A glucose-stimulated BOLD fMRI study of hypothalamic dysfunction in mice fed a high-fat and high-sucrose diet

The hypothalamus is the central regulator of energy homeostasis. Hypothalamic neuronal circuits are disrupted upon overfeeding, and play a role in the development of metabolic disorders. While mouse models have been extensively employed for understanding the mechanisms of hypothalamic dysfunction, functional magnetic resonance imaging (fMRI) on hypothalamic nuclei has been challenging. We implemented a robust glucose-induced fMRI paradigm that allows to repeatedly investigate hypothalamic responses to glucose. This approach was used to test the hypothesis that hypothalamic nuclei functioning is impaired in mice exposed to a high-fat and high-sucrose diet (HFHSD) for seven days. The blood oxygen level-dependent (BOLD) fMRI signal was measured from brains of mice under light isoflurane anaesthesia, during which a 2.6 g/kg glucose load was administered. The mouse hypothalamus responded to glucose but not saline administration with a biphasic BOLD fMRI signal reduction. Relative to controls, HFHSD-fed mice showed attenuated or blunted responses in arcuate nucleus, lateral hypothalamus, ventromedial nucleus and dorsomedial nucleus, but not in paraventricular nucleus. In sum, we have developed an fMRI paradigm that is able to determine dysfunction of glucose-sensing neuronal circuits within the mouse hypothalamus in a non-invasive manner.


Introduction
Brain function requires continuous glucose supply, 1 which is ensured by a glucose sensing network across several brain regions. 2 The hypothalamus is pivotal in the central glucose sensing, in addition to controlling other basic life functions such as feeding behaviour, thermoregulation, sleep or fear response. 2,3 Different hypothalamic nuclei are involved in regulating peripheral metabolism to maintain glucose homeostasis (Figure 1), namely the arcuate nucleus (ARC), the lateral hypothalamus (LH), the ventromedial nucleus (VMN), the dorsomedial nucleus (DMN) and the paraventricular nucleus (PVN). 4 The ARC and VMN, as well as brainstem and corticolimbic structures, contain glucose-sensing neurons: high glucose levels depolarize glucose-excited neurons via glucokinase (GK) and ATP-mediated closure of ATP-sensitive K þ channels, and low glucose concentration depolarises glucoseinhibited neurons namely via AMPK-dependent closure of Cl À channels. 2 The hormonal regulation of the melanocortin system, which consists of two functionally antagonistic neuronal populations in the ARC, is key to the control of appetite, energy expenditure, and glucose homeostasis: one subset of neurons expresses the orexigenic neuropeptides (appetite-stimulating) agouti-related peptide (AgRP) and neuropeptide Y (NPY), the second subset expresses the anorexigenic peptides (appetite-suppressing) pro-opiomelanocortin (POMC) and cocaine and amphetamine-regulated transcript (CART). 2,3 These signals are then integrated by secondary neurons in other hypothalamic nuclei, namely the PVN, VMN, DMN, and LH, as well as in extrahypothalamic areas. 3 ARC neurons also receive feedback from other hypothalamic nuclei, resulting in a finely tuned response to control food intake. 5 Hypothalamic function has been measured noninvasively by magnetic resonance imaging (MRI). Smeets et al. 6 reported a prolonged dose-dependent decrease of the functional MRI (fMRI) signal in the hypothalamus after a glucose load. This response started shortly after subjects drank the glucose solution and lasted for at least 30 min, and was most pronounced in the upper anterior hypothalamus. In the upper posterior hypothalamus, the signal decrease was independent of the glucose load, and no effect was found in the lower hypothalamus. Such hypothalamic responses have also been reported elsewhere. 7 Conducting such studies in animal models has been challenging due to breathing or uncontrolled animal movement that generates image artifacts, or biased by the employment of anaesthesia. Nevertheless, Mahankali et al. 8 found a negative BOLD fMRI response elicited by i.p. glucose administration in the hypothalamus of rats under a-chloralose anaesthesia. Lizarbe et al. 9 have also previously observed T 2 *weighted contrast changes in the hypothalamus of mice in an image acquired during glucose infusion versus baseline. While this study lacked temporal resolution to evaluate dynamic responses of hypothalamic nuclei, we aimed at establishing an experimental setup for evaluating the function of the hypothalamus in real time using glucose-induced fMRI.
Metabolic syndrome and obesity are associated to hypothalamic dysfunction through mechanisms that include neuroinflammation and the subsequent insulin and leptin resistance of neurons, which disrupts sensing of metabolic cues and further promotes food intake and body weight gain. 3 Hypothalamic inflammation is actually an early event in the development of the metabolic syndrome upon overfeeding, and mice exposed to a high-fat diet show an inflammatory response within one day, which is proposed to play an important role on the subsequent hypothalamic neurodegenerative process. 10 In fact, exposure of rodents to diets is a widely used approach to investigate metabolic syndrome components and its impact on brain function. 11 Namely, diets rich in fat and sugar were reported to impact the central nervous system within only one week. 12,13 Therefore, we further investigated the impact of short-term high-fat and high-sucrose diet (HFHSD) exposure on the fMRIdetected hypothalamic response to glucose.
In sum, the goal of this work is two-fold: (i) to develop a paradigm for non-invasively assessing hypothalamic functioning in mice; (ii) to map hypothalamic dysfunction upon short-term high-fat and highsucrose feeding.

