Dorsomedial hypothalamic BDNF neurons integrate thermal afferent signals to control energy expenditure

Mutations in the gene brain-derived neurotrophic factor (BDNF) cause obesity in humans. BDNF signaling and its expressing neurons in the hypothalamus help control feeding, energy expenditure (EE), and physical activity. However, whether the BDNF neurons interact with another EE-regulating system, the thermoregulation circuitry, remains unclear. Here, we show that BDNF neurons in the dorsomedial hypothalamus (DMH) are activated by environmental cooling and sufficient to induce body temperature increases and brown adipose tissue (BAT) thermogenesis. Conversely, blocking these neurons impairs BAT thermogenesis and cold defense, causing body weight gain and glucose intolerance. DMH BDNF neurons are therefore an important type of thermoregulatory neuron, integrating thermal afferent signals to control EE during cold defense. This reveals a critical intersection between the BDNF circuitry and the thermoregulatory system.


Introduction
Genes encoding brain-derived neurotrophic factor (BDNF) and its receptor, tyrosine receptor kinase B (TrkB), have been linked to human obesity, as loss-of-function mutations in either gene cause severe obesity (Wen et al. 2012;Gray et al. 2006;Yeo et al. 2004). Studies in rodents have shown that BDNF and its neural circuitry in the hypothalamus control feeding, adaptive thermogenesis, and physical activity (Xu & Xie 2016; Wang et al. 2020;An et al. 2015;Unger et al. 2007). For example, genetic deletion of BDNF in the mouse ventromedial hypothalamus (VMH) and the dorsomedial hypothalamus (DMH) causes hyperphagic behavior and obesity (Unger et al. 2007). BDNF neurons within the anterior and posterior part of the PVN control feeding and thermogenesis, respectively . These data suggest that the BDNF circuitry is a crucial player in energy homeostasis. However, whether and how the BDNF circuitry integrates thermal afferent signals to rapidly and efficiently regulate energy expenditure (EE) is unknown. This is a critical gap in our understanding of how BDNF regulates obesity.  (Yang et al. 2020). Signals are integrated in the preoptic area (POA) and then POA neurons directly target neurons in the DMH or the raphe pallidus nucleus (RPa) to drive different aspects of thermal effector activities, such as brown adipose tissue (BAT) thermogenesis and vasodilation (Tan & Knight 2018;Morrison & Nakamura 2019). Blocking Pdyn neurons in the LPB in the warm afferent pathway increases BAT thermogenesis and EE and is protective against weight gain driven by a high-fat diet (Yang et al. 2020). Several POA neuron types can induce severe hypothermia and substantially reduce EE (Morrison & Nakamura 2019;Zhao et al. 2017;Yu et al. 2016). For example, activation of preoptic neurons that express the leptin receptor reduces core temperature (Tcore) to 30 o C and lowers EE by 75% (Yu et al. 2016). This also lowers food intake to reduce body weight (Yu et al. 2016), which may be explained by lower energy demand in the warm environment. However, whether and how the BDNF circuitry interacts with the thermoregulatory pathway has not been examined.
Here, we found that BDNF neurons in the DMH (DMH BDNF ) were quickly activated by ambient cold temperatures. Activation of these neurons was sufficient to increase body temperature, EE, and physical activity. Chronic blocking of these neurons impaired glucose metabolism and BAT thermogenesis, leading to weight gain on normal chow. Therefore, our data indicate that DMH BDNF neurons function as a rapid cooling sensor in the thermoregulatory pathway, quickly detecting ambient thermal inputs and then adjusting EE to meet the energy demand required to maintain core temperature (Tcore).  (Luo et al. 2019) to a Cre-dependent GFP reporter line (GFPL10). ~35% glutamate neurons and ~15% GABA neurons in the DMH were BDNF-positive, and most of BDNF neurons were glutamatergic (~60%) ( Figure   1A-C). BDNF neurons overlapped with those that expressed cFos following cold or warm exposure ( Figure 1D). Quantification of this overlap revealed that in response to cold, ~ 34.1% of cFos-positive neurons were BDNF neurons, whereas this overlap was only ~ 13.3% following warm exposure (Figure 1E-F). Expressed in a different way, 22.2% of the BDNF neurons in the DMH responded to cold, whereas only 4.7% responded to warmth (Figure 1E-F). These results indicate that DMH BDNF neurons responded preferentially to cold rather than warm temperature.

