgery; EWL excess weight loss; F ANOVA F-statistic; H Kruskal–Wallis H-statistic; HC hormonal contraceptive; NW normal-weight women; OB women with obesity; pBS^success women with a successful response to BS; RYGB Roux-en-Y Gastric Bypass; SG sleeve gastrectomy; TWL total body weight loss. Non-categorical data are expressed as mean ± standard deviation, range and categorical data as the number of individuals (percentage).

ANOVA or the non-parametric alternative, the Kruskal-Wallis test, found significant differences between groups for all anthropometric metrics except for the waist/hip circumference ratio. See Table 2 for details. Post hoc Bonferroni tests for multiple comparisons reveal statistically significant differences between all groups except when comparing pBS^success to NW for the neck circumference (p = 0.143) and fat percentage (p = 0.081). See Supplementary Table S3 for details. These results reveal that the pBS^success group is closer to the NW group than to the OB group for these measures, reflecting improvements in anthropometric metrics after surgery.

Regarding behavior self-report measures, for the EDE-Q, we found significant differences between groups for all subscales and the global score. See Supplementary Table S1 for details. In terms of global score, the OB group, ranging from 1.51 to 4.15 (mean = 2.52; standard deviation [SD] = 0.79), was the only group to score above the cut-off for the non-clinical portuguese population⁴². For the TEFQ, we found differences for uncontrolled eating (p = 8.456$x{10}^{-4}$) and emotional eating (p = 0.032) but not for cognitive restraint (p = 0.575). See Supplementary Table S1 for details. In this questionnaire, the OB group also exceed the non-clinical portuguese population ranges for the subscale of uncontrolled eating⁴¹ and had the highest scores across all subscales. Finally, in DASS, there were differences between groups for anxiety (p = 1.364$x{10}^{-4}$) and stress (p = 0.018) but not for depression (p = 0.223). See Supplementary Table S1 for details. Even though the pBS^success group showed mean values in the mild ranges of depression, anxiety and stress, the OB group exhibited higher levels of anxiety and stress, falling within the moderate range⁴⁰. In the Post hoc Bonferroni tests for multiple comparisons, no differences were found in any behavior self-report measures when comparing pBS^success to OB. Still, post hoc tests also revealed that statistically significant differences were more pronounced when comparing the OB group to the NW group than when comparing the pBS^success to NW. Furthermore, no statistically significant differences were observed between pBS^success and NW for EDE-Q restraint (p = 0.242), TEFQ uncontrolled eating (p = 0.306), TEFQ emotional eating (p = 0.171) and DASS stress (p = 0.082). See Supplementary Table S4 for details. Altogether, these results show that although anthropometric changes are more pronounced, there is also an improvement in behavior self-report measures after a successful WL through BS.

PET Analysis

Mean BP_ND [¹¹C]raclopride PET images of each group showing the D2/3 R availability are displayed in Fig. 2. As expected, parametric maps reveal the high intensity of [¹¹C]raclopride signal in striatal brain regions⁵¹. In addition, these maps also reveal that D2/3 R BP_ND is similar in pBS^success and NW groups and higher in the OB group, suggesting that obesity-related effects on D2/3 R availability may be reversible within region of the reward system.

Fig. 2. Mean D2/3 R BP_ND [¹¹C]raclopride PET images.

D2/3 R availability for each group: pBS^success (n = 19); OB (n = 12) and NW (n = 17). BP_ND non-displaceable binding potential; D2/3 R dopamine type 2 and 3 receptors; NW normal-weight women; OB women with obesity; pBS^success women who have successfully responded to bariatric surgery; PET Positron Emission Tomography.

Region of interest (ROI)-based PET analysis

MANCOVA found a statistically significant D2/3 R BP_ND difference between groups in the ventral striatum (F(2,42) = 3.286, p = 0.046) and no difference in putamen (F(2,42) = 2.996, p = 0.061) and in caudate (F(2,42) = 0.568, p = 0.571), while adjusting for age and menstrual cycle phase. Post hoc Bonferroni tests for multiple comparisons only found a significant difference between OB and NW (p = 0.04). These results support the notion that the mean D2/3 R BP_ND values in the pBS^success and NW groups are similar within the ventral striatum, suggesting that obesity-related effects in the binding values of this reward region may be reversible after BS. See Fig. 3 and Table 3 for details.