Animals
All procedures on animals were approved by the Malm€ o/Lund Committee for Animal Experiment Ethics (permit number 994/2018), conducted according to EU Directive 2010/63/EU, and are reported following the ARRIVE guidelines (Animal Research: Reporting In Vivo Experiments, NC3Rs initiative, UK). Male C57BL/6J mice were obtained from Taconic (Ry, Denmark) at eight weeks of age and allowed to acclimatise to the animal facility for one week. Mice were housed in groups of 4-5 on a 12-h light-dark cycle with lights on at 07:00, room temperature at 21-23 C, humidity at 55-60%, and access to tap water and food ad libitum. Controls were fed a lowfat diet with 10% kcal from fat, 20% kcal from protein, and 70% kcal from carbohydrates (D12450J, Research Diets, New Brunswick, NJ, USA). HFHSD-exposed mice were fed a high-fat diet with 60% kcal from fat, 20% kcal from protein and 20% kcal from carbohydrates (D12492, Research Diets), and were also supplemented with a 20% (w/v) sucrose solution in addition

MRI experiments
On the day of scanning, access to food was terminated at 8:00, and experiments took place from 14:00 to 16:00. Mice were anesthetized with isoflurane by induction at 3.5% and maintenance at 2% vaporized in a 1:2 O 2 :N 2 O gas mixture. A PE50 line (Warner Instruments, Hamden, CT-USA) was inserted in the mouse trachea for mechanical ventilation, and a cannula was placed i.p for glucose or vehicle infusion. Then mice were positioned on the MR-compatible bed with a teeth bar and two ear bars for stereotaxic head fixation. The mouse was then fitted into a homebuild mask that connected the tracheal line to the MRcompatible mechanical ventilator (MRI-1, CWE, Ardmore, PA, USA). When breathing was assisted by mechanical ventilation, mice received an i.p. injection of 0.1 g/kg of pancuronium bromide (Sigma-Aldrich, Schnelldorf, Germany) diluted in sterile saline. The ventilation rate was kept at 90 breaths per minute with a volume of 1.9-2.2 mL and 50% inspiration rate. Isoflurane level was then decreased to 0.7%, 16 and the mice were then inserted into the MRI scanner. Body temperature was continuously monitored with a rectal temperature probe (SA Instruments, Stony Brook, NY, USA) and maintained at 36-37 C with a warm water circulation system. Breathing rate was also confirmed by the SA Instruments monitoring system. MRI experiments were performed on a Bruker BioSpec 9.4 T equipped with a 1 H quadrature transmit/receive MRI cryoprobe. 17 A localizer MRI sequence was first used to confirm good positioning of the animal in the scanner and allow for eventual adjustments. Then a set of T 2 -weighted images over the whole brain was acquired for anatomical reference using rapid imaging with refocused echoes (RARE) sequence with repetition time (TR) of 3.5 s, echo time (TE) of 33 ms, 0.5 mm slice thickness, image size of 320 Â 320 and field of view (FOV) of 14 Â 14 mm 2 .
These images were used to position the slices for fMRI. BOLD fMRI data were acquired during 20 min using gradient echo-echo planar imaging (GE-EPI) sequence with TR ¼ 2 s and TE ¼ 18 ms. 18 Signal was acquired from four 0.5-mm slices encompassing the whole hypothalamus, resolution of 0.104 Â 0.109 mm 2 and FOV of 14 Â 10 mm 2 . After 5 min of fMRI scanning, either 2.6 g/kg of glucose (from a 30% solution in saline) or equivalent saline volume was injected i.p. during 3 min. Saline infusion was usually performed in the same mouse before or after the glucose infusion. The order of the two infusions was randomly chosen to eliminate systematic errors, and there was an interval of at least 30 min between the two infusions. After the experiment, mechanical ventilation and isoflurane anaesthesia were stopped when spontaneous breeding reflexes were observed, and mice returned to their cages.
Mock experiments out of the scanner were conducted in two mice to determine the typical glycaemic profile during the glucose-induced fMRI paradigm. Glucose was determined from tail-tip blood samples.
fMRI data analysis EPI images were reconstructed and corrected for motion using SPM12 (Statistical Parametric Mapping, London, UK), and the average signal from predefined regions of interest (ROI) in the hypothalamus was extracted. 15 Namely, the hypothalamus was separated in different ROIs based on their relative position in the mouse brain atlas by Paxinos and Franklin 19 : the dorsomedial nucleus (DMN), the paraventricular nucleus (PVN), the arcuate nucleus (ARC) and finally the lateral hypothalamus (LH) plus the ventromedial nucleus (VMN), from which the average signal was extracted. Signals from LH and VMN were analysed as a single ROI due to lack of contrast on anatomical and functional images (LH þ VMN). For each experiment, average signal was extracted from the four ROI using the SPM MarsBar toolbox, 20 and normalized to the average signal during the 2 min before glucose infusion (baseline signal).
Brains for microscopy were processed as previously described. 22 Briefly, brains were fixed in phosphatebuffered formaldehyde, and then stored at 4 C in a 30% sucrose solution in PBS. Immunostaining was carried out on 20-mm cryostat-sectioned coronal brain slices. Slices were incubated for 1 h at room temperature with blocking buffer (PBS containing 5% normal goat serum, 1% BSA and 0.3% Triton X-100), and then overnight at 4 C with the AlexaFluor488-conjugated antibody against phosphor-CREB (1:50 in blocking buffer). After washing in PBS, slices were mounted with ProLong Glass Antifade (Invitrogen), and examined under a Nikon A1RHD confocal microscope with a 20Â Plan Apo objective, NA 0.75 (Nikon Instruments, Tokyo, Japan). Images were acquired with NIS-elements, version: 5.20.01 (Laboratory Imaging, Nikon). Images were processed for nuclei counting in ImageJ (NIH, Bethesda, MD, USA).

