Feeding impulse in the brain results from a complex balance between orexigenic and anorexigenic stimulations in the hypothalamus, reward centers, brain stem and cortex, driven by peripheral and intracerebral effectors¹. Despite enormous progress during the last decades on the molecular and cellular determinants of appetite, the in vivo organization of cerebral responses to feeding paradigms remains poorly understood. Non-invasive imaging techniques have previously contributed important progress towards this end². We previously showed that functional Diffusion Weighted Imaging (fDWI) may detect areas of cerebral activation by feeding/fasting paradigms, through changes in the water diffusion parameters³. However, our analysis focussed previously only in hypothalamic and cortical structures. Here, we present the results of a study on global cerebral response to appetite in the mouse brain.

A total of fourteen adult male mice C57BL/6, receiving normal chow diet (A04 http://www.safe-diets.com/eng/home/home.html, SAFE Augy, France, 2900 kcal/kg) and water ad libitum, were investigated by fDWI under fed (n=14), 48h (n=8) or 16h (n=6) fasted conditions. Diffusion Weighted Imaging performed in anesthetized mice (isofluorane 2% / oxygen 99.9% mixture, 1mL/min), at 7T (Bruker Avance III) used a mouse head resonator (23 mm) and the Stejskal-Tanner spin-echo sequence in three orthogonal directions (left-right L-R, anterior-posterior A-P, and head-foot H-F) with a 4 shot EPI read-out train. Acquisition conditions were: δ = 4 ms, Δ = 20 ms, TR = 3000 ms, TE = 51 ms, Av = 3, FOV = 21mm x 21mm, acquisition matrix = 128 x 128, corresponding to an in-plane resolution of 164μm x 164μm and a slice thickness = 1.5 mm. We obtained five high b value acquisitions (200<b<1200 s.mm^-2) and six low b value acquisitions (10<b<90 s.mm^-2). Therefore, every pixel is described by a 33 component vector (eleven b values in three orthogonal directions) representing signal intensity values along the three orthogonal directions. fDWI datasets were analysed by a model-free classification method based on Linear Discriminant Analysis (LDA), classified between the fed and fasted conditions and the accuracy of classification determined using the leave-one-out strategy (Figure 1, left). Also, a biexponential model was fitted to the same fDWI datasets, using the following expression: $$\frac{S(b)}{S(0)} = SDP \cdot exp(-b \cdot D_{slow}) + FDP \cdot exp(-b \cdot D_{fast})$$ where, S(b) and S(0) represent the individual pixel intensities in the presence and absence of diffusion gradient, SDP and FDP represent the slow and fast diffusion phases (FDP = 1-SDP), D_slow and D_fast the corresponding apparent diffusion coefficients (Figure 1, right).

Figure 2 illustrates the application of model-free classification method based on LDA in a representative mouse under both conditions (left: fasted, right: fed). LDA was trained to find the linear projection that optimally discriminates individual pixels as belonging to the fed or fasted states. We named this projection “Appetite Index” (AI), and used it to classify the test mice as fed or fasted. Model-free analysis allowed the correct classification of the mouse brain between the two states (Figure 2, top panels), as well as AI maps representation (Figure 2, central panels). AI was higher through the whole brain in the fasted condition, as revealed by the shift to red colours. Using this algorithm and a leave-one-out strategy, we obtained a 81% correct classification hit rate with the group of eight mice and a 75% with the group of six mice. Next, we tested lateralization of the appetite index in the left and right hemispheres separately, under the fed or fasted conditions. Notably, the AI is higher in the right cerebral hemisphere in all cases (Figure 2, bottom panels).

To confirm further the lateralization of appetite, we used biexponential fitting of the same datasets. Figure 3 summarizes the results obtained after fitting the biexponential model on the fDWI datasets from eight mice fasted 48h. Top panel presents the mean values, standard deviations and comparative statistical significances of the SDP, D_slow, and D_fast diffusion parameters in the whole brain, from all mice in this group, under fed or fasted conditions, along the three orthogonal directions investigated (L-R, A-P, and H-F). The three diffusion parameters increased significantly with fasting in the three directions. Bottom panels show significant differences in the diffusion parameters between the right and left cerebral hemispheres suggesting that these changes determine those observed in AI.

In summary, we report that the cerebral activation by appetite can be detected by a global analysis of the mouse brain using fDWI. Our results reveal that appetite stimulation is asymmetrically distributed through the brain, predominantly affecting the right hemisphere.