Fig. 3. ROI-based PET analysis.

Scatter plot (mean $\pm$ SD) showing the brain D2/3 R BP_ND values in three anatomical brain ROIs for each group: NW (n = 17) is depicted using gray rhombuses, OB (n = 11 for putamen and n = 12 for other ROIs) with orange triangles, and pBS^success (n = 19) with blue circles. BP_ND non-displaceable binding potential; D2/3 R dopamine type 2 and 3 receptors; NW normal-weight women; OB women with obesity; pBS^success women who have successfully responded to bariatric surgery; PET Positron Emission Tomography; ROIs regions of interest.

Table 3.

Region of interest (ROI)-based Positron Emission Tomography (PET) analysis results

Brain regions	PET with [11C]raclopride (BPND)			Statistical analyses
	pBSsuccess (n = 19)	OB (n = 12)	NW (n = 17)
Ventral striatum	2.02±0.30, 1.58 – 2.62	2.25±0.38, 1.69 – 2.85	1.89±0.23, 1.51 – 2.29	F(2,42) = 3.286, p = 0.046
Putamen	3.31$\pm$0.32, 2.52 – 3.94	3.53$\pm$0.24, 3.07 – 3.96	3.27$\pm$0.24, 2.86 – 3.80	F(2,42) = 2.996, p = 0.061
Caudate	2.35$\pm$0.30, 1.84 – 3.01	2.54$\pm$0.39, 1.97 – 3.55	2.35$\pm$0.27, 1.95 – 2.81	F(2,42) = 0.568, p = 0.571

Regional values of brain dopamine type 2 and 3 receptors (D2/3 R) non-displaceable binding potential (BP_ND) for each group. F = Multivariate analysis of covariance. NW  = normal-weight women; OB = women with obesity; pBS^success  = women who have successfully responded to bariatric surgery. Data are expressed as mean ± standard deviation, range.

The rank-based non-parametric alternative to MANCOVA, whilst also adjusting for age and menstrual cycle phase, did not find any statistically significant D2/3 R BP_ND difference between the type of BS for any ROI: ventral striatum (F(1,15) = 0.718, p = 0.410); putamen (F(1,15) = 2.214, p = 0.158); and caudate (F(1,15) = 1.413, p = 0.253), ruling out the possibility that BS type may influence the above PET results obtained. Mean regional D2/3 R BP_ND values for each type of BS is presented in Supplementary Table S5 and displayed in Supplementary Fig. S1.

Voxel-based PET Analysis

ANCOVA also using the menstrual cycle and age as covariates found cluster-level significant D2/3 R BP_ND alterations between the groups over the whole striatum. T-contrasts confirm and extend to ROI-based PET analysis the result of differences between pBS^success and OB, reinforcing the possibility that obesity-related effects on D2/3 R availability in striatal reward regions may be reversible following successful BS outcomes. See Figs. 4 and 5 for details.

Fig. 4. Voxel-based PET analysis.

F-Contrast —comparison of brain D2/3 R BP_ND between groups: pBS^success (n = 19); OB (n = 12) and NW (n = 17). Statistical analysis was performed using the FDR correction (p < 0.05) and an extent threshold of 226 voxels. BP_ND non-displaceable binding potential; D2/3 R dopamine type 2 and 3 receptors; FDR false discovery rate; PET Positron Emission Tomography.

Fig. 5. Voxel-based PET analysis.

T-Contrasts —comparison of brain D2/3 R BP_ND between pBS^success (n = 19), OB (n = 12) and NW (n = 17) groups. Statistical analysis was performed using the FDR correction (p < 0.05) and an extent threshold of 522 voxels for OB $>$ pBS^success and 489 for OB $>$ NW. BP_ND non-displaceable binding potential; D2/3 R dopamine type 2 and 3 receptors; FDR false discovery rate; NW normal-weight women; OB women with obesity; pBS^success surgery; PET Positron Emission Tomography.

fMRI Analysis

Whole-brain RFX ANCOVA testing the contrast (Food $>$ Non-food Objects, p = 0.05) revealed significantly higher activation within striatal brain regions for the OB compared to the other groups. This fMRI finding shows additional insights into the PET results by revealing differences in the ventral striatum, suggesting functional and molecular disruption in obesity. See Fig. 6 for details.

Fig. 6. Whole-brain fMRI analysis.

