A cohort of 589 healthy, young Caucasians enrolled in the University of Cologne were recruited and genotyped for the rs9939609 T/A variant of the FTO gene (hereafter: FTO ^– group (TT) and FTO⁺ group (AT, AA)) as previously described (30, 31).

Participants were invited for further partially overlapping sub-studies. Eighty-seven participants were included in this sub-study, in which all participants underwent an impulsivity assessment using the Barrattt Impulsivity Scale-11 among other questionnaires (Becks Depression Inventory (BDI-2), Sensitivity to Punishment and Reward Questionnaire (SPSRQ), Alcohol Use Disorders Identification Test (AUDIT)), a buccal swab for DNA isolation and genotyping, a blood draw to assess the metabolic status and MRI imaging. The assessment of the Barrattt Impulsivity Scale and DNA isolation from the buccal swabs were performed as part of a previously published study (31). Inclusion criteria for the analysis performed in this study comprised: fully answered BIS-11 questionnaire and complete MRI imaging, no known diseases, fasting glucose and insulin within normal range, BDI-2 within normal range. 6 participants had to be excluded from further analyses due to unanswered BIS-11 questionnaire (5) and suspected diabetes mellitus (1) as indicated by a Homeostasis Model Assessment of Insulin Resistance (HOMA-IR, (41)) >5. In total, eighty-one participants were included in the final analysis ( Table 1 ), 39 non-carriers (FTO^–; 24.8 ± 3.5 years, 22.6 ± 2.6 kg/m²) and 42 risk allele carriers (FTO⁺; 25.7 ± 4.4 years; 23.2 ± 3.2 kg/m²).

All participants gave written informed consent to participate in the experiment, which had been approved by the local ethics committee of the Medical Faculty of the University of Cologne (Cologne, Germany, 10-226).

Impulsivity in participants was assessed by the German version of the BIS-11 questionnaire (42). The BIS-11 is the most often used self-report questionnaire to assess trait impulsivity (43). It consists of 30 items and is divided into 6 first-order (Attention, Motor, Self-control, Perseverance, Cognitive Instability, Cognitive Complexity) and 3 second-order factors (Attentional impulsiveness, Motor impulsiveness, and Non-planning impulsiveness). Each item is measured on a 4-point ‘Likert Scale’ ranging from “rarely/never” to “almost always/always”. The total score is calculated by summing all the item ratings, of which 8 cover attentional impulsiveness, 11 motor impulsiveness, and 11 non-planning impulsiveness, with a possible total range from 30 to 120.

Isolation of DNA from buccal swabs was performed using the QIAamp DNA Blood Mini Kit (# 51106, Qiagen GmbH, Hilden, Germany) according to the manufacturer’s instructions. The concentration and quality of the DNA were determined with a ND-1000 UV/Vis-Spectrophotometer (Peqlab GmbH, Erlangen, Germany). SNP genotyping for rs9939609 (FTO) was performed with 20 ng of DNA in triplicates using allelic discrimination assays (TaqMan SNP Genotyping Assays, Applied Biosystems by Life Technology GmbH, Darmstadt, Germany). The genotyping PCR was carried out on a 7900HT Fast Real-Time PCR System (Applied Biosystems) and the resulting fluorescence data were analyzed with Sequence Detection Software (SDS) 2.3 (Applied Biosystems) as already previously described for this cohort of patients.

Blood was drawn after an overnight fasting period of at least 8 hours to determine insulin, glucose, cholesterol and triglyceride values to compare the metabolic state of the participants. Insulin resistance was calculated according to the homeostatic model of insulin resistance (HOMA-IR) (41).

