The rapid worldwide spread of obesity and associated comorbidities such as diabetes, cardiovascular diseases, and cancer (Dixon, 2010) as well as its complex etiology and inter-individual variability in response to intervention demand the development of new therapeutic strategies (Roman et al., 2015). The majority of currently available weight loss programs are based on dieting and physical activity (Jakicic and Davis, 2011; Amorim Adegboye et al., 2013; Soeliman and Azadbakht, 2014), which on the one hand are accessible and affordable, but on the other hand often lead to timely restricted effects followed by rapid weight regain after the program has ended (known as the yo-yo effect).

Bariatric surgery (weight loss surgery) is currently seen as the gold standard in the therapy of obesity, since it proved to be effective in inducing lasting weight loss (Sjöström, 2000; Buchwald and Oien, 2013). However, high costs and surgery-associated risks leave it to be an exceptional option for only morbidly obese individuals. Since most weight-loss programs approach obesity on the symptomatic level, the treatment of underlying causes, repeatedly shown to be brain-dependent, appear indispensable for developing new therapeutic strategies leading to lasting weight loss and healthier living.

Converging evidences agree on the notion that obesity affects the structure (Hollmann et al., 2012; Cone et al., 2014) and function of the central nervous system (Hollmann et al., 2012) in relation to dysregulated hormonal feedback from the digestive system (Schlögl et al., 2016). Whereas, homeostatic control sites in the hypothalamus integrate and process information from the body's periphery to ensure energy balance (Morton et al., 2006), food cues such as sight, smell and taste affect hedonic brain regions involved in goal-directed and habitual behavior, such as the ventral and dorsal striatum, respectively (Wang et al., 2001; Saper et al., 2002; Stoeckel et al., 2008). The obesity-related, dysregulated feedback from the digestive system to those homeostatic and hedonic brain sites as well as the attenuated reward responsivity to food intake (Stice et al., 2008a) seem to maintain overeating behavior.

Especially high-caloric foods seem to affect the brain's reward responses like drugs of abuse (Volkow et al., 2013). Like drug addicts, obese individuals present increased craving as well as attenuated reward responses to high-calorie foods, probably supporting compensatory overeating (Wang et al., 2001; Stice et al., 2008b; Johnson and Kenny, 2010). Dopamine seems to have a central role in mediating these effects. In obese rats, for instance, electrically evoked dopamine release in slice preparations was significantly attenuated, not only in the nucleus accumbens but also in additional terminal sites of dopamine neurons such as the accumbens shell, dorsal striatum, and medial prefrontal cortex, suggesting that there may be a widespread dysfunction in mechanisms regulating dopamine release in obesity (Geiger et al., 2008, 2009; Zhang et al., 2015).

However, the interplay between the hypothalamus and reward-related regions alone cannot explain the complex neurobiological mechanisms involved in food choices, such as those underpinning the appraisal or reappraisal of healthy and unhealthy food. Associated brain mechanisms are of potential interest for the development of new therapeutic strategies, such as neurofeedback training or non-invasive brain stimulation. In previous studies, we aimed to identify brain regions involved in those processes with functional magnetic resonance imaging (fMRI; Hollmann et al., 2012) and electroencephalography (EEG; Kumar et al., 2016). Findings suggest that an active reappraisal of tasty but unhealthy food recruits the brains valuation system in combination with cognitive control areas, such as the dorsolateral prefrontal cortex (i.e., DLPFC), and gustatory areas such as the frontal operculum, which together with the neighboring anterior insular cortex is assumed to host primary gustatory processes, such as taste perception (Rolls et al., 1988; Zatorre et al., 1992; Small et al., 1999). The DLPFC's activity increased during admitting the desire for high and low calorie food (Kumar et al., 2016), supporting the notion of the DLPFC's decisive influence on self-control (Hare et al., 2009) and cognitive reappraisal (Kober et al., 2010). The right frontal operculum's activity increased when regulating food desire, assigning higher cognitive functions, such as food imagery, to the primary gustatory cortex (Kumar et al., 2016). Based on these findings, neurofeedback training, or non-invasive brain stimulation based on the DLPFC's and frontal operculum's state-dependent activation levels could strengthen executive top-down control on food choices and food-related reward processing through modulating the DLPFC's and frontal operculum's functional implementation in an individualized manner.