Whole-brain RFX ANCOVA (Food > Non-Food, p = 0.05) revealing significantly higher activation within the striatum for the OB (n = 11) compared to the NW (n = 17) and pBS^success (n = 18) groups. ANCOVA analysis of covariance; NW normal-weight women; OB women with obesity; pBS^success women who have successfully responded to bariatric surgery; RFX Random Effects Analysis.

Correlation analysis

Anthropometric data

Only the OB group exhibited correlations between mean regional D2/3 R BP_ND values and anthropometric data. In obesity, we found a moderate negative correlation between mean D2/3 R BP_ND in the ventral striatum and fat percentage [r = -0.615, p = 0.033] and BMI [r = -0.609, p = 0.035]. See Supplementary Fig. S2 for details. To better illustrate this significant correlation, as well as the relationship between BMI and BP_ND in the ventral striatum, we provide Fig. 7. This figure demonstrates that there is no evident pattern in the relationship between BMI and BP_ND in the ventral striatum across all subjects, supporting the correlation analysis results [r = 0.260, p = 0.074]. Moreover, pBS^success and NW groups show a similar pattern, as visually assessed by the distribution of data points, suggesting no correlation. These visual observations also corroborate the correlation analysis results for the pBS^success and NW groups, with [r = -0.173, p = 0.478] and [r = -0.077, p = 0.769], respectively. In contrast, the OB group shows a clearer drop in the distribution of data points, confirming the moderate negative correlation found in the correlation analysis [r = -0.609, p = 0.035] (Supplementary Fig. S2).

Fig. 7. Association between BMI and vSTR (BP_ND).

Scatter plot showing the relationship between BMI and vSTR (BP_ND) for each group: NW (n = 17) is depicted using gray rhombuses, OB (n = 12) with orange triangles, and pBS^success (n = 19) with blue circles. BMI Body Mass Index; BP_ND non-displaceable binding potential; NW normal-weight women; OB women with obesity; pBS^success women who have successfully responded to bariatric surgery; vSTR ventral striatum.

Self-report behavior Three Eating Factor Questionnaire (TEFQ)

For TEFQ, across all subjects, we found a weak positive correlation between uncontrolled eating and D2/3 R BP_ND in the three ROIs used: ventral striatum [r = 0.388, p = 0.006], putamen [r = 0.297, p = 0.043] and caudate [r = 0.291, p = 0.045]. For ventral striatum was also found a weak positive correlation between D2/3 R BP_ND in this region and emotional eating [r = 0.312, p = 0.031]. In contrast, for cognitive restraint it was found a weak negative correlation with D2/3 R BP_ND in caudate [r = -0.298, p = 0.040]. For the OB group, we found a positive strong correlation between D2/3 R BP_ND in caudate and uncontrolled eating [r_s(10) = 0.588, p = 0.048] and emotional eating [r_s(10) = 0.710, p = 0.012]. Finally, in the NW group there was a moderate negative correlation between D2/3 R BP_ND in caudate and cognitive restraint [r(15) = -0.544, p = 0.024]. See Supplementary Fig. S3 for details.

Measures of fMRI paradigm

In terms of fMRI measures related to food cue processing, we first investigated the mean $\beta$-values extracted from the cluster within the anatomical boundaries of the striatum (Fig. 1, striatum in green). We only found significant results for OB group, a positive correlation between the mean β-values within striatum from Food > Baseline contrast and D2/3 R BP_ND in the ventral striatum [r = 0.604, p = 0.049]. See Supplementary Fig. S4 for details.

Second, we correlate the mean $\beta$-values extracted from the clusters within the anatomical boundaries of the NAc (Fig. 1, NAc in orange) and GPe (Fig. 1, GPe in yellow). Across all subjects, a weak positive correlation was identified between D2/3 R BP_ND in the ventral striatum and the mean $\beta$-values within NAc from the Food > Baseline contrast [r = 0.310, p = 0.036] and Food > Non-food contrast [r = 0.305, p = 0.040]. See Supplementary Fig. S5 for details. Interestingly, in the OB group, the correlations between mean $\beta$-values extracted from the NAc and GPe clusters and D2/3 R BP_ND in dorsal striatum were negative. Specifically, we found the following negative correlations: 1) between the mean $\beta$-values within NAc from High-calorie Foods > Low-calorie Foods contrast and D2/3 R BP_ND in the caudate [r = -0.633, p = 0.037]; 2) between the mean β-values within GPe from High-calorie Foods > Non-food objects contrast and D2/3 R BP_ND in the putamen [r = -0.661, p = 0.044]. See Supplementary Figs. S5 and S6 for details.