Imaging was performed on a 3 Tesla MRI scanner (Siemens TIM Trio, Erlangen, Germany). First, for each subject structural MRI data sets were acquired in a 12-channel array head coil and a maximum gradient strength 40 mT/m with a whole-brain field of view (T1-weighted: MDEFT3D; TR = 1930 ms, TE = 5.8 ms, resolution 1.0 × 1.0 × 1.25 mm, flip angle 18°, sagittal; T2-weighted: RARE; TR = 3200 ms, TE = 458 ms, 176 sagittal slices, resolution 1.0 × 1.0 × 1.0 mm³). Subsequently diffusion-weighted magnetic resonance imaging (dMRI) was conducted using spin echo planar imaging [SE-EPI; TR = 12,000 ms, TE = 100 ms, resolution 1.7 × 1.7 × 1.7 mm³, flip angle: 90°, Field of View (FoV): 220 × 220 × 122 mm³, bandwidth: 1345 Hz/pixel, orientation: axial, data matrix: 128 × 128 × 72, PAT factor 2, partial Fourier 6/8] with double-spin echo preparation (44). Diffusion-weighting was isotropically distributed along 60 diffusion-weighted directions (b value = 1000 s/mm²). Additionally, in each subject, seven data sets with no diffusion weighting (B0) were acquired initially and interleaved after each block of 10 diffusion-weighted images as anatomical reference for motion correction. To optimize the signal-to-noise ratio, when available scans from three assessments were averaged.

To create seed masks informing the tractography approach, we used anatomical constraints to outline the regions of VTA/SN, NAc, and caudate nucleus. Specifically, the mask location comprising the midbrain (VTA/SN) was chosen upon comparison across standard anatomical atlases (45, 46) as detailed in Pelzer et al. (47). For the midbrain mask, the following structures marked the border: ventromedially the red nucleus, latero-dorsally the subthalamic nucleus, and anteriorly the cerebral peduncles. The SN/VTA mask was labeled on an average T2-image in standard space in axial plane, and subsequently checked in coronal and sagittal plane. The masks for NAc and the caudate were retrieved from the Harvard-Oxford subcortical atlas (48, 49). Initially labeled in MNI-152-1mm space, the masks were afterwards transformed to individual diffusion space using the FSL’s applywarp function (50) and visually checked for each participant ( Figure 1 ).

Preprocessing and data analysis were performed using tools from the FMRIB software library (FSL, version 5.0.4, https://www.fmrib.ox.ac.uk/fsl): First, we applied FSL’s brain extraction tool (BET) to T1-weighted (T1w) images and applied the resulting brain mask to T2-weighted (T2w) images, that had been aligned to T1-weighted images using rigid body alignment. Diffusion-weighted images were corrected for head motion using affine registration [FLIRT; 51] while using the first acquired B0 as a reference volume. Based on this transformation, the original b-vectors were corrected for the rotational components. To further enhance the reliability of the upcoming tractography algorithm, the preprocessed diffusion-weighted data were interpolated to 1mm (51, 52). T1-, T2- weighted images were then linearly co-registered using FSL’s FLIRT tool (53) to the MNI-152 1mm isotropic brain (12 DOF-affine registration). Diffusion-weighted images were co-registered to standard space using non-linear registration (FNIRT). The anatomical masks were nonlinearly transformed from standard space to diffusion space by applying the transformation matrices acquired with FSL’s FNIRT function (54). FSL’s BedpostX program was applied to model diffusion parameters for crossing fibers in native space. Afterwards, probabilistic tractography was performed in diffusion space using FSL’s Probtrackx2 (53, 55, 56). Probabilistic tractography is leveraged to reconstruct white matter structural connectivity in diffusion data. Starting from a voxel in the seed mask the tracking is performed in a stepwise manner from a distribution of possible directions. The distributions of principal diffusion directions calculated through BedpostX were used at each voxel to compute a streamline through these local samples and generate a probabilistic streamline. By taking many such samples, the FMRIB’s diffusion toolbox (FDT) is able to build upon the posterior distribution on the streamline location or the connectivity distribution (55). For each subject and seed region, the following parameters were chosen: The number of samples was 5000 for each seed voxel with a curvature threshold of 0.2, a maximum number of steps of 2000 with a step length of 0.5 mm. A path distribution function was added in order to correct for differences in distance between seed and target regions. The tractography was performed twice between seed and target in either direction. The ventricles were included as exclusion masks.