Statistics
Statistical analyses were performed using Prism 8.4.2 (GraphPad, San Diego, California, USA). Results are shown as mean AE SD unless otherwise stated. Data were assumed to follow normality, as confirmed by the Shapiro-Wilk test. Comparison of results was carried out with unpaired, two-tailed Student t-tests, or with ANOVA followed by Fisher's LSD post-testing. BOLD fMRI time-courses were analysed using a twoway repeated-measures ANOVA with time and diet as factors, and with Geisser-Greenhouse correction for non-sphericity. Significant differences were considered for P < 0.05.

Glucose-induced fMRI
In pilot experiments, functional MRI experiments with freely breathing mice under 1.5% isoflurane anaesthesia allowed to obtain stable images with limited motion (data not shown). However, we observed unexpected and unpredictable signal variations in bottom areas of the hypothalamus that are closer to the trachea, which were considered to be breathing-associated artifacts. Such artifacts remained during fMRI acquisitions triggered to the breathing rate and, furthermore, variable TR would result in magnetization variations that require additional TR-dependent signal corrections. BOLD fMRI time courses with minimal artifacts were obtained from mice under mechanical ventilation with pancuronium bromide-induced paralysis, as reported by others. 17 This also allowed us to lower isoflurane anaesthesia level to 0.7%, thus reducing its effects on cerebrovascular reactivity as in previous work. 16 Anatomical images of the rat brain were used to identify the hypothalamus and approximate location hypothalamic nuclei by comparison to a mouse brain atlas (Figure 2(a)). T 2 *-weighted images were obtained from four coronal slices encompassing the hypothalamus (Figure 2(b)), and were used to calculate relative BOLD signal variations. These images showed minimal distortions and also provided sufficient contrast to identify the hypothalamus. Upon administration of glucose, the hypothalamus responded promptly with a BOLD signal reduction, which was not observed in other brain areas, such as the cortex (Figure 2(c)), or the lateral ventricles (not shown). This hypothalamic response appeared to be biphasic, reaching a plateau $5 min after the beginning of the glucose infusion, and decaying again after $10 min.
Confirming an effective infusion, blood glucose measured from the tail tip increased from 11.5 AE 1.6 mmol/L before the fMRI experiment to 27.2 AE 2.1 mmol/L at the end of the scan (range of 2-3-fold increase; P ¼ 0.002). This increase in blood glucose upon i.p. administration is likely linear during the 15 min of experiment, as reproduced in two bench tests (Figure 2(d)).
A continuous BOLD signal drift ($1% per 15 min) was observed across the mouse brain (Figure 2(c)). This was of much lower magnitude than the signal drop after glucose administration, and was also observed in the hypothalamus upon saline infusion (Figure 3  (a)). Therefore, it might be that a small fraction of the estimated glucose-induced hypothalamic response includes a glucose-independent signal drift.

