*3.4. Post-hocs for Di*ff*usion Imaging Measures*

Because MD is defined as the average diffusivity, it is not as biologically specific as axial diffusivity (AD) or radial diffusivity (RD). Therefore, we ran post-hoc analyses using Model 4 only to determine whether the observed association of increased fructose intake on MD of the right cingulum, prefrontal connections, was driven by axial diffusivity (AD), radial diffusivity (RD), or both. The AD for the right cingulum, prefrontal connections was significantly associated with increased percentage of calories from fructose (ß = 4.62 <sup>×</sup> 10<sup>−</sup>6, sr = 0.27, *p* < 0.005; Model 4). The RD for the right cingulum, prefrontal connections was significantly associated with increased percentage of calories from fructose (ß <sup>=</sup> 2.89 <sup>×</sup> <sup>10</sup><sup>−</sup>6, sr = 0.23, *p* < 0.01; Model 4). This pattern was quite similar to the pattern observed with MD. Comparing the effect sizes of sr = 0.27 for AD and sr = 0.23 for RD, we conclude that the effects were similar, but slightly larger for AD.

#### **4. Discussion**

In this study, we examined the influence of dietary sugar intake on hippocampal neuroanatomy, both gray matter and white matter, in children. We found that increased volume in the right hippocampal CA2/3 subfield was associated with increased consumption of the monosaccharides, fructose or glucose, and/or added sugars in general, while increased MD in the right cingulate-prefrontal cortex connections were only associated with increased dietary intake of fructose. We found associations on the right but not the left hemisphere, which is broadly consistent with prior work showing lateralities on the impact of environmental insults in the hippocampus [58,59]. These associations between dietary sugar intake and hippocampal volume and fructose intake and cingulate-prefrontal cortex connections remained significant after adjusting for child's intracranial volume, age, sex, BMI, SES, and prenatal exposures suggesting an effect of dietary fructose, added sugar, and glucose intake on hippocampal volume and an effect of dietary fructose on cingulate-prefrontal cortex white matter connectivity that is independent of a number of potential confounding factors.

The observed association between dietary sugar intake and increased hippocampal volume during childhood may be driven by a few factors. A larger hippocampal CA2/3 volume during childhood could be due to a delay in synaptic pruning, a process that typically occurs during early adolescence. During adolescence, the hippocampus reaches peak volume, and begins to undergo synaptic pruning, eliminating unused connections [60,61], and prior studies have shown that the CA3 hippocampal subfield gradually begins to decrease in volume during mid childhood [61,62]. Work by Vuong et al. found that synaptic pruning is delayed in juvenile rats exposed in utero to obese mothers with GDM [63]. In their findings, inflammation and recruitment of microglial cells occurred during post-natal development in the hippocampus, along with a reduction in synaptic pruning, and the animals presented with altered hippocampal morphology. These findings suggest that prenatal environmental insults can result in hippocampal inflammation, reductions in synaptic pruning and altered hippocampal development [63]. Interestingly, Hershey et al., found that children with Type 1 diabetes had increased hippocampal volume, despite the commonly reported decreased volume in adults with Type 1 and Type 2 diabetes [64]. There is substantial research that a diet high in added sugar contributes to a pro-inflammatory environment [13,19,65]. Therefore, increased volume in the CA2/3 hippocampal subfield may be due to inflammation and/or a delay in synaptic pruning. Future studies that examine inflammation specifically, such as T2 scans for gliosis, could potentially address this possibility [66].

Notably, we found that the CA2/3 hippocampal subfield was preferentially impacted by dietary sugar intake. Coincidentally, the CA3 subfield is among the last subfields to undergo postnatal maturation, paralleling DG development [67,68]. Additionally, substantial work in animals has shown that the CA3 subfield is altered by a host of prenatal and early life environmental insults, such as prenatal exposure to GDM [69,70] or postnatal exposure to chronic stress [71–73]. Therefore, our findings are in line with prior research in animals indicating a preferential sensitivity of the CA3 subfield. It is worth noting, due to the predetermined boundaries delineated by FreeSurfer, we were unable to decipher the boundary between the CA2 and CA3 subfield. Future studies that incorporate manual tracing should be considered to confirm if excessive added sugar intake preferentially impacts the CA3 subfield of the hippocampus.

Many studies have observed that MD decreases while FA increases over the course of development [52–55]. We found that increased dietary fructose intake was associated with increased mean diffusivity (MD) in the white matter cingulum tract that connects the right hippocampus and the right prefrontal cortex. MD has been shown to increase in many different white matter diseases that lead to demyelination, dysmyelination, and/or wallerian degeneration [74–76]. For example, increases in MD have been observed prior to the appearance of lesions on a gadolinium-enhanced scan in multiple sclerosis, which is a complex disease that involves not only demyelination and wallerian degeneration, but also edema and inflammation, indicating that changes in the MD may be sensitive to pre-lesion changes in the blood brain barrier [77]. Patients with chronic epilepsy and hippocampal

sclerosis show increased MD values in the hippocampus [78,79]. Together, these results are consistent with the idea that the increased MD values associated with fructose could reflect an inflammatory process, possibly associated with a loss of myelin, or delayed axonal pruning.