Structural connectivity between the SN/VTA and the NAc was analyzed. In order to test for construct validity, we also tested for differences in the structural connectivity between the DA midbrain and caudate, which is more implicated in attentional impulsivity than motor impulsivity and hence served as a validator for the specificity of the expected alterations in VTA/SN-NAc connectivity. The connectivity measure was defined as described in Stephan et al. (57). From the seed region (S) 5000 samples were initiated, and the proportion of the samples (K_S) that intersected with target (T) defined the connectivity between these areas. Hence, this quantity reflects a relative connection density yielding numbers between 0 and 1 and does not inform about the absolute number of connections between these areas (58). It is important to note that this measure differs depending on whether tractography is sourced from S or T. Therefore, we performed the tractography in both directions and normalized by the size of the source region (N). Hence the relative connectivity density (φ_ST) was defined by the following formula:

For statistical analyses this connectivity measure was considered to assess connectivity between the VTA/SN and NAc. This connectivity measure was also used to assess connectivity between the VTA/SN and the caudate in order to demonstrate the specificity of the effect of FTO polymorphism on structural connectivity between VTA/SN and NAc. For both VTA/SN-NAc and VTA/SN-caudate, connectivity measures were analyzed averaged across hemispheres.

Data was analyzed using RStudio [version 1.4.1717; (59)] and R [version 4.0.0; (60)] using the functions of R’s ‘Stats’ package (version 3.6.2) if not stated otherwise. Frequency distribution was tested with the Shapiro-Wilk-Test. For the connectivity data the Shapiro-Wilk-Test resulted in a deviation (p< 0.001) from the normal distribution. Due to this deviation from the normal distribution, we performed a logarithmic transformation of the connectivity values before further parametric testing. For further statistical assessment, linear regression models were performed using the “lm”-function in R. First, a linear regression model was used to assess differences in the total score of the BIS-11 scale between both groups (FTO ⁺ vs FTO ^–, Table S1 ). To test for differences in the second-order factors of the BIS-11 (attentional impulsiveness, motor impulsiveness and non-planning impulsiveness) between the FTO ⁺ and the FTO ^– group we also applied linear regression models ( Table S2 ). We controlled for multiple comparisons using Bonferroni correction. Then we analyzed differences in (log) connectivity between the two groups ( Table S3 ) and tested for the effect of FTO risk allele carrier status and (log) connectivity on motor impulsiveness in a combined model as required for the following mediation analysis ( Table S4 ) using linear regression models respectively. Last, a mediation analysis was performed to test if (log) connectivity mediates the effect of risk allele carrier status on motor impulsiveness using R’s ‘Mediation’ package [version 4.5.0; (61); Table S5 ]. Effect sizes were calculated using R’s package ‘Effectsize’ [version 0.4.5; (62)]. Statistical significance was reported at a level of p< 0.05 if not otherwise stated.

For all analyses, we report the formula of the model using the Wilkinson notation in addition to the degrees of freedom, t- or F-values, significance of the respective regressors and their effect size in the Supplementary Material . A post hoc power analysis with an alpha-error of 0.05 was performed for each t-test and linear regression using the respective effect sizes and number of predictors (see Supplemental Tables 1–4 ). The power analysis was performed in G*Power [version 3.1; (63)].

In total, eighty-one participants were included in the final analysis ( Table 1 ), 39 non-carriers (FTO^–; 24.8 ± 3.5 years, 22.6 ± 2.6 kg/m²) and 42 risk allele carriers (FTO⁺; 25.7 ± 4.4 years; 23.2 ± 3.2 kg/m²), who were matched for age, BMI and sex. All participants self-reported to be healthy without any known diagnoses. Accordingly, participants in both groups did not differ in their metabolic status as measured by fasting insulin, glucose, cholesterol, and triglyceride values (all p > 0.4; Table 1 ). Neither risk allele carriers nor non-carriers reported depressive symptoms according to the BDI-2 (65) nor alcohol use disorder according to the alcohol use disorder identification test (66). Sensitivity to punishment and reward as well as eating behavior as assessed using the German version of the three-factor eating questionnaire were comparable across groups (all p > 0.05, Table 1 ).

