Relation of Regional Grey and White Matter Volumes to Current BMI and Future Increases in BMI: A Prospective MRI Study

Introduction

In the US 65% of adults are overweight or obese (1) and obesity rates in children and adolescents have increased dramatically over the past decennia (2). Obesity is associated with increased risk for mortality, atherosclerotic cerebrovascular disease, coronary heart disease, colorectal cancer, hyperlipidemia, hypertension, gallbladder disease, and diabetes mellitus, resulting in over 111,000 deaths annually in the US (3). Obesity is also associated with poor cognitive function and neuro-degenerative disorders, such as dementia (4,5).

There is evidence that obesity is associated with anomalous regional metabolism and neural activity, particularly in areas of the brain involved in reward and taste. Functional magnetic resonance imaging (fMRI) studies have found that obese versus lean humans show greater responsivity in gustatory regions (insula/frontal operculum), somatosensory regions (e.g., Rolandic operculum), and reward processing regions (e.g., caudate, putamen, amygdala, orbitofrontal cortex [OFC]) in response to palatable food cues (6–9). Further, there is limited evidence that this hyper-responsivity increases risk for future weight gain (10). Lean adolescents at high- versus low-risk for obesity show greater responsivity in the dorsal striatum and parietal/frontal operculum during palatable food receipt (11). These findings collectively suggest that elevated responsivity of reward and somatosensory regions increase risk for obesity.

Yet, obese versus lean humans have fewer dopamine (DA) receptors in striatal regions (12,13) and show reduced striatal response to palatable food intake (14,7). Further, blunted dorsal striatal response to chocolate milkshake receipt also predicts future increases in BMI for young women with the A1 allele of the single nucleotide polymorphism referred to as the TaqIA (15), which is associated with lower D2 striatal receptor availability and reduced striatal resting metabolism (16,17). Interestingly, blunted response of the dorsal striatum and frontal operculum to images of palatable foods also predicts future increases in BMI for those with a TaqIA A1 allele (8).

One explanation for these seemingly inconsistent findings is that some of these effects reflect initial vulnerability factors, whereas others result from obesity. Firing of DA neurons in reward regions shifts from food intake to cues that predict food intake after conditioning in rats (18,19), implying that overeating is related to increased responsivity of regions that encode the reward value of food images and cues. Further, overeating leads to reduced D2 receptor density, D2 sensitivity, and reward sensitivity in rats (20–22) and reduced striatal response to food in humans (23), suggesting that overeating leads to receptor down-regulation in dopamine-based reward regions.

There is also evidence that obesity is related to structural/volumetric brain differences. BMI has correlated with low global brain volume (24,25) as well as with regional grey matter (GM) and white matter (WM) volume differences (e.g, 26). The function of GM, which consists of neuronal cell bodies, neuropil, glial cells and capillaries, is to route sensory or motor stimuli to interneurons of the central nervous system to create a response to the stimuli through chemical synapse activity (27). WM contains myelinated axon tracts that connect different areas of GM in the brain. Myelin acts as an insulator increasing the speed of transmission of all nerve signals (27). Pannacciulli et al. (26) has found that obese versus lean individuals have significantly lower GM volume in regions implicated in taste processing (postcentral gyrus and frontal operculum/extended insula) and reward (putamen). These regional volumes were negatively correlated with fasting plasma leptin concentration, a hormone that is produced in proportion to the amount of body fat (28). Obese versus lean showed high GM volume in the middle occipital gyrus, inferior frontal gyrus, and cuneus and higher WM volume in the striatum (26). Taki et al. (25) has found negative correlations between BMI and GM volume in the fusiform gyrus, superior parietal lobe, precentral gyrus, inferior- and superior frontal gyri, midbrain, uncus, cerebellum, and precuneus as well as positive correlations between BMI and GM volume in the posterior lobe of cerebellum, temporal gyrus, thalamus, cingulate gyrus, and caudate. Walther et al. (29) has found negative correlations between BMI and GM volumes in the OFC, inferior- and middle frontal gyri, parahippocampal gyrus, lingual gyrus, and cerebellum. Additionally, Horstmann et al. (30) has found a negative correlation between BMI and GM volume in the dorsolateral prefrontal cortex, but a positive correlation between BMI and GM volume in the OFC, putamen, and hypothalamus. Maayan et al. (31) has found that obese versus lean adolescents have significantly lower GM volume in the OFC. Although Haltia and colleagues (32) did not find significant differences in GM volume, WM volume was greater in the temporal gyrus, fusiform gyrus, parahippocampal gyrus, brain stem and cerebellum in obese versus lean individuals. Finally, Raji et al. (33) has found negative correlations between BMI and GM and WM volume in the OFC, ACC, medial temporal lobe, hippocampus, putamen, globus pallidus, and thalamus.