In the present study, we investigated the role of the DLPFC and the frontal operculum in the active reappraisal of high or low calorie food, as well as their consumption using non-invasive transcranial direct current brain stimulation (tDCS). TDCS is a secure procedure for subliminal, tonic electric stimulation of the brain (Nitsche and Paulus, 2000; Brunoni et al., 2011). A weak direct current of 1–2 mA is applied to generate regional changes in cortical excitability, which, depending on the duration and the polarity, can last for several minutes up to a few hours after stimulation (Nitsche and Paulus, 2001; Hummel and Cohen, 2005). Whereas the neuronal effects during tDCS are characterized by a shift of membrane potentials in cortical neurons that lead to a modification in the regional neuronal activity, sustainable effects (i.e., the following 20 min after application) seem to be mediated by changes in the efficiency of synaptic transmission (Clark et al., 2011; Rahman et al., 2013). Studies in animals as well as humans indicate that anodal stimulation leads to an increase in neuronal excitability, whereas cathodal tDCS leads to hyperpolarization of the membranes and therefore causes decrease in neuronal excitability (Paulus, 2004; Been et al., 2007). However, this clear dichotomy seems to describe the effects in the motor cortex, which cannot be transferred to cognitive tasks per-se (Boehringer et al., 2013; Macher et al., 2014; Taubert et al., 2016). Previous studies investigating the effect of tDCS on the DLPFC suggest that anodal stimulation in lean individuals reduces food craving (Goldman et al., 2011; Montenegro et al., 2012; Kekic et al., 2014; Ljubisavljevic et al., 2016) and caloric intake (Fregni et al., 2008; Jauch-Chara et al., 2014) immediate to tDCS. The tDCS study by Gluck et al. appears to be the only study to date, probing repetitive application of tDCS to the DLPFC in an solely obese cohort, resulting in decreased daily kilocalories consumed and greater percentage of weight loss as compared to cathodal and sham (placebo) stimulation (Gluck et al., 2015).

Based on our recent EEG findings (Kumar et al., 2016) and the well-described dichotomic tDCS influences on the motor cortex (Paulus, 2004; Been et al., 2007), we here hypothesized, that tDCS, with the cathode placed over the left DLPFC, downregulates the DLPFC's activity, whereas the simultaneous anodal stimulation of the right frontal operculum simultaneously upregulates the frontal operculum's activity. The potential up- or downregulation in the two targeted areas (left DLPFC, right operculum), whose influence on food desire modulation in obese was evinced in our previous study (Kumar et al., 2016), could also trigger alterations of the obesity associated reduced dopamine response in central reward related regions such as ventral and dorsal striatum as well as nucleus accumbens (Geiger et al., 2008, 2009; Zhang et al., 2015) due to their tight interconnectedness. We expected that these tDCS influences strengthen the ability to regulate food desire and reduce calorie consumption as compared to the inverse tDCS polarity (i.e., anodal stimulation over left DLPFC, cathodal stimulation over right frontal operculum) as well as sham tDCS.

All women tolerated tDCS well. Reported side effects like headache, dizziness, or burning sensations were only temporary and did not lead to premature discontinuation of an experiment session.