Because mean diffusivity is a weighted average of axial (AD) and radial diffusivity (RD), we conducted post-hoc analyses on these measures to obtain more biological specificity about what might be underlying the observed associations with MD. In general, AD is thought to reflect intracellular water mobility and is influenced by the integrity and arrangement of axonal membranes and cytoskeletal proteins, while RD is thought to reflect more of the extracellular water mobility and is primarily influenced by myelin [80]. Previous studies have shown that over the course of normal neurodevelopment, both RD and AD decrease with age [81–85]. Decreases in RD are typically attributed to myelination, while the decreases in AD are thought to correspond more to axonal pruning [86]. In our study, we found that both AD and RD were increased in the cingulum tract in children with increased dietary fructose intake, consistent with the idea that increases in fructose intake could be associated with a decrease or delay in the myelination process and the normal pruning process. Future studies could use multi-compartment diffusion imaging models, such as Neurite Orientation Dispersion and Density Imaging (NODDI), to further investigate this tract to determine whether these changes in diffusivity are more related to cellular density and/or orientation dispersion, thus adding more biological specificity [87].

Self-reported dietary assessments have known limitations, including under-reporting of dietary intake. Given that eating behavior can vary from day to day, the collection of multiple 24-hr recalls on non-consecutive days is recommended to improve the accuracy of habitual dietary intake estimates [88]. Some studies have recommended obtaining at least three 24-hr dietary recalls on non-consecutive days to provide better accuracy of energy intake [36]. Dietary assessments in our study were obtained from repeated 24-hr recalls obtained during in-person visits on non-consecutive days. We obtained two days of dietary assessments on the majority of participants, and four days of dietary assessments on a subset of 35 participants. We implemented techniques to improve the accuracy of our dietary data collection method, including the Multiple Pass 24-h recall method, which was previously shown to significantly reduce levels of under-reporting [89]. We also used a consensus recall method, in which the child and parent were interviewed together, which has been suggested to improve the accuracy of dietary assessments compared to interviews with the child or parent alone [90]. Estimates of total energy intake in our cohort are similar to national averages for children in this age range [91], while the average percent calories from added sugar in our cohort was 2.5% lower than the average of 16.4% reported in the 2009 to 2012 National Health and Nutrition Examination Survey for children age 6–11 [92]. Thus, while total energy intake estimates in our cohort are in line with those of national averages of children in the U.S., it is possible that our cohort under-reported intake of foods and beverages that contain added sugars. It is also possible that the children in our cohort consumed, on average, slightly lower amounts of added sugars than average children in the U.S. population.

In summary, our findings suggest that increases in dietary fructose are associated with alterations in hippocampal structure and connectivity in children. These findings should be interpreted cautiously given the limitations of self-reported dietary intake assessments, and it is important to note that our observations are correlational and do not confer causality. Future studies that include experimental designs that manipulate dietary intake of fructose and/or added sugars are necessary to determine the effects of fructose and added sugar on hippocampal structure and connectivity during childhood. Moreover, the potential cognitive consequences of the observed associations between dietary fructose and alterations in hippocampal structure and connectivity remains an important consideration. Our findings support the need for future studies that include cognitive testing in addition to neuroimaging to examine whether increased dietary fructose intake in childhood is associated with altered hippocampal structure and hippocampal function in childhood.

*Nutrients* **2020**, *12*, 909

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6643/12/4/909/s1, Figure S1: Raw data plots of added sugar consumption and the mean diffusivity of the right cingulum, prefrontal connections, Figure S2: Raw data plots of added sugar consumption and the volume of the right CA2/3 subfield of the hippocampus, Figure S3: Raw data plots of glucose consumption and the mean diffusivity of the right cingulum, prefrontal connections, Figure S4: Raw data plots of glucose consumption and the volume of the right CA2/3 subfield of the hippocampus, Table S1: Table of hippocampal and total brain volumes.

**Author Contributions:** K.A.C., K.A.P., J.M.A., S.J. were responsible for conceptualization of the study; K.A.C., J.M.A., R.P.C., B.A. contributed to methodology and formal analysis; J.M.A., A.G.Y., B.A., S.L. were responsible for management and coordination of the study execution; K.A.P. and A.H.X. were responsible for supervision of the research activities; K.A.C., K.A.P., S.J., J.M.A. wrote the original draft; K.A.C., K.A.P., S.J., J.M.A., S.L., A.G.Y., R.C., A.H.X. provided critical review, commentary and revisions to the manuscript; K.A.P., A.H.X., S.L., J.M.A. provided funding for this study. Prior Presentation. Some of the data from the BrainChild Cohort were used to address a separate research question related to prenatal exposure to maternal obesity and hippocampal volume in children, which was accepted for publication in *Brain and Behavior* on December 8, 2019. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by an American Diabetes Association Pathway Accelerator Award (#1-14-ACE-36; principal investigator K.A.P.) and in part by National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health (NIH), R01-DK-116858 (principal investigators K.A.P. and A.H.X.), K01-DK-115638 (principal investigator S.L.), and the National Institute of Mental Health, F31MH115640 (PI: J.M.A). A Research Electronic Data Capture (REDCap) database was used for this study, which is supported by the Southern California Clinical and Translational Science Institute through NIH grant UL1-TR-001855.

**Acknowledgments:** The authors would like to thank the families who participate in the BrainChild Study. The authors would also like to thank Ana Romero for managing the BrainChild study, Mayra Martinez and Janet Mora-Marquez for recruiting volunteers, Alexis Defendis for helping with study execution, and the staff at Dana and David Dornsife Cognitive Neuroimaging Center at USC and at the USC Diabetes and Obesity Research Institute for their assistance with these studies.

**Conflicts of Interest:** The authors declare no conflict of interest.