The abovementioned cross-sectional data provide evidence of a complex pattern of volumetric differences in obesity. Most studies show that obesity is associated with reduced GM volume in regions involved in taste (postcentral gyrus, frontal operculum/insula) and inhibitory control (inferior-, middle-, and superior frontal gyri). Findings with regard to GM volume in regions involved in reward (OFC, caudate, putamen, midbrain) are somewhat mixed. The few studies that examined the association between BMI and regional WM volume have also provided inconsistent results. Further, it is unclear whether structural/volumetric brain differences reflect initial vulnerability factors or are the result of obesity. One hypothesis is that an increase in BMI results in volumetric changes in the brain. For example, inflammatory cytokines/adiopokines such as fibrinogen, IL-1β, IL-6, and C-reactive protein are associated with excess adipose tissue (e.g., 34,35) and elevated levels of such inflammatory markers are positively correlated with insulin resistance, metabolic syndrome, and type 2 diabetes (e.g., 36,37). IL-6 levels are negatively correlated with global brain volume and regional GM volume in the hippocampus and medial PFC (38). In fact, IL-6 levels mediate the association between body fat and hippocampal grey matter volume (39). Further, monkeys on a long-term, calorie-restricted diet show reduced levels of IL-6 and decreased IL-6-related global GM and WM atrophy, as well as GM atrophy in parietal and temporal regions (40). Both rat and human studies have shown that a low-calorie diet restricts protein expression of IL-6 (e.g., 41). In sum, inflammatory markers may be mechanisms by which obesity is related to low GM volume in the brain. Moreover, Haltia et al. (32) has hypothesized that an increase in body fat may enhance the density of myelin, which increases WM volume. This is supported by their findings that obese versus lean individuals show significantly higher concentrations of serum free acids and that serum free fatty acid concentration and regional WM volume in obese subjects were positively associated. This is also supported by rodent studies indicating that the hypothalamic metabolism of fatty acids can modify feeding behavior (42). Haltia et al. (32) also found that global WM as well as regional WM volumes in the temporal lobe were partially reduced by a controlled very low-calorie diet for 6 weeks in obese subjects. Changes in global or regional gray matter were non-significant (32). Animal studies have also found that overeating is related to cognitive decline while caloric restriction slows brain aging (43). These findings suggest that WM brain volumetric changes could, at least partly, be secondary to weight change.

An alternative hypothesis is that low GM volume and/or high WM volume increases risk for obesity. Comparable to the reward deficit model of obesity (44), it is possible that low GM and/or high WM volume in reward regions attenuates reward from food, prompting increased intake in a compensatory fashion. It has also been theorized that impulsive individuals are more sensitive to cues for reward and more vulnerable to the omnipresent temptation of appetizing foods (45,46), increasing risk for unhealthy weight gain. Trait impulsivity is thought to result in greater sensitivity to reward-predictive cues, which may contribute to compulsive food-seeking behavior (47). Thus, it is possible that low GM and/or high WM volume in regions involved in inhibitory control increases the likelihood of giving into tempting palatable foods. Both impulsivity and obesity have found to be related to less GM volume in the OFC (31,48). Obese versus lean adults also show less GM volume in the prefrontal cortex (26), a region that modulates inhibitory control. Obese versus lean young women show less activation of prefrontal regions (e.g., middle frontal gyrus, ventrolateral prefrontal cortex) when trying to inhibit responses to unhealthy food images and show behavioral evidence of reduced inhibitory control (49). Relative to obese controls, adults who achieve lasting weight loss show greater activation of inhibitory regions (superior frontal- and middle temporal gyri) in response to palatable food images (50). Reduced activation in the dorsolateral prefrontal cortex in response to palatable food images also predicts increased ad lib food intake over the next 3 days in men and women (51). To our knowledge, no other study has tested whether reduced GM volume and/or increased WM volume is related to future weight gain.

We think it is important to use voxel-based morphometry to test whether volumetric differences in certain brain regions are related to future BMI as this may provide us with a better understanding of the risk processes that give rise to obesity. We hypothesized that reduced GM and increased WM in regions involved in taste (anterior insula/frontal operculum, Rolandic operculum), reward (OFC, caudate, putamen), and inhibitory control (inferior-, middle-, and superior frontal gyri) would predict future increases in BMI.