In the present study, we investigated the effect of tDCS over the left DLPFC and right frontal operculum in obese women, firstly on the active reappraisal of visually presented high and low caloric food items and secondly on subsequent food consumption. We hypothesized that the self-reported ability to either regulate or admit the desire for presented food pictures would vary between cathodal and anodal/sham tDCS. Furthermore, we expected that the consumption of palatable food, offered in an appealing buffet, would be down-regulated by previous cathodal tDCS in comparison to anodal/sham stimulation. The present findings could not confirm an effect of tDCS, neither on regulating/admitting the desire for foods, nor on calorie consumption. In all three tDCS conditions, participants rated the ability to admit the desire for the visually presented foods as significantly easier than the reappraisal of the same foods. This was to be expected considering the fasting state of 5 h prior to the experiments.

The DLPFC is believed to represent an essential component of the complex network controlling eating behavior, particularly the cognitive control over presented food cues in everyday life and food reward processing (Hollmann et al., 2012, 2013). The frontal operculum, instead, is believed to primarily host gustatory processes (Rolls et al., 1988; Zatorre et al., 1992; Small et al., 1999), but recent findings also suggest its contribution to higher cognitive processes, such as regulating food desire (Kumar et al., 2016). Numerous studies have started investigating brain areas, mainly implied in the suppression of palatable foods (Schlögl et al., 2016). Their findings highlight an almost disclosed brain circuit involving mesocorticolimbic regions as well as the frontal operculum and the DLPFC (Hollmann et al., 2012; Siep et al., 2012; Kumar et al., 2016). Our present results suggest that endogenous suppression techniques (i.e., reappraisal strategies, see Table 6) associated with one-time extraneous modulation via non-invasive brain stimulation, directed to just two of the circuit's components is not sufficient to modulate food reappraisal abilities or food consumption in obese females. The interpretation of such null findings is generally problematic, since effects may become significant with increasing the sample size or number of tDCS sessions. Nevertheless, as compared to previous studies on tDCS and food choices (Fregni et al., 2008; Goldman et al., 2011; Montenegro et al., 2012; Jauch-Chara et al., 2014; Kekic et al., 2014; Gluck et al., 2015; Ljubisavljevic et al., 2016), we here targeted distinct brain sites derived from recent pilot experiments that were conducted with the same food picture task and hence with the same task demands as in the present study (Kumar et al., 2016). Furthermore, our tDCS study included the largest human obese sample so far. To avoid gender variability as a well-known confounding effect (Horstmann et al., 2011; Mueller K. et al., 2011; Melasch et al., 2016), we focused on the female brain. A major limitation of our study is, however, that menstrual cycle was not inquired. Since, menstrual cycle has effects on food perception and intake (Bryant et al., 2017), it should be considered for future studies investigating the effects of non-invasive stimulation of the female brain on food evaluation and consumption.

In obese individuals, the right prefrontal cortex (PFC) seems to present crucial functional differences when compared to lean individuals, founding the “right brain hypothesis” (Alonso-Alonso and Pascual-Leone, 2007). It suggests that dysregulation, in particular reduced activity in the right PFC, contributes to enhanced probability of overeating, inactivity, reduced capacity of self-reflection on dietary choices and general deficits in decision-making. Also, the left DLPFC was shown to be critically involved in food choices and hence in the control of eating behavior (Gluck et al., 2015), adding evidence to the strong assumption of a prefrontal imbalance in obesity (Carnell et al., 2012; Brooks et al., 2013; Vainik et al., 2013). In our recently published EEG study (Kumar et al., 2016), we also found such a hemispheric imbalance, however, not between both prefrontal cortices, but between the left prefrontal (i.e., DLPFC) and the right frontal cortex (i.e., frontal operculum). Activity in the left DLPFC increased while allowing the desire for food, whereas activity in the right frontal operculum increased with the ability to regulate the desire for food (Kumar et al., 2016). Based on these findings, we here hypothesized that tDCS with its two polarities (i.e., anodal and cathodal) downregulates activity in the left DLPFC, via cathodal stimulation, while simultaneously upregulating activity in the right frontal operculum, via simultaneous anodal stimulation. We expected that this tDCS effect on prefrontal/frontal activity levels strengthens the ability to regulate food desire as well as reduces calorie consumption, which we could not confirm. This could imply that tDCS failed to provoke up- or downregulation of activity in targeted regions (i.e., left DLPFC and right frontal operculum), and possibly also food-related dopamine responses in central reward regions, such as ventral and dorsal striatum as well as nucleus accumbens, connected with the targeted regions. It is therefore questionable whether other regions such as tDCS of left and right DLPFC, as targeted in former studies (Fregni et al., 2008; Goldman et al., 2011; Kekic et al., 2014; Lapenta et al., 2014), would have provoked a feasible alteration of dopaminergic response arising from structures such as striatum or mediofrontal system (Geiger et al., 2008, 2009) and hence initiated indirect dopamine-dependent behavior modification. Due to the three envisaged tDCS conditions (anodal, cathodal, sham) we renounced further expansion with additional potential control brain regions, since we perceived three experiment sessions for each participant as only just acceptable. However, considering the negative results of our study, the possibility of alternating dopaminergic response in regions deep within the brain (e.g., striatum) through non-invasive stimulation of superficially located areas remains questionable. A directly through tDCS mediated interference of dopamine response in the striatum is rather unlikely, since tDCS as well as alternative transcranial magnetic stimulation (TMS) with likewise implemented procedure protocols lack in necessary depth of penetration.