We first tested, whether FTO risk allele carriers differ in their impulsive phenotype in comparison to the non-carriers. Trait impulsivity in general, as measured by the BIS-11 total score, did not differ between the FTO ⁺ and the FTO^– group (F = 2.08, p = 0.156, Supplemental Table S1 ). As motor and attentional impulsivity have been particularly associated with weight gain, with motor impulsiveness being strongly linked to DA functioning in the ventral striatum and attentional impulsiveness in the caudate (64), we next assessed if motor impulsiveness differed between the FTO ⁺ and FTO ^– groups. Compared to controls, FTO risk allele carriers demonstrated increased motor impulsiveness (F = 6.63, p = 0.030 after adjustment for multiple comparisons using Bonferroni correction), while no significant differences between FTO risk allele carriers and non-carriers could be observed for the other second-order factors of the BIS-11 including attentional impulsiveness ( Supplemental Table S2 , Figure 2A ). Neither motor impulsivity nor the effect of FTO risk allele status on motor impulsivity differed between sexes ( Supplemental Tables S1, S2 ).

We next analyzed, if structural connectivity is changed in FTO risk allele carriers compared to non-carriers. We particularly considered connectional changes between VTA/SN and NAc due to their relevance in encoding motor impulsiveness (40) and the link of FTO to DA functioning. Additionally, we also measured structural connectivity between VTA/SN and the caudate nucleus, which is implicated in attentional impulsivity (64) but not primarily in motor impulsiveness, to verify the specificity of the expected effect of gene variant on connectivity between the midbrain and the NAc.

We found that connectivity between the VTA/SN and the NAc was increased in FTO risk allele carriers compared to non-carriers (F = 4.06, p = 0.044, Supplemental Table S3 , Figure 2B ) while we did not find a significant difference in structural connectivity between VTA/SN and the caudate (FTO ⁺ vs FTO ^- group; F = 1.17, p = 0.283, Supplemental Table S3 ). Neither VTA/SN – NAc nor VTA/SN-caudate connectivity nor the effect of FTO risk allele carrier status on either connectivity measurement differed between sexes ( Supplemental Table S3 ).

To assess, whether increased connectivity in meso-striatal pathways in FTO risk allele carriers contributes to the greater motor impulsivity also observed in the FTO ⁺ group, we applied a mediation analysis. We demonstrate that increased structural connectivity between VTA/SN and NAc mediates the effect of allele status on motor impulsivity. Specifically, 21% of the FTO effect on motor impulsivity was mediated by the structural connectivity between the VTA/SN and NAc ( Supplemental Tables S4, S5 , Figure 2C ), indicating a hardwired predisposition, that may well relate to neurodevelopmental alterations in FTO risk allele carriers and contribute to their impulsive phenotype.

While the selection of our participants ensures the attribution of the observed effects to FTO risk allele carrier status, inviting exclusively non-obese participants might at the same time introduce a sampling bias, by including a specific group of participants, who despite carrying risk alleles, are able to avoid weight gain. Second, even though we here aimed for conclusive evidence to support our molecularly well-grounded hypothesis, as in all candidate gene studies, the sample-sized dependent achieved power (0.73, Table S4 ) might represent a noteworthy limitation (86). Future particularly molecular studies will validate whether the proposed theories and concepts about the neurodevelopmental role of FTO as a contributor to obesity-promoting behavioral traits hold true in various scenarios. Moreover, probabilistic tractography only provides an indirect measure of structural connectivity and is not sufficient to detect alterations on the axonal level (87). It further needs to be considered that the here investigated polymorphism rs9939609 is in strong linkage disequilibrium with two further SNPs [rs8050136 and rs9930506] in Europeans, rendering it impossible to fully attribute the observed phenotype to the rs9939609 SNP (88, 89). Intriguingly, in individuals of African descent, results linking rs9939609 to increased BMI are more heterogenous so that further studies a required to disentangle the effect of FTO polymorphisms in different ethnicities (90, 91). These limitations and possibly differential effects of the various FTO polymorphisms on neurodevelopment might contribute to the inconsistent findings so far linking FTO polymorphism with either altered structural connectivity (92, 93) or no connectivity changes (94). With increasing data availability and computational power, future (genome wide) association studies will be required to replicate the demonstrated relationship between allelic variation and brain connectivity across a more heterogenous sample including further (cortical) brain circuitries involved in encoding of impulsivity.

Overall, our data reveals that increased motor impulsiveness in FTO risk allele carriers is partially mediated by enhanced structural connectivity between the midbrain and ventral striatum indicating that FTO polymorphism affects obesity promoting behavior via structural alterations in addition to complementary mechanisms. Further studies are needed to scrutinize the underlying molecular mechanisms.