In an attempt to replicate previous findings, we also tested the correlation of BMI to global and regional brain matter volumes. Based on the abovementioned studies, we hypothesized that BMI is correlated with reduced overall GM volume and with reduced GM volume in brain regions involved in taste, reward, and inhibitory control and increased WM volume in the caudate, putamen, inferior-, middle- and superior temporal gyri, and parahippocampal gyrus.

Methods

Results

Discussion

In the current study, obese female adolescents show significantly lower overall GM volume compared to overweight and lean female adolescents, suggesting greater global brain atrophy. This result converges with findings of previous studies in middle-aged obese adults (24,25). There was no difference between overweight and lean individuals in global GM volume suggesting that the relation between BMI and global GM volume is non-linear. A possible explanation could be that obese individuals compared to individuals with a lower weight experience many more biological changes (e.g., hypertension, insulin resistance) that negatively affect global GM volume.

Although obese individuals had lower overall GM volume compared to lean and overweight individuals, we did not find significant negative associations between BMI and regional GM volumes. This result is in contrast with the findings of two previous studies (26,29), but comparable with those from another (32). A possible explanation for these inconsistent findings is that participants in both the current study and in Haltia et al. (32) were overall younger and less obese compared to those in the other two studies. It is possible that only more severe and chronic obesity negatively influences regional GM volume, in line with the thesis that volumetric changes are at least in part secondary to elevated adipose tissue levels. Further, Taki and colleagues (25) found a significant correlation between BMI and reduced regional GM in men, but not women, suggesting possible sex differences in the relation between BMI and regional GM volume. Taki et al. (25) posit that the null findings in women may be the result of gender differences in fat distribution, with visceral fat predominating in men and subcutaneous fat predominating in women (60). Visceral fat is likely indicative of metabolic syndrome (61,62), which is associated with elevated serum levels of inflammatory markers. As discussed earlier, inflammatory markers have been associated with changes in GM and WM volume (38,39). However, Walther et al. (29) did find significant correlations between BMI and GM volume in women, suggesting that other sample or study characteristics may play a role in the difference in findings. Additional studies are needed to ascertain whether severity of obesity, gender and types of fat distribution affect regional GM volume differently.

Interestingly, we found a positive correlation between BMI and regional GM volume in the right middle occipital gyrus. This result was not an a priori defined region of interest, but is comparable with the findings of a previous study (26), in which GM density in the middle occipital gyrus was greater in obese compared to lean individuals. Occipital regions are typically involved in visual processing such as object recognition, color perception, and selective attention (63,64). Using a food-based visual attention task, one neuroimaging study has found that BMI positively correlates with selective attention to appetizing food and greater activation in reward processing regions including the anterior insula, ventrolateral PFC and lateral OFC (10). Further, a meta-analysis of visual processing of food and non-food cues found that exposure to food images resulted in elevated activation in the lateral of GM complex (a region extending from the posterior fusiform gyrus to the inferior occipital gyrus) (65). Given that individuals with a higher BMI show increased selective attention toward appetitive stimuli, it is possible that greater GM in the visual cortex (e.g., occipital region) reflects this difference in neural activity. However, it is important to note that increased GM volume measured with VBM can also result from several factors that we were unable to test, including differences at the cellular level and different folding patterns (66,67). Therefore, it will be important for future studies to define which of these possibilities underlies the observed positive relation.

Comparable to a previous study (32), we did not find significant differences in global WM volume between obese and lean individuals or between overweight and lean individuals. However, considering that we did find greater overall WM in overweight compared to obese individuals, it is also possible that our null findings can be explained by a small sample size. Only 17 participants in our sample were obese, which potentially limited our statistical power to detect small effects. It will important for future studies to explore the global WM volume differences in larger samples sizes.

BMI correlated positively with WM volumes in bilateral middle temporal gyrus, left fusiform, and bilateral parahippocampal gyrus, left Rolandic operculum, and right dorsal striatum (putamen and caudate), all regions previously found to be involved in palatable food receipt, food cues, and reward (68–72). Further, we also found a positive correlation between BMI and WM volume in the middle occipital gyrus. Overall, these results converge with previous findings (32,26). Moreover, Haltia et al. (32) found that after obese participants went on a low calorie diet for 6 weeks, global WM volume as well as regional WM volume in the left temporal lobe was reduced. Haltia et al. (32) hypothesized that fatty acid excess in obesity could result in pathological lipid metabolism in the brain, and this may increase brain WM volume. It is therefore possible that WM expansion is secondary to fat accumulation in obesity. Although we found positive relations between BMI and regional WM volumes, it is important to consider that it is unknown whether the higher WM volume actually reflects increased myelinization. Previous studies have found that higher BMI is associated with regional WM damage (73,74). Further, obese mice appear to have a lower amount of myelin compared to normal mice (75). Diffusion tensor imaging (DTI) is a neuroimaging technique able to examine the microstructural integrity of white matter. Increased myelinization could be indicated by a higher fractional anisotropy value while white matter damage by a lower one. Further studies determining whether the higher volume of white matter observed in obese individuals actually reflects increased myelinization are warranted.

