Hypothalamic gene transfer of BDNF promotes healthy aging in mice

Abstract The aging process and age‐related diseases all involve perturbed energy adaption and impaired ability to cope with adversity. Brain‐derived neurotrophic factor (BDNF) in the hypothalamus plays important role in regulation of energy balance. Our previous studies show that recombinant adeno‐associated virus (AAV)‐mediated hypothalamic BDNF gene transfer alleviates obesity, diabetes, and metabolic syndromes in both diet‐induced and genetic models. Here we examined the efficacy and safety of a built‐in autoregulatory system to control transgene BDNF expression mimicking the body's natural feedback systems in middle‐aged mice. Twelve‐month‐old mice were treated with either autoregulatory BDNF vector or yellow fluorescence protein (YFP) control, maintained on normal diet, and monitored for 28 weeks. BDNF gene transfer prevented the development of aging‐associated metabolic declines characterized by: preventing aging‐associated weight gain, reducing adiposity, reversing the decline of brown fat activity, increasing adiponectin while reducing leptin and insulin in circulation, improving glucose tolerance, increasing energy expenditure, alleviating hepatic steatosis, and suppressing inflammatory genes in the hypothalamus and adipose tissues. Moreover, BDNF treatment reduced anxiety‐like and depression‐like behaviors. These safety and efficacy data provide evidence that hypothalamic BDNF is a target for promoting healthy aging.

stress response, and cardiovascular function (Mattson, Maudsley, & Martin, 2004). BDNF has diverse functions in brain development and plasticity (Lu, Pang, & Woo, 2005). It is neuroprotective in many different brain areas against dysfunctions and insults (Lindsay, 1994). In addition, BDNF is an important component of the hypothalamic pathway that controls energy homeostasis (Xu & Xie, 2016). We have previously demonstrated that an enriched environment (EE) improves brain function and the body's overall state of health. Our mechanistic studies lead to the characterization of a novel brain-fat axis-the hypothalamic-sympathoneural-adipocyte (HSA) axis, and the development of molecular therapy for obesity, diabetes, and cancer. The complex stimuli (physical, social, and cognitive) provided by EE induce hypothalamic BDNF and elevate sympathetic tone to the adipose tissue. Increased sympathetic tone remodels the adipose tissue, inducing browning of white fat and suppression of leptin, leading to an anti-obesity and anticancer phenotype (Cao & During, 2012;Cao et al., 2011Cao et al., , 2009Cao et al., , 2010. Furthermore, hypothalamic BDNF modulates secondary lymphoid tissues (spleen and lymph nodes) and enhances CD8 T cell immunity, contributing to the anticancer effects of EE (Xiao et al., 2016). A reduction in BDNF signaling has been documented during normal aging and decreased BDNF levels are associated with vulnerable neuronal populations in several neurodegenerative disorders including Alzheimer's, Parkinson's and Huntington's diseases, demonstrating the need for further therapeutic research on components of the BDNF signaling pathway (Tapia-Arancibia, Aliaga, Silhol, & Arancibia, 2008). Some physiologic or pathologic age-related changes in the CNS could be offset by the administration of exogenous BDNF and/or by stimulation of its receptor expression (Tapia-Arancibia et al., 2008). In addition, BDNF signaling in the brain is thought to mediate at least some of the anti-aging effects of an intermittent fasting regiment (Duan et al., 2003;Lee, Duan, & Mattson, 2002) although data on the hypothalamus are not reported. Moreover, it is unclear how BDNF signaling in neurons is transferred to the periphery to improve the healthspan of many different organ systems. Our characterization of the HSA axis and the critical role of BDNF in this brain-fat axis suggest a mechanism whereby hypothalamic BDNF, highly responsive to environmental stimuli, controls the HSA axis activity and thereby influences body composition, metabolism, immune function, and cancer via its preferential regulation of the phenotype and functions of adipose tissue. Here we investigated the long-term effects of hypothalamic gene transfer of BDNF in middle-aged mice using an autoregulatory rAAV vector. The single rAAV vector harbors two cassettes, one expresses human BDNF driven by a constitutive promoter, the other expresses a microRNA targeting BDNF under the control of agouti-related peptide (AGRP) promoter that is activated by weight loss and fat depletion. This dual-cassette vector mimics the body's natural feedback system to achieve autoregulation of the transgene and its efficacy has been examined in genetic models of obesity and diabetes such as db/db mice (Cao et al., 2009) and melanocortin-4 receptor (MC4R) deficient mice (Siu et al., 2017).