HFHSD effect on hypothalamic responses
Mice exposed to seven days of HFHSD were of similar weight than controls (27 AE 1 g vs. 24 AE 1 g; P ¼ 0.082), had higher blood glucose (in mmol/L: 9. BOLD fMRI traces were different between control and HFHSD-exposed mice ( Table 1). The glucoseinduced hypothalamic response was dampened in HFHSD-exposed mice, when compared to controls (Figure 3(a)). That was mainly due to an impairment of the initial response to glucose. When measured at 5 min from infusion start, the BOLD signal relative to baseline decreased by 3.0 AE 0.5% in controls and by 1.0 AE 0.3% in HFHSD (P ¼ 0.017). After saline infusion, BOLD signal relative to baseline was reduced by only 0.5 AE 0.4% in controls and 0.3 AE 0.2% in HFHSD (P ¼ 0.775). When measured at the end of the fMRI experiment (last minute), the BOLD signal relative to baseline decreased by 5.3 AE 1.0% in controls and by 3.0 AE 0.9% in HFHSD (P ¼ 0.118). After saline infusion, final BOLD signal was À1.1 AE 0.3% in controls and À1.1 AE 0.5% in HFHSD (P ¼ 0.951) relative to baseline.
The influx of Ca 2þ into stimulated neurons triggers CREB phosphorylation, which acts as a transcription factor increasing the expression of c-fos mRNA that is then translated into a functional regulator of gene expression. 24 Levels of pCREB and c-fos were used for confirming neuronal activation (Figure 4). Twenty minutes after the glucose load, HFHSD-exposed mice had less pCREB-positive nuclei than controls (repeated measures ANOVA: diet F 1,3 ¼ 40.03, P ¼ 0.008; ROI F 2,6 ¼ 15.09, P ¼ 0.005; interaction F 2,6 ¼ 0.26, P ¼ 0.779; Figure 4(a)). Post hoc testing identified a HFHSD-induced reduction in the number of pCREBpositive nuclei in all analysed areas (ARC P ¼ 0.018, VMN P ¼ 0.029, DMN P ¼ 0.008). Immunoblotting  experiments found no significant change in CREB phosphorylation rate (P ¼ 0.067; Figure 4(b)), and a 26 AE 9% reduction of c-fos levels (P ¼ 0.029; Figure 4 (c)) in the whole hypothalamus of HFHSD-exposed mice, relative to controls.