The field of non-invasive brain stimulation experiments is fairly young and current data in the area of eating behavior is few and controversial (Horvath et al., 2015). As Horvath et al. (2015) reported in a meta-analysis of tDCS outcomes, a majority of study designs were lacking in valid control conditions or double-blinding. The only reliable effect of tDCS, that could be shown in this review, was the modulation of motor evoked potentials after tDCS was applied to the motor cortex. The discrepancy in outcomes between studies with however comparable tDCS protocols suggests that there are several influencing variables, such as the thickness and fat proportion of the scalp (Truong et al., 2012), whose impact on brain stimulation are neither fully investigated nor understood. A recently presented trial by György Buzsáki of New York University (NYU) even demonstrated that hardly 10% of alternating current applied via tDCS to cadaver scalps, measured by implanted electrodes, reached brain tissue (Underwood, 2016). This trial is to be anticipated with interest, serves it more doubt on our knowledge of tDCS effects and its applications. Assuming an actual effect of inhibitory or facilitating tDCS to targeted brain regions, it is conceivable that inhibitory stimulation led to suppression of inhibitory GABAergic interneurons and hence excitation of interconnected neural networks. Consequently, behavioral pattern could have been reinforced through inhibitory tDCS, contrary to hypothesized suppression. Since, we could not evince neither facilitation nor suppression of behavioral patterns under inhibitory tDCS concerning the modulation of food desire as well as actual consumption, inhibition of inhibitory GABAergic interneurons in the context of our work seems unlikely.

Although, food-craving as an addiction-like behavior has been shown to decrease after tDCS (Fregni et al., 2008; Goldman et al., 2011; Kekic et al., 2014; Ljubisavljevic et al., 2016), tDCS effects on food choices, especially with respect to palatable but unhealthy food, in hungry obese females or males remain largely unexplored. Most studies were restricted to participants within a lean BMI range (i.e., BMI < 30; Fregni et al., 2008; Goldman et al., 2011; Jauch-Chara et al., 2014; Kekic et al., 2014). To the best of our knowledge, Gluck et al. (2015) has published the only study so far showing that the repetitive application of anodal prefrontal tDCS to the left DLPFC decreases caloric intake in a small cohort consisting of nine obese males and female, whereas Ljubisavljevic et al. (2016) suggested reduced food craving after anodal stimulation of the right DLPFC in a cohort of normal as well as obese young adults. In our study, we used a single session tDCS because of its well-described effects on food-craving (Jansen et al., 2013; Kekic et al., 2014; Lapenta et al., 2014). Our findings however suggest that a single tDCS session of 20 min is not sufficient to modulate reappraisal strategies as well as calorie intake. Another striking difference to the study by Gluck et al. (2015) is, that we applied tDCS during the reappraisal task and not prior to the task. We decided for this online tDCS design, since Nitsche et al. previously reported that tDCS effects occur already within minutes after tDCS initiation (Nitsche and Paulus, 2000, 2001). Thus, different tasks in combination with different online tDCS protocols may have different immediate influences on participants' actual brain function and behavior. Future methodological tDCS studies are necessary to compare tDCS in combination with online or offline tasks in order to assess their specific advantages or disadvantages.