| Short-term hypothalamic gene transfer of autoregulatory BDNF vector
To test the efficacy of hypothalamic BDNF gene transfer in middleaged mice, we performed a short-term study. The autoregulatory dual-cassette construct expressing the human BDNF (autoBDNF) or destabilized yellow fluorescent protein (YFP) control were packaged into serotype 1 AAV capsids (Cao et al., 2009; Figure 1a). Tenmonth-old female C57BL/6 mice were randomized into the two treatment groups, to receive bilateral hypothalamic injections of either AAV-autoBDNF or AAV-YFP into the arcuate (ARC)/ventralmedial (VMH) nuclei of the hypothalamus (Figure 1f). BDNF-treated mice showed lower body weight compared to YFP mice (Figure 1b).
A glucose tolerance test was performed 57 days post-AAV injection and BDNF-treated mice displayed improved glycemic control (Figure 1c,d). Mice were sacrificed 63 days post-AAV injection and the intrascapular brown adipose tissue (BAT) and various white adipose tissue (WAT) including inguinal (iWAT), gonadal (gWAT), and retroperitoneal (rWAT) depots were dissected. BDNF treatment reduced adiposity with intra-abdominal fat depots displaying the most reduction ( Figure 1e). Profiling of serum biomarkers at the sacrifice showed a significantly higher adiponectin level and a strong trend toward lower leptin in BDNF mice (Figure 1f). YFP fluorescence confirmed that transgene expression was mainly in ARC/VMH nuclei ( Figure 1g). The BDNF protein levels in the hypothalamus block dissections were measured using ELISA. The autoBDNF mice showed eightfold higher hypothalamic BDNF level than YFP mice ( Figure 1h).

| Systemic metabolic effects of long-term hypothalamic gene transfer of BDNF
Next, we conducted a long-term study to assess the effects of hypothalamic BDNF overexpression on normal aging and more comprehensively characterize the metabolic and behavioral implications ( Figure 2a). Twelve-month-old female C57BL/6 mice were randomized to receive AAV-autoBDNF or YFP and monitored for 7 months.
YFP-treated mice gradually gained weight. In contrast, BDNF treatment completely prevented aging-related weight gain ( Figure 2b).
Moreover, autoBDNF-treated mice maintained stable body weight throughout the 7-month duration of the study (Figure 2b). Food intake was monitored between week 3 and 10 postsurgery. The absolute food intake of BDNF-treated mice was lower than YFP mice while the relative consumption calibrated to body weight was not different (Figure 2c). Rectal temperature measured at 12-weeks post injection revealed no significant differences between the two groups ( Figure 2d). At 13-weeks postsurgery, BDNF-treated mice performed better in a glucose tolerance test (Figure 2e,f).
To assess energy expenditure, mice were subjected to indirect calorimetry beginning 20 weeks post-AAV injection over a 24-hr period after habituation. Oxygen consumption in BDNF-treated mice Area under the curve (AUC) of (f). (f) Glucose tolerance test at 13 weeks post injection. n = 8-9 for YFP, n = 9-10 for autoBDNF (b) to (f). (g) CLAMS assessment at 20 weeks post injection. Oxygen consumption, respiratory exchange ratio (RER), and physical activity in a 24-hr period; Shaded area, dark phase. n = 6 per group. Error bars represent mean ± SEM. *p < 0.05. **p < 0.01 was significantly increased in both dark and light phases compared to YFP mice ( Figure 2g). The respiratory exchange ratio (RER) was slightly increased in the BDNF group during the dark phase but was not statistically significant. (Figure 2g). Surprisingly, physical activity was significantly decreased in BDNF-treated mice ( Figure 2g). Food intake was not different during this period in the metabolic chambers (data not shown). The higher oxygen consumption concurrent with lower physical activity indicated that BDNF treatment elevated the resting metabolic rate.

| Behavioral assessments of long-term BDNF gene transfer
In addition to assessing metabolic efficacy, we were interested in the long-term safety of this approach in aged mice on normal diet.
Thus, we performed various assays to screen for changes in anxietyor depression-like behavior from 15-17 weeks post-AAV injection. In an anxiety behavior test, cold-induced defecation (CID; Barone et al., 2008), BDNF mice showed significant reduction in fecal boli compared to YFP mice ( Figure 3a). The novelty-suppressed feeding (NSF) test assesses hyponeophagia, in which exposure to a novel environment suppresses feeding behavior (Samuels & Hen, 2011). NSF has been used to study anxiety-and depression-related behaviors because it is sensitive to anxiolytic and chronic antidepressant treatments. BDNF treatment shortened the latency to feed (Figure 3b).
Two assays for depression were used, the tail suspension (TST) and forced swim tests (FSTs). For the TST test, the time being immobile was significantly diminished in 5 min of the 6-min test for BDNFtreated mice ( Figure 3d). The total amount of immobile time was also significantly reduced in the BDNF mice ( Figure 3e). The FST is one of the most commonly used rodent behavioral tests for screening antidepressant drugs (Cryan & Mombereau, 2004). The time being immobile was significantly decreased in the last 2 min of the 6-min test for BDNF-treated mice ( Figure 3f). The total immobile time of the FST revealed a trend toward lower immobility in BDNF mice ( Figure 3g).