Calcium dynamics of DMH BDNF neurons to temperature stimuli
To study the speed and dynamics of neural responses to temperature, we injected Credependent AAVs expressing GCaMP6s into the DMH of BDNF-Cre mice (Figure 1H), and then used fiber photometry to record calcium activity of DMH BDNF neurons ( Figure   1G-H) following temperature manipulations. Floor cooling induced a sharp increase in calcium activity in DMH BDNF neurons (Figure 1I-J). Interestingly, these responses mainly occurred during the cooling phase, which has also been reported for sensory neurons and the spinal cord (Yarmolinsky et al. 2016;Wang et al. 2018;Ran et al. 2016). Once the temperature stopped changing, responses diminished to baseline. This is different from the overall activity of glutamatergic and GABAergic neurons, where there is a plateau of neural activity after temperature stopped changing (Zhao et al. 2017). In contrast, warming slightly inhibited the activity of DMH BDNF neurons ( Figure 1K). These data suggest that DMH BDNF neurons function as a rapid and sensitive detector for cooling, but not for static cold temperatures.
Activation of DMH BDNF neurons elevates core body temperature, energy expenditure, and physical activity while reducing RER The selective response to cooling predicts that these BDNF neurons may be necessary for cold-induced thermoregulation. To test this hypothesis, we used the DREADD-hM3Dq system to activate DMH BDNF neurons (Figure 2A). As expected, intraperitoneal injection of the hM3Dq agonist CNO (2.5 mg/kg) dramatically increased Tcore, EE, and physical activity (Figure 2B-D). The respiratory exchange ratio (RER) dropped significantly following CNO injection (Figure 2E), indicating that activation of DMH BDNF neurons increased fat fuel utilization. Daily injection of CNO did not alter food intake ( Figure 2F), suggesting that DMH BDNF neurons do not regulate feeding. These data indicate that DMH BDNF neurons are sufficient to induce hyperthermia, and elevate EE and physical activity, all of which are consistent with these neurons' response to cold temperature.

DMH BDNF neurons are required for BAT thermogenesis during cold defense
To test whether DMH BDNF neurons are necessary for thermoregulation and the regulation of EE, we used neural toxin TeNT to block their synaptic transmission ( Figure   3A). Consistent with DREADD activation studies, mice in which these neurons were chronically inhibited could not maintain Tcore in a cold environment than GFP controls ( Figure 3B). Chronic inhibition of DMH BDNF neurons also reduced EE, RER, and physical activity, both in baseline and cold temperatures ( Figure 3C-E), suggesting long-term effects on both basal metabolism and cold-induced thermogenesis. In contrast, Tcore, EE, RER, and physical activity were not significantly affected in these mice following exposure to warmth (compared to GFP controls) ( Figure 3F-I). This suggests that blocking DMH BDNF neurons selectively affected thermoregulation, EE, and physical activity in cold defense.
Based on these defects in cold defense, we hypothesized that BAT thermogenesis would also be affected, since it contributes up to 60% of cold adaptive thermogenesis (Abreu-Vieira et al. 2015). We first measured BAT temperature using a small telemetry probe embedded between the two interscapular fat pads. Following cold exposure, BAT temperature was significantly lower in mice lacking DMH BDNF activity compared with controls ( Figure 3J). Next, we measured changes in the major thermogenin, UCP1.
We found significant reductions in both UCP1 mRNA and protein ( Figure 3K-L), suggesting the BAT thermogenesis was impaired when DMH BDNF neurons were blocked.