The laboratory settings of previous (Gluck et al., 2015; Ljubisavljevic et al., 2016) and the present study, just as much as the placing of our standardized buffet in an investigation room, stresses the distance to habitual eating behavior in a familiar social environment. Heightened awareness of observation under lab conditions, for instance, was shown to cause obese females to reduce their calorie consumption (Robinson et al., 2016), suggesting social modeling as an important influencing factor on eating behavior. Cruwys et al. (2015) reviewed several studies published between 1974 and 2014 and found that social modeling of eating seems at least to be partially mediated through behavioral mimicry, which occurs without conscious awareness. Since participants in our study were alone at the buffet, there was no “ideal model” they may have desired to affiliate with. This makes social modeling rather unlikely to account for potential reluctance in eating. Each participant was furthermore investigated three times (i.e., sham, anodal, cathodal tDCS) in random order under similar conditions, canceling out possible influences of social modeling, if tDCS conditions are statistically compared against each other.

Another potentially influencing factor on the present findings is the heterogeneity of obesity phenotypes together with heterogeneous eating behavior traits, specifically with respect to impulsivity and disinhibition. To assess the influence of those traits on our results, we investigated the relation between grade of obesity and impulsiveness as well as temper control through correlation analysis of BMI and the BIS-15 as well as AT questionnaires, which revealed no verifiable coherence. The BDI was used for initial screening of indicators for depression and participants scoring more than 15 points were excluded from the study. However, no further diagnostic tool was instated, which also applies for anxiety disorders that could be associated to binge eating behavior (Mitchell et al., 1999; Ostrovsky et al., 2013). Another aspect of the present study design, that may have accounted for data heterogeneity and hence the lack of tDCS effects, is the strategy participants used to regulate their food desire. As in previous studies (Hollmann et al., 2012; Kumar et al., 2016), we specifically allowed participants to freely choose the best strategies. We hypothesized that this approach supports optimal individual food regulation abilities. However, Siep et al. (2012) found that short-term suppression of food desire is more successful in inhibiting corresponding brain activation then cognitive restraint, hence thinking of long-term consequences. Women of our study however used either short-term or long-term consequences as strategies for regulating their food desire (Table 6).

A number of previous studies investigated the effect of non-invasive brain stimulation linked with cognitive restraint training and even physical activity on food choices, eating behavior, or calorie consumption. Controversial findings due to diverse study designs, sample sizes, and stimulation response in individuals called for supplemental research (Barth et al., 2011; Horvath et al., 2015). Our fully randomized, within-subject, placebo-controlled, and double-blinded study does not support the notion of tDCS as a promising method to improve the regulation of food desire or food consumption in obese women. However, our results are in disagreement with previous pilot trials by Gluck et al. who used repetitive application of tDCS (Gluck et al., 2015). This discrepancy between the present and previous findings demands further studies combining a comparable study design as in the present study with repetitive tDCS instead of just one single session.

This study was carried out in accordance with the recommendations of the ethics committee of the medical faculty of the University of Leipzig with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the ethics committee of the medical faculty of the University of Leipzig.

Conception and design of study: FG, CB, SK, JM, BP. Acquisition of data: FG, CB. Analysis and/or interpretation of data: FG, CB, MR, JM, BP. Drafting the manuscript: FG, BP. Approval of the version of the manuscript to be published: FG, CB, SK, MR, JM, BP.