Trend-level reduced regional GM volumes in the prefrontal cortex were associated with an increase in slope BMI from baseline to 1-year follow-up, controlling for initial BMI. Although studies have explored the associations between BMI and brain volume (e.g., 26,25) and the effects of dieting on regional GM/WM volumes in obese subjects (32), this is the first prospective study to examine the relation between regional GM/WM volumes and future weight gain. These findings dovetail with evidence that obese versus lean adults show less GM volume in the prefrontal cortex (26), less activation of prefrontal regions (e.g., middle frontal gyrus, ventrolateral prefrontal cortex) when trying to inhibit responses to unhealthy food images, and show behavioral evidence of reduced inhibitory control (49). The prefrontal cortex has been commonly implicated in the inhibition of inappropriate responses, control of goal-directed behaviors, and ability of error detection/correction (76) and could have a role in decisions to terminate feeding (77) and whether or not to eat unhealthy food. Collectively, these data may suggest that reduced GM volume in regions involved in inhibitory control may increase the risk for giving into tempting high-fat/sugar foods, resulting in overeating and weight gain.

In contrast to our hypothesis, we did not find a significant relation between WM volume and future increases in BMI. Haltia and colleagues (32) found that after following a very-low-calorie-diet over a period of 6 weeks, obese individuals showed a significant reduction in global and regional WM volume in the temporal lobe while changes in global and regional GM were non-significant. It is therefore possible that abnormalities in regional GM volumes, but not WM volumes increase risk for future weight gain and abnormalities in regional WM volumes but not GM volumes are secondary to weight gain. An alternative possibility is that there are reciprocal relations between BMI and regional GM/WM volumes. For example, reduced GM in the prefrontal cortex may lead to less inhibitory control towards eating high-fat/sugar foods, which may lead to overeating and weight gain, and consequently may lead to higher regional WM due to fat accumulation. Unfortunately, we were not able to test the effect of change in BMI on brain volume, as we did not have repeat scans for these subjects at 1-year follow-up. Either possibility suggests that examining bi-directional relations between BMI and regional brain volume in prospective repeated-measure studies may provide important information above and beyond the current findings.

It is important to consider the limitations of this study. First, due to possible registration errors and smoothing, it cannot be excluded that some GM signal is included in the total WM signal and vice versa. Second, the current study was conducted solely with young females, thus results should be generalized with caution to males and to adults. Third, due to our relatively small sample size, the current study was sufficient to detect large and medium effects, but was not adequate to detect small effects. Fourth, although BMI is widely used to assess excess adiposity, is inexpensive, and shows high test-retest reliability, it does not distinguish between increased mass in the form of fat, lean tissue, or bone and hence can lead to significant misclassification (78,79). However, it is important to note that BMI correlates highly with direct measures of total body fat such as dual energy x-ray absorptiometry in female (r = 0.85 – 0.87) and male adolescents (r = 0.82 – 0.89) (53,54). Nonetheless, it is possible that by using a more accurate metric of access fat mass (e.g., dual energy x-ray absorptiometry), additional or stronger correlations with volumetric brain differences would have been found. Future studies testing the correlations between fat mass and volumetric brain differences are needed to examine these relations more closely.

Despite the aforementioned limitations, the current findings suggest that elevated weight is associated with low global GM volume and high WM volumes in regions previously found to be involved in palatable food receipt, food cues, and reward. The current findings also suggest that reduced GM volumes in regions related to inhibitory control are associated with future increases in BMI. Undoubtedly, the relations between BMI and GM/WM volumes are complex and further consideration of other variables (e.g., medical conditions, genetics) is warranted. However, while the specific mechanisms underlying these volumetric differences remain to be investigated, our findings have important public health implications because they suggest that regional and global brain volume abnormalities are related to BMI and increases in BMI at a relatively young age, potentially resulting in greater risk for future declines in cognition or other brain functions.