Finally, we also test the correlation between mean regional D2/3 R BP_ND and the results of the food preference rating task, which was performed using pictures of food from the fMRI paradigm. Across all subjects we found a weak positive correlation between high/low-calorie food ratio preferences and D2/3 R BP_ND values in the putamen [r = 0.349, p = 0.018] and in the ventral striatum [r = 0.402, p = 0.005]. For the OB group, we found a stronger positive correlation between D2/3 R BP_ND in the putamen and high-calorie food [r = 0.803, p = 0.006] and high/low-calorie food ratio [r = 0.791, p = 0.005] preferences. See Supplementary Fig. S7 for details.

Discussion

We have performed a neuroimaging multimodal (PET and fMRI) study to investigate central mechanisms in patients with successful WL after BS. This cross-sectional study in women addresses our main research question, suggesting that successful WL after BS may reverse the striatal dopaminergic molecular and functional dysfunction observed in obesity.

Compared to normal-weight controls and patients who underwent BS, individuals with obesity exhibited significantly higher D2/3 R BP_ND values and increased brain activation to Food vs Non-food cues within striatal reward-related brain regions. Regarding fMRI, although two meta-analyses investigating brain responses to food content using conservative multiple comparison correction methods found no differences between individuals with obesity and lean^65,66, a meta-analysis employing less conservative methods and a similar fMRI paradigm to ours—comparing Food vs. Non-food cues—was consistent with our findings⁶⁷. The authors proposed that the increased activation in reward-related areas observed in individuals with overweight and obesity in response to food cues supports the incentive salience theory for this population⁶⁷. In terms of PET analyses, our results lends direct support to the overall consensus on this topic, which proposes higher striatal D2/3 R BP_ND for class I/II obesity (30-39.9 kg/m²)^68,69. Since [¹¹C]raclopride seems to be more sensitive in reflecting dopamine tone⁷⁰, the higher striatal D2/3 R BP_ND values found in obesity may reflect a lower dopaminergic tone. This interpretation also resonates with the results of other authors who associate mild obesity with lower striatal dopaminergic tone^31,71. Our PET results also reveal no differences in striatal D2/3 R BP_ND values between patients who underwent BS and normal-weight controls, suggesting that BS may reverse striatal dopaminergic dysfunction observed in obesity. A recent cross-sectional single-photon emission computed tomography (SPECT) study also reported altered D2/3 R availability in obesity and no differences between patients who underwent BS and normal-weight controls⁷². Although the authors did not specify whether the BS patients were classified as successful or unsuccessful, they assessed long-term outcomes and reported a BMI for this group below the obesity threshold⁷². Longitudinal PET and SPECT studies investigating brain dopamine receptor availability changes after BS have mixed results (with the latter being only semi-quantitative). Some studies have reported higher BP of these receptors after BS^30,73, while others have found lower BP⁷⁴ or even no differences^75,76. We highlight that postoperative follow-up time and the extent of WL or BMI changes— particularly whether patients who underwent BS are successful or unsuccessful responders—may impact D2/3 R assessment. Considering that only one study⁷³ included patients with a follow-up time longer than 1 year and that the mean postoperative BMI in all studies remained within the obesity range, these mixed results do not contradict our hypothesis on patients who underwent BS with successful WL outcomes. Finally, PET results comparing the types of bariatric surgery, suggest that among individuals with successful BS outcomes, the type of BS did not display differential striatal dopamine receptor binding values. As such, the potential striatal dopamine receptor availability changes after successful BS in women appear to be attributed to successful WL rather than the specific type of surgery performed. Furthermore, our results indicate an improvement in anthropometric and behavior self-report measures after a successful WL following BS, suggesting that physical and behavioral changes accompanied central changes after successful WL outcomes. However, it is important to note that behavioral improvements seem to occur more gradually compared to anthropometric changes. Several studies corroborated our findings on reductions in BMI^73,77–79, waist circumference⁷⁹ and body fat⁷³ after surgery. In terms of self-report behavior measures, a systematic review and meta-analysis also concluded that emotional eating scores decreased up to 12 months after surgery⁸⁰. However, results were mixed for larger follow-up times⁸⁰. The authors proposed that emotional eating behavior may improve in the first year after surgery due to immediate postoperative clinical education, support, and physiological intolerance to high-fat or high-sugar foods⁸⁰. We note that other factors, such as weight regain after BS, may also contribute to the mixed results found after the first year of surgery. Finally, several studies have been using the DASS instrument in the bariatric population. Cross-sectional studies found that the majority of patients after surgery experienced normal self-report measures of stress, anxiety, and depression^81,82. Two longitudinal studies reported controversial conclusions. One study found that while depression levels improved 1 year after BS, anxiety and stress levels increased over the same period⁸³. In contrast, the other study found improvements in all the subscales of the DASS instrument⁸⁴. While the first study⁸³ did not report WL data for the BS patients studied, the latter study⁸⁴ included patients with a median TWL of 26% and EWL of 64%. As such, the latter study, with a BS population similar to ours—comprising individuals with a successful response to BS (EWL$\ge$50% or TWL$\ge$20%)—supports our results.