Long-term blocking of DMH BDNF neurons increases body weight and impairs glucose metabolism
We reasoned that impaired BAT thermogenesis in response to long-term blocking of DMH BDNF neurons might impact body weight and glucose metabolism. Indeed, inhibiting these neurons resulted in increased body weight and fat mass compared with controls two months after viral injection (Figure 4A-B). This overweight phenotype did not result from increased food intake as the TeNT group and controls consumed the same amount of food one month after viral injection ( Figure 4C). We continued this line of analysis by measuring glucose clearance and insulin tolerance using the glucose tolerance test (GTT) and insulin tolerance test (ITT), respectively.
Intraperitoneal injection of glucose (2 g/kg) quickly increased blood glucose, reaching a peak within 30 minutes ( Figure 4D). Compared with controls, TeNT mice exhibited higher blood glucose and a larger area under curve, suggesting impairment in the ability to clear blood glucose after neural blocking ( Figure 4D). Whole-body insulin sensitivity was also significantly lowered ( Figure 4E). Taken together, these results show that long-term blocking of DMH BDNF neurons reduced BAT thermogenesis resulting in reduced EE and body weight gain.

Conclusion
These analyses reveal that DMH BDNF neurons function as cooling detectors in the thermoregulatory pathway to control body temperature, EE, BAT thermogenesis, and physical activity, with links to body weight control and glucose metabolism. Given that the human BDNF circuitry has been linked to the development of obesity (Wen et al.

2012; Gray et al. 2006; Yeo et al. 2004), our findings provide key mechanistic insights
toward understanding the physiology of human obesity.

Experimental animals
All experiments were performed on male adult mice housed under constant temperature (22-25 °C) in a light/dark cycle (light time, 9 PM to 9 AM) with ad libitum food and water. BDNF-IRES-Cre mice were made through the CRISPR/Cas9 genome targeting reported previously (Luo et al. 2019). EEF1A1-LSL.EGFPL10 (L10 is a ribosomal resident protein), were gifted by Dr. Jeff Friedman (Stanley et al. 2013). All procedures conducted on animals were conformed to institutional guidelines of ShanghaiTech University, Shanghai Biomodel Organism Co., the Animal Facility at the National Facility for Protein Science in Shanghai (NFPS), and governmental regulations.

Calcium fiber photometry
The optical fiber (Inper Inc.) was implanted 100 μm above the viral injection site. For temperature stimuli, the floor temperature was controlled by a Peltier controller (#5R7-001, Oven Industry) with customized Labview code (National Instrument) and was monitored using T-type thermocouples. Floor temperature was converted to a voltage signal and was simultaneously acquired with fluorescence signals in a fiber photometry system (Fscope, Biolinkoptics, China).
Data analysis referred to previous reports (Zhao et

Bodyweight and body composition analyses
Bodyweights were continuously measured for all animals before mice sacrificing. Body composition was determined using a Minispec whole-body composition analyzer (Burker Minispec CMR LF50).

Metabolic measurement
The energy expenditure, respiratory exchange ratio, body temperature, and locomotor activity were monitored by the Comprehensive Lab Animal Monitoring System with Temperature Telemetry Transmitter (CLAMS; G2 E-Mitter). Temperature transponders were implanted into the abdominal cavity 3-5 days before testing. Mice were adapted in the metabolic chambers for two days before giving temperature stimuli or CNO (i.p., 2.5 mg/kg body weight, ENZO, #BML-NS105-0025).

Cold tolerance test (CTT)
TeNT (as to their controls) injected mice were exposed to a 4-h cold exposure and recording interscapular brown adipose tissue temperature simultaneously to assess BAT thermogenesis capacity, which is defined by cold tolerance test (CTT) in our study.
The IBAT temperature was measured three times for average using subcutaneous implanted thermal probes (IPTT-300, Biomedic Data Systems). The probes were implanted at the midline in the interscapular region at least one week before testing.

Statistical analyses
Excel and GraphPad Prism 8 were used to plot data and calculate statistical significance.