| BDNF treatment promotes a lean phenotype
Mice were sacrificed 194 days post-AAV injection. Significant reductions in absolute mass were observed across all fat depots in the BDNF-treated mice. BDNF treatment decreased adiposity: fat mass, relative to body weight, by 68% for iWAT, 79% for rWAT and gWAT ( Figure 4a). Although the absolute weight of liver was lower in BDNF group, the relative liver mass when normalized to body weight was significantly increased compared to YFP mice ( Figure 4b).
The lean phenotype of BDNF-treated mice was associated with a serum biomarker profile featured as sharp drop of leptin and insulin, and rise of adiponectin and IGF-1 (Figure 4c).

| Adipose remodeling
Aging is associated with a decline in BAT activity (Enerback, 2010).
The BATs of 19 months old YFP mice appeared pale whereas the BAT in BDNF mice was darker. H&E staining revealed the BAT of BDNF mice maintained typical BAT morphology of younger mice and was devoid of white adipocyte infiltration often associated with aging ( Figure 5a). The morphological changes of BAT were associated with robust regulation of BAT gene expression (Figure 5b). Leptin expression was reduced by over 90% while adiponectin expression was upregulated. Insulin receptor expression was also significantly upregulated whereas glucose transporter type 4 (Glut4), the major type of glucose transporter in adipose tissue, was not different between the two groups ( Figure 5b). Both the lipolytic gene Hsl (encoding hormone-sensitive lipase) and the lipogenic gene Srebp1c (encoding sterol regulatory element-binding protein 1c) were upregulated in BDNF mice. BAT dissipates energy via releasing chemical energy from mitochondria in the form of heat. This process is primarily mediated by uncoupling protein-1 (UCP1) that is a specific BAT marker (Enerback et al., 1997). UCP1 was significantly upregulated by BDNF treatment suggesting the preservation of proper BAT functions against aging-related loss (Figure 5a,b, Supporting Information Figure S1). The transcriptional coactivator peroxisome proliferator-activated receptor γ coactivator 1-α (PGC-1α) switches cells from energy storage to energy expenditure by inducing mitochondrial biogenesis and genes involved in thermogenesis (Puigserver et al., 1998). Leptin expression was decreased by 70% in both depots. Inflammation-modulatory genes Il1b, Il6 and Saa3 (encoding serum amyloid 3A) were highly suppressed in gWAT adipocyte (Figure 5e). Histology and quantification showed that the size of white adipocyte in BDNF mice was much smaller than that in YFP mice (Figure 5a, Supporting Information Figure S2).

| BDNF treatment inhibits liver steatosis
Aging is associated with hepatic steatosis (Sheedfar, Biase, Koonen, & Vinciguerra, 2013 Figure 6c). Limited changes were found in skeletal muscle (Figure 6d).   (Wang, Wang, Clark, & Sferra, 2003) and therefore BDNF was primarily overexpressed in neurons in this study. However, the current study is unable to distinguish between the autocrine and paracrine effects of BDNF since BDNF can be secreted from transduced neurons. We are currently using AAV1 to deliver a dominant-negative TrkB receptor (Cao et al., 2011)  Meek and colleagues report that infusion of BDNF to either the lateral cerebral ventricle or the VMH attenuates diabetic hyperglycemia via an insulin-independent inhibition of hepatic glucose production (Meek et al., 2013). Future studies will elucidate whether a hypothalamic BDNF-liver axis directly modulates liver glucose and lipid metabolism and thereby contributes to glycemic control in aged animals.