BMI measure has been the most common metric used to reflect altered D2R function. However, the findings across studies are contradictory. Some studies report a positive correlation between dopamine receptor binding and BMI^85–88, one indicated a negative correlation⁸⁹ and others found no correlation^90–93. A study reported that in subjects with a BMI of 25 kg/m² or higher, D2R availability was negatively associated with BMI for [¹¹C]raclopride⁹⁴. Accordingly, we found a negative correlation between D2/3 R BP_ND in the ventral striatum and BMI of 30 kg/m² or higher—OB group. As the correlation between D2/3 R BP_ND and BMI found for obesity was negative, this result is consistent with the inverted U-shaped relationship between BMI values and dopaminergic tone³¹, where greater obesity severity is associated with lower D2/3 R BP_ND values. We also found statistically significant negative correlations between D2/3 R BP_ND and fat percentage for the OB group, indicating that, in addition to BMI, other anthropometric metrics are associated with the availability of these receptors in this group. This last correlation raises the question of whether the negative association between D2/3 R BP_ND and BMI for obesity relies on excessive body fat. This distinction is relevant since higher BMI values do not always reflect body fat⁹⁵. Higher BMI values can be due to increased muscle mass, influenced by several factors such as sex, age, genetics, exercise, and ethnicity⁹⁵. Finally, the lack of correlation between D2/3 R BP_ND and BMI, EWL and TWL for the pBS^success group, suggests that BMI and WL metrics does not have an impact on the availability of these receptors among successful BS responders. Regarding behavioral self-report measures, previous studies have suggested that emotional eating behavior and discounting monetary reward tasks correlate positively with striatal D2R binding values across normal-weight⁹⁶ and individuals with obesity^90,96. Our study also found a positive correlation between D2/3 R BP_ND and uncontrolled and emotional eating across all subjects and in the OB group. These results suggest that higher D2/3 R binding values are associated with greater self-report measures of uncontrolled and emotional eating behaviors, particularly in obesity, where the correlation strength was stronger compared to all subjects. Moreover, since the interval of uncontrolled eating and emotional eating subscales exceeded the non-clinical portuguese population range⁴¹ for the OB group, these results indicate that obesity has the least control over their food intake and the highest tendency to overeat and use food to deal with their emotions. Altogether, these results show the existence of an association between altered higher D2/3 R BP_ND and dysfunctional self-report measures of eating behaviors. For the NW group, as the mean value of cognitive restraint was in the non-clinical range⁴¹, and the correlation found was negative, we can associate lower D2/3 R binding values with functional strategies of eating behaviors. Finally, in the OB group, we also observed that mean $\beta$-values within striatum from Food > Baseline contrast and high-calorie food preferences correlated positively with mean D2/3 R binding values in the ventral striatum and in the putamen, respectively. These correlations highlight that, in obesity, brain responses to food-related cues may be driven by immediate reward gratification and impulsive behaviors (ventral striatum)⁹⁷, while food preferences may reflect habitual learning processes (putamen)⁹⁸. When using the NAc and GPe as regions of interest to extract the mean $\beta$-values, across all subjects we observed a weak positive correlation between D2/3 R BP_ND in the ventral striatum and the mean $\beta$-values within NAc. These results suggest that, generally, higher D2/3 R binding values in the ventral striatum are associated with increased brain activity in the NAc in response to food cue visual stimuli. Interestingly, in the OB group, the correlations between mean β-values extracted from the NAc and GPe clusters and D2/3 R BP_ND in dorsal striatum were negative. In the indirect basal ganglia pathway, D2R-expressing medium spiny neurons that originate in the dorsal striatum innervate the GPe and NAc⁹⁹. Since this pathway regulates motivation by inhibiting reward-seeking behavior^26,99, the observed negative correlation found in these regions also indicates incentive salience to food-associated stimuli in obesity. Altogether, these results show that altered higher D2/3 R BP_ND values in obesity may enhance motivation toward food cues, driving cravings and activating brain circuits involved in compulsive behavior regulation and reward-seeking processes. The only two studies we found that explored the relationship between central dopamine binding values and brain responses to food cues and food preferences align with and support our findings^100,101.