Discussion
Hypothalamic neurons sense the nutrient status and integrate signals from peripheral hormones including pancreas-derived insulin and adipocyte-derived leptin to regulate appetite, metabolism and energy expenditure. 2,3 Although understanding the functioning of hypothalamic circuits is key for metabolic diseases research, hypothalamic functional imaging in mouse models remains challenging. In this work, we developed an experimental set-up for non-invasively assessing the function of the mouse hypothalamus using fMRI, and found that short-term high-fat and highsucrose feeding impairs hypothalamic neuronal circuits, particularly in ARC, DMN and VMNþLH. MRI experiments were performed under light isoflurane anaesthesia, and image artifacts at the level of the hypothalamus could only be prevented by paralysis and mechanical ventilation. 13,25 Despite efforts to improve image quality, responses within the ARC remained particularly sensitive to background noise, likely due to its proximity to the trachea. Under these conditions, the hypothalamus responded to glucose with a biphasic BOLD signal reduction -there was a first plateau 5 min after glucose infusion followed by a subsequent signal decay. This biphasic response might be caused by different onsets of activation across hypothalamic nuclei and their interactions (Figure 1), by signal cancellations due to inhibitory and excitatory responses involving negative and positive BOLD effects, 26 or by subsequent hormonal actions, such as insulin. 27 We further analysed sub-hypothalamic VOIs in order to dissect their temporal activation. Interestingly, all analysed areas displayed an immediate response upon glucose administration, which was stronger in ARC and weaker in PVN. Since all areas display the typical biphasic response, it is likely that it represents the counter-regulatory hormonal action following the glucose load.
Negative glucose-induced hypothalamic responses have also been observed in humans 6,7 and anesthetised rats. 8 In mice, others have observed a signal increase in LH and DMN, but not ARC or VMN, in a T 2 *-weighted image acquired during glucose administration relative to baseline. 9 Interestingly, the authors reported signal reductions in ARC and VMN upon saline administration, which might be attributed to the infusion protocol. One should also note that mice in Lizarbe's study were exposed to a prolonged fasting period prior experimentation, which is not the case in our work.
To the best of our knowledge, this is the first study investigating the direct response of the hypothalamus after short-term diabetogenic diet feeding in a rodent model. The biphasic responses observed in controls were strikingly attenuated in the hypothalamus of HFHSD-fed mice, especially in the ARC and VMN. This functional impairment was confirmed by reduced molecular markers of neuronal activation (pCREB and c-fos). The ARC is a key element in food intake regulation, since its border with the blood-brain barrier is semi-permeable and facilitates passage of blood-stream hormones, and its neurons project to other hypothalamic areas for signal processing and integration. 2 The loss of immediate ARC neuronal responses in HFHSD might thus preclude adequate activation of the neuronal circuit that allows the hypothalamus to elicit a normal response to glucose. The measured HFHSD-induced dampening of the BOLD fMRI response to glucose cannot be simply attributed to impaired glucose sensing. The neurovascular coupling that underlies the BOLD signal involves tight metabolic interactions between neurons and glial cells, namely astrocytes. 1 The early neuroinflammatory process of the hypothalamus also involves astrogliosis. 10,29 During this astrogliosis process, the metabolic support from astrocytes to neurons is likely disrupted. For example, our recent work on T2D rats showed altered astrocytic metabolism and impaired glutamine synthesis for glutamatergic neurons, 17,30 which account for a substantial fraction of hypothalamic cells. 31 Therefore, it is plausible that disrupted neuronastrocyte metabolic interactions lead to blunted neurovascular coupling, even with preserved glucose sensing at cellular level.
On the other hand, astrocytes are also direct sensors for glucose and fatty acids, 32 leptin 33 and insulin. 34 Importantly, astrocytic insulin signalling plays a role in glucose-induced neuronal activation, 35 and obesity impacts astrocyte functions. [33][34][35][36] Loss of astrocytic glucose sensing in the hypothalamus may play a role in obesity and T2D development. 32

Limitations
In order to avoid any hemodynamic changes caused by an intravenous infusion, glucose was given by i.p. infusion of a hyperosmotic solution. Although the hypothalamus responds to blood osmolality changes, we believe that the observed hypothalamic response results mainly from glucose sensing. Accordingly, continuous glucose infusions have been shown to result in vasopressin responses much smaller than those triggered by hypertonic saline, urea or mannitol. 37 While the present experiments were designed to identify glucose-induced hypothalamic responses, other brainstem and corticolimbic structures contain glucose-sensing neurons. 2 Future studies should implement a whole brain fMRI acquisition in order to provide insight into physiological responses of those areas. Moreover, since other brain regions participate in the response to feeding or food cues, 38 as well as in the integration of hypothalamic signals, 3 both restingstate and glucose-stimulated functional connectivity studies might be considered for an exhaustive characterization of appetite and satiety regulation. 25 Given the biological and experimental variability observed in this study, such complex analyses would require increased sensitivity in BOLD fMRI acquisition, and larger sample sizes. Despite the limited number of mice, this proof-of-concept study identified a significant glucose-induced hypothalamic response in every control mouse (see Figure 2(c)).
The voxels composing the hypothalamus or hypothalamic nuclei were carefully selected according to visible anatomical landmarks in 0.5-mm thick slices and the drawings in a mouse brain atlas. 19 This approach does not preclude that extra-hypothalamic volumes are included in the VOIs, and any partial volume effects might reduce the magnitude of our observations, but do not impact our conclusion.

Conclusion
Mouse fMRI experiments are especially challenging in the hypothalamus, which is near an airway that causes major image artifacts. While previous functional studies in the mouse hypothalamus lacked temporal resolution to investigate dynamic responses, 9 we now developed an experimental paradigm for real-time glucose-induced fMRI tasks. This approach is fully noninvasive and can thus be employed for longitudinal mouse studies, 22 such as following trajectories of hypothalamic dysfunction.