| DISCUSSION
With respect to the effects on behavior, most studies associate a reduction in BDNF with cognitive deficits. Postnatal knockout of Bdnf leads to increased anxiety along with obesity (Rios et al., 2001), while forebrain-specific deletion results in impaired spatial learning and certain discrimination tasks (Gorski, Balogh, Wehner, & Jones, 2003). Moreover, low serum levels of BDNF are correlated with depression in human patients (Karege et al., 2005;Shimizu et al., 2003). However, scarce evidence is available regarding the role of hypothalamic BDNF in emotionality. To our knowledge, this study is the first assessing hypothalamic BDNF-induced behavioral adaptations in aging. We examined a battery of anxiety and depression behavior tests and showed that hypothalamic BDNF treatment significantly reduced anxiety-and depression-like behaviors. Transgene was expressed mainly in the ARC and VMH nuclei of the hypothalamus in this study, which might increase the BDNF protein level in the adjacent dorsomedial hypothalamus (DMH). The DMH is a brain area not only involved in physiological functions such as metabolism and environmental threats but is also critically involved in behavioral regulation, particularly fear, anxiety, and panic-like disorders (Shekhar, Sims, & Bowsher, 1993;Silva et al., 2014). Recently, it was reported that loss of corticotropin-releasing hormone (Crh) in the paraventricular hypothalamus (PVH) results in reduced anxiety behaviors (Zhang et al., 2016). Future research is required to elucidate the role of BDNF in these specific hypothalamic nuclei regarding anxiety. Alternatively, we have reported that hypothalamic BDNF modulates the hypothalamic-pituitary-adrenal (HPA) axis partially mediating the EE's regulation of T cell immunity (Xiao et al., 2016). It will be interesting to investigate whether hypothalamic BDNF's modulation of the HPA axis contributes to the anti-depression effect in the aged animals whose stress response system becomes less agile and dysfunctional. Although no change in BDNF expression was found in either the hippocampus or the amygdala, it is still possible that global improvement of metabolism induced by hypothalamic BDNF overexpression indirectly influences other limbic structures, including the prefrontal cortex, hippocampus, nucleus accumbens, ventral striatum, amygdala, and hypothalamus (Russo, Murrough, Han, Charney, & Nestler, 2012) and thereby affects brain functions and behaviors (de Noronha et al., 2016).
In conclusion, hypothalamic BDNF gene transfer with an autoregulatory AAV vector prevents aging-related weight gain, reduces adiposity, increases energy expenditure, improves glycemic control, alleviates liver steatosis, suppresses inflammatory genes in the hypothalamus and adipose tissues, and decreases anxiety-and depression-like behaviors. This long-term study provides efficacy and safety evidence targeting hypothalamic BDNF for healthy aging.

| Animals
National Institute on Aging, Aged Rodent Colonies, provided female C57Bl/6 mice, 12 months of age. Mouse litters were group housed (no more than five per cage) in a 12:12 light:dark cycle with ad libitum access to standard rodent chow and water in a humidity-and temperature-controlled environment. All animal experiments were carried out in compliance and conform to the regulatory standards of the Ohio State University Institutional Animal Care and Use Committee.

| rAAV vector constructs
The

| Stereotaxic surgery
The 10 or 12-month-old female C57Bl/6 mice were randomly assigned to receive AAV-autoBDNF or AAV-YFP. Mice were anaesthetized with a single intraperitoneal dose of ketamine/xylazine (100 mg/kg and 20 mg/kg) and secured via ear bars and incisor bar on a Kopf stereotaxic frame (Tujunga, CA). A single midline incision was made through the scalp to expose the skull and two small holes were made with a dental drill above the injection sites.

| Food intake and body weight
Following surgeries, body weight and food intake-on normal chow diet-were recorded every 5-7 days. Animals injected with the same vector remained housed together postsurgery. Food intake was averaged per mouse per week in each cage. Mice were monitored up to 63 days postsurgery for short-term study while 194 days post injection for long-term study.

| Body temperature
Rectal temperature was measured at 2 p.m. for all mice after 5 min of sedation with 2.5% isoflurane. The Physitemp BAT −12 rectal thermometer (Clifton, NJ) remained in place for 30 s to allow temperature to stabilize before being recorded. Mice were then returned to their home cages to recover.

| Serum harvest and analysis
Truncal blood was collected at 10 a.m. following decapitation at sacrifice. Serum was allowed to clot on ice for at least 30 min before MCMURPHY ET AL.

| Hepatic triglyceride measurement
Lipid was extracted from liver by chloroform/methanol (

| BDNF ELISA
Hypothalamic block dissections were homogenized in ice-cold Pierce RIPA buffer containing Calbiochem protease inhibitor cocktail III (San Diego, CA). The homogenates were centrifuged and the protein content was measured using BCA kit (Pierce Biotechnology, Rockford, IL). BDNF protein levels were measured using R&D DuoSet ELISA Development Systems (Minneapolis, MN).

| Statistical analysis
Data are expressed as mean ± SEM. We used Prism Mac version 6.0f software (GraphPad, La Jolla, CA) and SPSS Statistics v24.0.0.0 (IBM, Armonk, NY) to analyze the following: student's t test for body weight or food intake at single time points, adiposity, body temperature, organ weights, serum ELISAs, behavior, and quantitative RT-PCR data. Mixed analysis of variance was performed on time course measurements (body weight, VO 2 , RER, physical activity, GTT).

ACKNOWLEDG MENTS
This work was supported by NIH grants CA163640, CA166590, CA178227, and AG041250 to LC. L.C. is an inventor of US patent 9,265,843 B2 on the autoregulatory BDNF vector. All the other authors declare no conflict of financial interests.

AUTHOR CONTRIBU TI ONS
T.M, W.H., X.L., J.J.S., N.J.Q, and R.X.: carried out the research and interpreted the results. L.C.: conceived the concept, designed the studies, interpreted the results, and wrote the manuscript. All authors approved the manuscript.