In summary, we propose that the higher striatal D2/3 R BP_ND values found in obesity, which can be interpreted to indicate lower dopaminergic tone, might play a key role in the overconsumption of food under positive energy balance conditions. We also demonstrate that striatal regions show critical associations between altered higher D2/3 R BP_ND values and dysfunctional self-report measures of eating behaviors, incentive salience to food cues, and high-calorie food preferences in obesity. Specifically, higher binding values in obesity may attribute incentive salience to food-associated stimuli so that food cues elicit craving and trigger neural networks linked to compulsive behavior control and reward-seeking processes. Moreover, successful WL after BS may reverse the striatal molecular and functional dopaminergic disruptions observed in obesity. The underlying mechanisms of this response appear to be complex, involving a combination of central dopaminergic neurotransmission, as well as changes in anthropometric factors and behavior following surgery.

The results of this work should be interpreted according to its limitations. Reduced dopamine receptor availability has been associated with caloric restriction, even in the absence of BMI effects¹⁰². However, a group of individuals who did not undergo BS matched for BMI with the BS group is needed to confirm that the observed results were not primarily driven by BMI differences. Given the cross-sectional design, causal inferences cannot be drawn. A longitudinal study including patients with a successful response to BS is needed to confirm the findings and conclusion of this cross-sectional study design. This work is also limited to the small sample size per group and women only, in whom the menstrual cycle phase was used as a covariate. Finally, the radiotracer used in this work, [¹¹C]raclopride, is susceptible to endogenous dopamine. As such, it does not distinguish between receptor density, affinity and the amount of endogenous neurotransmitter occupancy. Nevertheless, dopaminergic tone seems to have a greater impact on the obesity-related differences found in central dopamine receptors⁷⁰. As such, radiotracers with displaceable binding like [¹¹C]raclopride can reveal these differences due to displacement, whereas other radiotracers may not^87,103. Future work should also investigate individuals with unsuccessful BS outcomes. Despite the lack of studies, we expect individuals who experience insufficient WL or weight regain after surgery to present similar regional mean D2/3 R BP_ND values as individuals with obesity. Moreover, similarly to individuals with obesity, we also expect this group to show a positive correlation between measures of dysfunctional eating behavior and altered higher D2/3 R BP_ND values, with no improvement in anthropometric and disordered eating attitudes and behaviors following surgery.

Author contributions

M.L.P., J.C., and M.C.-B. designed the study protocol. J.C. and M.C.-B. designed the protocol of the MRI experiment. M.L.P. and J.C. performed the acquisition and analysis of MRI data. M.L.P. performed clinical, anthropometric and self-report measures assessment and PET data analysis. M.C.-B. and J.C. supervised the project, M.C.-B. provided the funding for PET and MRI acquisition, and A.A. provided expertise in [¹¹C]raclopride synthesis. M.L.P. wrote the first draft of the manuscript. M.L.P., J.C., A.A. and M.C.-B. critically revised the manuscript and approved the final version.

Peer review

Peer review information

Communications Medicine thanks the anonymous reviewers for their contribution to the peer review of this work.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. To protect participant privacy, this information is not publicly available. Supplementary Data 1 contains the source data for Figs. 3–7 and for all Supplementary Figures.

Code availability

The statistical analyses were performed IBM SPSS Statistics for Windows, version 25 (IBM Corp., Armonk, N.Y., USA) and RStudio: Integrated Development Environment for R. Posit Software, PBC, Boston, MA.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s43856-025-01079-z.